Cerebral organoids containing an AUTS2 missense variant model microcephaly

Abstract

Variants in the AUTS2 gene are associated with a broad spectrum of neurological conditions characterized by intellectual disability, microcephaly, and congenital brain malformations. Here, we use a human cerebral organoid model to investigate the pathophysiology of a heterozygous de novo missense AUTS2 variant identified in a patient with multiple neurological impairments including primary microcephaly and profound intellectual disability. Proband cerebral organoids exhibit reduced growth, deficits in neural progenitor cell (NPC) proliferation and disrupted NPC polarity within ventricular zone-like regions compared to control cerebral organoids. We used CRISPR-Cas9-mediated gene editing to correct this variant and demonstrate rescue of impaired organoid growth and NPC proliferative deficits. Single-cell RNA sequencing revealed a marked reduction of G1/S transition gene expression and alterations in WNT-β-catenin signalling within proband NPCs, uncovering a novel role for AUTS2 in NPCs during human cortical development. Collectively, these results underscore the value of cerebral organoids to investigate molecular mechanisms underlying AUTS2 syndrome.

Keywords: AUTS2 syndrome; CRISPR-Cas9; cerebral organoids; microcephaly; neurodevelopmental disorder.

Figure 1

Identification of a de novo AUTS2 variant within a conserved histidine-rich domain. (A) Family pedigree. Circle: female; square: male; filled: affected. (B and C) MRI of affected individual showing reduced cortical area, ventriculomegaly (asterisk) and cerebellar atrophy (arrow). MRI images are arranged in sagittal and coronal planes, respectively. Scale bar = 2 cm. (D) Sanger-based DNA sequence chromatogram of the c.1600A>C AUTS2 variant in exon 9 in the affected patient and absent in her unaffected, healthy parents. (E) AUTS2 genomic organization showing disease-causing variants, key domains and exon structure of the AUTS2 gene. Pathogenic and likely pathogenic variants from the ClinVar database as of January 2021 are represented by circles and shaded according to variant classes. Proline-rich domains (PR1/PR2; containing nonsense and frameshift variants) and histidine-rich domains (H1/H2; containing predominately missense variants) were obtained from UniProt (entry Q8WXX7). The curated AUTS2 protein family domain from PFAM is denoted as AUTS2). Exon locations and numbering in the bottom panel reflect the canonical full-length transcript (NM_015570.4). (F) AUTS2 amino acid conservation within the H1 domain and spanning position 534 across multiple species. Arrow represents the AUTS2 T534P alteration in the affected patient.

Figure 2

AUTS2^T534P patient COs show reduced growth and proliferative deficits. (A) Overview of protocol used to generate COs. Briefly, PBMCs were isolated from patient and control blood samples and underwent Sendai virus transduction with Yamanaka factors (OCT3/4, SOX2, KLF4, c-MYC) to produce iPSCs. These cultures were characterized and then used to generate COs through undirected, spontaneous differentiation. A was created with BioRender.com. (B and C) Representative images of parental control COs and proband COs at Day 34 showing a significant growth reduction in proband COs. Scale bar = 1 mm. (D) Growth trajectory analysis of parental control and proband COs from Days 16 to 37 in CO development. (E and F) Parental control COs show proliferating neural progenitors identified as EdU+ and phospho-Histone H3+ (pHH3; mitosis marker) compared to proband COs, which show a reduction (quantified in G and H). Scale bar = 50 µm. DAPI was used to stain all nuclei (shown in blue). All data are shown as the mean ± standard deviation (SD). Statistical analyses in D were performed using Mann–Whitney U-tests and those in G and H were performed using one-way ANOVA with Tukey’s multiple comparisons test (n = 4 independent organoids per group and two independent experiments performed). Statistically significant differences between parental control and proband COs were observed between groups commencing at Day 20. *P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001.

Figure 3

AUTS2^T534P patient COs show functional and molecular deficits in NPCs. (A) The fate of neural progenitor daughter cells correlates with their mitotic division class. Symmetric divisions (top) give rise to two daughter apical progenitors whereas asymmetric divisions (bottom) give rise to a single apical progenitor and one mature cell type (intermediate progenitor or neuron). A was created with BioRender.com. (B and C) Representative mitotic divisions at the apical surface of parental control and proband rosettes of COs. The division class is determined by the acute angle, θ_a, that forms between the cleavage plane (defined by TPX2+ mitotic poles) and the apical surface (marked by ZO1, membrane protein seen as a mostly continuous cell layer). NPCs are stained for SOX2. Scale bar = 50 and 10 µm for lower and higher magnification images, respectively. (D) Asymmetric oblique and vertical divisions are overrepresented in proband COs (65%) compared to parental control COs (36%). (E and F) Acetylated tubulin (AcTub) staining reveals normal rosette microtubular organization within parental control COs, but severely disrupted organization in proband COs (G and H). Scale bar = 250 µm. (I) SOX2+ neural progenitors within rosettes of parental control COs form robust, uniform ARLB13B+ cilia at the apical surface (central ring-like structure), whereas those within rosettes of proband COs show shortened and irregular arrangement (J). Scale bar = 10 µm. (K) Proband cilia show a statistically significant reduction in length compared to parental control cilia. All data are shown as the mean ± SD. Statistical analysis of cilia quantification data was performed using an unpaired t-test (n = 540 parental control cilia and n = 191 proband cilia quantified across four independent organoids per group and one independent experiment performed). ****P ≤ 0.0001.

Figure 4

AUTS2^T534P gene editing with CRISPR-Cas9 rescues proband COs phenotypes. (A) CRISPR-Cas9-mediated homology-directed repair gene correction strategy. (B) Chromatogram generated by Sanger sequencing of affected patient (AUTS2 c.1600A>C, proband) and GC hiPSC line. First arrow denotes the C>A base pair change at c.1600 and the second arrow denotes the silent gene edit (G>A) at c.1608. (C–G) Schematic showing major steps in an optimized protocol to generate reproducible COs from hiPSCs with representative culture phase contrast microscopic images below each step. Scale bar = 200 and 500 µm for D and E–G, respectively. A and C were created with BioRender.com. (H, L and P) Representative images of parental control, proband and GC control COs at Day 30, cross-sectional area of each CO group quantified in K. (I, M and Q) Decreased percentage of EdU+ and phospho-Histone H3+ (pHH3; magenta, mitosis marker) progenitors in proband CO rosettes compared to parental and GC controls. Scale bar = 200 µm. Magnified in J, N and R, scale bar = 50 µm; quantified in O and S. All data are shown as the mean ± SD. Statistical analyses were performed using one-way ANOVA with Tukey’s multiple comparisons test (n = 14 rosettes quantified across a minimum of four independent organoids per group and one independent experiment performed). ***P ≤ 0.001; ****P ≤ 0.0001; ns = not significant.

Figure 5

The AUTS2 T534P variant disrupts interaction with P300. (A) Schematic of Duolink proximity ligation assay (PLA) depicting detection of AUTS2 interacting with P300. (B) Representative images of CO rosettes with nuclear staining shown in the first column (DAPI), PLA signal (demonstrating AUTS2-P300 interaction) in the second column, and merged DAPI and PLA signal in the third column. (C) Quantification of the PLA intensity per nucleus. (D) Model showing the results from the Duolink PLA assay in which AUTS2-P300 interaction is disrupted in the proband, leading to the decreased PLA signal intensity per nucleus. All data are shown as the mean ± SD. Statistical analyses were performed using one-way ANOVA with Tukey’s multiple comparisons test (n = a minimum of 150 nuclei quantified across at least three rosettes per group).

Figure 6

Single-cell RNA sequencing reveals an under-represented population of proliferative NPCs in proband COs. (A) UMAP plot of key cell types organized into five major clusters: Intermediate progenitors, immature neurons and three classes of NPCs containing 17 752 cells. (B) Select canonical markers used to determine cluster identities. MKI67 and TOP2A show enriched expression within type 1 NPCs, HOXA2 and OLIG3 show enriched expression in type 2 neural progenitor cells and SOX9 shows enriched expression in type 3 NPCs; SOX2 is a pan-progenitor marker; EOMES (TBR2) labels intermediate progenitors; STMN2 and TBR1 label immature neurons. (C, F and I) UMAP plots of cells from parental control COs (8523 cells), proband COs (3754 cells) and GC control COs (5475 cells). Type 2 neural progenitors are outlined in each plot to highlight their under-representation in proband COs. (D, G and J) Percentages of cell types per group. (E) GO terms identified in type 2 neural progenitors enriched for genes associated with G1/S cell cycle phase transition (source: Metascape). (H) Feature plots showing expression of G1/S cell cycle genes in proband and GC control COs.

Figure 7

Deficits in WNT-β-catenin pathway gene expression in AUTS2 patient COs. (A) GO terms identified in type 1 neural progenitors enriched for genes associated with the WNT-β-catenin signalling among others. (B) Violin plots of DEGs in type 1 neural progenitors shows chromatin-modifying genes: HIST1H2AG, KCNQ1OT1 and HMGA1; transcriptional regulator cortical layer 5 gene, TSHZ2, identified as an AUTS2 target gene; and genes associated with neural and glial differentiation: FGFR3, NFIB, MAP2; and cell adhesion: CNTNAP2, also identified as an AUTS2 target gene. (C) Heat maps of type 1 neural progenitor DEGs in proband and control COs show alterations in gene expression associated with WNT-β-catenin signalling. Feature and violin plots of CTNNB1 gene expression in types 1 and 3 neural progenitors, and immature neurons shows reduced expression in proband COs compared to controls. (D) Schematic showing the WNT-β-catenin pathway and colour-coded type 1 neural progenitor DEGs (dashed border lines): upregulated = RSPO3, SULF2, GPC4; and downregulated = SFRP2, HSP90AB1, CTNNB1, ZIC1 , CCND1. (E) GO terms identified in type 3 neural progenitors enriched for genes associated with the canonical WNT signalling pathway, regulation of glial cell differentiation and others. (F) Violin plots of DEGs in type 3 neural progenitors show NOTCH signalling transcription factor, HES1, and ID3, which regulate glial differentiation; genes regulating cellular division: CENPF and CENPW; chromatin-modifying gene, HMGA1; and WNT signalling regulators: RSPO3, SFRP2 and ZIC1. (G) GO terms identified in immature neurons enriched for genes associated with synapse maturation, cholesterol metabolism, cytoskeletal organization and others. (H) Violin plots of DEGs in immature neurons shows genes associated with synapse maturation: NEFM, RELN and NRXN1; cholesterol biosynthesis: ACAT2 and HMGCS1; and genes associated with cytoskeletal organization: SMC1A and NEFM; and neurite morphogenesis: DCC and MDK. Of note, NRXN1 (AUTS2 target gene), DCC, RELN (AUTS2 target gene) and SMC1A are also designated SFARI genes implicated in autism spectrum disorder and intellectual disability. Log fold changes (FC) are shown under each violin plot and denote FC relative to either the parental control or gene-corrected control.