Mutations in ANKLE2, a ZIKA virus target, disrupt an asymmetric cell division pathway in Drosophila neuroblasts to cause microcephaly.

SUMMARY

The apical Par complex, which contains aPKC, Bazooka (Par-3), and Par-6, is required for establishing polarity during asymmetric division of neuroblasts in Drosophila, and its activity depends on L(2)gl. We show that loss of Ankle2, a protein associated with microcephaly in humans and known to interact with Zika protein NS4A, reduces brain volume in flies and impacts the function of the Par complex. Reducing Ankle2 levels disrupts ER and nuclear envelope morphology, releasing the kinase Ballchen/VRK1 into the cytosol. These defects are associated with reduced phosphorylation of aPKC, disruption of Par complex localization, and spindle alignment defects. Importantly, removal of one copy of ballchen/VRK1 or l(2)gl suppresses Ankle2 mutant phenotypes and restores viability and brain size. Human mutational studies implicate the above mentioned genes in microcephaly and motor neuron disease. We suggest that NS4A, ANKLE2, VRK1 and LLGL1 define a pathway impinging on asymmetric determinants of neural stem cell division.

Graphical Abstract

eTOC Blurb

The Zika virus protein NS4A interacts with ANKLE2, a protein linked to hereditary microcephaly. Mutations in ANKLE2 also cause microcephaly-like phenotypes in flies. Link et al. now connect these phenotypes with disruption of an asymmetric cell division pathway in fly neuroblasts via an interaction between ANKLE2 and the kinase Bällchen/VRK1.

INTRODUCTION

Proper development of the human brain requires an exquisitely coordinated series of steps and is disrupted in disorders associated with congenital microcephaly. Congenital microcephaly in humans is characterized by reduced brain size (using occipital frontal circumference, OFC, as a surrogate measure) more than two standard deviations below the mean (Z-score < −2) at birth. It is associated with neurodevelopmental disorders, such as developmental delay and intellectual disability and can be caused by external exposures to toxins, in utero infections, or gene mutations. Pathogenic gene variants for microcephaly have been identified through targeted genetic testing, genomic copy number studies, and exome sequencing (ES) (Brunetti-Pierri et al., 2008; Dumas et al., 2012; Lupski, 2015; Shinawi et al., 2010; Shaheen et al., 2018), identifying 18 primary microcephaly loci. Many syndromes significantly overlap with classic microcephaly phenotypes, and together, these disorders can be caused by defects in a wide variety of biological processes, including centriole biogenesis, DNA replication, DNA repair, cell cycle and cytokinesis, genome stability, as well as multiple cell signaling pathways (Jayaraman et al., 2018). In flies, we refer to microcephalic phenotypes when the third instar brain lobes are reduced in size (Yamamoto et al., 2014) or when adult flies exhibit small heads relative to the their body size (Oláhová et al., 2018). As in humans, microcephaly in flies can be a result of mutations that affect cell division and centrosome biology as demonstrated with mutations in WDR62 (Ramdas Nair et al., 2016; Lim et al., 2017) and ASPM/ASP (Rujano et al., 2013), but also those that affect the spindle assembly checkpoint (Poulton et al., 2017) and neuroblast proliferation (Kanai et al., 2018).

A forward, mosaic screen for neurodevelopmental and neurodegenerative phenotypes associated with lethal mutations on the X-chromosome in Drosophila identified 165 loci, many with corresponding human genetic disease trait phenotypes (Yamamoto et al., 2014). Among them, a mutation in Ankle2 (Ankryin repeat and LEM domain containing 2) causes loss of Peripheral Nervous System (PNS) organs in adult mutant clones and severely reduced brain size in hemizygous third instar larvae. To identify patients with pathogenic variants in ANKLE2, we surveyed the exome database of the Baylor-Hopkins Center for Mendelian Genomics (BHCMG) (Bamshad et al., 2012; Posey et al., 2019) and identified compound heterozygous mutations in ANKLE2 in two siblings. Both infants exhibited severe microcephaly (Z-score = −9), and the surviving patient displayed cognitive and neurological deficits alongside extensive intellectual and developmental disabilities. We showed that mutations in Ankle2 led to cell loss of neuroblasts and affected neuroblast division in the developing third instar larval brain. Remarkably, expression of the wild type human ANKLE2 in flies rescued the observed mutant phenotypes (Yamamoto et al., 2014). Here we explore the molecular pathways and proteins that are affected by Ankle2 loss.

ANKLE2 belongs to a family of proteins containing LEM (LAP2, Emerin, MAN1) domains that typically associate with the inner nuclear membrane (Lin et al., 2000; Barton et al., 2015). Conventional LEM proteins have been shown to interact with BAF (Barrier to Autointegration Factor), which binds to both DNA and the nuclear lamina (Segura-Totten et al., 2002) to organize nuclear and chromatin structure. However, the LEM domain in Drosophila and C. elegans Ankle2 is not obviously conserved (Marchler-Bauer et al., 2017). Studies in C. elegans indicate that a homolog of ANKLE2 regulates nuclear envelope morphology and functions in mitosis to promote reassembly of the nuclear envelope upon mitotic exit (Asencio et al., 2012; Snyers et al., 2018). During this process, ANKLE2 modulates the activities of VRK1 (Vaccina Related Kinase 1) and PP2A (Protein Phosphatase 2A) (Asencio et al., 2012). However, all experiments in worms were performed at the embryonic two-cell stage, and no other phenotypes were reported except early lethality. Whilst mutations in ANKLE2 have been associated with severe microcephaly (OFC z-sore = −2.5 to −9), human VRK1 pathogenic variant alleles can cause a neurological disease trait consisting of complex motor and sensory axonal neuropathy and microcephaly (Gonzaga-Jauregui et al., 2013).

Mutations in both Ankle2 and the fly homologue of VRK1, ballchen, cause a loss of neuroblasts in 3^rd instar larval brains in Drosophila (Yamamoto et al., 2014; Yakulov et al., 2014). Neuroblasts (NBs) divide asymmetrically and are often used as a model to investigate stem cell biology (Homem and Knoblich, 2012) and asymmetric cell division (Gallaud et al., 2017). Most NBs in the larval central brain give rise to another NB and a smaller ganglion mother cell (GMC), which then divides once again to produce neurons or glia. Proper NB maintenance and regulation is essential for precise development of the adult nervous system, and misregulation of NB number or function can lead to defects in brain size (Wang et al., 2009; Gateff and Schneiderman, 1974).

Congenital Zika virus infection in humans during pregnancy has been associated with severe microcephaly that can be as dramatic as certain genetic forms of microcephaly including phenotypes associated with bi-allelic mutations in MCPH16/ANKLE2 (Moore et al., 2017; Yamamoto et al., 2014). Recently, we showed that a Zika virus protein, NS4A, physically interacts with ANKLE2 in human cells. Expression of NS4A in larval brains causes microcephaly, induces apoptosis, and reduces proliferation. Importantly, expression of human ANKLE2 in flies that express NS4A suppresses the associated phenotypes, demonstrating that NS4A interacts with the ANKLE2 protein and inhibits its function (Shah et al., 2018). Interestingly, Zika virus crosses the blood brain barrier and targets radial glial cells, the neural progenitors in the vertebrate cortex (Devhare et al., 2017; Tang et al., 2016).

Here, we show that Ankle2 is localized to the endoplasmic reticulum and nuclear envelope, similar to NS4A, and genetically interacts with ball/VRK1 to regulate brain size in flies. An allelic series at the ANKLE2 and VRK1 loci shows that perturbation of this pathway results in neurological disease including microcephaly. Our data indicate that the Ankle2-Ball/VRK1 pathway is required for proper localization of asymmetric proteins and spindle alignment during NB cell division by affecting two proteins, aPKC and L(2)gl, that play critical roles in the asymmetric segregation of cell fate determinants. In addition, NS4A expression in neuroblasts mimics phenotypes seen in Ankle2 mutants, and NS4A induced microcephaly is suppressed by removing a single copy of ball/VRK1 or l(2)gl. Human genomics variant data and disease trait correlations extend this asymmetric cell division pathway from proteins identified in flies and reveal insights into neurological disease. In summary, NS4A hijacks the Ankle2-Ball/VRK1 pathway, which regulates progenitor stem cell asymmetric division during brain development and defines a human microcephaly pathway.

RESULTS

Human ANKLE2 variants cause microcephaly

We previously reported that compound heterozygous variants in ANKLE2 are associated with microcephaly (Z-score = −9) (MCPH16, MIM#616681) in two affected siblings (Yamamoto et al., 2014). Here, we report two additional probands carrying unique variants in ANKLE2 identified in Seattle (LR17–511 and LR18–033; Figure S1 and S2, Table S1). Brain MRIs of an age matched control (Figure S1A) and a proband with microcephaly from family LR17–511 document one of the more severe cases of microcephaly (Z-score = −8) (Figure S1B). To investigate potential genotype-phenotype correlations, we explored the spectrum of reported neurological disease trait manifestations associated with ANKLE2 present in the Baylor Genetics (BG) Laboratories databases. These contain results of clinical exome sequencing (ES) of patients with presumed genetic disorders that remain unsolved for a molecular diagnosis but patients/families have consented to research analyses. We screened for rare biallelic variants, predicted damaging, in ANKLE2 that fulfill Mendelian expectations for a recessive disease trait. Three families were found to fulfill these criteria in probands with neurologically associated phenotypes (Figure S1D and S2, Table S2). These cases suggest that a diverse set of variants in ANKLE2 may be associated with a spectrum of neurologic disease and reveal either sporadic disease, apparent vertical transmission, and in some cases, consanguineous parentage (Yamamoto et al., 2014; Shaheen et al., 2018). The identified mutations are missense, nonsense, or splicing variants that lead to premature stop codons; all subjects have biallelic variants, either compound heterozygous or homozygous alleles (Figure S1 and S2). Probands exhibit congenital microcephaly, but some also present with severe brain MRI abnormalities and skin pigmentation abnormalities (Figure S1D). These aggregate data demonstrate that mutations in ANKLE2 cause autosomal recessive microcephaly.

Null alleles of Ankle2 are associated with reduced brain size in flies

Given the human genetic implications noted above, we used Drosophila to elucidate molecular mechanisms underlying ANKLE2 associated microcephaly. The mutation originally identified in flies, Ankle2^A (L326H), causes reduced brain size in third instar larvae and leads to pupal lethality at temperatures ≥ 22°C (Figure S3A). It results in decreased neuroblast number, reduced cell divisions when assessed in MARCM clones, and a high incidence of apoptotic cell death throughout the brain, including neuroblasts and neurons (Yamamoto et al., 2014). However, first instar larvae are born with a slight, but not significant, reduction in the number of neuroblasts or Dpn positive cells (Figure S3B and C) indicating that Ankle2 loss causes a premature differentiation, loss of stem cell markers, or death of neuroblasts.

To create a severe loss of function allele for Ankle2, we integrated a CRIMIC construct containing attP-FRT-SA-3XSTOP-polyA-3xP3-EGFP-FRT-attP sequences using CRISPR-Cas9 in the fifth intron shared by all isoforms (Figure 1A; pM14; (Lee et al., 2018). The construct leads to a truncated transcript that likely corresponds to a null allele (Ankle2^CRIMIC, Figure 1A). These animals die as 3^rd instar larvae (Figure 1B), are smaller than wild type and Ankle2^A animals, and show a very severely reduced brain volume (Figure 1E versus I, and M) with complete disruption of brain morphology, especially the optic lobe (Figure 1I). Since the severe phenotypes make it difficult to assess the biological questions presented below, we assayed the majority of phenotypes with the hypomorphic Ankle2^A missense allele.

Figure 1: Mutations in ANKLE2 cause microcephaly in humans and Drosophila.

The structure and transcripts of the Ankle2 gene with insertional mutations or tags are shown in (A). Ankle2^A, noted by *, is an EMS generated L326H mutation, while Ankle2^CRIMIC represents a Crispr-Cas9 mediated MIMiC-like (CRIMIC) insertion in the 4^th coding intron (Lee et al., 2018). ΦC31 mediated cassette exchange replaced the stop-polyA in Ankle2^CRIMIC with an artificial exon cassette consisting of SA-GFP-SD (Venken et al., 2011) to produce Ankle2^IGFP. A C-terminal GFP tagged Ankle2 genomic rescue construct (Ankle2-GFPR) was generated using recombineering (Venken et al., 2008). (B) Lethal stages and rescue by Ankle2-GFPR or (C) complementation tests of Ankle2^A, Ankle2^CRIMIC, and Ankle2^IGFP. (D) Ankle2 gene structure and genomic rescue construct (CH321–85N12; Venken et al., 2009). (E-L) Partial projections of 3^rd instar larval brains stained with Deadpan (Dpn) (purple, neuroblasts) a neuroblast marker (Bier et al., 1992) and Pros (green, GMC, neurons), a neuronal lineage marker (Campbell et al., 1994), to document overall brain structure is shown in (E) wild type (y w FRT19a), (F) y w Ankle2^A FRT19a (G) y w Ankle2^A FRT19a; Ankle2-GFPR, (H) y w Ankle2^A FRT19a; da::hANKLE2^wt, (I) Ankle2^CRIMIC, (J) Ankle2^CRIMIC; Ankle2-GFPR (K) y w Ankle2^A FRT19a; da:: hANKLE2 p.L573V and (L) y w Ankle2^A FRT19a; da:: hANKLE2 p.Q782* animals. The C-terminal GFP-tagged rescue P[acman] clone CH321–85N12, (Ankle2-GFPR, diagramed in A and B) rescues brain morphology, brain size, and lethality in both y w Ankle2^A FRT19a (G) and Ankle2^CRIMIC (J) animals. Quantification of brain size is shown in (M). Box plots hinges represent the 25^th to 75^th percentiles, the central line is the median, and whiskers represent min to max. Note that hANKLE2^wt and hANKLE2 p.L573V rescue brain size and lethality of Ankle2^A mutants, whereas hANKLE2 p.Q782*, hANKLE2 p.A109P, and hANKLE2 p.G201W do not. Here, Ankle2-GFPR is reported as GFPR. (N) Neuroblast specific expression of wild type human ANKLE2 (y w Ankle2^A FRT19a; insc::hANKLE2^wt) partially rescues brain size of Ankle2^A mutants. For quantifications, the total number counted (n) is noted below each graph. See also Figure S1–S3.

To determine whether Ankle2 is expressed in the brain, we used the CRIMIC allele to introduce an artificial exon that contains SA-GFP-SD in frame which produces a tagged fusion protein (Ankle2^IGFP, Figure 1A–B) (Lee et al., 2018). We readily detect Ankle2^IGFP protein in brains of heterozygous animals (Figure S4A). However, homozygous animals are lethal and exhibit very small brains, indicating that integration of this exon disrupts protein function. Based on complementation tests, the strength of the allelic series is Ankle2^A < Ankle2^CRIMIC= Ankle2^IGFP (Figure 1B and C). We therefore used recombineering (Venken et al., 2006) to add a C-terminal GFP tag to Ankle2 in a bacterial artificial chromosome (BAC CH321–85N12; referred to as Ankle2-GFPR, Figure 1A and D) (Venken et al., 2008; Venken et al., 2009). When this P[acman] clone was introduced in all three Ankle2 mutant backgrounds, Ankle2-GFPR rescued brain phenotypes and lethality of these alleles (Figure 1B, G, and J). Hence, the chromosomes carrying the three Ankle2 alleles do not carry second site mutations that affect brain size or viability and the tagged protein is likely to reflect the endogenous Ankle2 protein distribution.

The human reference ANKLE2 gene rescues lethality and small brain phenotypes of Ankle2^A animals when expressed ubiquitously (da-GAL4 >UAS-hANKLE2, Figure 1H, M) or in neuroblasts only (insc-GAL4>UAS-hANKLE2, Figure 1N). To determine whether the microcephaly associated mutations in human ANKLE2 are loss of function alleles, we next expressed ANKLE2 p.L573V, ANKLE2 p.Q782*, ANKLE2 p.A109P, ANKLE2 p.G201W, in Ankle2^A mutant animals. The p.Q782*, p.A109P, and p.G201W variants failed to rescue lethality or reduced brain sizes (Figure 1L–M) consistent with them being severe loss-of-function variant alleles. However, p.L573V restored both viability and brain size (Figure 1K, M) in some Ankle2^A animals (Figure 1M), suggesting that this variant is a mild hypomorphic allele.

Ankle2 localizes to the ER and nuclear envelope and is required for their integrity

The endogenously tagged Ankle2^IGFP (Figure S4A) as well as the tagged genomic rescue construct, Ankle2-GFPR (Figure 2A–B, Figure S4B–G) show that Ankle2 is expressed in most tissues including larval discs (Figure S4B–C), embryos (Figure S4D–F), and cells of the third instar larval brain. Neuroblasts, ganglion mother cells, and neurons (Figure S4G) all show high expression of Ankle2. The protein appears to be localized to the cytoplasm of all cells including neuroblasts (arrows in Figure 2A–B). However, in a subset of cells, the protein is clearly enriched at the nuclear envelope (arrowhead). To determine the dynamics of Ankle2 protein localization, we performed live imaging. As shown in Movie S1, the protein is recruited to the nuclear envelope at the initiation of mitosis and remains associated with the nuclear envelope until briefly after cytokinesis. Indeed, Ankle2 is enhanced at the nuclear envelope in cells positive for phospho-histone H3 (pHH3), a mitotic marker (Figure S4H–I, arrowhead). Hence, Ankle2 is localized to the nuclear envelope in neuroblasts undergoing mitosis.

Figure 2: Ankle2 localizes to the ER and is dynamically expressed in the brain.

(A-B) 3^rd instar larval brains from Ankle2-GFPR animals stained for GFP to document Ankle2 expression and localization in (A-B) single slices (arrows point to neuroblasts). (C-E) Live imaging of Ankle2 (D, green) and the ER labeled with Sec61B (E, da-GAL4, UAS-Sec61β-tdTomato, purple) (Summerville et al., 2016) shows strong colocalization. (F) Live Ankle2^A mutant neuroblast display aberrant Sec61β-tdTomato expression. (G-I) Fixed Ankle2-GFPR animals highlighting Ankle2-GFP (green) and another ER marker (Calnexin 99a, purple) (Riedel et al., 2016). (J) Fixed Ankle2 mutant animals (Ankle2^A) display aberrant ER structures (Calnexin 99a, purple). (K-M) Ankle2 (L, green) colocalizes with some portions of the nuclear envelope (Lamin Dm0, purple). (N) Ankle2 mutant animals (Ankle2^A) display disrupted nuclear envelope structure (Lamin Dm0, purple). See also Figure S4 and Movie S1.

To determine precisely where Ankle2 is localized, we performed live imaging of brains from animals carrying Ankle2-GFPR and a transgene that labels the ER: da-Gal4>UAS-Sec61β-tdTomato (Summerville et al., 2016). In neuroblasts (large cell in Figure 2C–E), the Ankle2 protein fully colocalizes with Sec61β at the nuclear envelope as well as the ER. In the surrounding neurons (small cells), much of the cytoplasm is co-labeled. We also counter-stained fixed samples with Calnexin 99a, another ER marker (Riedel et al., 2016). Again, Ankle2 localizes to the nuclear envelope and the ER, but in fixed samples, the ER structure is less obvious than in live imaging (Figure 2G–I).

To determine if Ankle2 is required for proper ER structure, we performed live imaging of Ankle2^A mutant neuroblasts expressing Sec61β-tdTomato (Figure 2F). When compared to wild type (Figure 2E), Ankle2^A mutants display highly aberrant Sec61β localization in many NBs (25°C). The neuroblast in Figure 2F appears larger than normal and displays Sec61β folds within the nucleus. In addition, we stained fixed Ankle2^A mutant neuroblasts with Calnexin 99a and found that Ankle2^A mutants also display irregular Calnexin 99a localization (Figure 2J versus 2I), similar to defects shown in Figure 2F, suggesting that even a partial loss of Ankle2 disrupts ER and possibly nuclear envelope structure. Indeed, the morphology of the nuclear envelope is aberrant and convoluted in some Ankle2^A mutant cells when stained with Lamin Dm0 (Riemer et al., 1995), a nuclear envelope marker (compare Figure 2N with Figure 2K–M). Hence, Ankle2 is required for proper ER and nuclear envelope morphology.

Ankle2 mutations affect the asymmetric localization of neuroblast determinants

Due to the reduced cell proliferation and reduced neuroblast number in Ankle2^A third instar brains (Yamamoto et al., 2014), we sought to explore neuroblast division in more detail. Neuroblast polarity during division relies on the function of the highly conserved apically localized Par complex, which consists of Bazooka (Baz, Par-3) (Schober et al., 1999), Par-6 (Petronczki and Knoblich, 2001), and atypical Protein Kinase C (aPKC) (Rolls et al., 2003). Once activated, the Par complex is responsible for restricting Miranda (Mira) and other cell fate determinants to the basal domain of neuroblasts. After division in most neuroblasts, the basal domain will become the ganglion mother cell, which divides again to produce neurons or glia (Betschinger et al., 2003; Atwood and Prehoda, 2009). Several proteins have been implicated in regulating the Par complex (Chabu and Doe, 2009; Andersen et al., 2012; Bonaccorsi et al., 2007; Atwood et al., 2007), including those associated with cell cycle regulation (Chabu and Doe, 2008; Lee et al., 2006; Wang et al., 2007; Wang et al., 2006).

A previous inspection of dividing neuroblasts revealed no obvious defects in Mira staining (Yamamoto et al., 2014). However, a quantitative analysis of dividing Ankle2^A neuroblasts stained with anti-Baz, Par-6, aPKC, and Mira revealed severe localization defects of these proteins in greater than 40% of metaphase neuroblasts during asymmetric division (Figure 3A–L, quantified in Figure 3M–P) in both Ankle2^A and trans-heterozygous animals (Ankle2^A/Ankle2^CRIMIC). These defects are rescued by the genomic construct (Figure 1D), Ankle2-GFPR (Figure 3D, H, L, M–P). Finally, we performed live imaging of 3^rd instar larval brains of wild type (Movie S2) and Ankle2^A mutants labeled with Mira-RFP and Histone-GFP (Movies S3–S5). As shown in Movies S3–S5, neuroblasts exhibit abnormal Mira localization as well as instances of failed division including DNA segregation defects, chromatin bridges, and cytokinesis defects (Movies S3–S4).

Figure 3: Ankle2 mutations affect asymmetric division, spindle alignment and centrosomes.

Metaphase neuroblasts stained with Baz (green, A-D), Par-6 (green, E-H), aPKC (green, I-L) and Mira (red, I-L) are shown in wild type (A, E, I), Ankle2^A/Y hemizygous (B, F, J), Ankle2^A/Ankle2^CRIMIC transheterozygous (C, G, K) and rescued (D, H, L) animals. Phospho-Histone H3 (pHH3, green, A-C, E-K) was used to identify cell cycle stage. * notes samples where pHH3 was not used. (M-P) quantification of phenotype severity demonstrate that Ankle2 is required for protein localization during asymmetric division in numerous cells. Below each graph, the # of neuroblasts counted for each genotype (n) is noted. (Q-U) Metaphase neuroblasts stained with aPKC (green) and Mira (red) to mark the polarity axis and DNA (white) and CNN (green puncta) to highlight the spindle axis from (Q) wild type (FRT19a) and (R-S) Ankle2 mutant neuroblasts. The angle between the spindle axis and polarity axis is measured and % of metaphase neuroblasts is plotted in 15° in tervals and is shown in (T-U). Below each group, the # of neuroblasts counted for each genotype (n) is noted. See also Movie S2–S5.

For proper neuroblast division to occur, cells must not only asymmetrically localize Par complex members and cell fate determinants, they must also align the mitotic spindle so that divisions segregate cell fate determinants to the proper daughter cell (Cabernard and Doe, 2009). In wild type neuroblasts, the mitotic spindle is aligned parallel to the polarity axis (Figure 3Q, T). Our initial observations in Ankle2^A mutants suggested no defects in mitotic spindle alignment in many cells (Yamamoto et al., 2014), but after a quantitative analysis of metaphase neuroblasts, we noted that spindle alignment appeared disrupted in some Ankle2^A mutant cells. To quantify these defects, we measured the axis of division using DNA and Centrosomin (CNN) (Lucas and Raff, 2007) to highlight centrosome placement relative to the localization of cell polarity proteins aPKC and Mira (Figure 3Q–U). Surprisingly, we found that nearly 40% of Ankle2^A mutant neuroblasts contained supernumerary centrosomes (Figure 3S). In the remaining 60% of Ankle2^A mutant metaphase neuroblasts with obvious aPKC/Mira localization, we also found varying degrees of mitotic spindle alignment defects (compare Figure 3Q–R and 3T–U), showing that Ankle2 is also required for proper spindle alignment in neuroblast division. Together, these results show that Ankle2 plays a prominent role in asymmetric protein localization, spindle alignment, and cell division of neuroblasts.

Ankle2 interacts with VRK1/Ballchen

A C. elegans homologue of Ankle2, Lem4L, was previously shown to physically and genetically interact with VRK1, the homologue of Ballchen (Ball) in flies (Asencio et al., 2012). Lem4L and VRK1 in worms localize to the nuclear envelope of the 2-cell stage embryo (Asencio et al., 2012). In contrast, Ball appears to be nuclear during interphase/prophase in all Drosophila cells (Figure S3D) (Yakulov et al., 2014) as well as in mammalian cells (Figure 4G–H), and both fly and human proteins have nuclear localization signals. Interestingly, human VRK1 pathogenic variants cause reduced brain size and microcephaly as well as axonal neuropathy in affected patients (Gonzaga-Jauregui et al., 2013; Renbaum et al., 2009). Hence, to characterize the relationship between Ankle2 and Ball/VRK1, we analyzed the expression and localization of Ball and Ankle2 during neuroblast cell division (Figure 4A–D). During interphase, Ankle2 and Ball do not co-localize as Ankle2 is in the cytoplasm and ER whereas Ball is in the nucleus (Figure 4A). During the mitotic prophase, Ankle2 accumulates at the nuclear envelope but the proteins do not seem to co-localize (Figure 4B). However, at metaphase, the nuclear envelope becomes fragmented but it does not completely dissociate in flies (Katsani et al., 2008). Ankle2 can still be seen localized to the fragmented nuclear lamina, and Ball is briefly present throughout the cytoplasm (Figure 4C). Yet, after telophase, Ball is quickly recruited back to the nucleus and briefly enriched at the nuclear envelope (Figure 4D; Movie S6). After mitosis and once the chromatin is no longer condensed, Ball is nuclear. Interestingly, the spatial restriction of Ball in Ankle2^A mutants during interphase and prophase is abolished in many neuroblasts as Ball localizes throughout the cell, a phenotype that is not observed in wild type brains (Figure 4E–F, quantified in Figure S3E). In summary, Ankle2 is required for proper nuclear localization of Ball in Drosophila.

Figure 4: Ankle2 controls Ballchen/VRK1 localization and function.

(A-D) Immunostaining of Ankle2-GFPR (green) and Ball (red) show dynamic localizations during the cell cycle. (B) Ankle2 is localized at the nuclear envelope at prophase. Ball is nuclear (A-B) until nuclear envelope breakdown (C) and then localizes to the cytoplasm through the end of mitosis (D). (E-F) Immunostaining of Ball (red) and Lamin (white) in Ankle2^A/+ heterozygotes and Ankle2^A/Y hemizygous mutants. Note that Lamin is disrupted and Ball becomes mislocalized throughout the cytoplasm during interphase in Ankle2^A/Y hemizygous mutants. Quantification is shown in Figure S3E. (G-H) Confocal projections of immunostaining of VRK1 (green) and DNA (purple) in primary human fibroblasts from (G) parental unaffected (p.L573V/+) and (H) an ANKLE2 compound heterozygous patient (p.L573V/p.Q782*). VRK1 is mislocalized in fibroblasts carrying microcephaly associated ANKLE2 variants (p.L573V/p.Q782* and p.V229G/p.V229G), quantified as nuclear intensity in (I). Arrows in (H) indicate cytoplasmic VRK1 staining, which is minimal in (G) control fibroblasts. (J-L) Partial projections of 3^rd instar larval brains stained with apical marker aPKC (green) and basal marker Miranda (red) in (J) wild type (y,w, FRT19a), (K) Ankle2^A, and (L) Ankle2^A;;ball^e107/+ animals. Note that removal of a single copy of ball rescues the phenotypes of Ankle2^A. (M) Quantification of 3^rd instar larval brain size (as shown in (J-L)). N≥6. One-way ANOVA with multi-comparison post-test. ****p<0.0001, ***p<0.001. Box plots hinges represent the 25^th to 75^th percentiles, a line is at the median, and whiskers represent min to max. (N) Ankle2^A, but not Ankle2^CRIMIC, lethality is rescued with introduction of multiple ball heterozygous mutations. (O-P) Quantification of (O) aPKC or (P) Mira crescent intensity in 3^rd instar metaphase neuroblasts in wild type (y w FRT19a), Ankle2^A and rescued Ankle2^A; ball^e107/+ animals, demonstrating that Ankle2 asymmetric division phenotypes are rescued with ball heterozygosity. Note that wild type and Ankle2^A quantifications were shown in Fig 3. Below each graph, the # of neuroblasts or brains counted for each genotype (n) is noted. See also Figure S3 and S5 and Movie S6.

To determine whether ANKLE2 regulates Ball/VRK1 subcellular localization in human cells, we assayed VRK1 localization in human fibroblasts. In reference human primary fibroblasts (parental variant p.L573V/+), VRK1 is localized to the nucleus (Figure 4G) consistent with published literature indicating that VRK1 is nuclear localized (Nichols and Traktman, 2004). However, fibroblasts from microcephaly patients carrying compound heterozygous variants in ANKLE2 (p.L573V/p.Q782* and p.V229G/p.V229G) display significantly reduced VRK1 intensity in the nucleus (Figure 4H, quantified in Figure 4I) and increased cytoplasmic staining in nondividing cells (arrows in Figure 4H) with no significant change in overall VRK1 intensity (Figure S3F). These data argue for a conserved role between fruit flies and human for ANKLE2 in restricting VRK1 to the nucleus.

Given that Ankle2 is required to maintain Ball/VRK1 in the nucleus during interphase, it is possible that Ball/VRK1 is ectopically active in the cytoplasm of Ankle2^A mutants and inhibits or promotes phosphorylation of proteins not normally encountered in the biological homeostatic state. Reducing the level of Ball/VRK1 may therefore alleviate the phenotype associated with the reduction in Ankle2 protein. Indeed, we observe evidence for strong dominant interactions between Ankle2^A and ball (multiple alleles). Ankle2^A animals are pupal lethal and have reduced brain volumes (compare Figure 4J to 4K). However, removal of one copy of ball, akin to a heterozygous deletion CNV resulting in haploinsufficiency in human, restores brain development (Figure 4L–M) and suppresses the lethality of Ankle2^A mutants (Figure 4N). Importantly, loss of one copy of ball (ball^e107) in Ankle2^A mutants also restores the asymmetric protein localization of aPKC and Mira crescents in metaphase neuroblasts (Figure 4O–P). Hence, a partial reduction of Ball/VRK1 activity rescues Ankle2^A mutants, providing strong evidence for a gene dosage sensitive locus. However, removing both copies of ball in wild type animals leads to pupal lethality (Cullen et al., 2005), causes severely reduced brain volumes in 3^rd instar larvae (Herzig et al., 2014), and does not rescue Ankle2^A animals, emphasizing that the gene dosage and balance of the protein levels is critical. Indeed, a severe loss of function allele, Ankle2^CRIMIC, cannot be suppressed by reducing Ball/VRK1 activity (Figure 4N). In summary, these data demonstrate that both Ankle2 and Ball/VRK1 control the distribution of asymmetric determinants, and experimental evidence reveals an antagonistic relationship between both proteins.

The Ankle2-Ball pathway modulates aPKC and L(2)gl

Due to the similarities in defects observed with loss of Ankle2 or aPKC, including mislocalization of Par-6 and Mira (Kim et al., 2009), decreased cell divisions, and reduced neuroblast clone volume (Rolls et al., 2003), we hypothesized that the activity of aPKC, an important mediator of neuroblast asymmetric division (Figure 5A), might be affected. aPKC phosphorylation (Kim et al., 2009) or abundance could be modulated by Ankle2. We therefore assessed both total and phosphorylated aPKC levels in third instar larval brains using an antibody specific for human p-aPKC T410 (T422 in flies). This phosphorylation site is located in its activation loop and was shown to be important for its kinase activity (Kim et al., 2009). Phosphorylation of aPKC (T422) relative to total aPKC is decreased in Ankle2 mutants (Figure 5B) and is restored with either addition of Ankle2-GFPR or reduction of ball (Figure 5B), consistent with the data presented in Figure 4. However, overexpression of aPKC or constitutively active aPKC (aPKC^ΔN) (Betschinger et al., 2003) in Ankle2^A mutants did not rescue brain size or viability (data not shown).

Figure 5: Ankle2 affects asymmetric division through aPKC and L(2)gl.

(A) Ankle2 and Ball regulate asymmetric division. (B) Western analysis of phosphorylated aPKC in larval brains (mammalian T410 corresponds to T422 in Drosophila) from genomic rescue (Ankle2^A; Ankle2-GFPR), Ankle2^A, and Ankle2^A;; ball^e107/+ animals or total aPKC levels in genomic rescue (Ankle2^A; Ankle2-GFPR) and Ankle2^A mutants. Note that p-aPKC is reduced in Ankle2 mutants but restored with introduction of genomic rescue (GR) or reduction of Ball and is quantified as the ratio of p-aPKC to total aPKC. N=3 replicates. (C) In vivo immunoprecipitation of L(2)gl-GFP using GFP-nAb in L(2)gl^{MI07575-GFSTF} larvae demonstrates L(2)gl can interact with Ball in vivo. (D,F) Partial projections of third instar larval brains stained for Dpn (purple, neuroblasts) and Pros (green, daughter cells and neurons) of (D) Ankle2^A and (E) Ankle2^A/+;l(2)gl^ts3 mutant animals raised at 22°C with brain volume qua ntified in (F). Note that reduction of L(2)gl in an Ankle2^A hemizygous animals rescues brain size defects and lethality at 22°C. Box plots hinges represent the 25^th to 75^th percentiles, a line is at the median, and whiskers represent min to max. Below the graph, the # of brains counted for each genotype (n) is noted. See also Figure S6.

aPKC has been shown to physically interact with L(2)gl (Betschinger et al., 2003), a regulator of apico-basal polarity that inhibits the function of aPKC (Atwood and Prehoda, 2009; Wirtz-Peitz et al., 2008). aPKC and l(2)gl genetically interact as removal of one copy of aPKC suppresses l(2)gl loss of function phenotypes (Rolls et al., 2003), and aPKC has been shown to phosphorylate L(2)gl to control its plasma membrane or cortical release (Betschinger et al., 2003). When aPKC is active, L(2)gl is phosphorylated and released from the cortex; once released, it no longer binds to aPKC or inhibits its function. Because aPKC and L(2)gl interact, the Ankle2-Ball pathway may affect L(2)gl. We therefore assessed whether L(2)gl physically interacts with the Ankle2-Ball pathway using immunoprecipitation of a GFP-tagged L(2)gl from third instar larval brains and found that Ball indeed interacts with L(2)gl (Figure 5C).

The reduced aPKC activity that we observe may be associated with a gain of function of L(2)gl. Therefore, to determine whether removal of one copy of l(2)gl suppresses Ankle2 associated phenotypes, we introduced a temperature sensitive mutation of l(2)gl (l(2)gl^ts3) into the Ankle2^A mutant background and found that reducing l(2)gl in Ankle2^A mutants at 22°C (Figure 5D–F) and 25°C indeed partially restored brain size. Ankle2^A is pupal lethal at 22°C, but when combined with a heterozygous l(2)gl mutant allele, some Ankle2^A animals survive to adulthood. However, unlike the removal of one copy of ball, these animals die a few days after eclosion. In summary, Ankle2 and Ball interact with the apical-basal polarity regulators aPKC and L(2)gl (Figure 5A) and affect aPKC and L(2)gl activity by disturbing the asymmetric segregation of apical-basal polarity factors in neuroblasts. Our data suggest that in Ankle2 mutants, L(2)gl acts as a gain of function (increased activity), and aPKC activity is reduced. Therefore, by reducing L(2)gl activity, aPKC function is restored.

Disease associated variants in VRK1 and its paralogs

Ten families have been described with biallelic variants in VRK1 that cause a spectrum of neurologic diseases including 6 individuals with microcephaly (Feng et al., 2018; Gonzaga-Jauregui et al., 2013; Najmabadi et al., 2011; Nguyen et al., 2015; Renbaum et al., 2009; Shaheen et al., 2018; Stoll et al., 2016) (Table S1; Figure S5). The family structures suggest either a sporadic or recessive neurological disease trait; historical consanguinity in 3/10 pedigrees implicate an autosomal recessive locus. Screening the BHCMG and BG databases identified two additional families with potentially biallelic variants in VRK1. (Table S2, Figure S5). These cases suggest that like ANKLE2 (Figure S1), a heterogenous set of variant alleles in VRK1 are associated with neurologic disease and microcephaly.

It was previously shown that fly genes with more than one human homolog, especially those that are evolutionarily conserved, have an enriched association with Online Mendelian Inheritance in Man (OMIM) disease phenotypes (Yamamoto et al., 2014), which is a comprehensive and authoritative catalogue of human genes and genetic phenotypes associated with mendelian disorders. We searched the BHCMG database to establish if damaging variants in paralogs of VRK1 are associated with disease. Predicted deleterious, biallelic variants were found in two paralogs of VRK1: VRK2 is associated with very small eyes and VRK3 with severe microcephaly (Table S2, Figure S6).

NS4A targets the Ankle2 pathway

Drosophila has been developed as a model of viral infection (Harsh et al., 2018; Liu et al., 2018), and we recently showed that expression of the Zika virus protein NS4A results in reduced brain size in Drosophila (Shah et al., 2018). Strikingly, NS4A expression in Ankle2^A/+ heterozygous animals leads to a more severe phenotype than NS4A expression in a wild type background, and these animals display brain phenotypes that mimic Ankle2^CRIMIC null mutants (Shah et al., 2018). These data again suggest that levels of Ankle2 protein are critical. In our previous experiments, we were unable to detect NS4A using immunohistochemistry. We therefore generated an additional NS4A construct with a C-terminal HA tag (UAS-NS4A-HA) and drove expression with insc-GAL4>UAS-NS4A-HA. As shown in Figure 6A, NS4A is localized to the nuclear envelope and ER (Figure 6A), similar to the subcellular localization of Ankle2 (Figure 2). Expression of NS4A may cause brain defects by affecting aPKC and Miranda localization. Indeed, expression of NS4A in neuroblasts (insc-GAL4>UAS-NS4A) affects the apical aPKC localization and leads to an expansion of the Mira domain (Figure 6, compare B to C-D and quantified in Figure 6F–G). In the metaphase neuroblasts that express NS4A, we also note spindle orientation defects in some cells (Figure 6J–K), similar to Ankle2^A animals shown in Figure 3. These data indicate that NS4A targets the Ankle2 pathway; this is further strengthened with the observation that when NS4A is expressed in neuroblasts of ball heterozygous animals, aPKC and Mira crescents are restored to their wild type patterns (Figure 6E–G) and spindle orientation defects are rescued (compare Figure 6L with K). Finally, ubiquitous expression of NS4A-HA (Figure 6H–I) using two different Act-GAL4 insertions causes reduced brain volume (Figure 6H) that is rescued by removal of one copy of ball, (Figure 6I) co-expression of human ANKLE2, or (Figure 6I) removal of one copy of l(2)gl. In summary, the Zika virus protein NS4A targets the Ankle2 pathway and affects asymmetric distribution of cell fate determinants, leading to defects in neuroblast division and brain development.

Figure 6: Zika virus NS4A targets the Ankle2 pathway.

(A) 3^rd instar brains with neuroblast specific (insc-GAL4) NS4A-HA expression stained for HA (green) and DNA (white) shows that NS4A localizes in a pattern similar to Ankle2 (Figure 2). (B-E) Metaphase neuroblasts stained for aPKC (green) and Mira (purple) in brains with neuroblast specific (insc-GAL4) expression of (B) CD8-GFP, (C-D) NS4A, or (E) NS4A in ball^e107 heterozygous animals. aPKC and Mira crescent intensities are quantified in (F-G). (H) Brain volume quantification from animals with ubiquitous expression of CD8-GFP (control), NS4A-HA, or NS4A-HA in ball^e107 heterozygous animals using Act-GAL4 on the 2^nd chromosome. (I) Act-GAL4 (3^rd chromosome) ubiquitous expression of CD8-GFP (control), NS4A-HA, NS4A-HA and human ANKLE2, or NS4A-HA in l(2)gl^ts3 heterozygous animals. Note that human ANKLE2 expression or reduction of Ball or L(2)gl activity rescues NS4A induced brain defects. One-way ANOVA with multi-comparison post-test. ****p<0.0001, **p<0.01, *p<0.05. Box plots hinges represent the 25^th to 75^th percentiles, a line is at the median, and whiskers represent min to max. (J-L) aPKC (green), CNN (green), and Mira (purple) staining of metaphase neuroblasts with insc-GAL4 expression of (J) CD8-GFP, (K) NS4A, and (M) NS4A in ball^e107 heterozygous animals. The angle between the spindle axis and polarity axis is measured and % of total metaphase neuroblasts is noted at 15° intervals. Expression of NS4A causes localizati on defects of aPKC and Mira and spindle orientation defects in metaphase neuroblasts. (M) Zika virus NS4A inhibits the ANKLE2/VRK1 pathway, which regulates asymmetric determinant localization as well as the division axes.

DISCUSSION

We investigated the biological basis for ANKLE2 associated microcephaly. We report six additional patients with microcephaly that carry mutations in ANKLE2 and show that three variants identified in probands cause a loss of ANKLE2 function when tested in flies (Figure 1 and S1), providing compelling evidence that its loss causes reduced brain size in flies and severe microcephaly (Z-score < −2.5) in humans. Ankle2 is a dosage sensitive locus whose product is inhibited by the Zika virus protein NS4A. We show that Ankle2, like NS4A, is localized to the ER, and that it targets the nuclear envelope during mitosis. Loss of Ankle2 affects the nuclear envelope and ER distribution and results in a redistribution of Ball/VRK1, a kinase that is normally localized to the nucleus except when the nuclear envelope breaks down during mitosis (Figure 4). Loss of Ankle2 disrupts the localization of neuroblast apical-basal polarity determinants such as aPKC, Par-6, Baz, and Miranda, and aPKC phosphorylation is reduced by Ankle2 mutations. Importantly, loss of one copy of ball or l(2)gl suppresses the reduced brain volume associated with a partial loss of Ankle2, suggesting that much of the biological function of Ankle2 is modulated by aPKC and L(2)gl. Finally, the negative influence of NS4A on the activity of ANKLE2 can also be suppressed by removal of one copy of ball or l(2)gl, suggesting the following pathway: NS4A ┤ ANKLE2 ┤ Ball/VRK1 → L(2)gl/LLGL1 ┤ aPKC. This pathway regulated by ANKLE2 plays an important role in neuroblast stem cell divisions in flies and microcephaly and potentially other neurological disease phenotypes in humans.

Interestingly, the above pathway links environmental cues with several genetic causes of sporadic and autosomal recessive microcephaly in human; moreover, it implicates this pathway in microcephaly accompanying congenital infection. As one example of the latter, the Zika virus has been shown to cross the infant Blood Brain Barrier (Mlakar et al., 2016) and has been identified in radial glial cells (Li et al., 2016), as well as intermediate progenitor cells and neurons (Lin et al., 2017). We propose that NS4A affects the function of Ankle2 leading to the release of Ball/VRK1 from the nucleus. We speculate that this in turn affects the phosphorylation of aPKC and L(2)gl directly by masking phosphorylation sites or indirectly by promoting the activity of one or more phosphatases. Loss of VRK1 has been shown to cause microcephaly and some variant alleles are also associated with pontocerebellar hypoplasia (PCH) in humans (Gonzaga-Jauregui et al., 2013; Renbaum et al., 2009), consistent with the loss of ball in flies that causes a severe reduction in brain size (Yakulov et al., 2014). Note that ANKLE2, VRK1, LLGL1, and aPKC, as well as other components of the apical complex like PARD3 are all present in radial glial cells during cortical development (Ayoub et al., 2011). These data suggest that ANKLE2 and its partners such as LLGL1 and asymmetric determinants are important proteins during neural cell proliferation and that the proper levels and relative amounts of these proteins determine how many neurons will eventually be formed in vertebrates. Our data also indicate that variant alleles at either ANKLE2 or VRK1 are responsible for some causes of embryonic lethality and severe congenital microcephaly.

LLGL1 has recently been shown to play an important role in radial glia in mice during neurogenesis, and its loss in clones increases the number of divisions (Beattie et al., 2017). In addition, aPKCζ/λ localizes at the apical membrane of proliferating neural stem cells in chicken embryos during division and has been shown to provide an instructive signal for apical assembly of adherens junctions (Ghosh et al., 2008). Mouse knockouts of aPKCλ (Soloff et al., 2004) and aPKCι (Seidl et al., 2013) are embryonic lethal; however aPKCζ knockouts are viable (Leitges et al., 2001), perhaps suggesting redundant functions within the atypical PKC family. These proteins have not been linked to microcephaly in mice, but conditional removal of an apical complex protein Pals1 in cortical progenitors resulted in complete cortex loss (Kim et al., 2010). Finally, Numb is asymmetrically localized by the Par complex protein in Drosophila, segregated to the daughter cell during asymmetric cell division (Wirtz-Peitz et al., 2008), and essential for daughter cells to adopt distinct fates (Bhalerao et al., 2005). In mice, Numb localization is also asymmetric and null mutations exhibit embryonic lethality, neural tube closure defects, and premature neuron development (Zhong et al., 2000). These data indicate that asymmetric division may be important for vertebrate neuronal development, but microcephaly is not a phenotype that typically associates with loss of the mice homologues of asymmetric localized determinants identified in Drosophila. However, the observations reported here indicate that the ANKLE2/PAR complex pathway is evolutionarily conserved from flies to human, although the precise mechanisms remain to be determined as different cells may use this pathway in different contexts (Suzuki and Ohno, 2006).

In order to determine whether predicted deleterious biallelic variants in PAR complex encoding genes or their paralogs associated with a neurologic disease trait, we searched the BHCMG database for mutations associated with neurological disease. We found homozygous predicted deleterious missense variants in PARD3B (c.1222G>A, p.G408S) in a patient that has microcephaly (Table S2, Figure S6) and compound heterozygous mutations in PARD3B (c.1654G>A;p.A552T) that are associated with other neurological defects (Table S2, Figure S6). The human orthologue of L(2)gl, LLGL1, is deleted in Smith-Magenis syndrome (SMS) (Smith et al., 1986) and 86–89% of the SMS patients have brachycephaly (Greenberg et al., 1996). These observations extend the mutational load beyond ANKLE2 and VRK1 and suggest an association between congenital disease and variants within the PAR complex (Table S2, Figure S6) potentially by a compound inheritance gene dosage model (Wu et al., 2015).

Aurora A (AurA) kinase has been shown to phosphorylate the Par complex (Wirtz-Peitz, Nishimura and Knoblich, 2008) as well as L(2)gl (Carvalho et al., 2015) and regulates cortical polarity and spindle orientation in neuroblasts (Lee et al., 2006; Wang et al., 2006). The aberrant localization of Ball/VRK1 in Ankle2 mutants may lead to gain of function phenotypes that are highly dosage sensitive, as they can be repressed by removing a single copy of Ball/VRK1 in Ankle2^A. Mislocalized Ball/VRK1 may mask or interfere with the function of AurA in neuroblast asymmetric division as they share similar kinase substrate consensus sequences (Sanz-García et al., 2011; Ferrari et al., 2005). Future studies are needed to assess Ball/VRK1 redundancy or interference with AurA function.

Another possible evolutionarily parallel with implications in multicellular organismal development is the genetic interaction between the C. elegans homologue of VRK1 and an ANKLE2-like protein at the two cell stage (Asencio et al., 2012). Whereas VRK1 in both Drosophila and humans (Nichols and Traktman, 2004) is localized to the nucleus, except during mitosis when the nuclear envelope is broken down (Figure 4), the worm VRK1 protein is localized to the nuclear envelope. The worm ANKLE2-like protein, Lem-4L, also interacts with the phosphatase PP2A (Asencio et al., 2012), and the fly PP2A regulates neuroblast asymmetric division by interacting with aPKC and excluding it from the basal cortex (Chabu and Doe, 2009; Ogawa et al., 2009; Wang et al., 2009). PP2A also antagonizes the phosphorylation of Baz by PAR-1 to control apical-basal polarity in dividing embryonic neuroblasts (Krahn et al., 2009) and regulates Baz localization in other cells such as neurons (Nam et al., 2007). This raises the possibility that the Ankle2 pathway also acts with PP2A in neuroblast asymmetric division.

Here, we identified a pathway that plays a significant role in neuroblast asymmetric division. By combining functional studies in Drosophila together with human subject data, we have linked several microcephaly-associated genes and congenital infection to a single genetic pathway. These studies allowed us to highlight conserved functions of the ANKLE2 pathway, and provide mechanistic insight into how a Zika infection might affect asymmetric division. This ANKLE2-VRK1 gene dosage sensitive pathway can be perturbed by genetic variants that disturb biological homeostasis resulting in neurological disease traits or by environmental insults such as Zika virus impinging on neurodevelopment. Hence, lessons learned from the study of rare diseases such as MCPH16/ANKLE2 can provide insights into more common disease and potential gene by environmental interactions.

STAR Methods

Lead Contact and Materials Availability

All unique/stable reagents generated in this study are available from the Lead Contact without restriction. Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Hugo J. Bellen (hbellen@bcm.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Drosophila melanogaster

The following fly lines were used: FRT19a (Yamamoto et al., 2014), Ankle2^A (Yamamoto et al., 2014), Ankle2^CRIMIC (this study), Ankle2^IGFP(this study), Ankle2-GFPR (this study), P{UASt-hANKLE2}VK37 (Yamamoto et al., 2014), P{UASt-hANKLE2 p.L537V}VK37 (this study), P{UASt-hANKLE2 p.Q782*}VK37 (this study), P{UASt-hANKLE2 p.A109P}VK37 (this study), P{UASt-hANKLE2 p.G201W}VK37 (this study), P{20XUAS-tdTomato-Sec61beta}attP2 (Summerville et al., 2016), ball^e107(Cullen et al., 2005), ball² (Herzig et al., 2014), l(2)glMI07575-GFSTF.0 (Nagarkar-Jaiswal et al., 2015a), l(2)gl^ts3cn¹sp¹³ (Manfruelli et al., 1996), P{UASt-NS4Aug} (Shah et al., 2018), P{GW UAS-NS4Aug-HA} (this study), P{UASt-CD8-GFP} (Lee and Luo, 2001), Actin-GAL4 (P{Act5C-GAL4}17bFO1) (Ito et al., 1997), inscuteable-GAL4 (P{w[+mW.hs]=GawB}insc[Mz1407]) (Luo et al., 1994), daughterless-GAL4 (P{w[+mW.hs]=GAL4-da.G32}UH1) (Wodarz et al., 1995), ball-GFP (fTRG-823) (Sarov et al., 2016), wor-mira-cherry P{w[+mC]=wor.GAL4.A}2,P{w[+mC]=UAS-mira.cherry}2/CyO) (Cabernard and Doe, 2009), P{His2Av[T:Avic\GFP-S65T]}62A (Clarkson and Saint, 1999), P{w[+mC]=UAS-aurA.Exel}2, M{ UAS-aPKC.ORF.3xHA}ZH-86Fb (Bischof et al., 2013), P{w[+mC]=UAS-aPKC.DeltaN}3 (Drier et al., 2002). All flies were maintained at 22°C and grown on standard cornmeal and molasses medium in plastic vials. Crosses were performed at temperature indicated (18°C, 22°C, 25°C, or 29°C). Hemizygous males were analyzed as Ankle2 mutants (which is on the X chromosome) and females were used for Ankle2 heterozygous studies. All other studies contained males and females. Brain volume measurements were conducted in late 3^rd instar larvae (gut clearance, extruding spiracles). Act-GAL4 was use to ubiquitously express Zika virus NS4A and Sec61β, da-GAL4 was used to express human ANKLE2 constructs and aPKC, insc-GAL4 was used to express NS4A in neuroblasts.

Human studies

All study subjects enrolled into the Baylor-Hopkins Center for Mendelian Genomics (BHCMG) provided informed consent for exome sequencing and study participation under the Baylor College of Medicine Institutional Review Board-approved protocol H-29697. BAB701 provided informed consent for molecular and genomic analysis under the Baylor College of Medicine Institutional Review Board-approved protocol H-9170. All study subjects enrolled through Baylor Genetics Laboratory (BGL) were analyzed on a retrospective basis, and only deidentified information is provided under the Baylor College of Medicine Institutional Board approved protocol H-41191. Patients were ascertained from the 7148 sequenced individuals in BHCMG or the ~12500 sequenced individuals in the BGL by searching for biallelic variants with CADD scores >15 in conjunction with phenotypes of interest. Six male and two female patients were ascertained from the BHCMG database. The age of the patient is known for 3 individuals (2y, 7y, 32y). Two female and three male patients were ascertained from the BGL database. Ages at referral were 2 months, 7y, 12y, 20y, and 41y.

Human Clinical Information

BAB4821

A homozygous missense variant in VRK2 (c.1234G>A;p.D412N; Table S2, Figure S6) was observed in a proband, BAB4821 born to consanguineous Turkish parents, with nanophthalmos. The variant was in a region of absence of heterozygosity (AOH) ~7.7Mb in length gleaned from unphased ES data (Karaca et al., 2018) and consistent with identity-by-descent.

BAB7812

In BAB7812, a patient with severe microcephaly (Z = −6.7) and additional brain malformations including cortical dysplasia and agenesis of the corpus callosum, a homozygous variant with a predicted detrimental effect on splicing in a canonical splice site within VRK3 (c.139+2 T>G) was found (Table S2, Figure S6). Additional homozygous variants were found in the same (PNKP) and an adjacent (RNT2) AOH block as VRK3 (Table S2, Figure S6). All three homozygous variants are predicted deleterious. PNKP has a known disease association with microcephaly, seizures, and developmental delay (MIM# 613402) while RNT2 has a known disease association with Spastic paraplegia 12 (MIM# 604805). Multilocus variation may contribute to the phenotype in BAB7812. The proband was born full term to consanguineous Turkish parents. At 5 months of age, he developed seizures. At 32 months, he was unable to walk, and his weight was 9.8kg (−3.34SD), height was 85cm (−2.03SD), and head circumference was 36.7cm (−6.7SD). Physical examination revealed thick eyebrows, protuberant ears, and almond shaped eyes. Metabolic and ocular examinations were within normal limits. He received continued follow-up care for microcephaly and refractory epilepsy, and special education for associated intellectual disability and developmental delay. His occipital frontal circumference was below the 3^rd percentile when measured at 8–11 years of age. Parents and an unaffected sibling were heterozygous for the variant, demonstrating segregation of the variant with disease.

BAB10531

A BHCMG search database found a homozygous predicted deleterious missense variant in PARD3B (c.1222G>A, p.G408S) in BAB10531 who has microcephaly (Table S2, Figure S6). At 15 months, weight he was 9kg (−2.1SD), height was 79cm (0.4SD), and head circumference was 44.5cm (Z = −2.2 measured at 15m and and 18m). Both unaffected consanguineous parents of Turkish descent were heterozygous for the variant that mapped to a region of AOH ~22Mb in length. BAB10531 has had frequent upper respiratory tract infections and was admitted at 17 months of age to investigate a cause for with no clear etiology found. Follow-up head circumference was performed at 18 months, which demonstrated microcephaly was nonprogressive (45cm, −2.2SD). Notably, no follow-up for intellectual disability was described on recruitment to the BHCMG.

BAB8223

In BAB8223, a proband with distal arthrogryposis, a homozygous missense variant in PARD3B (c.1654G>A;p.A552T) which segregated with disease was found (Table S2, Figure S6). The proband was born to consanguineous Turkish parents and the homozygous variant was found within a region of AOH ~37Mb in length. At age 5.5y head circumference was 49cm (13^th percentile, ~ −1SD). Cognitive and motor development is within normal limits.

BGL – 1

Referral diagnosis: Microcephaly, intrauterine growth restriction, radial microbrain and immature gyral pattern.

BGL – 2

Referral diagnosis: Global developmental delay with aphasia, microcephaly and seizures, progressive weakness, lethargy with episodes of irregular breathing and respiratory acidosis, MRI showing progressive diffuse central white matter atrophy and gliosis, hypomyelination, atrophy of thalami and pons. (Table S2, Figure S2)

BGL – 3

Referral diagnosis: Intellectual disability, ataxia, spasticity, autism spectrum disorder, speech delay, and anxiety disorder. (Table S2, Figure S2)

BGL – 4

Referral diagnosis: Spinal muscular atrophy complicated by failure to thrive and respiratory insufficiency, microcephaly, short stature, scoliosis, delayed motor milestones, progressive weakness, hypotonia, history of prematurity and intrauterine growth retardation (Table S2, Figure S5).

BGL – 5

Referral diagnosis: Prematurity, spasticity of inferior limbs, seizures, contractures of the Achilles, myopia (Table S2, Figure S5).

Human primary cultures

Fibroblasts were cultured in flasks containing Gibco DMEM (1x) with 4.5g/L D-Glucose, L-Glutamine, 25mM HEPES, HyClone FBS (10%), and Gibco Anti-Anti (1%) at 37 degrees Celsius with 5% CO₂. Cultures of p.L573V/+ and p.L573V/p.Q782* were male; p.V229G/p.V229G was female.

METHOD DETAILS

Generation of Ankle2 mutations and constructs

To generate Ankle2^CRIMIC by CRISPR-Cas9, two guide RNAs targeting Ankle2, 5’-ATAAAGTATTTTCTTAACGGTGG-3’ and 5’-TAATAATTTTAAATTCTCATTGG-3’ with PAM sites underlined, were cloned into pCDF3 (Port et al., 2014). Regions of homology targeting the 4^th coding intron were cloned into PM14 (Lee et al., 2018) using 5’- ccatagctatggGCAATTCCTCAATGTCGAATTTACTGCTCA-3’ and 5’- ttatgcatATTTTCTTAACGGTGGGAAATTATAC-3’ to amplify the left arm for BstXI/NsiI cloning and 5’- tagcatgcATACTTTATTATTGCATTTGTTATAAGTATGAGA −3’ and 5’- tactcgagGCAAAGTTCCAGACCGTTTCTGATTTATC −3’ to amplify the right arm for conventional cloning with SphI/XhoI. This donor construct and two guide RNA constructs were injected into ,w;attP40(y+){nos-Cas9(v+)}/CyO (Kondo and Ueda, 2013) embryos, and positive expression of 3XP3-GFP was used to isolate animals with targeted events. PCR and genomic sequencing of surrounding regions validated the Ankle2^CRIMIC allele. Ankle^IGFP was generated by RMCE by injecting a plasmid expressing integrase with pBS-KS-attB1–2-PT-SA-SD-EGFP-FlAsH-StrepII-TEV-3xFlag (Nagarkar-Jaiswal et al., 2015b) into Ankle2^CRIMIC animals. Animals with 3XP3-GFP loss were screened using PCR for targeted cassette exchange. Regions flanking the targeted event were sequenced to verify the allele. The Ankle2-GFPR was created using recombineering (Venken et al., 2008) of BAC CH321–85N12 (Venken et al., 2006). A GFP donor construct was generated by amplifying the GFP coding region and a selection cassette from plasmid PL-452 C-EGFP with primers containing 50bp homology with the C-terminal end of Ankle2 (5’-GGGATCAACGGTCCTATAACGAGGGGGACACGCCGCTGGGCAATCGGAACGCAGCCCAATTCCGATCATATTC-3’ and 5’-CATCAATCAGTCGCTGTTTCTGTTTCTGTTTCCGGGCCGATT CCGTTTCATTACTTGTACAGCTCGTCCATG-3’. Regions matching PL-452 C-EGFP are underlined. CH321–85N12 was transformed into DY380 cells using electroporation. Stable colonies were grown overnight at 30°C, induced for recombination functions at 42°C for 15 min, and transformed using electroporation (1.8kV, 200Ohm, 25μFD) with the amplified donor construct. Colonies were selected for both the BAC (chloramphenicol) and insertion of GFP (kanamycin). Resulting colonies were verified using PCR, restriction enzyme digestion, and sequencing. The GFP tag selection cassette was removed using Cre mediated excision by transforming the Ankle2-GFP BAC into induced EL350 cells. Properly excised events were verified by PCR, absence of growth on kanamycin selection plates, and sequencing.

Generation of human ANKLE2 and NS4A-HA expression constructs

NEB Q5 Site-directed mutagenesis was performed on P{UASt- hANKLE2}. Each plasmid was sequence verified and injected into VK37 flies with a plasmid expressing integrase for site-specific integration. NS4A, including the 2K peptide, from strain MR-766 was PCR amplified from UASt-NS4A (Shah et al., 2018) and Gibson assembly was used to insert NS4A into pGW-HA.attB linearized with AgeI and KpnI to remove the ccdB region. Resulting vector was injected into VK37 flies as above.

Brain immunostaining

Late 3^rd instar (based on gut clearance and extruding spiracles) larval brains were dissected in PBS and fixed with 4% PFA/PBS/0.3%Triton for 20 minutes. For immunostaining, brains were blocked in PBS/0.3%Triton/1%BSA/5% normal goat serum and incubated in primary antibody in PBS/0.3%Triton/1%BSA overnight. Primary antibodies include rat anti-Deadpan (Abcam Cat# ab195172, 1:250 or 1:500), mouse anti-Prospero MR1A (Developmental Studies Hybridoma Bank, 1:1000), rat anti-Miranda (1:500, Abcam Cat# ab197788), rabbit anti-aPKC (1:1000, PKCz (C-20) Santa Cruz, discontinued), rabbit anti-GFP (1:1000, Invitrogen Cat# A11122), mouse anti-Calnexin 99a (Developmental Studies Hybridoma Bank, 1:100), mouse anti-Lamin Dm0 ADL67.10 (Developmental Studies Hybridoma Bank, 1:250), guinea pig anti-Bazooka (1:1000) (Siller et al., 2006), rat anti-Par-6 (1:50) (Rolls et al., 2003), rabbit antiphospho-Histone H3 (1:1000, Millipore Cat# 06–570), rabbit anti-Ball (1:1000) (Yakulov et al., 2014), rabbit anti-VRK1 (1:1000, Abcam Cat# ab151706), rabbit anti-CNN (1:1000) (Lucas and Raff, 2007), and mouse anti-Strep (Qiagen Cat# 34850, 1:500) with goat or donkey secondary antibodies from Jackson ImmunoResearch used 1:500. Brains were mounted with double sided tape spacers and imaged using a Leica Sp8 with 2 μm or 3 μm sections through the entire brain lobe.

Live imaging

3^rd instar larvae were dissected in sterile PBS supplemented with 1% FBS and 0.5mM ascorbic acid, fine dissected on an inverted Sarstedt lumox dish 50 in a petroleum jelly well. Samples were imaged on a Leica Sp8 with optimized settings for high quality images without bleaching or a Zeiss 880 with Airy scan (wild type Ankle2-GFP and Sec61β colocalizaiton).

Protein immunoprecipitation and western analysis

3^rd instar larvae or dissected larval brains from l(2)glMI07575-GFSTF.0 animals were dissociated in 0.1% CHAPS buffer supplemented with protease and phosphatase inhibitors for at least 30 min on ice, centrifuged for 10 min at 4°C, and supernatant was used for immunoprecipitation or western analysis. For immunopreciptation, 25ul of Allele Biotechnology GFP nanoantibody agarose (nAb, Cat# ABP-NAB-GFPA100) was equilibrated and incubated with lysate 2hrs - overnight at 4°C with rotation. Agarose was spun down for 1 min at 1000 x g at 4°C, supernatant was removed, and pellet was was hed 3X (1X binding buffer (10mM Tris-HCl pH 7.5,150mM NaCl), 2X wash buffer (10mM Tris-HCl pH7.5, 500mM NaCl)). Remaining agarose pellet was eluted for western analysis in loading buffer. For western analysis, larval brains were dissected and dissociated as stated above, and were lysed in 0.1% CHAPS buffer [[50mM Nacl, 200mM HEPES, 1mM EDTA and protease inhibitor cocktail (Roche)] Loading input was adjusted for brain size and protein concentration. Primary antibodies include rabbit anti-GFP (1:2500, Invitrogen Cat# A11122), rabbit anti-Ball (1:1000) (Herzig et al., 2014), rat anti-L(2)gl (Peng et al., 2000), rabbit anti-aPKC c-20 (1:1000, Santa Cruz, discontinued), rabbit anti-aPKC phosphoT410 (1:1000, Santa Cruz, discontinued), and mouse anti-Actin-c4 (1:5000, Millipore Cat# MAB1501). Secondary antibodies include Rockland DyLight 600 and 800 (1:1000), BioRad Star Bright Blue 700 (1:1000) and Jackson ImmunoResearch HRP conjugated (1:5000). Blots were imaged on a Bio-Rad ChemiDocMP.

Food ingestion assay

Ankle²A/Y or Ankle2^A/+ third instar lavae from the same vial were washed in PBS and placed on yeast paste containing 0.08% Brilliant Blue R dye in 10mm petri dishes for 20 minutes. Larvae were transferred, rinsed, boiled for 10 sec, and aligned on a glass slide ventral side up. Larvae from both genotypes were imaged at the same time using a Biorad ChemiDoc MP Imaging system using the 715/30 (far red/680) filter. Total intensity of blue food in the gut was measured using Biorad’s ImageLab software and plotted as total intensity per lavae.

Human cell immunohistochemistry

Cells were detached using Trypsin-EDTA 0.05% and plated onto 18mm glass cover slips in 6 well plates. Cells were cultured for an additional 3 days under the same conditions before fixing and staining. Cells were rinsed with PBS followed by fixing in 4% paraformaldehyde in PBS. Cells were rinsed and washed 3x in PBST, washed 2x in PBST + 1% BSA (PBSTB), and then blocked in PBSTB + 5% normal goat serum. They were then incubated in PBSTB and primary antibody overnight at 4 degrees Celsius. Cells were then washed 3x in PBSTB, incubated in anti-rabbit Cy5 secondary antibody (1:500) for 2 hours, and washed 3x in PBST. The cells were given a final wash in PBST with DAPI (1:1000) for 30m before mounting using SlowFade glow on glass slides and sealing with nail polish.

Exome and Sanger Sequencing

Exome sequencing was performed under the Baylor Hopkins Center for Mendelian Genomics (BHCMG) research initiative as previously described (Lupski et al., 2013). Exome capture was performed with Nimblegen reagents and a custom capture reagent, VCRome2.1. Raw data was processed using the Mercury pipeline, available on DNANexus (http://blog.dnanexus.com/2013-10-22-run-mercury-variant-calling-pipeline/) (Reid et al., 2014) and the ATLAS2 method was used for variant calling followed by an in-house Cassandra annotation pipeline based on Annotation of Genetic Variants (ANNOVAR). The LLGL1 variant was orthogonally validated and segregated with disease by dideoxy Sanger sequencing of PCR amplicons (Sanger et al., 1977).

QUANTIFICATION AND STATISTICAL ANALYSIS

Brain volume

Brains from third instar larvae were stained and mounted with tape spacers and imaged using a Leica Sp8 with 2 μm or 3 μm sections through the entire brain lobe. Resulting stacks were analyzed using the Surfaces function in Imaris (Bitplane) to quantify brain lobe volume as total microns cubed. One lobe from each brain was imaged and a total of 5–10 brains were analyzed per genotype or condition. Brain lobe volumes are displayed as box plots with hinges representing the 25^th to 75^th percentiles, a line represents the median, and whiskers represent min to max. Statistical significance was determined using one-way ANOVA with multiple comparisons post-test calculated using GraphPad Prism. Brain volumes in Figure 1 are normalized to wild type (FRT19a). Average volume from wild type is set to 100%, and each mutant or condition is normalized as percentage of wild type volume. Brain volumes from Figure 4–5 are displayed as total brain volume (μm³).

Asymmetric phenotypes

3^rd instar larvae were immunostained for pH3, Baz, Par-6, aPKC, or Mira as described above. Metaphase neuroblasts (pH3 positive, chromosomes aligned at the metaphase plate) were imaged on a Leica Sp8 (63X). Only metaphase neuroblasts in the correct plane for imaging were analyzed. Mild disruption refers to weak or incomplete crescent localization, and strong disruption indicates no crescent localization. To quantify spindle orientation, CNN was used to mark the plane of division, and aPKC, and Mira were used to establish cortex polarity. Only metaphase neuroblasts in the correct plane for imaging were analyzed. The angle between spindle orientation and cortex polarity was measured using the angle function of ImageJ. Phenotypes are portrayed as percentage of total counted metaphase neuroblasts. For all samples, n is noted in the figure.

VRK1 intensity

Human fibroblasts were stained as described above, imaged on a Zeiss 710 as Z-stacks with equivalent laser power and confocal settings in the same imaging session. Resulting images were analyzed in Imaris (Bitplane) using the Surfaces function to mark nuclear volume. Total intensity sum of the VRK1 channel within the nucleus and nuclear volume were recorded. VRK1 intensity is displayed as intensity sum normalized to volume. One-way ANOVA with multi-comparisons post-test from GraphPad Prism was used to assess significance. Each dot represents one nucleus. Three fields from each cell line were assessed.

DATA AND CODE AVAILABILITY

The dbGaP accession number for all exome sequences reported in this paper and for which informed consent for data sharing in controlled-access databases has been provided under the Baylor College of Medicine Institutional Review Board Protocol Number H-29697 is dbGaP: phs000711.v5.p1. BAB701 was consented under the Baylor College of Medicine Institutional Review Board Protocol Number H-9170 and subjects enrolled through Baylor Genetics Laboratories were consented under the Baylor College of Medicine Institutional Review Board Protocol Number H-41191 which do not allow for data sharing through public repository.

Supplementary Material

2

Movie S1: Ankle2-GFPR localizes to the nuclear envelope during division. Related to Figure 2. Time lapse images of live 3^rd instar larval brains from Ankle2-GFPR animals showing subcellular localization of Ankle2 protein (green). Large cells are neuroblasts. Ankle2 is cytoplasmic during interphase but is recruited to the nuclear envelope during cell division.

3

Movie S2: Miranda localizes to the basal membrane in wild type dividing neuroblasts. Related to Figure 3. Time lapse imaging of live 3^rd instar larval brains from wild type animals expressing Miranda-GFP (green) and Histone-RFP (DNA, white). Large cells are neuroblasts. Note the localization of Miranda at the basal membrane during mitosis.

4

Movie S3: Ankle2^A mutant neuroblasts fail to segregate DNA. Related to Figure 3. Time lapse imaging of live 3^rd instar larval brains from Ankle2^A animals expressing Miranda-GFP (green) and Histone-RFP (DNA, white). Large cells are neuroblasts. Note reduced localization of Miranda at the basal membrane and DNA segregation failures during mitosis.

5

Movie S4: Ankle2^A mutant neuroblasts fail to undergo cytokinesis. Related to Figure 3. Time lapse imaging of live 3^rd instar larval brains from Ankle2^A animals expressing Miranda-GFP (green). Large cells are neuroblasts. Note reduced localization of Miranda at the basal membrane and cytokinesis failure during mitosis.

6

Movie S5: Ankle2^A mutant neuroblasts have defective Miranda localization. Related to Figure 3. Time lapse imaging of live 3^rd instar larval brains from Ankle2^A animals expressing Miranda-GFP (green) and Histone-RFP (DNA, white). Large cells are neuroblasts. Note reduced localization of Miranda at the basal membrane.

7

Movie S6: Ballchen localizes to the nucleus during interphase and redistributes throughout the entire cell during mitosis. Related to Figure 4. Time lapse imaging of live 3^rd instar larval brains from wild type animals expressing Ballchen-GFP (green). Large cells are neuroblasts. Note the nuclear localization of Ball during interphase, redistribution during mitosis, and recruitment back to the nucleus upon completion of mitosis.

Highlights.

Ankle2 interacts with Ballchen/VRK1 to regulate brain development in Drosophila.

Ankle2/VRK1 pathway affects asymmetric protein localization during neuroblast division.

ANKLE2 and VRK1 are associated with human microcephaly.

The Zika virus protein NS4A interacts with and inhibits the ANKLE2 pathway.