Thalamocortical organoids enable in vitro modeling of 22q11.2 microdeletion associated with neuropsychiatric disorders

SUMMARY

Thalamic dysfunction has been implicated in multiple psychiatric disorders. We sought to study the mechanisms by which abnormalities emerge in the context of the 22q11.2 microdeletion, which confers significant genetic risk for psychiatric disorders. We investigated early stages of human thalamus development using human pluripotent stem cell-derived organoids and show that the 22q11.2 microdeletion underlies widespread transcriptional dysregulation associated with psychiatric disorders in thalamic neurons and glia, including elevated expression of FOXP2. Using an organoid co-culture model, we demonstrate that the 22q11.2 microdeletion mediates an overgrowth of thalamic axons in a FOXP2-dependent manner. Finally, we identify ROBO2 as a candidate molecular mediator of the effects of FOXP2 overexpression on thalamic axon overgrowth. Together, our study suggests that early steps in thalamic development are dysregulated in a model of genetic risk for schizophrenia and contribute to neural phenotypes in 22q11.2 Deletion Syndrome.

Graphical Abstract:

eTOC Blurb

The 22q11.2 microdeletion is a strong risk factor for neuropsychiatric disorders, such as schizophrenia and autism spectrum disorder. Nowakowski and colleagues identified that this copy number variant mediates abnormal axonal outgrowth in an organoid model of thalamocortical pathway development, in part by dysregulating a high conflidence autism risk gene, FOXP2.

INTRODUCTION

The thalamus is a subcortical structure that relays sensory information to the cortex via the thalamocortical pathway^1,2. Dysfunction of this pathway has been implicated in sensory and cognitive deficits observed in neuropsychiatric disorders including schizophrenia, autism spectrum disorder (ASD), bipolar disorder, and attention deficit hyperactivity disorder (ADHD)^3–11. While these disorders have a complex genetic risk architecture, copy number variation at the 22q11.2 locus represents a major risk factor^12–15. The 22Q11 Deletion Syndrome (‘22q11DS’) is caused by a microdeletion of a 1.5–3 Mb part of chromosome 22 spanning 46 protein-coding genes, and affects approximately 1 in 4000 live births¹⁴. 22q11DS patients exhibit a 20-fold risk for developing psychosis^14,16 and a 30–40% risk for developing ASD¹⁷. Imaging studies have identified increased connectivity between thalamic and sensory areas of the cortex in 22q11DS patients^18,19, consistent with alterations observed in a broad range of psychiatric conditions^3,19–21 and may contribute to disease phenotypes in 22q11DS. Clinical manifestations of 22q11DS can be detected in early life^22,23, indicating that symptoms may originate from early neurodevelopmental alterations.

Brain organoids are three-dimensional in vitro models of developing brain tissue, and can be differentiated from pluripotent stem cells to study early phases of brain development^24–27. Changes in neuronal excitability have been identified in 22q11DS cortical organoids²⁸ and 2D neurons²⁹, but subcortical structures have not been examined. The thalamocortical pathway develops when thalamic glutamatergic neurons send long-range projections to the cerebral cortex, and deep layer cortical glutamatergic neurons project to the thalamus¹. Co-culture assays of thalamic and cortical tissue have been instrumental for identifying basic mechanisms of thalamocortical pathway development³⁰. This experimental approach has recently been applied to regionalized brain organoids, demonstrating the feasibility of examining early steps in human thalamocortical pathway formation^31,32.

To study how early human thalamus development might be influenced by the 22q11.2 microdeletion, we used thalamic organoid differentiations. We identified transcriptional differences between control and 22q11DS organoids using single-cell sequencing and found an enrichment for genes associated with neuropsychiatric disorders. To begin to understand how transcriptional changes in 22q11DS might influence normal thalamus development, we examined the role of elevated FOXP2 expression in 22q11DS thalamic glutamatergic neurons. We show that upregulated FOXP2 expression mediates thalamic axon overgrowth. To identify potential mechanisms of FOXP2-induced axonogenesis phenotypes, we profiled FOXP2 binding sites and identified several differentially expressed axon guidance cues. One of these genes, ROBO2, encodes for an axon repulsive receptor depleted in 22q11DS glutamatergic neurons. We demonstrate that ROBO2 knockdown in control thalamic organoids phenocopies FOXP2 overexpression. Together, this study identifies key molecular and cellular phenotypes during early human thalamocortical pathway development associated with 22q11DS and neurodevelopmental disorders, and uncover a molecular program controlled by FOXP2 that leads to thalamic axon overgrowth in 22q11DS.

RESULTS

Generation and Benchmarking of Human Thalamic Organoids

To assess the impact of 22q11.2 microdeletion on early development of human thalamic cell types, we utilized iPS lines from four 22q11DS patients and four karyotypically normal controls (Figure 1A, Figure S1A–S1D, Table S1) [STAR Methods]. We differentiated iPS lines into thalamic organoids³¹ and confirmed proper specification based on canonical marker expression (Figure 1B and 1C, Figure S1E) [STAR Methods]. To further benchmark thalamic organoids, we performed single-cell RNA sequencing (scRNAseq) at ten weeks of differentiation. This time point corresponds to the peak of neurogenesis in organoids³³ which enables recovery of diverse cell states, including progenitor and neuronal populations. We obtained single-cell transcriptomes for 69,625 cells from three control and four patient lines across four batches of differentiation (Figure S1F–S1J) [STAR Methods]. Thalamic organoids strongly resembled primary thalamus based on Voxhunt spatial similarity mapping³⁴ against E13.5 mouse brain and PCW15–24 human brain³⁵ (Figure S1K–S1L). To further corroborate thalamic identity and guide interpretation of cell type classification, we mapped the data onto a scRNAseq reference dataset of the second trimester human thalamus³³ (Figure 1D–1F and Figure S1M–S1P). Cells from all lines mapped to the major neuronal and progenitor populations in the human thalamus, although sampling biases associated with scRNAseq contributed to cell type proportion differences across iPS lines and batches (Figure 2A–2B and Figure S1H, S1I, S1N–S1P). We thus identified organoid cells that resembled glutamatergic neurons, GABAergic neurons, intermediate progenitors, astrocytes, glial progenitors, and dividing cells of the developing human thalamus (Figure 1F and 1G, Table S2).

Figure 1. Benchmarking of human thalamic organoids derived from 22q11DS and control iPS lines.

(A) SNP array imaging results for chr22 of control and 22q11DS patient iPS lines. (B) Immunostaining against pan-neuronal marker HuC/D and thalamic neuron marker GBX2 (top), GABAergic neuron markers GAD65/67 and LHX1 (middle), and pan-thalamic marker TCF7L2 and progenitor marker PAX6 (bottom). Scale bar, low magnification: 500 μm, high magnification: 50 μm. (C) Quantification of immunostainings in (B). Mean ± SEM is shown. (D) Schematic of scRNAseq analysis strategy. (E, F) Uniform Manifold Approximation and Projection (UMAP) plots depicting cell types in the primary reference (E) and reference-mapped thalamic organoid (F) datasets. (G) Gene expression plot for major cell type markers in thalamic organoids.

See also Figure S1, Table S1, and Table S2.

Figure 2. Transcriptomic changes in 22q11DS thalamic and cortical organoids.

(A) UMAPs by genotype. (B) Cell type proportions by genotype. (C) Number of 22q11DS-associated differentially expressed genes (DEGs) by cell type. (D, E) Overlap between ASD risk genes (SFARI, risk score = 1) and DEGs in at least one cell type (D) or per cell type (E). *** FDR < 0.0005; hypergeometric test. (F) Significance of differentially expressed transcription factors implicated in ASD (SFARI, risk score = 1). (G, H) Immunostaining for pan-neuronal marker HuC/D, pan-thalamic marker TCF7L2, and DEGs FOXP2 (G) or SOX2 (H) in week 10 thalamic organoids. Left: tilescan of the entire organoid, scale bar = 250 μm. Right: high magnification images represented by insets on the left. Scale bar = 50 μm. (I, J) Percentage of TCF7L2/HuC/D+ cells expressing FOXP2 (I) or SOX2 (J) in thalamic organoids. Mean ± SEM is shown (*** p < 0.0005; Wilcoxon rank-sum test; n = 9 organoids per genotype). (K, L) Immunostaining for TCF7L2, FOXP2 (K), and SOX2 (L) in mouse thalamus. Left: tilescan of the entire thalamus (dotted line). Scale bar = 1 mm. Right: Higher magnification images represented by insets on the left. Scale bar = 50 μm. (M) Percentage of TCF7L2+ cells expressing FOXP2. Mean ± SEM is shown (* p < 0.05; linear mixed model regression; N = 3 biological replicates per genotype) (N) Percentage of NeuN+ cells that express SOX2. Mean ± SEM is shown (* p < 0.05; linear mixed model regression; N = 3 biological replicates per genotype)

See also Figure S2 and Table S3.

Transcriptomic Changes in 22q11DS Cell Types

Transcriptomic changes associated with the 22q11.2 microdeletion have been profiled in cortical cells^28,29,36, but not in cells of subcortical identity including the thalamus. We sought to identify dysregulated gene signatures associated with the 22q11.2 microdeletion in thalamic organoids and compare them to those in cortical organoids (Figure S1Q–S1V and Table S3). We identified 5835 and 4292 differentially expressed genes (DEGs) in thalamic and cortical organoids, respectively, in at least one cell type (Figure 2C, Figure S2A and S2B, and Table S3) [STAR Methods]. DEGs included protein-coding genes within the 22q11.2 deletion, many of which were shared across regional identities (thalamic: 19/43, cortical: 20/43, shared: 15) (Figure S2C and S2D). However, a minority of DEGs outside of the 22q11.2 locus were shared between thalamic and cortical organoids, suggesting that transcriptional dysregulation outside of the 22q11.2 locus is often region-dependent (Figure S2A and S2B; Table S3).

While the 22q11.2 microdeletion carries high risk for schizophrenia, genes within the 22q11.2 locus have not been implicated by recent exome sequencing studies^37,38, suggesting complex contributions of this copy number variant to disease risk. We thus examined whether transcriptomic changes identified in patient-derived organoids converge upon pathways implicated in schizophrenia or other neurodevelopmental disorders (Figure 2D–2F and Figure S2E–S2M). We first compared DEGs to high confidence schizophrenia risk genes identified by the SCHEMA consortium³⁷. Of the 32 genes identified, 15 and 14 genes were dysregulated in at least one cell type in thalamic and cortical organoids, respectively, seven of which were shared (Figure S2E–S2G). Notably, the directionality of differential expression was often the opposite: primarily upregulated in thalamic organoids and downregulated in cortical organoids (Figure S2G). We also considered ASD high-confidence risk genes (SFARI, risk score = 1, 326 genes) and identified 130 and 100 genes that overlapped with thalamic and cortical DEGs, of which 64 were shared (Figure 2D and S2E). These results suggest that the 22q11.2 microdeletion mediates transcriptomic changes that are strongly convergent with neurodevelopmental disorder risk genes. Many of these changes are distinct to each model (Figure S2A–S2M), highlighting the need for more nuanced brain region-specific modeling for the 22q11.2 microdeletion.

Elevated FOXP2 expression in 22q11DS thalamic neurons mediates axon overgrowth

Overconnectivity between thalamus and somatosensory cortex has previously been identified in 22q11DS patients^3,18,19 (Figure S3A) [STAR Methods], including in infants³⁹. Therefore, we closely examined transcriptional changes in 22q11DS thalamic glutamatergic neurons, the primary cell class that participates in the formation of the thalamocortical pathway⁴⁰. We focused on differentially expressed transcription factors, because such genes orchestrate many developmental decisions in the mammalian brain, including patterning, differentiation, and axonogenesis^41–44 (Figure 2F and Table S3). DEG analysis identified FOXP2 as the most upregulated, and SOX2 as one of the most downregulated transcription factors in 22q11DS glutamatergic neurons in thalamic, but not cortical organoids (Table S3). We confirmed dysregulation of FOXP2 and SOX2 by immunostaining in thalamic organoids (Figure 2G–2J; Figure 3F and 3G) and a mouse model of 22q11.2 microdeletion (Figure 2K–2N), indicating that our finding is not a culture artifact.

Figure 3. 22q11.2 microdeletion mediates thalamocortical axon overgrowth via elevated FOXP2 expression.

(A) Thalamocortical organoid co-culture model for visualizing axon outgrowth. (B) Immunostaining against thalamus marker TCF7L2 and telencephalon marker FOXG1 in thalamocortical organoid co-cultures. Scale bar: 500 μm. (C) Visualization of thalamic (green) or cortical (magenta) axons crossing the fusion boundary after six days post-fusion (dpf). Scale bar: 500 μm. (D, E) Quantification of thalamocortical (D) or corticothalamic (E) axon outgrowth. *** adj. p < 0.0005, ns = not significant; Wilcoxon rank-sum test with Bonferroni-Holm correction. In (D), n = 21 control organoids, 9 CAG-FOXP2 organoids, and 29 patient organoids. In (E), n = 16 control organoids, 10 CAG-FOXP2 organoids, and 20 patient organoids. (F) Immunostaining against FOXP2, TCF7L2, and HuC/D in a CRISPR-engineered 22q11DS thalamic organoid and isogenic control. Scale bar, low magnification: 250 μm, high magnification: 50 μm. (G) Percentage of TCF7L2/HuC/D+ cells expressing FOXP2. * adj. p < 0.05; Wilcoxon Rank-sum test with Bonferroni-Holm correction; n = 5 control organoids and 6 CRISPR-engineered organoids. (H) Visualization of thalamic (green) or cortical (magenta) axons crossing the fusion boundary. Scale bar: 500 μm. (I, J) Quantification of thalamocortical (I) or corticothalamic (J) axon outgrowth. ** adj. p < 0.005, *** adj. p < 0.0005, ns = not significant; Wilcoxon rank-sum test with Bonferroni-Holm correction. In (I), n = 4 isogenic control organoids, 3 CAG-FOXP2 organoids, and 11 CRISPR-engineered organoids. In (J), n = 4 isogenic control organoids, 4 CAG-FOXP2 organoids, and 5 CRISPR-engineered organoids. (K) Immunostaining against thalamocortical axon marker PKCδ in parasagittal sections of P56 wildtype or Df(h22q11)/+ mouse brain. Scale bar, low magnification: 1 mm, high magnification: 500 μm. (L) Percent area in the striatum occupied by thalamocortical axons across eight sections per biological replicate, comparing wildtype (n = 3) or Df(h22q11)/+ mice (n = 3). (* adj. p < 0.05; linear mixed regression model).

See also Figure S3.

FOXP2 is a high-confidence ASD risk gene^45,46 (Figure 2F and Table S3) implicated in thalamocortical pathway development in mice⁴⁷. We therefore examined axonal projections in control versus FOXP2-overexpressing organoids using a line-matched co-culture assay (Figure 3A and 3B, Figure S3B–S3Z) [STAR Methods]. We chose one week of co-culture as our first time point, given prior descriptions indicating it as the earliest stage at which axonal outgrowth is apparent, and five weeks as a later time point by which axons have grown in^31,32,48. FOXP2 overexpression led to a significant increase in thalamocortical axon outgrowth compared to mock transduced control co-cultures at both one and five weeks post-fusion (Figure 3C and 3D, Figure S3E–S3H). We examined whether this phenotype is consistent with the 22q11.2 microdeletion by comparing control and patient co-cultures, as well as co-cultures generated using isogenic iPS lines engineered to carry the 22q11.2 microdeletion using CRISPR⁴⁹ (Figure S3L). We confirmed that co-cultures carrying the 22q11.2 microdeletion also displayed overgrowth of thalamocortical axons (Figure 3C and 3D, Figure S3E–S3H). Finally, we tested whether df(h22q11)/+ mice exhibit evidence of thalamocortical axon overgrowth, given that FOXP2 is elevated in the thalamus of mutant mice (Figure 2K–2N). We stained for PKCδ, a marker of thalamocortical axons⁵⁰, and identified a striking increase in PKCδ+ axons in the striatum in df(h22q11)/+ mice compared to wildtype controls (Figure 3K and 3L). This data suggests that the 22q11.2 microdeletion mediates excessive outgrowth of thalamocortical projections as a result of elevated FOXP2 expression during prenatal development.

Thalamocortical axons in the developing forebrain can facilitate corticothalamic axon guidance through fasciculation⁵¹. Consequently, it is possible that the initial overgrowth of thalamic axons could influence the development of reciprocal axons from cortical organoids. In co-cultures where control thalamic organoids overexpressing FOXP2 were fused to cortical organoids, we detected a significant increase in the number of cortical axons by week 5 post-fusion, although not at week 1 (Figure 3C and 3E, Figure S3I–S3K). Similar findings were made using organoids generated from CRISPR-engineered 22q11DS iPS lines (Figure 3H and 3J, Figure S3I–S3K) and patient-derived lines (Figure 3C and 3E, Figure S3I–S3K). To test the hypothesis that thalamic axon overgrowth is the primary driver of cortical axon overgrowth, we generated a panel of ‘chimeric’ organoids where thalamic organoids of either genotype were fused with organoids of the opposing genotype (Figure S3L). Consistent with this hypothesis, fusion of control thalamus organoids with 22q11DS cortical organoids did not result in overgrowth of thalamic or cortical axons, whereas fusion of 22q11DS thalamus organoids with control cortical organoids did result in both thalamic and cortical axon overgrowth (Figure S3M–S3P). Furthermore, co-culture of thalamic organoids with slice cultures of human second trimester cortex also displayed thalamic axon overgrowth from 22q11DS organoids compared to control (Figure S3Q). These results suggest that FOXP2 overexpression in thalamic organoids is sufficient to induce thalamic axon overgrowth, which leads to overgrowth of reciprocal axons from cortical organoids.

ROBO2 is a putative downstream target of FOXP2 that mediates overgrowth of thalamic axons in 22q11DS organoids

We tested whether elevated FOXP2 expression is necessary for axonogenesis phenotypes in 22q11DS neurons by performing FOXP2 knockdown in 22q11DS thalamus organoids (Figure 4A) [STAR Methods]. Consistent with our model that upregulated FOXP2 expression is necessary for axon overgrowth, FOXP2 knockdown in 22q11DS patient and CRISPR-derived co-cultures restored axon outgrowth to control levels (Figure 4B–4F and Figure S3R–S3X). To determine the downstream mechanisms of FOXP2-mediated axon overgrowth, we profiled the DNA binding sites of FOXP2 in thalamic organoids using CUT&Tag⁵² (Figure 4G) [STAR Methods]. We identified 2023 consensus peaks that resided in annotated genomic regions, and 4572 genes that resided within 500 kilobases of peak coordinates (Figure 4H and Table S4). 184 genes overlapped with DEGs identified in 22q11DS thalamic glutamatergic neurons, 21 of which were implicated in axon development^53,54 (Figure 4I and 4J, Figure S4A and S4B). One of these genes, ROBO2, was well-characterized for its role in regulating thalamocortical axon guidance^55,56. ROBO2 encodes for an axon guidance receptor that mediates repulsion upon contact with SLIT ligands, and is robustly expressed in the second trimester thalamus and control thalamic organoids (Figure S4C). Genes encoding ROBO2 ligands, SLIT1 and SLIT2, were robustly expressed in cortical organoids (Figure S4D). The expression of ROBO2 was downregulated in 22q11DS thalamic neurons (Figure 4J, Figure S4D, Table S3), which we hypothesized may slow down the initial outgrowth of genotypic control thalamic axons into cortical organoids. To test this, we performed ROBO2 knockdown in control thalamic organoids, which led to thalamocortical axon overgrowth compared to non-targeting control (Figure 4M and 4N). This data suggests that elevated FOXP2 expression in 22q11DS thalamic organoids leads to axon overgrowth by downregulating ROBO2 (Figure 4O–4Q).

Figure 4. ROBO2 is a putative downstream target of FOXP2 whose downregulation mediates thalamic axon overgrowth.

(A) FOXP2-targeting sgRNA validation. (* adj. p < 0.05; Wilcoxon rank-sum test with Bonferroni-Holm correction) (B) Schematic of FOXP2 knockdown in thalamic organoids followed by co-culture with cortical organoids. (C, D) Thalamocortical axon outgrowth with or without FOXP2 knockdown after six days post-fusion (dpf) in 22q11DS patient lines (C) and CRISPR-engineered lines (D). Scale bar: 500 μm. (E, F) Quantification of thalamocortical axon outgrowth after FOXP2 knockdown in patient (E) or CRISPR-engineered (F) lines. * adj. p < 0.05, ** adj. p < 0.005, *** adj. p < 0.0005, ns = not significant; Wilcoxon rank-sum test with Bonferroni-Holm correction. In (E): n = 12 control organoids with non-targeting sgRNA, 8 patient organoids with FOXP2 sgRNA #1, 11 patient organoids with FOXP2 sgRNA #2, and 11 patient organoids with non-targeting sgRNA. In (F): n = 4 control organoids with non-targeting sgRNA, 13 CRISPR-engineered organoids with FOXP2 sgRNA #2, and 13 CRISPR-engineered organoids with non-targeting sgRNA. (G) CUT&Tag was performed to identify FOXP2 binding sites in week 5 control thalamic organoids. (H) Number of detected CUT&Tag peaks by functional annotation. (I) Overlap between genes within 500 kb of CUT&Tag peaks and thalamic organoid DEGs. (J) Volcano plot visualizing DEGs implicated in axon outgrowth that overlap with genes within 500 kb of FOXP2 binding sites. (K) ROBO2-targeting sgRNA validation. (* p < 0.05; Wilcoxon rank-sum test). (L) Schematic of ROBO2 knockdown in thalamic organoids followed by co-culture with cortical organoids. (M) Visualization of thalamocortical axon outgrowth with or without ROBO2 knockdown in control organoids, compared to 22q11DS organoids. Scale bar: 500 μm. (N) Quantification of thalamocortical axon outgrowth after ROBO2 knockdown. (** adj. P < 0.005, *** adj. P < 0.0005; Wilcoxon rank-sum test with Bonferroni-Holm correction; n = 12 control organoids with non-targeting sgRNA, 5 control organoids with ROBO2 sgRNA, 11 patient organoids with non-targeting sgRNA, and 14 CRISPR-engineered organoids with non-targeting sgRNA). (O-Q) Working model for how 22q11.2 microdeletion mediates axon overgrowth phenotypes in thalamocortical organoid co-cultures.

See also Figure S4 and Table S4.

DISCUSSION

Thalamocortical pathway abnormalities have been strongly implicated in the etiology of several neuropsychiatric disorders based on patient neuroimaging studies^3,19–21. However, prior investigations into the molecular and cellular basis of these disorders have predominantly centered on the cerebral cortex^{28,29,49,57,58}.

This study addresses this gap by examining the impact of the 22q11.2 microdeletion – a mutation that confers strong risk for several neurodevelopmental psychiatric disorders – on the early development of the human thalamus and formation of the thalamocortical tract.

Overgrowth of thalamocortical axons is in line with a growing body of literature implicating elevated connectivity in the thalamocortical pathway in neuropsychiatric disorders, including idiopathic and genetic forms of schizophrenia and ASD^{3,11,18,19,59–69}. Alterations in functional connectivity in the thalamocortical pathway can be observed in idiopathic cases of ASD as early as 6 months after birth³⁹, and our study suggests that defects in this pathway may emerge during prenatal development. Interestingly, overgrowth of thalamocortical axon tracts has been observed in other neuropsychiatric disease models, including Tsc1 deletion in mice⁷⁰, and CACNA1G loss-of-function in iPS-derived organoids³². While many circuit-level mechanisms can lead to thalamocortical overconnectivity⁴, our data suggests one caused by thalamocortical tract formation abnormalities mediated by 22q11.2 microdeletion.

Our study highlights how 22q11DS-associated phenotypes are not necessarily convergent across brain regions. We compared transcriptomic changes associated with the 22q11.2 microdeletion in thalamic and cortical organoids and observed that while some genes were differentially expressed in both models, many others were distinct, including genes implicated in neurodevelopmental disorders. One salient example is FOXP2, which is upregulated in 22q11DS thalamic glutamatergic neurons but not cortical organoids, and mediates excessive outgrowth of thalamocortical and corticothalamic axons. This study showcases how the modularity of organoids can be used to dissect region-specific contributions of disease-associated mutations.

FOXP2 is a high-confidence ASD risk gene best known for its importance in speech and language development^45,46. This gene plays an important role in regulating the development of the cerebral cortex and striatum^71–73, but its role in thalamus development is relatively understudied. Recently, FOXP2 was shown to regulate thalamocortical neuron specification and pathway development⁴⁷. FOXP2 expression is enriched in posterior thalamic nuclei, including the pulvinar, which is involved in transthalamic cortico-cortical communication⁷⁴ implicated in neuropsychiatric disorder pathobiology^75–77. Our findings provide an example for how the 22q11.2 microdeletion can mediate phenotypes via trans effects on the expression of high confidence risk genes implicated in neuropsychiatric disorders.

In summary, our study exemplifies how copy number variants linked to neuropsychiatric disorders can give rise to disease-relevant phenotypes in the thalamus, at least in part by inducing dysregulation of high confidence risk genes for neurodevelopmental disorders located outside the deletion region.

Limitations of the study

Several questions remain regarding the effects of the 22q11.2 microdeletion on thalamocortical circuit development. For example, a previous study²⁸ has reported that the 22q11.2 microdeletion mediates excitability in cortical organoids, and it remains unclear whether this phenotype is present in thalamic organoids and how thalamocortical axonal overgrowth might affect cortical circuit development. Additionally, future studies are needed to systematically interrogate the molecular mechanisms leading to FOXP2 overexpression in 22q11DS thalamus but not cortical neurons, the mechanism and degree to which FOXP2 regulates ROBO2 expression, and the relative contributions of other FOXP2 targets on axon outgrowth.

STAR★METHODS

RESOURCE AVAILABILITY

Lead Contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Dr. Tomasz Nowakowski (tomasz.nowakowski@ucsf.edu) upon reasonable request.

Materials Availability

Plasmids and cell lines generated in this study are available upon request from the lead contact.

Data and Code Availability

scRNAseq data (GSE245719) from cortical and thalamic organoids is available through GEO. scRNAseq data from the second trimester human thalamus is available through NEMO (http://data.nemoarchive.org/biccn/lab/kriegstein/transcriptome/scell/processed/counts/). Genotyping data from deidentified iPS lines is available through dbGAP (phs002624.v3.p1). Processed data related to the figures are available through UCSC Cell Browser: https://cells-test.gi.ucsc.edu/?ds=organoid-22q11 for organoid data and https://dev-thal.cells.ucsc.edu/ for primary reference data. This paper does not report original code, but all code used for analysis has been deposited at Github and is publicly available as of the date of publication. DOIs are listed in the key resources table. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon reasonable request.

KEY RESOURCE TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
TCF7L2 (mouse IgG2a, 1:250)	Millipore	05-511
TCF7L2 (rabbit, 1:100)	Cell Signaling	C48H11
FOXG1 (rabbit, 1:500)	Abcam	ab196868
GBX2 (rabbit, 1:250)	Proteintech	21639-1-AP
PAX6 (rabbit, 1:300)	Biolegend	901301
PAX6 (sheep, 1:500)	R&D Systems	AF8150
EOMES (sheep, 1:300)	R&D Systems	AF6166
EOMES (mouse IgG1k, 1:500)	eBioscience	14-4877-82
TBR1 (rabbit, 1:200)	Abcam	ab31940
SATB2 (mouse, 1:250)	Santa Cruz Biotechnology	sc-81376
OTX2 (goat, 1:100)	R&D Systems	AF1979
FOXP2 (sheep, 1:1000)	R&D Systems	AF5647
FOXP2 (rabbit, 1:1000)	Abcam	ab16046
FOXP2 (mouse IgG1k, 1:250)	Millipore	MABE415
SOX2 (rat, 1:500)	Invitrogen	14-9811-82
HuC/D (mouse IgG2b, 1:250)	ThermoFisher	A21271
NeuN (chicken, 1:500)	Millipore	ABN91
VGluT2 (mouse, 1:200)	Millipore	MAB5504
GFP (chicken, 1:1000)	Aves	1020
dsRed (rabbit, 1:250)	Takara	632496
H3K27me3	Cell Signaling Technologies	9733
IgG	Epicypher	13-0042
Anti-IgG secondary antibody	Antibodies-Online	ABIN101961
Bacterial and Viral Strains
5-alpha Competent E. Coli	New England BioLabs	C2987H
NEB Stable Competent E. Coli	New England Biolabs	C3040H
Chemicals, Peptides, and Recombinant Proteins
FBS	HyClone	SH30071.03
Matrigel	Fisher Scientific	354234
PBS-EDTA, pH 7.5	Lonza	BE02-017F
Trans-Isrib	Tocris	5284
Chroman 1	Medchem Express	HY-15392
Emricasan	Selleckchem	S7775
Polyamine Supplement	Sigma-Aldrich	P8483
IWR1-ε	Cayman Chemical	13659
SB431542	Tocris	1614
Dorsomorphin	Sigma-Aldrich	P5499
LDN193189	Sigma-Aldrich	SML0559
Insulin	Sigma-Aldrich	I9278
PD0325901	Tocris	4192
BMP7	R&D Systems	354-BP-010
BDNF	Alomone	b250
Penicillin-Streptomycin	Gibco	15070-063
Antibiotic-Antimycotic	Gibco	15240-062
Glutamax	Gibco	35050-061
N2 supplement	Gibco	17502-048
B27 supplement w/o Vitamin A	Gibco	12587-001
B27 w/ Vitamin A	Gibco	17504-001
Neurobasal	Gibco	21103049
DMEM/F12 with Glutamax	Gibco	10565042
L-Ascorbic Acid 2-Phosphate	Wako Chemicals USA	321-44823
Insulin	Gibco	A11382ij
Optiferrin	Invitria	777TRF029
Sodium Selenite	Sigma-Aldrich	S5261
FGF2-G3	Northwestern Recombinant Protein Production Core	FGF2-G3
NRG1	Shenandoah	100-46
TGFB3	Qkine	Qk054-0100
DMEM/F12	Corning	10-092-CM
Donkey Serum	Jackson Immuno	017-000-121
Prolong Gold	ThermoFisher	P36934
Glycine	Millipore	410225
1M HEPES	ThermoFisher	H3537
5M NaCl	Sigma-Aldrich	71386
Spermidine	Sigma-Aldrich	85558
EDTA-free complete protease inhibitor tablet	Millipore	11873580001
Concanavalin A-coated beads	Fisher Scientific	NC1526856
1M CaCl2	VWR	B9000S
MnCl2	Sigma-Aldrich	M5005
0.5M EDTA	Fisher Scientific	AM9261
BSA	NEB	B9000S
Digitonin, 5%	Fisher Scientific	BN2006
pA-Tn5 with Nextera Adapters	Epicypher	15-1117
SDS, 10%	Fisher Scientific	AM9822
Proteinase K	Fisher Scientific	FEREO0492
Phenol:chloroform:isoamyl alcohol (25:24:1 v/v)	ThermoFisher	15593031
Chloroform	Fisher Scientific	60-047-878
NEBNext High Fidelity 2X Master Mix	NEB	M0541S
Critical Commercial Assays
10X Chromium V2	10X Genomics	PN-120237
10X Chromium V3	10X Genomics	PN-1000092
Deposited Data
iPS line high density genotyping data	This study, dbGAP	phs002624.v3.p1
Processed single cell RNA-seq data (Thalamic Organoids)	This study, GEO	GSE245719
Processed single cell RNA-seq data (Cortical Organoids)	This study, GEO	GSE245719
Experimental Models: Cell Lines
Human 22q11DS iPS cell line 3577-3 (male)	This study	N/A
Human 22q11DS iPS cell line 5401-35 (male)	This study	N/A
Human 22q11DS iPS cell line 7215-32 (female)	This study	N/A
Human 22q11DS iPS cell line 60C2 (male)	Lachmann Lab78	N/A
Human 22q11DS iPS cell line WTC11-22q11AB Deletion Clone 1 (WTC11-CRISPR1; male)	Wiita Lab79	N/A
Human 22q11DS iPS cell line WTC11-22q11AB Deletion Clone 2 (WTC11-CRISPR2; male)	Wiita Lab79	N/A
Human 22q11DS iPS cell line WTC11-22q11AB Deletion Clone 3 (WTC11-CRISPR3; male)	Wiita Lab79	N/A
Human Ctrl iPS cell line 28126 (male)	Gilad Lab80	N/A
Human Ctrl iPS cell line 20961B (male)	Gilad Lab80	N/A
Human Ctrl iPS cell line WTC-11 (male)	Conklin Lab81,82	CVCL_Y803
Human Ctrl iPS cell line 1323-4 (female)	Conklin Lab83	CVCL_0G84
Experimental Models: Organisms/strains
wt/+ mice, C57BL/6N (male, post-natal day 56)	Taconic Biosciences	B6-M
Df(h22q11/+) mice, C57BL/6N (male, post-natal day 56)	Taconic Biosciences	11026-M
Oligonucleotides
FOXP2 qPCR fwd primer	https://pga.mgh.harvard.edu/primerbank/	AATCTGCGACAGAGACAATAAGC
FOXP2 qPCR rev primer	https://pga.mgh.harvard.edu/primerbank/	TCCACTTGTTTGCTGCTGTAAA
ROBO2 qPCR fwd primer	https://pga.mgh.harvard.edu/primerbank/	CGAGCCCACGACTCTGAAC
ROBO2 qPCR rev primer	https://pga.mgh.harvard.edu/primerbank/	ACACAAACGTAGCTTCCTTCATC
UBC qPCR fwd primer	https://pga.mgh.harvard.edu/primerbank/	GGAGCCGAGTGACACCATTG
UBC qPCR rev primer	https://pga.mgh.harvard.edu/primerbank/	CAGGGTACGACCATCTTCCAG
GAPDH qPCR fwd primer	https://pga.mgh.harvard.edu/primerbank/	GGAGCGAGATCCCTCCAAAAT
GAPDH qPCR rev primer	https://pga.mgh.harvard.edu/primerbank/	GGCTGTTGTCATACTTCTCATGG
Recombinant DNA
pMD2.G	Addgene	12259
psPAX2	Addgene	12260
pcDNA 3.1 puro Nodamura B2	Addgene	17228
pSico-CAG::GFP	Nowakowski Lab	N/A
pSico-CAG::dTomato	Nowakowski Lab	N/A
pLV-U6::sgRNAFOXP2.1-UbC::dCas9-KRAB-2A-GFP	Nowakowski Lab	N/A
pLV-U6::sgRNAFOXP2.2-UbC::dCas9-KRAB-2A-GFP	Nowakowski Lab	N/A
pLV-U6::sgRNAROBO2-UbC::dCas9-KRAB-2A-GFP	Nowakowski Lab	N/A
pLV-U6::sgRNAnon-targeting-UbC::dCas9-KRAB-2A-GFP	Nowakowski Lab	N/A
pSico-CAG:: FOXP2-2A-GFP-WPRE	Nowakowski Lab	N/A
pSico-CAG:: DGCR8-2A-GFP-WPRE	Nowakowski Lab	N/A
Software and Algorithms
ImageJ (Fiji)	ImageJ84	https://imagej.net/Fiji
CellRanger v3.0	10X Genomics	https://support.10xgenomics.com/single-cell-gene-expression/software/pipelines/latest/what-is-cell-ranger
Seurat v3.0	Satija Lab85	https://satijalab.org/seurat/
Signac	Satija Lab86	https://github.com/timoast/signac
Affinity Designer	Affinity	https://affinity.serif.com/en-us/designer/
BioRender	BioRender	https://biorender.com/
Github	This Study	https://github.com/dmshin14/22q11DSProject
Other
6-well Ultralow Attachment Plates	Corning	3471
Cell Culture Inserts, 0.4 μm pore size	Millipore	PICM03050
Lipofectamine 3000	ThermoFisher	L3000015
Phase Lock Tube	Qiagen	129046
SPRIselect beads	Beckman Coulter	B23317
PhiX Control v3	Illumina	FC-110-3001
NextSeq 2000 P2 100 cycle	Illumina	20046811
Neon Transfection System 100μL Kit	ThermoFisher	MPK10096

EXPERIMENTAL MODELS AND STUDY PARTICIPANT DETAILS

Derivation and Culture of Human Induced Pluripotent Stem Cells

All of the work related to human iPS cells has been approved by the UCSF Committee on Human Research and the UCSF GESCR (Gamete, Embryo, and Stem Cell Research) Committee. We have reprogrammed three iPS cell lines (3577–3, 5401–35, 7215–32) from fibroblasts derived from 22q11DS patients that were purchased from the Coriell Cell Repository (GM03577, GM05401, GM07215). 3 × 10⁵ fibroblasts were electroporated with the Y4 mixture of episomal plasmids encoding the reprogramming factors OCT3/4, SOX2, KLF4, L-MYC, LIN28, and an shRNA for TP53 using the Neon Transfection System 100 μl kit. Electroporations were performed using three pulses of 1,650 V and pulse width of 10 ms. 10-cm dishes that contained a feeder layer of irradiated or mitomycin-C treated SNL cells were seeded with 1–3 × 10⁵ electroporated fibroblasts, and cultured in primate ESC medium (Reprocell) containing FGF2. Colonies that resembled stem cells based on morphology were chosen for expansion and evaluation by immunostaining and karyotyping 25 days post-electroporation. We acquired an additional iPS cell line derived from 22q11DS patient fibroblasts from the lab of Dr. Herbert Lachmann (60c2), three CRISPR-engineered iPS cell lines derived from the WTC11 control line from the lab of Arun Wiita⁷⁹, and four control iPS cell lines of typical karyotype (WTC11, 28126, 1323–4, 20961B) that had been similarly derived from fibroblasts using episomal plasmids from the labs of Dr. Bruce Conklin and Dr. Yoav Gilad. We have confirmed that all of the iPS cell lines stain positive for pluripotency markers and exhibit the expected genotype and karyotype (Figure S1, see also⁷⁹). All of our experiments were performed using human iPS cell lines cultured at 37°C in a humidified incubator on Matrigel-coated plates in B8 media supplemented with 2.5% BSA and 1% EmbryoMax Nucleosides⁸⁷ and passaged using Biowhittaker PBS-EDTA (Lonza). Sex and other relevant metadata is reported in Table S1.

Mice

All activities related to animal husbandry and mouse procedures were conducted at the Laboratory Animal Resource Center (LARC) facilities at UCSF Mission Bay, in compliance with the guidelines outlined by the Institutional Animal Care Use Committee (IACUC) protocol #AN192726–01G. df(h22q11)/+ and wt/+ male mice that were 8 weeks of age and on the C57BL/6N background were purchased from Taconic Biosciences. Upon arrival, mice were housed in a controlled environment (70°C, 50% rack humidity) in a barrier facility with 12-hour light and 12-hour dark cycles. Mice were anesthetized with 40mg/mL tribromoethanol (Avertin) diluted in saline and administered at 0.1mL/10g of body weight. After anesthetization, mice were transcardially perfused with ice-cold PBS followed by 10mL of ice-cold 4% PFA. Brains were dissected out of the skull and submerged in 4% PFA at 4C for 24 hours followed by sucrose dehydration and OCT embedding described in Method Details.

METHOD DETAILS

Genotyping Analysis of iPS Cells

High density genotyping was performed on iPS cells using the Infinium Global Screening Array at the UC Berkeley QB3 Genomics Core. Each sample’s B-allele frequency and probe-level log R ratio was calculated using the Illumina GenomeStudio software (version 2.0.4).

Polygenic Risk Score Analysis of iPS Cells

Pre-imputation marker quality control (QC):

Samples were genotyped on the Illumina Infinium Global Screening Array-24 v3.0. We used PLINK⁸⁸ v1.9 [www.cog-genomics.org/plink/1.9/] and PLINK 2.0 [www.cog-genomics.org/plink/2.0/] to conduct QC. Across all 47 samples, single nucleotide variants (SNPs) were screened on the basis of call rate < 90%, out of Hardy-Weinberg Equilibrium (P < 0.0001), or with a minor allele frequency (MAF) < 0.01. We then subset to nine samples of interest for polygenic scoring, and further removed palindromic/ambiguous SNPs, small insertions and deletions, duplicate SNPs, multiallelic variants, and sex chromosomes. Given the small number of samples, SNPs with any differential missingness between four cases and four controls were removed (N=2,610).

Pre-imputation sample QC:

All samples had a genotype call rate > 99%. We identified and removed one duplicate sample. We merged the remaining eight samples with unrelated samples from the 1,000 Genomes Project (N=2,581)⁸⁹ to conduct a principal component analysis (PCA) of genotypes, and assign ancestry. SNPs for PCA (N=196,642) were those overlapping between the eight samples and 1,000 Genomes samples and LD-pruned (----indep-pairwise 50 5 0.2), including for regions of long-range LD⁹⁰. We identified and removed two samples with non-European ancestry, WTC11 and 5401–35 (Table S1).

SNPs were aligned to the TOPMED imputation panel⁹¹ using the imputation preparation and checking tool (v4.3.0) (https://www.well.ox.ac.uk/~wrayner/tools/). After QC and imputation preparation, 369,960 SNPs and six samples remained for imputation.

Post-imputation filtering:

Imputed SNPs were filtered to those with an R² ≥ 0.8. SNPs were then subset to those in the 1,000 Genomes Project. SNPs were then lifted over to GRCh37 using liftOver⁹² to match GWAS summary statistics for polygenic scoring and filtered to those with a MAF > 5% (N=5,298,358 SNPs).

Scoring:

We conducted polygenic scoring of six samples and 525 unrelated 1,000 Genomes European super-population individuals using LD clumping and p-value thresholding, also known as C+T (PRSice 2.3.5)⁹³. Summary statistics for autism spectrum disorder (ASD) and schizophrenia were taken from Grove et al., 2019⁹⁴ and Trubetskoy et al. 2022³⁸, respectively. Summary statistics were LD-pruned for independent sites (r² < 0.1), and a GWAS p-value threshold of p < 0.1 for ASD and p < 0.05 for schizophrenia; these thresholds were previously identified as explaining the most variance in the respective traits^38,94. This left 32,738 SNPs for the ASD PGS and 30,502 SNPs for the schizophrenia PGS. To control for differences in PGS due to differences in ancestry, we regressed the PGSs on the first 20 principal components from PCA. The residuals were then centered and scaled using the mean and standard deviation of the 1,000 Genome European-ancestry controls PGS distribution (N=525).

Cerebral Organoid Differentiation and Culture

Cortical and thalamic organoids were generated from iPS cells using recently published protocols^31,95. To generate cortical organoids, iPS cells were dissociated with PBS-EDTA and plated in an ultra-low attachment 6-well plate containing a DMEM/F12-based induction media containing 15% KSR, 1% MEM-NEAA, 1% Glutamax, 1% Pen-Strep, 100 μM β-Mercaptoethanol, 5 μM Dorsomorphin, 5 μM SB-431542, 3 μM IWR1-endo, and 1X CEPT apoptosis inhibitor cocktail (50 nM chroman 1, 5 μM emricasan, 0.1% polyamine supplement, and 0.7 μM trans-ISRIB). Neural induction media was replaced every two days for eight days, and Y27632 was removed from the media on the fourth day. After neural induction, plates containing cortical organoids were transferred to a plate shaker rotating at 80 rpm. From Days 8–18, cortical organoids were cultured in an expansion media comprised of a 1:1 mixture of DMEM/F12 and Neurobasal medium containing 1% N2, 2% B27 without vitamin A, 1% Glutamax, 1% MEM-NEAA, 55 mM β-Mercaptoethanol, 1% Antibiotic-Antimycotic, 10 ng/mL FGF2, and 10 ng/mL EGF, with a media exchange every two days. After Day 18, cortical organoids were moved into a maintenance media comprised of a 1:1 mixture of DMEM/F12 and Neurobasal medium containing 1% N2, 2% B27 with vitamin A, 1% Glutamax, 1% MEM-NEAA, 55 mM β-Mercaptoethanol, 1% Antibiotic-Antimycotic, and 200 μM ascorbic acid.

To generate thalamic organoids, iPS cells were dissociated with PBS-EDTA and resuspended in a DMEM/F12-based induction media containing 15% KSR, 1% MEM-NEAA, 1% Glutamax, 1% Pen-Strep, 100 μM β-Mercaptoethanol, 100 nM LDN-193189, 10 μM SB-431542, 4 μg/mL Insulin, and 20 μM Y27632. Induction media was replaced every two days, and Y27632 was removed on the fourth day. After eight days of induction, plates containing thalamic organoids were transferred to a plate shaker rotating at 80 rpm. From Days 8–16, thalamic organoids were cultured in a patterning media comprised of a 1:1 mixture of DMEM/F12 and Neurobasal medium containing 1% N2, 2% B27 without vitamin A, 1% Glutamax, 1% MEM-NEAA, 55 mM β-Mercaptoethanol, 1% Antibiotic-Antimycotic, 30 ng/mL BMP7, and 1 μM PD325901, with a media exchange every two days. After Day 16, thalamic organoids were moved into the maintenance media described above for cortical organoids.

Molecular cloning of plasmid vectors

For knockdown experiments, CRISPRi sgRNAs were chosen based on their predicted specificity and efficacy. The sequences for sgRNAs are as follows: FOXP2 sgRNA #1 - GGAGCCGGGAGACCAGACAC; FOXP2 sgRNA #2 - GTCAGTCTGGGACGTGATCG; ROBO2 sgRNA - GCCGAACTCCTACGTGGTGA; Non-targeting control sgRNA from Horlbeck et al. 2016⁹⁶ - GCGCCAAACGTGCCCTGACGG. Oligonucleotides containing guide sequences were synthesized through a commercial source (IDT), annealed, and cloned into the plasmid pLV hU6-sgRNA hUbC-dCas9-KRAB-T2a-GFP (Addgene Plasmid #71237) that was digested using the BsmBI restriction enzyme. Lentivirus was produced as described in the previous section. sgRNAs were validated for knockdown efficiency by transducing HEK293T cells at an MOI of 3 with lentivirus and performing qPCR to compare expression of the gene of interest between cultures transduced with FOXP2- or ROBO2-targeting sgRNAs versus non-targeting control.

Lentivirus Production

HEK293T cells cultured in DMEM containing 10% FBS and 1% Pen-Strep were transfected with the second generation lentiviral constructs pMD2.G, psPAX2, and a transfer vector of interest (1:1:2 molar ratio for plasmids less than 10 kb, and 1:1:3 molar ratio for plasmids greater than 10 kb) using Lipofectamine 3000 according to the manufacturer’s protocol. The next day, media was exchanged with Ultraculture media (Lonza) supplemented with 2% FBS, 1% Glutamax, 1% Pen-Strep, 5 mM Sodium Butyrate, and 110 μg/mL Sodium Pyruvate. Viral supernatant was collected 48 hours later and concentrated by ultracentrifugation with a 30% sucrose cushion using the Beckman Ultra SW32Ti rotor at 22,000 rpm for two hours at 4°C. Lentivirus was titered by transducing HEK293T cells using a limiting dilution and quantifying the number of fluorescent cells in wells transduced between 10–20% efficiency.

Viral Labeling and Generation of Thalamocortical Organoids

At day 30 of differentiation, cortical and thalamic organoids were sectioned on an automated vibrating microtome (Leica VT1200S) at 300 μm thickness and plated onto organotypic slice culture inserts. The resulting slices were injected with ~1 μl of lentivirus (1 × 10⁸ TU/mL) using a glass micropipette and cultured with complete media exchange every 2–3 days. At day 35, thalamic and cortical organoid slices were rinsed three times in maintenance media, to ensure removal of residual virus. To generate thalamocortical organoids, thalamic and cortical organoid slices were cultured adjacent to each other on cell culture inserts (Millipore PICM03050) coated with rat tail collagen I (10 μg/cm², 1 hour at room temperature).

In axon outgrowth experiments described in Figures 1 and 2, thalamic and cortical organoids were transduced with CAG∷EGFP-WPRE and CAG∷dTomato-WPRE lentivirus, respectively. Isogenic thalamocortical organoids were generated by plating thalamic and cortical slices derived from the same line. Chimeric thalamocortical organoids were generated by plating together thalamic and cortical slices derived from opposing genotypes. In knockdown experiments, thalamic organoids were transduced with hU6-sgRNA hUbC-dCas9-KRAB-T2a-GFP lentivirus expressing targeting or non-targeting sgRNA control. Given that GFP expression was not visible without immunostaining, sgRNA-labeled thalamic organoids were co-transduced with CAG∷dTomato-WPRE lentivirus to visualize axons. After transduction, thalamic organoids were fused to unlabeled cortical organoids derived from the same cell line. In FOXP2 overexpression experiments, thalamic organoids were transduced with CAG∷GFP-2A-FOXP2-WPRE or CAG∷GFP-WPRE mock control lentivirus and fused to unlabeled or CAG∷dTomato-labeled cortical organoids derived from the same cell line.

Generation of Co-cultures of Thalamic Organoids and Slice Cultures from Human Cortex

300-μm thick sections of human cortex (GW18–21) were generated using an automated vibrating microtome (Leica VT1200S) and plated onto organotypic slice culture inserts. Two thalamic organoids, one carrying the 22q11 deletion and labeled with EGFP and one control labeled dTomato, were plated in close proximity to the ventricular zone of the primary cortical slice. These co-cultures were maintained in organoid maintenance media described above and fixed one week later using 4% paraformaldehyde (PFA) for one hour at 4°C.

Quantification of Axon Outgrowth in Thalamocortical Organoids

Tilescans of thalamocortical organoids were collected using a Leica SP8 laser scanning confocal microscope with a 10X air objective, with optical sections collected at every 2.4 μm. Imaging was performed under physiological conditions (37°C, 5% CO2) using an environmental chamber. Confocal microscopy images of axon outgrowth in thalamocortical organoids were processed in ImageJ (Fiji). To visualize axons, z-stacks were projected by taking the sum of fluorescence intensity in ImageJ (Fiji). The brightness of grayscale images were adjusted until all of the thalamocortical projections were clearly visible, and the colors were inverted. We identified and marked the border between the cortical and thalamic organoids on the GFP or dTomato image, which was clearly visible based on brightfield images. Then, we defined a boundary in the organoid that the axons were migrating into that was 500 μm (fused for 1 week) or 1mm (fused for 5 weeks) away from the cortex-thalamus boundary and ran parallel to the border. We manually counted the number of axons that crossed this boundary in each of our co-cultures using the CellCounter plug-in in Fiji. Axon counts were normalized to the number of fluorescently labeled cells in the organoid from which the axons originated, which were quantified using QuPath software.

Single Cell mRNA Sequencing

Organoids were incubated in 10 U/mL papain (Worthington) containing 500 ng/mL DNAse I for 30 minutes at 37°C, and dissociated using manual trituration with a fire-polished glass pipette in 10% FBS containing DMEM. Dissociated cells were passed through a 40 μm nylon mesh filter. 5×10⁵ cells were centrifuged at 300×g for four minutes at 4°C, washed once with 1mL ice-cold PBS without Calcium or Magnesium, and passed through a 40 μm nylon mesh filter. Barcoded samples were pooled and 5 × 10⁴ cells total were loaded for each microfluidic lane for droplet-based single cell RNA sequencing using the 10X Genomics Chromium Single Cell 3’ Reagent Kits (v3) according to manufacturer instructions.

Sample Demultiplexing, Dataset Integration, and Clustering Analysis

Read alignment and generation of feature-barcode matrices were generated using CellRanger⁹⁷. Pooled samples were demultiplexed using SNP genotype data following reference genotyping using the Vireo package⁹⁸. Feature-barcode matrices were processed using the Seurat package⁹⁹ in R for clustering. Prior to clustering, we first removed cells that contained fewer than 1000 genes or greater than 10% of reads that corresponded to mitochondrial genes. For each dataset corresponding to a single batch, which we define as scRNAseq experiments that were run on the same 10X Chromium chip, we performed principal components analysis to identify the major sources of variance, followed by nearest neighbor embedding in the space of significant principal components. Significant principal components were determined using the ElbowPlot method. To determine molecularly distinct clusters, we performed Louvain clustering for community detection based on Jaccard similarities between pairs of cells. To identify the top genes with enriched expression in every cluster, we performed differential gene expression analysis using Wilcoxon rank-sum test. We removed cells corresponding to clusters defined primarily by ribosomal or mitochondrial genes. In addition, we removed clusters associated with off-target brain regions, i.e. hindbrain, midbrain, and cell types that were not reproducibly generated across lines, i.e. endothelial cells, pericytes, OPCs. After this series of quality control filtering, we pooled data across multiple batches and performed dataset integration using the Harmony workflow¹⁰⁰, or reference mapping using the Symphony workflow¹⁰¹. We also incorporated previously published scRNAseq data generated from cortical organoids²⁸ (GSE145122) during dataset integration.

Differential Gene Expression Analysis and Gene Ontology Analysis

Differential gene expression analysis was performed to compare 22Q11DS and control cells within each cell type identified in the thalamic and cortical organoid datasets. Cells corresponding to the cell type of interest were subsetted from the rest of the data, and we identified differentially expressed genes in 22Q11DS cells using MAST¹⁰². We removed genes from the output of this analysis whose average log fold change was less than 0.1 and adjusted p-value was less than 0.05. Next, we performed a variancePartition¹⁰³ analysis to identify the contribution of Batch effect or effect of a single individual for each differentially expressed gene using the VariancePartition package in R. We removed all genes whose differential expression was driven by greater than 25% batch effect and 25% individual effect. We performed gene ontology analysis on the remaining genes. To identify aberrant molecular pathways or biological processes, we performed gene ontology analysis using the clusterProfiler package¹⁰⁴.

DEGs identified from our organoid scRNAseq datasets were compared with existing psychiatric disease genes. These include genes associated with rare coding variants from exome sequencing studies for schizophrenia³⁷ and ASD⁴⁶, GWAS variants for schizophrenia³⁸, ASD^94,105, and bipolar disorder¹⁰⁶, postmortem genes previously identified using bulk RNAseq for schizophrenia, ASD, and bipolar disorder¹⁰⁷, and genes identified in bulk RNAseq from an iPS-derived cortical neuron model of 22q11DS²⁹. For bulk RNAseq datasets, we considered a gene to be differential expressed based on cutoff of adjusted p-value less than 0.05 and the absolute value of average log fold change greater than 0.25.

Tissue processing and Immunohistochemistry

Organoids, primary human thalamus, and hemisected mouse brain hemispheres were fixed using 4% paraformaldehyde (PFA) for one hour at room temperature (organoids) or for 24 hours at 4°C (tissue). Organoids and tissue were subsequently incubated in PBS containing 30% sucrose for 48 hours, and embedded in a 1:1 mixture of 30% sucrose and OCT (Tissue-Tek). Organoids were serially sectioned at 12 μm onto glass slides using a cryostat. Human thalamus tissue and the right hemisphere of mouse brains were serially sectioned at 12 μm and 25 μm, respectively, in the sagittal orientation onto glass slides. Mouse brain sections containing thalamus were manually aligned to the Allen Institute Brain Atlas for P56 sagittal mouse brain (http://mouse.brain-map.org/) using the position of the thalamus in relation to the hippocampus as a reference, which was visualized using a stereomicroscope. Mouse brain sections corresponding to slides 15 and 16 on the reference atlas were selected at random for analysis.

To perform antigen retrieval, slides were incubated for 15 minutes in 10 mM sodium citrate (pH 6) that had been heated to boiling temperature in a pressure cooker. After antigen retrieval, slides were washed briefly in 1X PBS and incubated for one hour in a blocking buffer comprising PBS containing 10% Donkey Serum, 2% Gelatin, and 0.1% Triton-X. Slides were subsequently incubated overnight at 4°C in primary antibodies diluted in the blocking buffer. The next day, slides were washed three times over the course of 30 minutes in PBS containing 0.1% Triton-X, incubated for one hour at room temperature in secondary antibodies diluted at 1:1000 in the blocking buffer, and washed again three times in PBS containing 0.1% Triton-X. Slides were treated with 300 nM DAPI for five minutes and mounted in a Prolong Gold antifade mounting medium. Primary antibodies and their concentrations are listed in the Key Resource Table.

Quantification of Immunohistochemistry Images

All images were collected on a Leica SP8 confocal microscope and processed in ImageJ (Fiji) using the CellCounter plug-in. For organoids, images were collected at 20X magnification from 2–3 cryosections per organoid, with three to five fields-of-view collected per cryosection, across two batches of differentiation. Quantifications per field-of-view were averaged by organoid. For FOXP2 and SOX2 staining in mouse brain, a tilescan was collected across three cryosections per brain, while keeping the anatomical locations and regions of interest consistent across biological replicates. For PKCδ staining, a tilescan was collected across eight cryosections per brain. The background was corrected by thresholding and a region of interest was manually drawn in the thalamus (FOXP2 and SOX2 stain) or along the perimeter of the dorsal striatum (PKCδ). The area fraction was quantified to represent the percentage coverage of of PKCδ+ thalamocortical axons.

CUT&Tag

CUT&Tag was performed on dissociated organoids as previously described⁵² with some modifications. Briefly, 200,000 cells per reaction were pelleted at 600g for 3 minutes at room temperature and resuspended and fixed in PBS with 0.1% formaldehyde for 2 minutes. 1.25M glycine was added to double the molar concentration of formaldehyde and stop cross linking, and cells were spun down at 1300g at 4C for 3 minutes. Cells were resuspended in wash buffer (20mM HEPES pH 7.5, 150mM NaCl, 0.5mM spermidine, and 1 EDTA-free complete protease inhibitor tablet). Concanavalin A-coated beads (Fisher Scientific #NC1526856) were prepared by adding 10μL beads per reaction to bead-binding buffer (20mM HEPES pH 7.9, 10mM KCl, 1mM CaCl2, and 1mM MnCl2). Using a magnetic rack, binding buffer was removed, and beads were resuspended in binding buffer once more before removing the binding buffer again and finally resuspending in enough binding buffer for 10μL per reaction. 10μL beads were then added to cells and incubated on an end over end rotator for 10 minutes at room temperature. Wash buffer was removed from cells using a magnetic rack, and cells were resuspended in enough antibody buffer (wash buffer with 2mM EDTA, 0.1% BSA, and 0.05% digitonin) for 50μL per reaction. Primary antibodies (FOXP2 Abcam ab16046, H3K27me3 Cell Signaling Technologies #9733, IgG Epicypher #13-0042) were added at 1:50 dilution and samples were nutated overnight at 4C. Using a magnetic rack, primary antibody was removed, and cells were resuspended in 100μL secondary antibody (Antibodies Online #ABIN101961) diluted in wash buffer with 0.05% digitonin. Cells were then nutated for one hour at room temperature. Secondary antibody mix was removed using a magnetic rack, and cells were washed with wash buffer with 0.05% digitonin three times. pA-Tn5 preloaded with Nextera adapters (Epicypher #15–1117) was diluted 1:20 in dig-300 buffer (20mM HEPES pH7.5, 300mM NaCl, 0.5mM spermidine, 0.015% digitonin with 1 EDTA-free complete protease inhibitor tablet) and cells were resuspended in 50μL of pA-Tn5 mix. Cells were nutated for one hour at room temperature. Using a magnetic rack, pA-Tn5 mix was removed, and cells were washed three times in dig-wash buffer. Cells were resuspended in 300μL of tagmentation buffer (dig-300 buffer with 10mM MgCl2) and incubated in a 37C water bath for one hour. Cells were released from beads with addition of 10μL 0.5M EDTA, 3μL 10% SDS, and 2.5μL 20mg/ml proteinase K. Samples were vortexed and incubated in a heat block at 55C for one hour. Fragments were purified by adding 300μL phenol:chloroform:isoamyl alcohol (25:24:1 v/v) and sample to a phase lock tube (Qiagen #129046), and spun down at 16,000g for 3 minutes. 300μL chloroform was added to each sample and spun down once more at 16,000g for 3 minutes. The aqueous layer was added to a 1.5ml lo-bind tube with 750μL 100% ethanol and mixed by pipetting. Samples were cooled on ice and spun down at 16,000g for 15 minutes at 4C. Supernatant was decanted, and samples were washed once more with 1ml 100% ethanol and spun down at 16,000g for 1 minute at 4C. Ethanol was decanted, the pellet was air dried, and resuspended in 22μL water. Libraries were amplified by mixing 21μL sample, 2μL each i5 and i7 primers (Illumina #FC-131–2001), and 25μL NEBNext High Fidelity 2X Master Mix (NEB #M0541S) and using the following PCR cycle settings: 72C for 5 minutes, 98C for 30 seconds, 98C for 10 seconds, 61C for 10 seconds, repeat steps 3–4 15x, and a final 72C incubation for 1 minute. Libraries were purified using SPRI Select Reagent (Beckman Coulter #B23317) and eluted in 20μL water. Libraries were pooled to 2nM and diluted to a final concentration of 750pM with 2% PhiX spike in for sequencing using the Illumina NextSeq 2000 with a targeted read depth of ~10 million reads per sample. 2 technical replicates were used per sample.

CUT&Tag analysis

Reads were trimmed using TrimGalore, aligned to reference genomes for hg38 and E. coli using bowtie2 with parameters “--end-to-end --very-sensitive --no-mixed --no-discordant --phred33 -I 10 -X 700.” Duplicates were marked and removed with picard. Peaks were called using MACS2 using the corresponding IgG sample as a control, with q = 0.01.

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis was performed in RStudio. The statistical test, sample size, and p-value for each experiment are described in the figure legends and Results section of the manuscript. Statistical significance was defined as a p-value less than 0.05 after correction for multiple comparisons when warranted.

Supplementary Material

2

Table S2. Cell type markers for week 10 thalamic and cortical organoid scRNAseq data. Related to Figure 1 and Figure S1.

3

Table S3. Differentially expressed genes in 22q11DS thalamic and cortical organoid cell types with disease association annotations. Related to Figure 2 and Figure S2.

4

Table S4. Annotation of FOXP2 CUT&Tag peaks in week 5 thalamic organoids. Related to Figure 4.

Highlights:

• Gene signatures in iPS-derived thalamic organoids resemble those in human thalamus

• 22q11.2 deletion mediates transcriptomic changes linked to psychiatric disorders

• 22q11.2 deletion promotes thalamocortical axon overgrowth in organoids and mice

• Elevated FOXP2 in 22q11DS thalamic organoids causes thalamocortical axon overgrowth