MARK2 variants cause autism spectrum disorder via the downregulation of WNT/β-catenin signaling pathway

Abstract

Microtubule affinity-regulating kinase 2 (MARK2) contributes to establishing neuronal polarity and developing dendritic spines. Although large-scale sequencing studies have associated MARK2 variants with autism spectrum disorder (ASD), the clinical features and variant spectrum in affected individuals with MARK2 variants, early developmental phenotypes in mutant human neurons, and the pathogenic mechanism underlying effects on neuronal development have remained unclear. Here, we report 31 individuals with MARK2 variants and presenting with ASD, other neurodevelopmental disorders, and distinctive facial features. Loss-of-function (LoF) variants predominate (81%) in affected individuals, while computational analysis and in vitro expression assay of missense variants supported the effect of MARK2 loss. Using proband-derived and CRISPR-engineered isogenic induced pluripotent stem cells (iPSCs), we show that MARK2 loss leads to early neuronal developmental and functional deficits, including anomalous polarity and dis-organization in neural rosettes, as well as imbalanced proliferation and differentiation in neural progenitor cells (NPCs). Mark2^+/- mice showed abnormal cortical formation and partition and ASD-like behavior. Through the use of RNA sequencing (RNA-seq) and lithium treatment, we link MARK2 loss to downregulation of the WNT/β-catenin signaling pathway and identify lithium as a potential drug for treating MARK2-associated ASD.

Keywords: ASD; LoF; MARK2 variants; WNT/β-catenin signaling pathway; autism spectrum disorder; lithium; loss-of-function.

Graphical abstract

Figure 1

Overview of MARK2 variants and associated phenotypes (A) Locations of variants in functional domains of MARK2 predicted by UniProt are shown. p.Arg302^∗ and p.Gln747^∗ are recurrent variants that occurred in 2 individuals. Kinase, protein kinase domain; UBA, ubiquitin associated; PR, polar residue; KA1, kinase-associated domain 1. (B) Facial photos of 8 affected individuals. Ages at the time of photograph are P2, 5 years and 4 months; P4, 4 years and 2 months; P6, 3 years and 1 month; P9 left, 4 years; P9 right, 11 years; P10, 8 years; P15, 46 years; P17, 7 years and 7 months; P19 left, 4 years; P19 right, 4 years and 7 months. (C and D) The mRNA expression (C) and protein accumulation (D) of MARK2 in the PBMCs from 3 affected individuals (Mut: P4, P6, P19). Both mRNA and protein levels were normalized to GAPDH levels and compared with those of age-matched healthy children (CTRL). The data of at least 3 independent experiments were analyzed by Student’s t test, ^∗p < 0.05. (E) Minigene assays of 2 splicing variants (P4: c.337+1G>T; P14: c.1514+2T>G). RT‒PCR products from HEK 293T and HeLa cells transfected with either the wild-type (WT) or mutant (Mut) pcDNA3.1 vector were separated by electrophoresis (left). M, DNA marker; the sizes of the bands are 2,000, 1,000, 750, 500, 250, and 100 bp. Sequencing of the RT-PCR products (right) show c.337+1G>T variant results in a 49-bp deletion, while the c.1514+2T>G variant results in a 98-bp deletion.

Figure 2

Structural modeling of six MARK2 missense variants (A) Alignment of the protein sequences for 6 MARK2 missense variants across species from zebrafish to humans. (B) 3D structural models of MARK2 missense variants. The protein domains of MARK2 are shown in different colors. The 3D structures of the non-intrinsically disordered regions (non-IDRs) were used to compute the changes in structural stability due to genomic variation. Amino acids in the IDRs are shown in a disordered configuration; we expect that any individual configuration of the IDRs would be an inadequate representation of their diverse dynamics. The locations of the 6 missense variants are marked by black spheres. Four variants in the kinase domain (p.A80V, p.G135R, p.F194S, p.R302Q) and 2 variants in the KA1 domain (p.V752A, p.R764P) were all predicted to locate in the activation loop. (C and D) Representative western blot image (C) and quantification analysis of exogenous MARK2 accumulation that were normalized by total GAPDH levels (D). Human HEK 293T cells transfected with wild-type (CTRL) or mutant EGFP-MARK2-Neo vectors were lysed. The data of at least 3 independent experiments were analyzed by Student’s t test ^∗p < 0.05, ^∗∗p < 0.01 and ^∗∗∗p < 0.001.

Figure 3

MARK2 variant leads to aberrant polarity in iPSC-derived neural rosettes and imbalanced proliferation and differentiation in iPSC-derived NPCs (A) Representative immunofluorescence images of TUJ1 (red) and ZO1 (green) in different iPSC-derived neural rosettes on the 10th day after neural induction. CTRL1 and CTRL2, 2 independent healthy adults without MARK2 variants; p.C723Sfs and p.F271Sfs, 2 affected individuals with LOF MARK2 variants; CRISPR-Del1 and CRISPR-Del2, 2 isogenic MARK2 deletions produced by the CRISPR-Cas9 editing technology. (B) Representative immunofluorescence images of SOX1 (red) and NESTIN (green) in different iPSC-derived neural rosettes on the 5th day after neural induction. (C) Representative images of different iPSC-derived neurospheres on the 5th day after neural induction. (D) Representative immunofluorescence images of BrdU (red) and Ki67 (green) in different iPSC-derived neural rosettes on the 4th day after neural induction. (E) Representative immunofluorescence images of DCX (red) and TUJ1 (green) in different iPSC-derived NPCs on the 13th day after neural induction. Scale bar = 50 μm. The quantification and statistical analysis were showed in Figure S3.

Figure 4

Mark2 loss in mice leads to ASD-like behavior and impaired memory (A and E) Trajectories of Mark2^+/+ (WT) and Mark2^+/− (HET) mice in the three-chamber test (A, empty or novel; E, familiar or novel). (B and F) Quantification analysis of sniffing time in the three-chamber test (B, empty or novel; F, familiar or novel). (C, D, and K) Trajectories of WT and HET mice in the marble-burying test (C) and quantification analysis of the buried marbles (D) and grooming time (K). (H and I) Trajectories of WT and HET mice in the elevated plus maze test (H) and quantification analysis of time spent in the open arms (I). (G and J) Trajectories of WT and HET mice in the Y-maze test (G) and quantification analysis of the correct alternation (J). (L–N) Trajectories of WT and HET mice in the Barnes maze test (L), quantification analysis of the latency of entering the target hole during the training sessions (M), and the latency of visiting target hole and of entering target hole during the test session (N). (O–Q) Trajectories of WT and HET mice in the Morris water maze test (O), quantification analysis of the latency to find the platform during the training sessions (P), and the latency to finding the platform and staying platform quadrant during the test session (Q). WT = 9, HET = 9. The data were analyzed by Student’s t test; ^∗p < 0.05 and ^∗∗p < 0.01.

Figure 5

iPSC-derived NPCs revealed decreased WNT signaling pathway due to MARK2 loss (A) Gene Ontology (GO) analysis of differentially expressed genes (DEGs) between mutant iPSC-derived NPCs and CTRL ones show that the downregulated genes are enriched in multiple pathways related to early neuronal development. (B) GO pathway network analysis showed the association of WNT3A with neuronal development, including neuron fate specification, forebrain development, central nervous system neuron differentiation, axonogenesis, and axon development. Size means number of genes involved in the specific pathway. (C) The enrichment score and rank of the WNT/β-catenin signaling pathway from DEGs. (D–F) Representative western blot images (D) and quantification analysis of β-catenin (E) and WNT3A (F) accumulation in mutant and CRISPR-Del iPSC-derived NPCs compared with CTRLs. (G–I) Representative western blot images (G) and quantification analysis of β-catenin (H) and p-GSK3β-Ser9 (I) accumulation in mutant, CRISPR-Del iPSC-derived NPCs with/without rhWNT3A (100 ng/mL) or LiCI treatment (0.7 mM). The data of at least 3 independent experiments were analyzed by Student’s t test; ^∗p < 0.05 and ^∗∗p < 0.01(compared with CTRL1); ^#p < 0.05 and ^##p < 0.01 (compared with CTRL2).

Figure 6

Lithium reverses the imbalance in the proliferation and differentiation of MARK2 mutant iPSC-derived NPCs by activating the WNT/β-catenin signaling pathway (A) Representative immunofluorescence images of ZO1 (green) and TUJ1 (red) in different iPSC-derived neural rosettes. CTRL1 and CTRL2, 2 independent healthy adults without MARK2 variant; p.C723Sfs and p.F271Sfs, 2 affected individuals with LOF MARK2 variants; and CRISPR-Del1 and CRISPR-Del2, 2 isogenic MARK2 deletions produced by the CRISPR-Cas9 editing technology. Both CRISPR-Del and mutant iPSC-derived neural rosettes were treated with LiCl (0.7 mM) or rhWNT3A (100 ng/mL). (B) Representative immunofluorescence images of Ki67 (green) and BrdU (red) in different iPSC-derived NPCs. (C) Representative immunofluorescence images of TUJ1 (green) and DCX (red) in different iPSC-derived NPCs. Scale bar = 50 μm. The quantification and statistical analysis were showed in Figure S7.