hESC-derived thalamic organoids form reciprocal projections when fused with cortical organoids

SUMMARY

Human brain organoid techniques have rapidly advanced to facilitate investigating human brain development and diseases. These efforts have largely focused on generating telencephalon due to its direct relevance in a variety of forebrain disorders. Despite its importance as a relay hub between cortex and peripheral tissues, the investigation of three-dimensional (3D) organoid models for the human thalamus has not yet been explored. Here, we describe a method to differentiate human embryonic stem cells (hESCs) to thalamic organoids (hThOs) that specifically recapitulate the development of thalamus. Single-cell RNA sequencing revealed a formation of distinct thalamic lineages, which diverge from telencephalic fate. Importantly, we developed a 3D system to create the reciprocal projections between thalamus and cortex by fusing the two distinct region-specific organoids representing the developing thalamus or cortex. Our study provides a platform for understanding human thalamic development and modeling circuit organizations and related disorders in the brain.

eTOC Blurb

Xiang and colleagues report a method for generating human thalamus-like brain organoids (hThOs) that recapitulate the development of human thalamus. By fusing hThOs and corticallike brain organoids (hCOs), they establish a 3D system in a dish to create the reciprocal projections between thalamus and cortex.

Graphical Abstract

INTRODUCTION

Brain organoids has become an important experimental avenue to investigate human brain development and neurological disorders (Clevers, 2016; Lancaster and Knoblich, 2014). The generation of region-specific brain organoids (Jo et al., 2016; Muguruma et al., 2015; Qian et al., 2016; Sakaguchi et al., 2015) further facilitates modeling the defined regions of the brain. More recently, tangential migration of cortical interneurons was recapitulated in vitro by fusing the organoids resembling the cortex (hCO) and MGE/subpallium (hMGEO) of the brain to allow a functional integration (Bagley et al., 2017; Birey et al., 2017; Xiang et al., 2017). This approach demonstrates the importance of brain organoids as a model system to investigate the complex interaction between specific brain regions in a three-dimensional (3D) in vitro culture.

In a developing brain, extensive thalamocortical (TC) and corticothalamic (CT) axon projections occur between the cortex and thalamus, and are critically involved in sensory-motor processing, attention, and arousal (Lopez-Bendito and Molnar, 2003; Sherman and Guillery, 1996; Steriade et al., 1993). Nevertheless, there has been a lack of methods to create TC and CT connections in vitro except for a few organotypic culture models that are limited to rodents (Yamamoto et al., 1989; Yamamoto et al., 1992). Neither the generation of human thalamus-like organoids, nor a method for modeling human thalamocortical connections using brain organoids, has been reported.

Here, we developed a method for differentiating human embryonic stem cells (hESCs) into thalamus-like brain organoids (hThOs). We dissected a variety of cells arising during hThO development by single-cell transcriptome. Importantly, we established a 3D model to recapitulate the reciprocal thalamocortical projections between human thalamus and cortex by fusing hThOs with hCOs to form human fused thalamus-cortex organoids (hThCOs).

RESULTS

Generation of hThOs from hESCs

The generation of hThOs was based on a static-to-spinning culture strategy (Xiang et al., 2017) (Figure 1A). hESCs were dissociated into single cells to facilitate uniform formation of embryoid bodies (EBs). Dual SMAD inhibition was performed to drive the early neuroectoderm fate (Chambers et al., 2009). In a developing brain, the thalamus is generated from the caudal region of forebrain, i.e. the diencephalon (Martinez et al., 2012) (Figure 1B), and insulin is known as a caudalization factor (Muguruma et al., 2010; Shiraishi et al., 2017; Wataya et al., 2008). Thus, we supplemented hThOs with human insulin during dual SMAD inhibition period for caudalization. After neural induction, MEK/ERK signaling was blocked by PD0325901 treatment to prevent an excess caudalization towards a midbrain cell fate (Shiraishi et al., 2017). Concomitantly, human BMP7 was supplemented as it is expressed in the developing thalamus and adding BMP7 promotes thalamic differentiation in a rodent model (Shiraishi et al., 2017; Suzuki-Hirano et al., 2011). We referred to the period of cooperative treatment with MEK/ERK inhibition and BMP7 activation as a thalamic patterning period. Finally, patterned brain organoids were subjected to further neural differentiation and maturation.

Figure 1. Generation of Region-specific Human Brain Organoids.

(A) Schematic view of the methods for generating hThOs, hMGEOs, and hCOs.

(B) Schematic view of expression patterns of regional markers during thalamus, cortex, and MGE development.

(C) qPCR analysis for expressions of regional markers in developing hThOs, hMGEOs, and hCOs. Each data represents expressions in pooled batch of 3 to 4 organoids, and 3 batches were collected for analysis. Mean ± SD is shown. *p<0.05, **p<0.01, ***p<0.001.

(D) Immunostaining for MAP2 and thalamic marker TCF7L2 in day 41 hThO, hCO, and hMGEO. The scale bar represents 250 μm.

(E) Immunostaining for thalamic and cortical progenitor marker PAX6, and cortical marker TBR1 in day 41 hThO, hCO, and hMGEO. The scale bar represents 250 μm.

See also Figure S1.

The thalamic fate was defined by a combination of markers specifying the rostral-caudal axis and the thalamic primordium (Scholpp and Lumsden, 2010; Shiraishi et al., 2017) (Figure 1B). qPCR analysis of various regionally specified brain organoids revealed that expression of the caudal forebrain marker OTX2 was significantly higher in hThOs than in hCOs and hMGEOs at both earlier (day 18) and later (day 41) developmental stages (Figure 1C). Similarly, expression of the ventral thalamic marker DBX1 significantly increased in hThOs. GBX2, a marker for the marginal zone of thalamus, was significantly expressed in hThOs only after longer culture (day 41) (Figure 1B and 1C) consistent with the late onset of GBX2 induction in vivo (Scholpp and Lumsden, 2010). Both hThOs and hCOs exhibited a robust PAX6 expression during differentiation, consistent with the expression of PAX6 in developing thalamus as well as cortex of human brain (Hansen et al., 2013; Ma et al., 2013), (Figure 1C). In contrast, the MGE-specific transcriptional factors NKX2-1 and DLX2 were only enriched in hMGEOs, and the vesicular glutamate transporter 1 (VGLUT1), a predominant VGLUT in the cortex, significantly expressed only in hCOs (Figure 1C and Figure S1A). Furthermore, immunostaining demonstrated a widespread expression of the broad thalamus-specific marker TCF7L2 (Nagalski et al., 2016; Shiraishi et al., 2017) (Figure 1B) only in hThO, although hThO, hCO, and hMGEOs all showed expression of MAP2, a general marker for neuronal differentiation (Figure 1D). Immunostaining also demonstrated that PAX6⁺ progenitors were enriched in both hThOs and hCOs, while TBR1⁺ cortical neurons were only enriched in hCOs (Figure 1E). Whereas hMGEOs barely produced PAX6⁺ cells, they were enriched with NKX2-1⁺ cells (Figure S1B). Together, these results demonstrate that unlike the hCOs and hMGEOs that resemble the developing cortex and MGE, respectively, hThOs specifically resemble the developing thalamus.

Single-cell Map of Developing hThO

To examine a lineage specification during hThO development, we profiled single cell transcriptome of total 11,277 cells derived from hThOs at two different developmental time points (day 34 and day 89) (Figure 2A). A total of 17 clusters were detected from hThO samples, which were further categorized into 10 cell types, including mature neurons (Mneu, Figure S1C–S1G). We noted that our scRNA-seq library had a limited doublet rate and UMI bias among clusters (Figure S1F and S1H). Gene Set Enrichment Analysis (GSEA) validated the presence of neurons, neural progenitor cells (NPCs) and astrocytes (AS), whereas gene signatures of oligodendrocytes (OL) were limited in hThOs (Figure S1E) (Darmanis et al., 2015). In hThO-derived cells, we detected strong expression of thalamic markers (e.g. OTX2, TCF7L2, GBX2 and PAX6), but not MGE (e.g. DLX1 and NKX2-1) or cortical makers (e.g. TBR1 and BCL11B) (Figure 2B). One glia-like cluster (proteoglycan-expressing glia [PGG]) uniquely expressed small extracellular proteoglycans (BGN, biglycan and DCN, decorin) (Figure 2A and 2B). In spite of limited studies of BGN and DCN in brain function, several central nervous system (CNS)-specific proteoglycans are involved in neuronal migration, synaptogenesis and neuronal disorders (Cui et al., 2013). Despite high expression of neuroepithelial markers (Figure S1D), seven clusters (Inter 1–2, cilia bearing cell [CBC], BMP-related cells [BRC], and unfolded protein response-related cells 1–3 [UPRC 1–3]) showed relatively low expression of mature neuronal and glia markers. The CBC showed a high expression of genes involved in cilium development, which is essential for brain development and function (Guemez-Gamboa et al., 2014). The three UPRC clusters highly expressed genes related to UPR mediated by endoplasmic reticulum (ER) (Figure 2A and 2B, Figure S1F), which is known to be essential for brain development, synaptic plasticity and memory formation (Martinez et al., 2018). Although ER stress is also elicited by cellular damage (e.g. dissociation for single cell suspension), we did not detect enrichment of apoptotic genes and mitochondria-derived reads at UPRCs compared to other clusters (Figure S1F), suggesting that these cells are not dead or damaged cells, but rather, they represent functioning cell clusters. Comparative analysis with single-cell transcriptome of human thalamus suggests that these clusters (UPRC, CBC, BRC, and PGG) are also generated from thalamus in vivo (Figure S1I), albeit the fact that single-cell transcriptomic dynamics for developing human thalamus is still understudied. Further, while on day 34 over half of hThO-derived cells were at non-committed or partially committed states (e.g. NPC and immature neuron), a majority of cells in hThOs became fate-committed after longer development (day 89) (Figure 2C). Overall, our results revealed the progressive production of typical thalamic cell types and the previously under-characterized cell types in developing hThOs.

Figure 2. Single-cell Analysis of Region-specific Brain Organoids.

(A) tSNE plot of hThO-derived single cells distinguished by annotations (left) or time point (right). Early stage: day 34; late stage: day 89.

(B) Expression pattern of cell-type and brain-region specific markers in hThO-derived cells. Relative expression level is plotted from gray to red colors.

(C) Ratio of hThO-derived cells clustered into each annotation.

(D) Ratio of excitatory and inhibitory neurons late stage brain organoids.

(E) Differential expression analysis between different brain organoids. Representative genes are colored by red (hCO), blue (hMGEO) and purple (hThO).

(F) GSEA of region-specific genes in different brain organoids. Normalized enrichment scores (NES) are plotted from blue to green colors. M1C/S1C: primary moter-sensory cortex; PCX: parietal neocortex; OCX: occipital neocortex; URL: upper rhombic lip; DFC: dorsolateral prefrontal cortex; MFC: anterior cingulate cortex; OFC: orbital frontal cortex; AMY: amygdaloid complex; LGE: lateral ganglionic eminence; CGE: candal ganglionic eminence; MGE: medial ganglionic eminence; HIP: hippocampus; DTH: dorsal thalamus.

(G) tSNE plot of cells derived from hThOs, hCOs and hMGEOs. Early and late time point data are shown at top and bottom panel, respectively.

(H and I) 3D diffusion map of NPCs and neurons from different organoids. Branches (left) and origin of cells (right) are shown by different colors. Note hThO-derived neurons were uniquely clustered in branch NB5.

(J) GO enrichment of each branch of neurons derived from different organoids. Branch identities are the same as shown in (I).

(K) Enrichment of disease-related genes in each branch. Enrichment and depletion in each branch are shown by −log10(FDR) and log10(FDR), respectively. Branch identities are the same as shown in (I).

See also Figure S1.

To further understand lineage commitment in hThOs in relation to other region-specific brain organoids, we compared scRNA-seq data from hThOs with those from hMGEOs and hCOs (Xiang et al., 2017). hThO-derived neurons were almost uniformly glutamatergic, as represented by the presence of vGLUT2 and EAAT4, whereas the GABAergic, dopaminergic, serotonergic and cholinergic neurons were rarely detected (Figure 2D and Figure S2F). This is consistent with a recent finding that the majority of neurons from a mouse thalamic domain were excitatory and vGLUT2-positive (Kalish et al., 2018). hCOs produced both glutamatergic excitatory neurons and GABAergic inhibitory neurons. In contrast, hMGEOs predominantly produced GABAergic inhibitory neurons (Figure 2D). Next, we identified the differentially expressed genes among hThOs, hMGEOs, and hCOs (Figure 2E). Thalamic markers were significantly enriched in hThOs, whereas markers specifying the cortex or MGE regions were depleted. For instance, sonic hedgehog (SHH) and NKX2-1, which control the patterning of ventral telencephalon, were uniquely expressed in hMGEOs. Furthermore, comparative analysis between region-specific transcriptomes from human fetal brain and those from different brain organoid demonstrated that hThO transcriptome is the closest to that of the dorsal thalamus (DTH) (NES=4.04, FDR<0.001) compared to other brain domains (Figure 2F). In contrast, hCO and hMGEO transciptomes displayed highest similarities with parietal neocortex (PCX) (NES=3.43, FDR<0.001) and MGE (NES=3.80, FDR<0.001) domains, respectively (Figure 2F).

To clarify the potential divergence of cell fate decisions among different brain regions, we assigned cell types from hCOs and hMGEOs with those from the closest cells from hThOs (Figure 2G). Similarly with hThOs, substantial number of BRCs, CBCs and PGGs were generated from hCOs, but not from hMGEOs. We commonly observed UPRCs from three types of organoids, indicating that these cell types are not unique to hThOs. For further dissection, we constructed the developmental trajectory using NPCs and neurons from three types of organoids by drawing diffusion maps with a k-nearest neighbor graph (Figure 2H and 2I) (Haghverdi et al., 2015). From NPCs, the diffusion maps yielded only two branches. Approximately 30% of hThO-derived NPCs were categorized into a unique branch (PB2), with the remaining not distinguishable from those of hCOs and hMGEOs. PB2 branches displayed expression of developmental genes and negative regulators of cell cycle, indicating that these NPCs were more fate-committed compared to others. In contrast, neuronal diffusion maps clearly separated each neuronal type into distinct branches (Figure 2I and S1K). NB1 was uniquely composed of hMGEO-derived neurons and highly expressed vGAT GABAergic transporters (Figure 2I and S1L). hCO-derived neurons were separated into two branches, NB3 and NB4, which were characterized by excitatory and inhibitory neurons, respectively. hThO-derived neurons were mainly classified into NB5 branch. Differential expression analysis revealed that hThO-derived neurons express distinct glutamatergic transporter (vGLUT2) compared to hCOs (vGLUT1) (Figure S1L). In addition, several neurodevelopmental genes (CFL1 and SNCA) and transcription factor (ZIC1) were preferentially expressed in hThO-derived excitatory neurons. NB2 contained the mixture of hThO- and hCO-derived neurons and were enriched with metabolic genes compared to other clusters, suggesting that cells in this branch were still at immature state (Figure 2J). Overall, the trajectory analysis enabled a clear separation of thalamic neurons generated in hThOs from neurons produced in hCOs and hMGEOs.

We further investigated the expression of disease-related genes in each branch. Notably, depressive disorder-related genes and schizophrenia-related genes were enriched in thalamic excitatory neurons (Figure 2K), consistent with the abnormal thalamic activity and compromised connection between thalamus and cortex in major depressive disorder patients and schizophrenia patients (Andreasen, 1997; Brown et al., 2017). These results underscore the importance of cell-type-specific impairments in distinct neuronal pathogenesis.

Fusion of hThOs and hCOs Recapitulates Reciprocal Thalamocortical Projections

Reciprocal connections between thalamic nuclei and the cortex arise during development (Figure 3A). To test whether hThOs and hCOs were able to project axons like their in vivo counterparts, we first plated brain organoids onto tissue culture dishes. To facilitate visualization of axon growth, we labeled hESC line with ubiquitous mCherry or GFP using the AAVS1 locus (Figure S2A and S2B). Interestingly, we found that while hThOs and hCOs plated on Poly-D-Lysine-coated dishes did not extend axons (data not shown), axon projections were observed only when plating hThOs or hCOs on a layer of dissociated hCO cells as feeders (Figure S2C and S2D), or on mouse astrocytes as feeders (data not shown). Thus, non-cellautonomous guidance may be required for hThOs and hCOs to make efficient axon projections, and a direct contact between hThO and hCO may favor axon projection.

Figure 3. hThCOs Model Connections Between Thalamus and Cortex.

(A) Schematic view of axon connections between thalamus and cortex in the brain.

(B) Schematic view showing the generation of hThCOs using mCherry⁺ hThO and hCO without fluorescence reporter.

(C) Epifluorescence images showing TC projections in an intact hThCO. Arrows indicate axon bundles. The scale bar represents 250 μm.

(D) 3D confocal imaging revealing TC projections in an intact hThCO near the border of fusion (left panel), and on the hCO side (right panel). Arrows indicate axon bundles. Regions a and b in the right panel are enlarged, and axonal boutons are indicated by arrow heads. The scale bar represents 50 μm.

(E) Schematic view showing the generation of hThCOs using mCherry⁺ hThO and GFP⁺ hCO.

(F) Epifluorescence images showing reciprocal TC and CT projections in an intact hThCO generated from mCherry⁺ hThO and GFP⁺ hCO. TC projections are shown in the left panel, and CT projections are shown in the right panel. Arrows indicate axon bundles. The scale bar represents 250 μm.

(G and H) 3D confocal imaging revealing TC (G) and CT (H) projections in an intact hThCO generated from mCherry⁺ hThO and GFP⁺ hCO. Arrow heads indicate growth cones. The scale bar represents 50 μm.

(I) Schematic view of method for quantifying axon targeting.

(J and K) Quantifications of range index for TC (J) and CT (K) projections at 1, 2, and 3 days post-fusion (dpf).

See also Figure S2 and S3.

We previously developed an organoid-fusion method, which revealed efficient functional integration between regionally specified organoids (Xiang et al., 2017). To create axon connections between the thalamus and cortex in a 3D in vitro system, we fused hThOs and hCOs to produce hThCOs. mCherry⁺ hThOs were first fused with mCherry⁻ hCOs at day 18 in order to examine TC connections (Figure 3B). Notably, we found a robust extension of mCherry⁺ processes from hTO cells into the conjoined hCO as detected by epifluorescence microscope (Figure 3C). 3D (x, y, z) imaging of hThCOs further demonstrated that within 6 days post-fusion (dpf) mCherry⁺ axons were already extensively projected out from hThO across the fusion border, and were targeting the far side of the hCO (Figure 3D). During targeting, hThO-derived axons also coalesced into bundles (Figure 3C and 3D), which was a pattern similarly observed in vivo (Enriquez-Barreto et al., 2012; Harsan et al., 2013; Little et al., 2009). Axonal boutons were also detected (arrow heads in Figure 3D). To examine reciprocal axon projections between hThOs and hCOs, we then fused mCherry⁺ hThOs with GFP⁺ hCOs (Figure 3E and Figure S2E). Similar with the efficient TC targeting from mCherry⁺ hThOs to GFP⁺ hCOs, we detected GFP⁺ processes in the hThO side at 6 dpf (Figure 3F), indicating the presence of CT targeting in hThCOs. Interestingly, 3D imaging of mCherry⁺/GFP⁺ hThCOs further demonstrated that while extensive hThO-derived mCherry⁺ axons reached the most opposite side of GFP⁺ hCO at 5 dpf (Figure 3G), there were also abundant hCO-derived GFP⁺ axons in the most opposite side of mCherry⁺ hThO (Figure 3H). Typical growth cones were detected in both mCherry⁺ TC and GFP⁺ CT axons (arrow heads in Figure 3G and 3H), suggesting the presence of active axon growth inside hThCOs. With longer culture, the intensity of the mCherry⁺ axon innervation at GFP⁺ hCO and the GFP⁺ axon innervation at mCherry⁺ hThO both increased (Figure S2F–S2H), demonstrating a continuous reciprocal TC and CT targeting during organoid development.

We found that although axon extension became prominently detectable after ~2 weeks post fusion due to the increase in innervation intensity over time (Figure 3C, Figure 3G and 3H, Figure S2F–S2H), the onset of TC/CT targeting in fact occurred as early as 1 dpf. To further examine the targeting rate and the robustness of targeting in hThCOs, we quantified the range index of TC/CT targeting at different date after organoid fusion (Figure 3I). 50.00%±0.25% of hThOs (n=3 batches, 12 hThCOs in total; mean±sd) extended mCherry⁺ axons towards hCOs at day 1, all of which did not cross the midline of hCOs, and the remaining organoids did not show detectable axon targeting (Figure 3J). At day 2, all hThCOs displayed TC targeting, with the majority (67.67%±0.14%; n=3 batches, 12 hThCOs in total; mean±sd) of axon targeting crossing the midline (Figure 3J and Figure S3A). At the end of assay (day 3), 91.67%±0.14% of hThCOs (n=3 batches, 12 hThCOs in total; mean±sd) showed TC targeting to the most opposite tip in hCOs (Figure 3J and Figure S3B). A similar gradual increase of the range index (16.67%±0.29% peak r1 index at day 1, 83.33%±0.29% peak r2 index at day 2, and 83.33%±0.29% peak r3 index at day 3; n=3 batches, 12 hThCOs in total; mean±sd) was detected for CT targeting within the same hThCOs (Figure 3K; Figure S3A and S3B). Thus, both TC and CT targeting were robust, efficient, and reciprocal in hThCOs. Collectively, our results suggest that, by fusing hThO and hCO together, the reciprocal connections between the human thalamus and cortex can be recapitulated in an in vitro 3D context.

Functional Maturation of Thalamic Neurons in hThOs and hThCOs

To understand functional properties of thalamic neurons, we first examined the electrophysiological properties of neurons in non-fused hThOs. Whole-cell patch-clamp recordings revealed that 10 of 21 recorded neurons produced APs (Figure S4A). Increased injected current resulted in an increase in firing frequency (1s-long current steps from −10 pA to +20 pA with +5 pA increments) at holding potential (around −60 mV) (Figure S4B; one-way ANOVA, p<0.05; n = 7 cells). Thalamocortical connectivity might cause changes in intrinsic properties of thalamic neurons. Thus, we examined if there were differences in firing capabilities between thalamic neurons in hThCOs and those in non-fused hThOs. Incident rate of obtaining thalamic neurons that produced APs were similar between fused hThOs (60.0 %, 9/15 cells) and non-fused hThOs (47.6 %, 10/21 cells). Interestingly we found that thalamic neurons from hThCOs produced higher firing frequency than that of thalamic neurons from non-fused hThOs (two-way ANOVA, p<0.05; Figure S4B; +20 pA: Non-fused, 6.0±2.5 Hz, n=7; Fused, 16.3±2.7 Hz, n=9; Bonferroni post-tests, p<0.001). We also utilized the genetically encoded calcium indicator GCaMP6s under human synapsin promoter and found that both individual hThOs and fused hThOs with hCOs displayed area-scale synchronization of calcium surges without stimuli (Figure S4C; n = 4 hThOs and 4 hThCOs; with one to two synchronized areas observed for each organoid), indicating the presence of spontaneous neuronal activity at network levels. These data suggest that there might be functional interaction between hThOs and hCOs, involved in maturation of intrinsic properties of thalamic neurons.

Directed Axon Targeting and Synaptogenesis in hThCOs

We then examined whether the axon targeting to the hCO from the hThO was region-specific. In hCOs, differentiated neurons are present at the basal side of the ventricular/subventricular zone(VZ/SVZ)-like regions (Figure 1E) (Lancaster et al., 2013; Pasca et al., 2015; Qian et al., 2016; Xiang et al., 2017). To distinguish the differentiated regions from the VZ/SVZ region, we immunostained hThCO (mCherry⁺hThO/mCherry⁻hCO) for the radial glia (RG) marker PAX6. We found that mCherry⁺ axons from hThOs preferentially extended to regions outside of VZ/SVZ-like structure where PAX6⁺ RGs were enriched (Figure 4A and Figure S4D). Similarly, immunostaining for the synaptic vesicle protein synaptophysin (SYP) demonstrated that hThO-derived mCherry⁺ axons preferentially targeted the basal side of VZ/SVZ area in hCOs (Figure 4B and 4C). We quantified the distribution of hThO axons in the hCO side, which revealed a density of 6.42 μm±0.51 μm axons/100 μm² area outside of VZ/SVZ-like regions, and a density of 0.47 μm±0.04 μm axons/100 μm² area inside of VZ/SVZ-like regions (n=4 hThCOs; mean±sd) (Figure 4D). Thus, instead of random pathfinding, TC axons tended to avoid the proliferating region in hCOs and specifically innervate differentiated neurons. This periphery pattern of axon innervation highly resembled the TC targeting in developing brain in vivo (Lopez-Bendito and Molnar, 2003).

Figure 4. TC Targeting and Synaptogenesis in hThCOs.

(A) Immunostaining for PAX6 and mCherry in hThCO section produced by fusing mCherry⁺ hThO and mCherry⁻ hCO. The hCO side is shown. The scale bar represents 50 μm.

(B and C) Immunostaining for mCherry and SYP in hThCO section (B) and enlarged image of region a (C). hThCO was produced by fusing mCherry⁺ hThO and mCherry⁻ hCO. hCO side is shown. Arrows indicate regions enriched with mCherry⁺ axons. The scale bar represents 100 μm in (B) and 50 μm in (C).

(D) Quantification of distribution of mCherry⁺ axons within and outside of VZ/SVZ-areas in hCOs. Mean ± SD is shown. ***p<0.001.

(E) Immunostaining for mCherry and vGLUT2 in hThCO section. hThCO was produced by fusing mCherry⁺ hThO and mCherry⁻ hCO. hCO side is shown. The scale bar represents 25 μm.

(F) Quantification of axons expressing mCherry and vGLUT2 in hCOs. Mean ± SD is shown. ***p<0.001 (compared with vGLUT2⁺/mCherry⁺ group).

(G) Immunostaining for mCherry and MAP2 in hThCO section. hThCO was produced by fusing mCherry⁺ hThO and mCherry⁻ hCO. hCO side is shown. The region outlined by dashed line is further presented on the right. The scale bar represents 50 μm.

(H) Immunostaining for mCherry, SYP and PSD95 in hThCO section. hThCO was produced by fusing mCherry⁺ hThO and mCherry⁻ hCO. hCO side was examined for synaptic structures. Arrows indicate presynaptic SYP⁺ puncta and arrow heads indicate postsynaptic PSD95⁺ puncta. The scale bar represents 5 μm.

See also Figure S4.

To further examine the identity of hThO-derived mCherry⁺ axons in hCOs, we stained for vGLUT2, a marker enriched in thalamus and TC axons (De la Rossa et al., 2013; Moechars et al., 2006; Shiraishi et al., 2017). We found that 90.74%±3.65% (n=4 hThCOs; mean±sd) of mCherry⁺ axons innervated in hCOs were vGLUT2⁺. mCherry⁻/vGLUT2⁺ axons were rarely detected (8.67%±3.32%; n=4 hThCOs; mean±sd) in hCOs, whereas the mCherry⁺/vGLUT2⁻ axons were negligible (0.58%±1.16%; n=4 hThCOs; mean±sd) (Figure 4E and 4F). 96.58% ± 1.29% (n=4 hThCOs; mean±sd) of mCherry⁺ processings were MAP2⁻, further confirming that they were axons but not dendrites (Figure 4G and Figure S4E). These results demonstrate that hThOs projected thalamic axons to hCOs, with a manner similar as TC targeting in vivo. Immunostaining for the presynaptic marker SYP and postsynaptic marker PSD95 in mCherry⁻ hCOs fused with mCherry⁺ hThOs revealed that mCherry⁺ axons in hCOs either contained PSD95⁺ puncta adjacent to neighboring local SYP⁺ puncta (Figure 4H, panel a), or contained SYP⁺ puncta adjacent to local PSD95⁺ puncta (Figure 4H, panel b), suggesting the establishment of synaptic connectivity during TC targeting.

We also examined CT connection by immunostaining hThCOs fused from GFP⁻ hThOs and GFP⁺ hCOs. GFP⁺ axons originated from hCO were detected in the hThO side, where thalamic nuclei represented by TCF7L2 expression were specifically enriched (Figure S4F), demonstrating a successful thalamic targeting. Similar with TC targeting, we detected the presynaptic SYP⁺ puncta and postsynaptic PSD95⁺ puncta juxtaposed to GFP⁺ axons in the hThO side (Figure S4G and S4H), indicating that synaptogenesis also occurred during CT targeting. Taken together, our results show that the reciprocal TC and CT connection in hThCOs mimics the in vivo path and suggest their functional relevance.

DISCUSSION

Despite the importance of thalamus as an information relay hub in the brain and the critical impact of thalamic dysfunction in mediating neurological and psychiatric disorders such as autism spectrum disorder (ASD), schizophrenia and epilepsy (Barch and Ceaser, 2012; Crunelli and Leresche, 2002; Schuetze et al., 2016), there has been a lack of in-depth study to generate human thalamic neurons in vitro, either on monolayer or in 3D cultures. Only recently, a method to direct mouse ESCs to differentiate into thalamic neurons was reported (Shiraishi et al., 2017). However, whether and how hESCs can be steered to thalamic differentiation remained unknown, especially for the purpose of functional organoid formation. Our study to generate hThOs offers an opportunity to investigate human thalamic development and thalamus-related brain disorders in vitro.

We have also established a system for modeling reciprocal axon connectivity between human thalamus and cortex in 3D cultures by fusing the regionally specified brain organoids. The use of brain organoids in 3D and self-organized organ-like cultures has revolutionized the study of the human brain development, disorders, and, importantly, the interaction between distinct brain domains within physiologically-relevant environment (Bagley et al., 2017; Birey et al., 2017; Xiang et al., 2017). As demonstrated here, regionalized human brain organoids can readily be harnessed to recapitulate the structural interactions between thalamic and cortical regions in a dish. While the improved firing frequency in thalamic neurons from our hThCOs suggest a potential functional interaction between hThO and hCO, further studies are needed to examine the genesis and regulatory mechanisms of functional circuits in the fusion system. Overall, we foresee an unprecedented opportunity to model organization of thalamocortical circuits as well as other domains within the brain by fusing disparate regionally specified human brain organoids.

STAR METHODS

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for reagents may be directed to, and will be fulfilled by the lead contact, Dr. In-Hyun Park (inhyun.park@yale.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

hPSCs Culture

Male H1 (WiCell, WAe001-A) and female HES-3 NKX2-1^GFP/w (Monash University, kind gift from Dr. Andrew G. Elefanty) human ES cells isolated from the inner cell mass of human blastocysts and subsequently edited (for HES-3 NKX2-1^GFP/w cells), and all derivative clones from genome manipulation (H1-AAVS1-CAG-GFP hESCs and HES-3 NKX2-1^GFP/w; AAVS1-CAG-mCherry hESCs) were cultured on Matrigel coated tissue culture dish with mTeSR1 media in a humidified 37°C incubator with 5% CO ₂. The HES-3 NKX2-1^GFP/w line was authenticated by GFP expression from the modified NKX2-1-GFP allele, when the cells were differentiated into MGE-like cells. The H1-AAVS1-CAG-GFP and HES-3 NKX2-1^GFP/w; AAVS1-CAG-mCherry lines were authenticated by constitutive expressions of GFP and mCherry proteins driven by the CAG promoter, respectively. Cells were passaged every 7 days by Dispase (0.83 U/ml) treatment, and were tested negative for mycoplasma. All experiments involving hESCs were approved by the Yale Embryonic Stem Cell Research Oversight Committee (ESCRO). Although both male and female hESCs were used in the current study, we did not observe significant differences in forming axonal connections between hESC-derived thalamic and cortical organoids caused by sex identity.

Organoid Details

Experiments	Samples	Sex	Culture stages (days)
Figure 1, S1: qPCR	hThO, hMGEO, hCO	Female	18, 41
Figure 1, S1: staining	hThO, hMGEO, hCO	Female	41
Figure 2, S1: scRNAseq	hThO	Female	34, 89
Figure 3, S2, S3: live imaging	hThCO	Female and male	20~35
Figure 4, S4: staining	hThCO	Female and male	50~60
Figure S4: patch-clamp recording	hThO, hThCO	Female	90~100
Figure S4: calcium imaging	hThO, hThCO	Female	49

METHOD DETAILS

Generation of hThOs

Feeder-free hESC colonies were dissociated into single cells using Accutase treatment. Single cells re-suspended in induction media (DMEM-F12, 15% (v/v) KSR, 1% (v/v) MEM-NEAA, 1% (v/v) Glutamax, and 100 μM β-Mercaptoethanol) supplemented with 100 nM LDN-193189, 10 μM SB-431542, 4 μg/ml Insulin, 5% (v/v) heat-inactivated FBS, and 50 μM Y27632 were plated to ultra-low-attachment 96-well plate. On day 2, the above-mentioned media without the supplement of FBS was replenished. On day 4 and day 6, the above-mentioned media without both FBS and Y27632 was replenished. On day 8, organoids were transferred to spinning culture (80 rpm/min) in ultra-low-attachment 24 well plate (1 organoid/well). From day 8 to day 16, patterning media (DMEM-F12, 0.15% (w/v) Dextrose, 100 μM β-Mercaptoethanol, 1% (v/v) N2 supplement, and 2% (v/v) B27 supplement minus vitamin A) supplemented with 30 ng/ml BMP7 and 1 μM PD325901 was used (media replenished every other day). Beginning at day 16, differentiation media (1:1 mixture of DMEM-F12 and Neurobasal media, 0.5% (v/v) N2 supplement, 1% (v/v) B27 supplement, 0.5% (v/v) MEM-NEAA, 1% (v/v) Glutamax, 0.025% (v/v) Insulin, 50 μM β-Mercaptoethanol, and 1% (v/v) Penicillin/Streptomycin) supplemented with 20 ng/ml BDNF and 200 μM ascorbic acid was used. Media was replenished every other day before day 25, and every four days thereafter.

Generation of hCOs and hMGEOs

hCOs and hMGEOs were generated as previously described (Xiang et al., 2017). Briefly, single cell suspensions from hESCs were plated to ultra-low-attachment 96-well plate (9,000 cells/well) in induction media (DMEM-F12, 15% (v/v) KSR, 1% (v/v) MEM-NEAA, 1% (v/v) Glutamax, and 100 μM β-Mercaptoethanol) supplemented with 100 nM LDN-193189, 10 μM SB-431542, 2 μM XAV-939, 50 μM Y27632, and 5% (v/v) heat-inactivated FBS. The above-mentioned media was replenished every other day until day 10 (on day 2, FBS was removed; on day 4, Y27632 was removed). Organoids were transferred to spinning culture (80 rpm/min) in ultra-low-attachment 6-well plate beginning at day 10 (6–8 organoids/well). For hCOs generation, differentiation media without vitamin A (1:1 mixture of DMEM-F12 and Neurobasal media, 0.5% (v/v) N2 supplement, 1% (v/v) B27 supplement minus vitamin A, 0.5% (v/v) MEM-NEAA, 1% (v/v) Glutamax, 0.025% (v/v) Insulin, 50 μM β-Mercaptoethanol and 1% (v/v) Penicillin/Streptomycin) was used from day 10 to day 18 (media replenished every other day). For hMGEOs generation, patterning media (DMEM-F12, 0.15% (w/v) Dextrose, 100 μM β-Mercaptoethanol, 1% (v/v) N2 supplement, and 2% (v/v) B27 supplement minus vitamin A) supplemented with 100 ng/ml recombinant SHH and 1 μM purmorphamine was used from day 10 to day 18 (media replenished every other day). For both hCOs and hMGEOs, differentiation media with vitamin A (1:1 mixture of DMEM-F12 and Neurobasal media, 0.5% (v/v) N2 supplement, 1% (v/v) B27 supplement, 0.5% (v/v) MEM-NEAA, 1% (v/v) Glutamax, 0.025% (v/v) Insulin, 50 μM β-Mercaptoethanol, and 1% (v/v) Penicillin/Streptomycin) supplemented with 20 ng/ml BDNF and 200 μM ascorbic acid was used beginning at day 18 (media replenished every other day before day 25, and every four days thereafter). 20 ng/ml of FGF2 was added between day 18 and day 22 for hCOs culture if using H1 ESCs.

Genome Editing

For generation of GFP reporter line, 2×10⁶ of single cells dissociated from H1 hESCs were electroporated with 1 μg of AAVS1-TALEN-L plasmid, 1 μg of AAVS1-TALEN-R plasmid, and 8 μg of AAVS1-CAG-hrGFP donor plasmid, and then plated onto Matrigel coated 10-cm dish. 1 week of puromycin selection was performed beginning at 3 days after electroporation. After another 5~7 days of recovery, single GFP⁺ clones were picked and expanded. For generation of mCherry reporter line, 2×10⁶ of single cells dissociated from HES-3 NKX2-1^GFP/w hESCs were electroporated with 1 μg of AAVS1-TALEN-L plasmid, 1 μg of AAVS1-TALEN-R plasmid, and 8 μg of AAVS1-CAG-mCherry donor plasmid. The following steps were the same as described for generation of GFP reporter line.

Real Time Quantitative PCR

Organoids for each group were pulled down together, and total RNA was extracted using the RNeasy Mini Kit according to the manufacturer’s protocol. cDNA was generated using the iScript Select cDNA Synthesis Kit with 500 ng total RNA. Real time quantitative PCR was performed using the SsoFast EvaGreen Supermix in the CFX96 Real-Time PCR System. The PCR cycling conditions were: 95°C for 15 min, followed by 40 two-step cycles at 94°C for 10 s and 60°C for 45 s. Primers used were as follows: OTX2 forward: 5’-CAACACAGCCTCCACTGTGA-3’, reverse: 5’-GGTTCAGAGTCCTTGGTGGG-3’; DBX1 forward: 5’-GAGCAGTCTTCTCCGACGTG-3’, reverse: 5’-TTTCATGCGTCGGTTCTGGA-3’; GBX2 forward: 5’-AGCGAGGTGCAGGTGAAAAT-3’, reverse: 5’-GGCCTGTTCTAGCTGCTGAT-3’; PAX6 forward: 5’-TGTTCCAACTGATATCGTGCCT-3’, reverse: 5’-ATGGCTGTTAGAGCCGCTTC-3’; NKX2-1 forward: 5’-GAGTCCAGAGCCATGTCAGC-3’, reverse: 5’-GCATA AAACAGCTTTGGGGTGT-3’; vGLUT1 forward: 5’-AGCTGGGATCCAGAGACTGT-3’, reverse: 5’-CCGAAAACTCTGTTGGCTGC-3’; DLX2 forward: 5’-GCCTCAACAACGTCCCTTACT-3’, reverse: 5’-TCACTATCCGAATTTCAGGCTCA-3’. Each qPCR data represents expressions in pooled batch of 3 to 4 organoids, and 3 independent replicates were processed for analysis.

Library Preparation for scRNA-seq

hThOs were randomly collected from 3 different culture dishes for each time point (day 34: totally 12 hThOs were pooled together; day 89: totally 10 hThOs were pooled together). Organoids dissociation, cDNA library preparation and sequencing were performed as previously described (Xiang et al., 2017). Briefly, organoids were dissociated using the papain dissociation system according to the manufacturer’s instructions. All solutions used were oxygenated with 95% O₂:5% CO₂ for 5 min before use. Organoids were collected, washed once with HBSS, then dissected into small pieces in oxygenated papain solution. After dissection, tissue suspension was oxygenated again with 95% O₂:5% CO₂ for 5 min, and incubated in 37°C water bath for 1 hour, with gentle shaking every 20 min. Gentle ttrituration was then performed to obtain single cell suspension and papain was inactivated by albumin-ovomucoid inhibitor. Single cells dissociated from organoids were suspended in 1% BSA/PBS supplemented with 10 μM Y27632 and stained with propidium iodide (PI) for 15 min on ice. FACS was performed to sort out PI⁻ cells, which were then re-suspended in 0.04% BSA/PBS at a concentration of 128 cells/μl. cDNA libraries were generated with the Single Cell 3’ Reagent Kits according to the manufacturer’s instructions. Briefly, cells were partitioned into nanoliter-scale Gel Bead-In-Emulsions (GEMs), and using microfluidics cells were flowed at a limiting dilution into a stream of Single Cell 3’ Gel Beads, and then a stram of oil. Upon Cell lysis and dissociation of the Single Cell 3’ Gel Bead within the droplet, primers containing an Illumina P7 and R2 sequence, a 14 bp 10xBarcode, a 10 bp randomer, and a poly-dT primer sequence were released and mixed with the cell lysate and bead-derived Master Mix. Then barcoded, full-length cDNA from poly-adenylated mRNA was generated in each individual bead. Individual droplets were then broken and homogenized before the remaining non-cDNA components were removed with silane magnetic beads. The libraries were then size-selected, and R2, P5, P7 sequences were added to each selected cDNA during end repair and adptor ligation. After Illumina bridge amplification of the cDNA, each library was sequenced using the Illumina HiSeq4000 2×150 bp in Rapid Run Mode.

Data analysis of scRNA-seq

Mapping to human genome (version hg19), quality control and UMI counting of Ensembl genes and K-means clustering were performed by “count” function of cellranger software with default parameter (v2.1.0). Multiple libraries were subsequently combined by “aggr” function. Raw UMI count in each cell was normalized by sum to the median barcode by sum and converted to log2 scale using cellrangerRkit (v1.1.0). Differentially-expressed genes in each cluster were then identified by p-value < 0.05 with two-side T test and 1.25 fold change. GO enrichment was studied by GOstats packages (v2.24.0) in Bioconductor. False discovery rate (FDR) for each GO term was calculated by Benjamini-Hochberg method with p.adjust function in R. GO terms with less than 0.05 FDR were used as statistical significance.

Gene signatures of neuron, NPC, endothelial cells, astrocyte, oligodendrocyte was obtained from public scRNA-seq of fetal and adult brain (Darmanis et al., 2015) and processed as described previously (Xiang et al., 2017). In each cell, GSEA was performed by GSEAPY software (v0.9.3) with options ““--max-size 50000 --min-size 0 -n 1000” to pre-ranked genes, which were sorted by relative expression to average of all cells. Doublet frequency was estimated by expression pattern of NeuN (RBFOX3) and GFAP, which are exclusively expressed in neuron and astrocyte, respectively.

Cell clusters were annotated by unique marker expression, GO enrichment and GSEA of cell type-specific gene signatures (Figure S1H). First, 17 clusters were separated into five neuronal and 12 non-neuronal clusters were separated by expression pattern of early neurogenesis genes (VIM, HES1, SOX2) and neuronal growth cone genes (GAP43, STMN2 and DCX) (Figure S1E). Since one neuronal cluster did not show significant enrichment of some neuronal-related GO terms (e.g. synaptic vesicle (GO:0008021), neurotransmitter transport (GO:0006836)), we defined this cluster as immature neuron. Non-neuronal cluster with higher expression of cell cycle-related genes were annotated as NPC. Four glia cell clusters were extracted by the enrichment of “Gliogenesis (GO:0042063)”, and two out of them were grouped as astrocyte. One cluster was characterized as high expression of proteoglycans and labeled as proteoglycan-expressing glia (PGG). According to the enrichment of BMP signaling-related genes (Figure S1F), one non-neuronal cluster was defined as BMP-related cells (BRC). Although six non-neuronal clusters display no significant enrichment of neuronal and glia gene signatures, “cilium (GO:0005929)”-related genes were enriched in one non-neuronal clusters and “endoplasmic reticulum unfolded protein response (GO:0030968)” in three clusters. The other two clusters were categorized as intermediate progenitor cells.

scRNA-seq libraries from three distinct organoids were scaled by Seurat (v2.2.1) (Macosko et al., 2015). First, UMI count was normalized by the total UMI and multiplied by 10,000 scaling factor with log-transformation for each cell. We removed cells expressing less than 200 genes or genes expressed in less than three cells from subsequent analyses. The normalized expression signal was regressed out by linear model to remove intrinsic technical noises in each library. Variable genes were defined with from 0.1 to 8 average expression and more than one dispersion cutoff. Canonical correlation analysis (CCA) was implemented with common genes, which were included in top 1,000 genes with the highest dispersion in all libraries. CCA subspace was then aligned using from first to 25^th CC.

Single-cell transcriptome of hThOs was also compared with those from in vivo human thalamus in Allen Brain Atlas (http://celltypes.brain-map.org/download). After minimizing experimental biases by Seurat as described above, cells from hThOs and in vivo thalamus were plotted into the same tSNE dimension. The annotation of the in vivo thalamic cell was assigned as that of the nearest hThO cell by calculating Euclidian distance from first to 25^th CC.

Differential expression analysis was performed to the scaled gene expression matrix with 0.25 difference and 0.05 p-value cutoff by two-side T test. To verify the transcriptional similarity between organoid and in vivo human brain, gene expression in each organoid were compared to the other organoids and sorted by the relative expression ratio. Regionally-specific genes in in vivo human brain was then identified from transcriptome dataset in BrainSpan (http://www.brainspan.org/) as described previously (Xiang et al., 2017). Enrichment of the regionally-specific genes was evaluated by GSEA software (v2.2.2) in the pre-ranked genes without collapsing gene set.

Dimensional reduction was performed by non-linear transformation of single-cell gene expression matrix with destiny (v2.0.8) library in Bioconductor (Haghverdi et al., 2015). Diffusion map was constructed from the scaled expression level of the variable genes from Seurat using 100 nearest neighbors and Euclidean distance measurement. The diffused branches were assigned by DPT function with 0.025 window width. Large branches were further divided by branch_divide function until the number of cells was less than 100 in two out of their three subbranches. The subbranches with less than 100 cells were merged with the neighboring subbranches.

Collections of genes and variants associated with human diseases were downloaded from DisGeNET database (Piñero et al., 2017). Curated gene-disease association with more than 0.25 score were used as disease-related genes. GSEA was then performed to genes sorted by relative gene expression in each branch to the other branches with the same parameters described above. Cells expressing vGAT (SLC32A1) and either vGLUT1 or 2 (SLC17A7 or SLC17A6) were classified as inhibitory and excitatory neurons. Cells expressing both or none of GABAergic and glutamatergic transporters were categorized as “unclassified” neuron.

Cryrosectioning and Immunostaining

Organoid samples were fixed in 4% PFA at 4°C for 2~4 days, washed 3 times with PBS (10 min RT incubation for each wash), and incubated in 30% sucrose solution at 4°C for 3 days. Organoids were then incubated in O.C.T compound at RT for 15 min, transferred to tissue base molds, and embedded in O.C.T compound on dry ice. Embedded organoids were stored at −80°C or immediately used for cryosectioning (40 mm slices). Slices were washed with PBS, incubated with 0.1% Triton-100 at RT for 15 min, blocked with 3% BSA/PBS at RT for 2 h, and incubated with primary antibody diluted in 3% BSA/PBS at 4°C overnight. After PBS wash, slices were incubated with secondary antibody diluted in 3% BSA/PBS at RT for 1 h, stained by DAPI for nuclei detection, and mounted with ProLong Gold Antifade Reagent. For quantification of both axon distribution in VZ/SVZ-like areas and the percentage of vGLUT2⁺ axons, 4 hThCOs were randomly collected for staining.

Live Imaging

Organoids were transferred to a Poly-D-Lysine coated 35 mm cell culture dish, and immediately used for imaging. Live imaging was performed using the Leica TCS SP5 confocal microscope with 40x objective, under a controlled temperature (37°C) and CO₂ concentration (5%). A x, y, z scanning mode was used to obtain z stack images. 3D image reconstruction was performed using Leica LAS-X software. For TC and CT axon innervation intensity, the most opposite tip of targeted organoids were live-imaged and used for quantification. 5, 3, 3, and 3 randomly collected hThCOs were used for quantifications of 5–6 dpf TC, 5–6 dpf CT, 12–13 dpf TC and 12–13 dpf CT conditions, respectively.

Organoids Dissociation

Organoids were dissociated using the papain dissociation system according to the manufacturer’s instruction with minor modifications. Specifically, for each group, organoids were collected, pulled together, washed once with HBSS, quickly cut into small pieces in papain solution, and incubated in papain solution in 37°C water bath for 60 min. Gentle trituration was then performed to obtain single cell suspension, after which papain was inactivated by albumin-ovomucoid inhibitor. All solutions used during the process was oxygenated for 5 min with 95% O₂: 5% CO₂. To plate single cell suspensions, neural differentiation media supplemented with 20 ng/ml BDNF, 200 μM ascorbic acid, and 10 μ Y27632 was used. The second day, the above-mentioned media without Y27632 was replenished.

Organoid Culturing On Dish

Cell culture dishes were coated with Poly-D-Lysine for 4 hours in the incubator. Single cell dissociations from day 24 hCOs were prepared as described above and plated to the dishes. 4 days after plating the cells, day 35 mCherry⁺ hThOs or GFP⁺ hCOs were plated on top of the neural culture in the dishes. Neural differentiation media supplemented with 20 ng/ml BDNF and 200 μM ascorbic acid was used and replenished every 4 days and axon growth was monitored.

Organoid Fusion

Organoid fusion was performed using the same strategy as we reported previously (Xiang et al., 2017). To fuse mCherry⁺ hThOs with mCherry⁻ hCOs derived from the HES-3 NKX2-1^GFP/w hESC line, day 18 organoids were collected and a single hThO and single hCO were transferred to one well of ultra-low-attachment 96-well plate in order to produce one hThCO. 2 days after fusion, half of the media was changed with caution. 3 days after fusion, the fused organoids were transferred to ultra-low-attachment 6-well plate (maximum 2 hThCOs/well) for further development, with the culture condition the same with hThO and hCO culturing. To fuse mCherry⁺ hThOs with GFP⁺ hCOs derived from H1 hESCs, day 22 organoids were collected, and the following processing was the same as described above.

Range Index Examination

mCherry⁺ hThOs and GFP⁺ hCOs were fused to produce hThCOs. Axon targeting in both hThO and hCO side was examined under an epifluorescence microscope (Nikon inverted microscope; Eclipse TS100) at 1 dpf, 2 dpf, and 3 dpf. The ranges of axon targeting were classified into 3 categories: not approaching the midline of targeting organoid (r1), already crossing the midline of targeting organoid (r2), and already approaching the opposite tip of targeting organoid (r3). 3 batches of fusion were performed and 4 hThCOs were generated and examined for each batch. The percentage of various range index at different dates were quantified for each fusion batch. 3 independent batches were used for quantification, with each batch containing 4 hThCOs.

Electrophysiological Recordings

90~100 days old organoids were embedded in a 4% agarose block. Organoid slices (300 μm) were cut in chilled oxygenated slicing medium containing (in mM) 85 NaCl, 75 sucrose, 2.5 KCl, 25 glucose, 1.25 NaH₂PO₄, 4 MgCl₂, 0.5 CaCl₂, and 24 NaHCO₃ using a Vibratome slicer (VT 1200S; Leica Microsystems Inc.: Buffalo Grove, IL, USA), and incubated ≥ 1 hours before recording. Whole-cell patch-clamp recordings were obtained using artificial cerebrospinal fluid (ACSF; the rate of 2.5 ml/min) containing (in mM) 126 NaCl, 2.5 KCl, 26 NaHCO₃, 2 CaCl₂, 2 MgCl₂, 1.25 NaH₂PO₄, and 10 glucose) gasses with 95% O₂/5% CO₂. Thalamic cells were visualized with an upright microscope (Eclipse FN1; Nikon Instrumets Inc.: Melville, NY, USA) with infrared differential interference contrast optics and with a high-power water immersion 60X objective. Whole-cell patch-clamp recordings were obtained from thalamic cells with borosilicate glass pipettes (5–7 MΩ) when filled with intracellular solution containing (in mM) 126 K-gluconate, 4 KCl, 10 HEPES, 4 ATP-Mg, 0.3 GTP-Na, and 10 phosphocreatine (pH 7.2 and osmolality of 290 mOsm). MultiClamp700B amplifier (Molecular Devices: San Jose, CA, USA) was used for recordings. Voltage signals were filtered at 3 kHz using a Bessel filter and digitized at 10 kHz with Digidata 1440A digitizer (Molecular Devices: San Jose, CA, USA). Thalamic neurons were injected with hyperpolarizing and depolarizing current steps (−10 to +20 pA, 5 pA increments, 1 s from around −60 mV) in order to characterize firing properties. The Clampfit 10 software (Molecular Devices: San Jose, CA, USA) was used to analyze the electrophysiological data. 7 (hThOs) and 9 (hThCOs) recorded cells showing typical APs were included for quantification of firing frequency in response to increased injected current.

Calcium Imaging

hThOs were transduced with AAV1.syn.GCaMP6s.WPRE.SV40 (Penn Vector Core) at day 34, and then either cultured individually, or fused with non-transduced hCOs on day 35. Calcium imaging were performed 2 weeks after fusion. Nikon inverted microscope (Eclipse TS100) was used to observe calcium surges with 10X objectives at 488-nm excitation. Time lapse images were captured using a Digital CCD camera (QICAM: FAST 1394) and Qcapture Pro7 software (QICAM) at a speed of 1 frame/sec (FPS). Area-scale image series were analyzed using Fiji software (Schindelin et al., 2012). 4 hThOs and 4 hThCOs were used for calcium imaging.

QUANTIFICATION AND STATISTICAL ANALYSES

For quantification of immunostaining, qPCR, axon identity and distribution, unpaired t test was used to determine the statistical significance. Mean values ± SD are shown unless otherwise stated, and n represents 3 independent replicates of samples or more. For patch-clamp recording, one-way ANOVA analysis was used to test significance within cells from non-fused hThOs, and two-way ANOVA analysis was used to test significance between cells from hThCOs and non-fused hThOs. Statistical analyses for processing sequencing data can be found in detailed methods (in the section of Data analysis of scRNA-seq). For all statistical analyses, p value less than 0.05 was interpreted as statistically significant. The statistical details of experiments can also be found in the figure legends and corresponding results.

DATA AND SOFTWARE AVAILABILITY

The accession number of data generated in this study is GEO: GSE122342. Other published datasets used are listed in the Key Resource Table.

Key Resources Table.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
MAP2	Millipore	Cat# MAB3418; RRID:AB_94856
TCF7L2	Cell Signaling	Cat# 2569; RRID:AB_2199816
TBR1	Abcam	Cat# ab31940; RRID:AB_2200219
PAX6	DSHB	Cat# AB528427; RRID:AB_528427
GFP	Sigma	Cat# SAB4600051; RRID:AB_2753203
GFP	Cell Signaling	Cat# 2956; RRID:AB_1196615
mCherry	Novus Biologicals	Cat# NBP2–25157; RRID:AB_2753204
vGLUT2	Millipore	Cat# MAB5504; RRID:AB_2187552
Synaptophysin	Abcam	Cat# ab8049; RRID:AB_2198854
PSD95	Sigma	Cat# P-246; RRID:AB_260911
PSD95	Abcam	Cat# ab12093; RRID:AB_298846
Chemicals, Peptides, and Recombinant Proteins
mTeSR1	Stem Cell Technologies	Cat# 05875
DMEM-F12	Life Technologies	Cat# 11330057
Neurobasal Media	Life Technologies	Cat# 2110349
FBS	Life Technologies	Cat# 10437028
Amino acids, non-essential	Life Technologies	Cat# 11140050
Penicillin/Streptomycin	Life Technologies	Cat# 15140–122
Glutamax	Life Technologies	Ca# 35050
Insulin	Sigma	Ca# I9278
β-Mercaptoethanol	Sigma	Ca# M7522
N2	Life Technologies	Cat# 17502–048
B27	Life Technologies	Cat# 17504–044
B27 supplement without vitamin A	Life Technologies	Cat# 12587010
bFGF	Millipore	Cat# GF003AF
KnockOut Serum Replacement	Life Technologies	Cat# 10828–028
HBSS	Life Technologies	Cat# 14170112
Matrigel	BD	Cat# 354230
Dextrose	Sigma	Cat# G7021
Poly-D-Lysine	Xona	Cat# XC PDL
Y-27632	Stem Cell Technologies	Cat# 72304
Dispase (100ml)	Stem Cell Technologies	Cat# 07913
Accutase (100ml)	Stem Cell Technologies	Cat# AT104
LDN-193189	Sigma	Cat# SML0559
SB431542	Abcam	Cat# ab120163
XAV939	Sigma	Cat# X3004
BMP7	Gibco	Cat# PHC9544
PD0325901	Axon Medchem	Cat# Axon 1408
SHH	R&D Systems	Cat# 464-SH-200
Purmorphamine	Stem Cell Biotech	Cat# 72204
Puromycin	Sigma	Cat# P8833
BDNF	Prepotech	Cat# 450–02
Ascorbic acid	Sigma	Cat# A92902
O.C.T compound	Tissue-Tek	Cat# 4583
Bovine serum albumin	American Bioanalytical	Cat# AB01088
ProLong Gold Antifade Reagent	ThermoFisher	Cat# P36930
Critical Commercial Assays
Papain Dissociation System	Worthington Biochemical Corporation	Cat# LK003150
Human Stem Cell Nucleofector Kit 1	Lonza	Cat# VPH-5012
RNeasy mini kit	QIAGEN	Cat# 74104
RNase-Free DNase Set	QIAGEN	Cat# 79254
iScript cDNA synthesis kit	Biorad	Cat# 1708891
SsoFast EvaGreen Supermix	Biorad	Cat# 1725201
Deposited Data
Raw and proposed scRNA-seq	This paper	GSE122342
Experimental Models: Cell Lines
H1 hESC line	WiCell	https://www.ncbi.nlm.nih.gov/pubmed/9804556/
H1-AAVS1-CAG-GFP	This paper
HES-3 NKX2–1GFP/w	Elefanty lab	https://www.ncbi.nlm.nih.gov/pubmed/21425409
HES-3 NKX2–1GFP/w;AAVS1-CAGmCherry	This paper
Recombinant DNA
AAVS1-TALEN-L	Addgene	RRID:Addgene_59025
AAVS1-TALEN-R	Addgene	RRID:Addgene_59026
AAVS1-CAG-hrGFP	Addgene	RRID:Addgene_52344
AAVS1-CAG-mCherry	Addgene	RRID:Addgene_80946
AAV1.syn.GCaMP6s.WPRE.SV40	Penn Vector Core	AV-1-PV2824
Software and Algorithms
Cellranger (v2.1.0)	N/A	https://support.10xgenomics.com/
CellrangerRkit (v1.1.0)	N/A	https://support.10xgenomics.com/
R (v3.3.2)	N/A	https://www.rproject.org/
Seurat (v2.2.1)	Macosko et al., 2015	https://satijalab.org/seurat/
GSEA (v2.2.2)	Subramanian et al., 2005	https://software.broadinstitute.org/gsea/index.jsp
GOstats (v2.24.0)	Falcon and Gentleman, 2007	https://bioconductor.org/packages/release/bioc/html/GOstats.html
GSEAPY (v0.9.3)	N/A	https://pypi.org/project/gseapy/
Destiny (v2.0.8)	Haghverdi et al., 2015	https://bioconductor.org/packages/release/bioc/html/destiny.html
Fiji	N/A	https://fiji.sc
Clampfit 10	Molecular Devices
Other
Reference transcriptome for hg19	N/A	https://support.10xgenomics.com/
RNA-seq for in vivo embryonic brain regions	Brainspan	http://www.brainspan.org/
scRNA-seq for human thalamus	Allen Brain Atlas	http://celltypes.brainmap.org/download
scRNA-seq for fetal and adult brains	Darmanis et al., 2015	SRA:SRP057196
U-bottom ultra-low-attachment 96-well plate	Corning	CLS7007–24EA
Orbital shaker	IKA	KS260
Nucleofector	Lonza	AAB-1001

HIGHLIGHTS.

• Thalamus-like brain organoids (hThOs) develop from hESCs.

• Single-cell transcriptome distinguishes hThOs from hCOs and hMGEOs.

• Reciprocal thalamocortical projections establish by fusing hThO and hCO.

• Axons exhibit directed targeting and synaptogenesis in hThCOs.