The neurodevelopmental disorder risk gene DYRK1A is required for ciliogenesis and control of brain size in Xenopus embryos

DYRK1A [dual specificity tyrosine-(Y)-phosphorylation-regulated kinase 1 A] encodes a highly conserved serine-threonine protein kinase expressed during human embryonic brain development (van Bon et al., 2016). Large-scale human genetics efforts have rigorously associated DYRK1A mutations with numerous developmental disorders (van Bon et al., 2015; Feki and Hibaoui, 2018). Specifically, DYRK1A is a high-confidence risk gene for autism spectrum disorders (ASDs) (Satterstrom et al., 2020; Iossifov et al., 2012). More recently, DYRK1A haploinsufficiency has been proposed to cause a distinct syndrome, characterized by ASD along with microcephaly, intellectual disability, developmental delay, and/or congenital anomalies of the kidney and urinary tract (van Bon et al., 2016; Earl et al., 2017; Ji et al., 2015; Blackburn et al., 2019). DYRK1A is located within the Down syndrome critical region on chromosome 21, and increased doses of DYRK1A have been implicated in Down syndrome pathobiology (Arron et al., 2006). Accordingly, DYRK1A kinase inhibitors are currently being explored for their therapeutic potential (Neumann et al., 2018; Feki and Hibaoui, 2018; Duchon and Herault, 2016).

The function of DYRK1A in brain development has been studied in a wide range of model systems. The microcephaly observed in humans with DYRK1A haploinsufficiency is recapitulated in both Drosophila and mouse heterozygous mutant animals (Fotaki et al., 2002; Tejedor et al., 1995). Mice with Dyrk1a heterozygous mutations also display motor deficits, developmental delay and altered behavior (Fotaki et al., 2002; Dierssen and de Lagrán, 2006; Arqué et al., 2009). At the cellular level, there is evidence that DYRK1A is required for cell cycle control, differentiation and dendritic spine development (Shaikh et al., 2016; Ori-McKenney et al., 2016; Dang et al., 2018). Molecularly, DYRK1A has been shown to phosphorylate a wide range of target proteins, including mRNA splicing factors (de Graaf et al., 2006; Shi et al., 2008), transcription factors (Fernandez-Martinez et al., 2009; Ehe et al., 2017; Mao et al., 2002), cyclins (Smith and Calegari, 2015; Soppa et al., 2014; Najas et al., 2015; Chen et al., 2013), and β-tubulin (Ori-McKenney et al., 2016). However, it is unknown which of these functions are central to the role of DYRK1A in brain development and which underlie pathobiology of related conditions.

Here we characterize the expression, localization and function of Dyrk1a in Xenopus tropicalis embryos to better understand its role in development. We uncover a novel localization of Dyrk1a to ciliary components and a corresponding novel requirement for Dyrk1a in ciliogenesis. Furthermore, we describe a novel localization of Dyrk1a to mitotic spindles, and demonstrate that inhibition alters cell cycle progression and cell survival in the developing brain. These observations suggest underlying deficits in microtubule dynamics, for which Dyrk1a has a known role (Ori-McKenney et al., 2016). Finally, we observe a marked reduction in forebrain size following dyrk1a loss of function, a phenotype consistent with human DYRK1A haploinsufficiency and deficits in cell cycle progression.

By whole-mount RNA in situ hybridization, we detected dyrk1a mRNA throughout X. tropicalis embryonic development, spanning gastrulation (Fig. 1A), neurulation (Fig. 1B) and organogenesis (Fig. 1C-F). Expression is strong in neural and ciliated tissues, including the embryonic epidermis (Fig. 1B), the developing brain and eye, the otic vesicle, and the pronephros (Fig. 1C-E; Blackburn et al., 2019). It is also expressed in the pharyngeal arches and developing heart (Fig. 1C,D). During neurogenesis, it is highly expressed in the proliferative cells lining the brain ventricles (Fig. 1E). Later, it is expressed throughout the tadpole brain, in the telencephalon (tel), diencephalon (di), mesencephalon (mes) and rhombencephalon (rhomb) (Fig. 1F).

By antibody staining of tailbud stages, we detected endogenous Xenopus Dyrk1a protein in puncta along ciliary axonemes, labeled by acetylated α-Tubulin (Fig. 1G). This staining is lost following dyrk1a depletion by morpholino oligonucleotides (MO) or by CRISPR/Cas9-mediated mutagenesis (Fig. S1). In tadpole stages, Dyrk1a antibody staining shows specific labeling throughout the brain (Fig. 1H), and especially in ciliated cells such as the olfactory epithelium (olf, Fig. 1I) and the roof of the fourth ventricle in the rhombencephalon (rhomb, Fig. 1J). We confirmed ciliary localization with a GFP-tagged human DYRK1A (hDYRK1A-GFP) expressed in the Xenopus embryonic epidermis. Human DYRK1A also localized in puncta along ciliary axonemes (Fig. 1K) and at ciliary basal bodies labeled with Centrin-CFP (Fig. 1L). These results demonstrate that Dyrk1a is expressed in neural and ciliated tissues during embryonic development and localizes to ciliary axonemes.

To investigate the function of Dyrk1a, we disrupted endogenous expression with either MO knockdown or CRISPR/Cas9 mutagenesis. This dyrk1a translation-blocking MO has previously been validated in Xenopus kidney development (Blackburn et al., 2019). We confirmed that the single guide RNA (sgRNA) targeting dyrk1a was efficient by Sanger sequencing and sequence deconvolution (mean efficiency 71%, standard deviation 18%; Fig. S2). In both cases, these loss of function strategies abolished Dyrk1a antibody staining in the ciliated epidermis in tailbud stages (Fig. S1). Therefore, we assayed whether ciliogenesis was disrupted following loss of dyrk1a by either strategy. In both cases, we observed that depletion of Dyrk1a during embryonic development leads to ciliogenesis defects (Fig. 2A-D, Fig. S3).

Because individuals with DYRK1A haploinsufficiency have microcephaly, we next assayed the gross brain anatomy of dyrk1a loss-of-function animals at the tadpole stages by labeling neurons with β-Tubulin antibody staining. We injected MO or CRISPR/Cas9 components into one cell at the two-cell stage, generating unilateral loss of dyrk1a function, where the uninjected half serves as a contralateral control (Willsey et al., 2018b; DeLay et al., 2018; Lasser et al., 2019). Loss of dyrk1a function by either method led to a reduction in telencephalon size without a dramatic alteration in gross regional anatomy (Fig. 2E-F,I, both P<0.05 by Mann–Whitney rank sum test compared with the negative control pigmentation gene slc45a2). In both this assay and in the ciliogenesis assay, the MO injection produced a stronger effect than the CRISPR injection, potentially because of the presence of maternal dyrk1a mRNA (Owens et al., 2016) that will only be targeted by the MO. We were able to rescue the reduction in telencephalon size following CRISPR injection by co-injection of hDYRK1A-GFP (Fig. 2H-I), suggesting conservation of function. Finally, we validated this loss-of-function phenotype by inhibition of Dyrk1a kinase activity using the validated pharmacological inhibitor harmine (Göckler et al., 2009) (Fig. 2G, P<0.05). Together, these results demonstrate that dyrk1a is required for ciliogenesis and regulation of telencephalon size control during Xenopus embryonic development.

To understand how dyrk1a loss of function leads to a smaller telencephalon, we sequenced RNA from dissected brains from control and dyrk1a CRISPR-injected animals at stage 46 (Fig. 3). After sequencing, we first confirmed that CRISPR/Cas9 mutagenesis resulted in mRNA disruption by analyzing sequence reads surrounding the protospacer adjacent motif (PAM) site within the target exon of dyrk1a. We observed a depletion of sequencing depth in all three injected replicates but not in control replicates (Fig. 3A), suggesting efficient targeting. Next, we identified 294 differentially expressed genes (Fig. 3B, Table S1), of which 172 were upregulated (orange points) and 122 were downregulated (blue points). Gene ontology enrichment analysis of the differentially expressed genes demonstrated strong enrichment in biological processes related to cell cycle control, DNA replication and microtubules (Fig. 3C, Table S2). Strikingly, canonical cell cycle genes such as cyclin-dependent kinase 1, aurora kinase B, cyclin B1, cyclin B2 and cyclin A2 are all increased greater than twofold (Fig. 3D, Table S1). Therefore, we hypothesized that the small telencephalon phenotype may be due to disruption of cell cycle progression.

It has previously been shown that loss of dyrk1a speeds the transition from gap 1 (G1) phase to synthesis (S) phase due to its role in regulating Cyclin D degradation (Chen et al., 2013; Najas et al., 2015; Soppa et al., 2014). However, our observations also suggest a novel role for Dyrk1a in regulating mitotic (M) phase. Specifically, we observed upregulation of cyclin B1 and cyclin B2, two genes known to reach maximum mRNA expression during the M phase of the cell cycle (Ito, 2000). We also observed enrichment of microtubule-related processes, including gene ontology terms related to the mitotic spindle. These data, combined with the fact that Dyrk1a localizes to ciliary axonemes and is required for ciliogenesis, a process that requires dynamic microtubule remodeling, led us to hypothesize that Dyrk1a might impact M phase via a role in mitotic spindle dynamics. This idea is consistent with a previously described function for Dyrk1a in regulating microtubule dynamics in developing dendrites (Ori-McKenney et al., 2016). Therefore, we stained blastula stage embryos, which are highly mitotic, for endogenous Dyrk1a. Remarkably, we observed a clear and novel localization near mitotic spindles marked by α-Tubulin (Fig. 4A,B). Together, these findings suggest that dyrk1a is important for cell cycle progression, possibly by modulating microtubule dynamics and/or the mitotic spindle, and that loss may cause stalling during the M phase of the cell cycle.

Next, we used the validated Dyrk1a kinase inhibitor harmine to determine whether cell cycle progression is indeed affected following loss of Dyrk1a function. This perturbation also results in the small telencephalon phenotype seen with the genetic inhibitors (Fig. 2G,I), but has the advantage of enhanced temporal control, and, thus, the ability to bypass early embryonic roles of Dyrk1a. Consistent with the RNA-seq data and the known role of Dyrk1a in regulating Cyclin D, we observed an increase in the number of cells expressing the S-phase marker PCNA (proliferating cell nuclear antigen) in the developing brain (Fig. 4C-E, P<0.0001 by Mann–Whitney rank sum test) as well as a relative increase in these cells compared with the rest of the telencephalon (Fig. 4F, P=0.0045). Dyrk1a inhibition also caused an increase in cells labeled by phospho-histone H3 (pHH3), a marker for M phase (Fig. 4G-I, P<0.0001, Fig. S4C,D), suggesting that more cells are in M phase. We also observed an increase in cleaved caspase 3 (CCP3) staining following Dyrk1a inhibition (Fig. 4J-L, P<0.0001), or following MO or CRISPR injection (Fig. S4A,B). These results generate hypotheses about the apparent paradox between a smaller brain (Fig. 2) and an increase in markers of proliferation (Fig. 3). It is possible that stalling in M phase, perhaps through disruption of the mitotic spindle, could increase the number of apparently proliferative cells, yet lead to a decrease in forebrain size, especially if stalling induces apoptosis. Alternatively, rapid progression into S phase could underlie the increase in M phase cells and cell death. Regardless, these data point to a clear role for Dyrk1a in cell cycle progression and survival during Xenopus neurogenesis.

In summary, this work reveals a novel role for Dyrk1a in embryonic ciliogenesis and a putative novel role at mitotic spindles during mitosis. Both processes require dynamic microtubule remodeling, and Dyrk1a has been shown to directly regulate microtubule dynamics during dendrite development (Ori-McKenney et al., 2016). Therefore, we hypothesize that the ciliogenesis and cell cycle phenotypes observed following loss of Dyrk1a both arise from an underlying deficit in microtubule dynamics. Future work should directly explore the role of Dyrk1a in mitotic spindle function. It should also determine whether cells are stalled in M phase rather than transitioning quickly from G1 to S, thereby resulting in an increase in the number of M-phase cells. It is also important to understand the mechanism underlying cell death, as well as its relative contribution to the observed small forebrain phenotype, compared with cell cycle progression defects. Furthermore, the observed defects in ciliogenesis suggest that potential disruptions of basal bodies and primary cilia should be investigated, as well as concomitant changes in patterning.

Overall, these findings provide insight into the pathobiology underlying DYRK1A haploinsufficiency. The identification of conserved phenotypes (microcephaly) between human DYRK1A haploinsufficiency and Xenopus dyrk1a loss of function reinforces the utility of this model organism for studying developmental disorder risk genes being identified by large-scale genetics efforts (Willsey et al., 2018a; Satterstrom et al., 2020), and informs current studies testing DYRK1A kinase inhibitors as potential therapeutic agents for Down syndrome and Alzheimer's disease (Branca et al., 2017; Stotani et al., 2016; Feki and Hibaoui, 2018; Duchon and Herault, 2016; Neumann et al., 2018), as it predicts side effects in ciliated and proliferative tissues. This work also suggests that the congenital kidney and urogenital tract abnormalities in individuals with DYRK1A haploinsufficiency may be due to underlying ciliogenesis defects and that these patients should be carefully evaluated for other cilia-related disorders, such as congenital heart disease, hydrocephalus and retinal degeneration.

Xenopus tropicalis adult animals were maintained and cared for according to established IACUC protocols. Animals were wild type and both sexes were used. Animals were ovulated using human chorionic gonadotropin (Sigma) according to Sive et al. (2000), and both in vitro fertilizations and natural matings were used. Embryos were staged according to Nieuwkoop and Faber (1958).

X. tropicalis dyrk1a cDNA IMAGE clone 7687837 (Morin et al., 2006) and digoxygenin-11-UTP were used to synthesize antisense probe according to standard protocol using SalI restriction enzyme and T7 polymerase (Sive et al., 2007). Embryos were fixed and stained according to a standard protocol (Harland, 1991), with the omission of the proteinase K step when assaying expression in the multiciliated cells on the epidermis.

Full-length human DYRK1A cDNA (transcript NM_001396) was obtained from the Harvard Plasmid repository (HsCD00082867) and cloned into pDONR221. LR cloning reactions added a N-terminal GFP using destination vector pCS-EGFP, a kind gift from John Wallingford (University of Texas at Austin, USA) (Tu et al., 2018). Sanger sequencing verified the sequences (ElimBio). Centrin-CFP was a kind gift from Peter Walentek (University of Freiburg, Germany) (Walentek et al., 2016; Antoniades et al., 2014; Park et al., 2008).

Embryos were injected with 2 nl per blastomere into one cell at the two-cell stage using a Narishige micromanipulator and a Parker Picospritzer III. Plasmid (hDYRK1A-GFP) was injected at 40 pg per embryo and mRNA (Centrin-CFP) was injected at 500 pg per embryo. Morpholino was injected at 2 ng per embryo; sgRNA was injected at 800 pg per embryo; and Cas9-NLS protein (MacroLabs, UC Berkeley; Lingeman et al., 2017) was injected at 1.5 ng per embryo. Fluorescent dextrans were co-injected to label the injected side of the embryo and animals were sorted left from right-injected at neurula stages.

Immunostaining was carried out according to Willsey et al. (2018b), with the omission of the bleaching step whenever phalloidin was included. Phalloidin was added during the secondary antibody incubation to visualize actin (1:50, Life Technologies A12380 or A22287). Mounting media with DAPI was used to visualize DNA (Vectashield, Fisher NC9524612). The following primary antibodies were used: DYRK1A (1:100, Abcam, ab69811), acetylated α-Tubulin (1:700, Sigma, T6793), β-Tubulin (1:100, DSHB, clone E7), PCNA (1:50, Life Technologies, clone PC10), polyglutamylated Tubulin (1:100, AdipoGen, GT335), phospho-histone H3 (1:100, Ser10, Sigma, 06-570) and cleaved caspase 3 (1:100, Asp175, Cell Signaling, 9661). Secondary fluorescence-conjugated antibodies were used at 1:250 (Life Technologies, A32723 and A32732).

Published and validated translation-blocking dyrk1a morpholino (5′ TGCATCGTCCTCTTTCAAGTCTCAT 3′) (Blackburn et al., 2019) was injected at 2 ng per embryo. sgRNA against X. tropicalis dyrk1a was designed using CRISPRscan (Moreno-Mateos et al., 2015) (target sequence: 5′-CGTTTAGGTTCTGCTGACGGCGG-3′, oligo sequence: 5′-taatacgactcactataGGTTTAGGTTCTGCTGACGGgttttagagctagaa-3′) and was synthesized in vitro (Engen) and purified (Zymo). Purified Cas9-NLS protein was acquired from MacroLab (UC Berkeley) (Lingeman, Jeans, and Corn, 2017). For genotyping, genomic DNA was isolated and the region around the protospacer adjacent motif (PAM) site was amplified by PCR using primers (F: 5′-GGAGAAATCCCTGACAATTGTATTAATTATAGCATTG-3′ and R: 5′-GTTCTTGACCGGTACTGACAAAATGAG-3′) and Sanger sequenced. Mutational efficiency was estimated by sequence deconvolution by tracking of INDELs (TIDE) (Brinkman et al., 2014; DeLay et al., 2018).

DYRK1A inhibitor harmine (Sigma Aldrich, 286044) (Göckler et al., 2009; Bain et al., 2007) was reconstituted in dimethylsulfoxide (DMSO) as a 1 mM stock solution. Animals were treated with 1.25 µM of harmine in 1/9 modified ringers or an equal volume of DMSO in 1/9 modified ringers at stage 30, unless otherwise indicated. The embryos were raised until fixation at stage 46 for immunofluorescence staining. Culture liquid was not refreshed during treatment for either condition. Dead animals were removed as soon as they were observed.

RNA in situ hybridization embryos were visualized on a Zeiss AxioZoom.V16 with a 1× objective and a Zeiss 105 color camera with extended depth of focus. Localization and cilia images were acquired on a Leica SP8 laser scanning confocal with a 63× objective. Tadpole nervous system images were acquired by Zeiss AxioZoom.V16 with a 1× objective with a Zeiss 506 monochrome camera and apotome. Whole-mount images are maximum intensity projections of optical sections. We have previously shown that our antibody staining protocol penetrates the brain and that this imaging strategy is sufficient to detect the same gross size changes as physical transverse sections (Willsey et al., 2018b).

Images were processed in FIJI (NIH) and compiled in Illustrator (Adobe). Differences in area were measured in FIJI using the free-hand selection function followed by the measure function. Cells with positive antibody staining were marked and counted manually in FIJI. Differences in mean area or cell number were tested for statistical significance by non-parametric Mann–Whitney rank sum test (GraphPad Prism 8).

Individual brains from control and bilaterally injected dyrk1a CRISPR tadpoles were dissected at stage 46 and immediately put in 200 µl of cold Trizol (Thermo Fisher), pipetted with a 30 gauge needle to dissociate and frozen at −80°C. Carcasses were genotyped to determine mutational efficiency, and the nine most mutagenized brains were selected and pooled into three independent replicates (three brains per replicate). Nine uninjected brains were pooled into three independent replicates (three brains per replicate). RNA extraction, polyA selection and low-input library preparation (500 bp size selection) were performed by the Functional Genomics Laboratory (UC Berkeley). Samples were processed together, barcoded and spread across a sequencing lane to reduce batch effects. Unstranded, 150 paired end sequencing was performed on an Illumina HiSeq 4000 by the Genomics Sequencing Laboratory (UC Berkeley).

RNA-seq reads were aligned to the XenBase X. tropicalis v9.1 reference genome using STAR v2.7.3 (Dobin et al., 2013) in gene annotation mode with default parameters. Read counts were converted into counts per million (cpm) and genes with more than 1 cpm in at least three samples were retained for differential gene expression (DEX) analysis. Filtered genes were tested for differential expression using DESeq2 v1.24.0 (Love, Huber, and Anders, 2014) with the shrinkage estimator apeglm (Zhu, Ibrahim, and Love, 2019). P-values were corrected for multiple testing using the Benjamini and Hochberg FDR correction. Significantly DEX genes are genes that pass a 0.05 significant threshold. Gene Ontology (GO) enrichment analysis of all annotated DEX genes (n=221) was performed using the online classification tool PANTHER (Mi et al., 2010) against the available Xenopus database using all expressed genes as a background (n=15,079).

Because DYRK1A has been shown to phosphorylate splicing factors (de Graaf et al., 2006; Shi et al., 2008), we also analyzed the data to identify differentially expressed exons. DEXseq (Anders, Reyes, and Huber, 2012) with default parameters was used to quantify exon-level counts and subsequently identify differential exon usage between wild-type and dyrk1a knockdown samples. Genes with total exon-level counts below 5 were removed before differential exon usage (DEU) analysis. P-values per exon were corrected for multiple testing using the Benjamini and Hochberg FDR correction. There were 97 identified DEU sites within 82 genes which passed a 0.05 significance threshold (Table S3).

We thank Mustafa Khokha and Michael Slocum for wild-type frogs and husbandry advice; Nolan Wong, Kelly Jensen, Gary Moulder, Louie Ramos and Shaun Coughlin for support at the UCSF Xenopus facility; James Evans for support at the UC Berkeley Xenopus facility; Edivinia Pangilinan for expert technical help; Xenbase (RRID:SCR_003280) for essential daily reference; the National Xenopus Resource (RRID:SCR_013731) for wild-type X. tropicalis and technical advice with husbandry and genotyping; the CSHL Xenopus Course (2015-2019); and John Wallingford for the pCS-EGFP Gateway vector. We acknowledge support from the Overlook International Foundation and the Weill Institute for Neurosciences. This work used the Vincent J. Coates Genomics Sequencing Laboratory at UC Berkeley, supported by a National Institutes of Health Instrumentation Grant (S10 OD018174).

The peer review history is available online at https://dev.biologists.org/lookup/doi/10.1242/dev.189290.reviewer-comments.pdf