Abstract
Heterozygosity for the TOR1A-Δgag mutation causes semi-penetrant childhood-onset dystonia (OMIM #128100). More recently, homozygous TOR1A mutations were shown to cause severe neurological dysfunction in infants. However, there is little known about the recessive cases, including whether existing reports define the full spectrum of recessive TOR1A disease. Here we describe abnormal brain morphogenesis in ∼30% of Tor1a−/− mouse embryos while, in contrast, this is not found in Tor1aΔgag/Δgag mice. The abnormal Tor1a−/− brains contain excess neural tissue, as well as proliferative zone cytoarchitectural defects related to radial glial cell polarity and cytoskeletal organization. In cultured cells torsinA effects the linker of nucleoskeleton and cytoskeleton (LINC) complex that couples the nucleus and cytoskeleton. Here we identify that torsinA loss elevates LINC complex levels in the proliferative zone, and that genetic reduction of LINC complexes prevents abnormal brain morphogenesis in Tor1a−/− embryos. These data show that Tor1a affects radial glial cells via a LINC complex mediated mechanism. They also predict human TOR1A disease will include incompletely penetrant defects in embryonic brain morphogenesis in cases where mutations ablate TOR1A function.
Introduction
TOR1A mutations cause dominant and recessive neurological disorders. The autosomal dominant disease is relatively well characterized and takes the form of a ∼30% penetrant childhood-onset form of dystonia that arises from a structurally normal brain (DYT1/DYT-TOR1A dystonia; OMIM #128100) (1). It is most commonly caused by heterozygosity for an in-frame three-base pair deletion that deletes a glutamic acid residue from the C-terminus of the torsinA protein (TOR1A-Δgag). More recently, recessive TOR1A disease has been described (2,3). All recessive TOR1A cases to date present with severe arthrogryposis at birth, and later developmental delay including intellectual disabilities. However, there is currently little information on the nature of the underlying neuropathology of any TOR1A disease.
The concept that homozygous TOR1A mutations would cause severe neurological disease was already predicted by Tor1a mutant mice. Mice that are homozygous for the Tor1a-Δgag mutation die within 48 h of birth under normal rearing conditions. They have structurally normal brains, but contain many neurons with abnormal internal membranes (4). A similar phenotype occurs when Tor1a is ‘knocked-out’ by a multi-exon deletion, thus establishing that Tor1a-Δgag is a strong loss-of-function allele in mice (4). This has been confirmed by structural and biochemical studies (5,6), although further comparison of the two mouse lines also revealed that Tor1a-Δgag consistently causes slightly less severe phenotypes than the full ‘knock-out’ allele (7).
To date, the children that survive with recessive TOR1A disease are homozygous or trans-heterozygous for TOR1A-Δgag or a G318S mutation that seems to similarly impact the C-terminus of torsinA (2,3). Consequently, it is likely that these patients retain a small amount of TOR1A/torsinA activity. In contrast, an infant failed to survive when homozygous for a TOR1A frameshift mutation that will delete a large section of torsinA (8). The case report of this infant described structural defects in brain imaging. This suggests that the clinical presentation of TOR1A disease may vary depending on how severely a mutation inhibits TOR1A activity (8).
The existence of Tor1a mouse models provides an opportunity to relatively rapidly learn more about the newly recognized TOR1A recessive disease. This includes better defining the spectrum of developmental defects, including the possibility of partially penetrant components as is well established for dominant TOR1A dystonia. Furthermore, mouse models allow work that investigates which neurodevelopmental stages and cell types are affected by Tor1a mutations, as well as the molecular pathology that link these to torsinA loss. In particular, while molecular studies have identified a number of torsinA interacting partners, there is little information about which are relevant for brain development (4,9–16). This includes questions about the neurodevelopmental relevance of several reports suggesting that torsinA regulates linker of nucleoskeleton and cytoskeleton (LINC) complex that couples the nucleus and cytoskeleton (11,14,16). This complex is comprised of two elements—Sun-proteins and Nesprin proteins—that have well-defined importance during neurodevelopment (17). However, to date, there is little overlap between the phenotypes of animals that lack torsin proteins versus those that lack LINC complexes (4,9,17–21), and a lack of clarity on the direction of the torsinA-LINC complex relationship given that torsinA is variably shown to increase, decrease or alter the location of LINC complexes in cultured cells (11,14,16).
Here, we identify morphologically abnormal brain development in ∼30% of Tor1a knock-out mice that are a genetic model for forms of recessive TOR1A disease caused by mutations that ablate gene function. The proliferative zones in the developing brains of these animals contain multiple defects associated with radial glial neuroprogenitor cells, including mislocalized mitosis and disorganized polarity markers. The defects in brain structure are also first observed in mid-embryogenesis, consistent with a radial glial cell origin. Furthermore, by later stages, the brains of abnormally developing Tor1a knock-out embryos contain almost 2-fold excess neuronal tissue. We also define excess LINC complexes as a molecular defect in the proliferative zone of abnormal Tor1a embryos, and show genetically that this contributes to this early neurodevelopmental pathology. This significantly extends in vitro work by defining the neurodevelopmentally relevant nature of the torsinA–LINC complex interaction. Considered together, these data predict that the full spectrum of TOR1A recessive disease will include a partially penetrant pathology where structural brain defects arise downstream of radial glial cell dysfunction.
Results
TorsinA loss causes partially penetrant neurodevelopmental defects that appear soon after neural tube closure
The finding that TOR1A loss-of-function mutations cause recessive congenital disease highlights the importance of defining how torsinA loss affects early neurodevelopment. We previously described that constitutive deletion of mouse Tor1a leads to early post-natal lethality, but that these pups have a normally appearing central nervous system (CNS) (4). More recently, conditional deletion of Tor1a starting at ∼ embryonic day (E) 11.5 had a less severe impact than constitutive deletion (7), suggesting that torsinA may have an undetected importance earlier in development. Indeed, torsinA is expressed in young as well as more mature neural tissues (4) (Supplementary Material, Fig. S1A and B).
We examined litters of embryos aged between E11.5 and E16.5 derived from crosses of Tor1a+/− mice bred more than 10 generations onto the C57Bl6 background. This identified that ∼30% of Tor1a−/− embryos had macroscopic defects in brain morphology (Fig. 1A). At E14.5, we most commonly saw protruding/exposed brain tissue outside the skull, which is the defect of exencephaly (Fig. 1B;Supplementary Material, Movies S1 and S2). In addition, we also saw some Tor1a−/− embryos with incomplete protrusion so that only one hemisphere was exposed (Fig. 1B, right panel). Histological analysis of coronal sections further confirmed that these Tor1a−/− embryos had grossly abnormal CNS morphology (Fig. 1C). Indeed, brain tissue from animals with fully or partially expressive exencephaly was disorganized to the point that we could rarely identify specific brain structures even after aligning sections by peripheral landmarks (Fig. 1C; Ey: eye). We also examined whether similar morphological defects occurred in mice with the Tor1a-Δgag allele by collecting E14.5 embryos from 13 litters from Tor1a+/Δgag parents that had also been crossed >10 generations onto the C57Bl6 background. This found that brain development appeared normal in all offspring, as well as that all genotypes were present at normal Mendelian ratios as previously described (4) (Fig. 1D). Thus, these data establish that Tor1a deletion causes partially penetrant defects in brain morphogenesis that are absent from mice homozygous for the Tor1a-Δgag mutation.
Abnormal brain morphology in 30% of Tor1a−/− embryos. (A) Frequency that a macroscopically abnormal brain is detected when embryos are examined under a widefield dissecting microscope. (B) Brightfield images show the head morphology of E14.5 embryos. Center panel: the cranial region of a Tor1a−/− embryo lacks obvious skin or skull, and instead neural-like material is exposed. Right panel: A Tor1a−/− embryo where only one hemisphere is disrupted. (C) Hematoxylin and eosin (H&E) stained coronal sections through the forebrain of the E14.5 control and Tor1a−/− embryos shown in (B). The plane of sections is indicated in (B) with a yellow line. Ey: eye. (D) No macroscopic brain defects are detected in Tor1aΔgag/Δgag embryos. Litters of E14.5 embryos were examined using a widefield dissecting microscope by an observer blind to genotype.
Exencephaly most commonly develops downstream of failed neural tube closure. In such cases, skull and skin are absent because they cannot develop above an open neural tube. Neural tube closure takes place between E8.5 and E10 in mice, and we therefore examined littermate control and Tor1a−/− embryos over this age range. This revealed normal Tor1a−/− embryo size (Fig. 2A and C) and somite numbers through early development (Supplementary Material, Fig. S1C). In addition, histology of E9.5 and E10.5 embryos detected a normally appearing neuroepithelium including that the neural tube was closed (Fig. 2B and D, arrowheads indicate points of neural tube closure). In contrast, at E11.5 there were clear macroscopic defects associated with the Tor1a−/− genotype (Fig. 2E and E’). These animals again had a neuroepithelium where the dorsal points of the neural tube were closed (Fig. 2F, arrow heads indicate points of neural tube closure). However, the neuroepithelium also appeared expanded and convoluted in these E11.5 Tor1a−/− embryos, while the ventricular volume appeared reduced (Fig. 2F). Thus, we conclude that the first morphological defects in Tor1a−/− mice appear at ∼E11.5, while earlier development appears normal. Thus, the exposed brain tissue that is present in later stage Tor1a−/− embryos is not explained by failed neural tube closure and an alternative mechanism must exist.
Morphological defects in Tor1a−/− embryos appear after neural tube closure. Tor1a−/− embryos appear normal during early development. Brightfield views of littermate control and Tor1a−/− embryos (from the side and from above) at E9.5 (A) and E10.5 (C). H&E stained horizontal sections through littermate control and Tor1a−/− E9.5 (B) and E10.5 (D) embryos. Arrowheads indicate points of neural tube closure. The plane of sections for each stage are indicated in (A) and (C) with a yellow line. (E) Brightfield views of the head of littermate E11.5 embryos. A morphological defect is clearly seen in the Tor1a−/− animal. (E’) Full view of the embryos shown in (E). (F) H&E stained horizontal sections through littermate control and Tor1a−/− embryos. Arrowheads indicate points of neural tube closure. The plane of sections is indicated in (E’) with a yellow line. N: neuroepithelium; V: ventricle; Ey: eye; G: trigeminal ganglion.
Radial glial cells are abnormal in Tor1a−/− embryos
At E11.5 the neuroepithelium that lines the ventricles is primarily comprised of proliferating neural progenitor cells. Neural progenitor cells undergo self-renewing or neurogenic divisions that amplify the neural progenitor pool or lead to committed neurons, respectively. E11.5 is also the time point when neurogenic divisions first begin in neuroepithelial regions that give rise to the brain, which have experienced little neurogenesis up to this time point. Thus, the fact that Tor1a-associated CNS defects are first detected at E11.5 points to abnormalities of a neural progenitor origin.
While there are several types of neural progenitor cell, nestin+/Pax6+ radial glia are the predominant type in the E11.5 neuroepithelium. We found that both markers were present in the neuroepithelium of all control and Tor1a−/− mice (Fig. 3A and B). Quantitation of the area occupied by Pax6+ cells also found that this was larger in abnormally developing Tor1a−/− mice compared to controls (Fig. 3C). Thus, we conclude that Tor1a−/− mice produce radial glia neural progenitor cells, and that these may be more abundant than in control animals.
Radial glia neuroprogenitor cells in the E11.5 neuroepithelium of control and Tor1a−/− mice. Images of horizontal sections through E11.5 control and Tor1a−/− (A) rostral CNS (telencephalon and diencephalon) (B) and caudal CNS (hindbrain) labeled with anti-Pax6 (red) and anti-nestin (green) markers of radial glial cells. DAPI shows nuclei (blue). (C) Quantitation finds a larger Pax6+ area in Tor1a−/− embryos compared to littermate controls. Bars show the average of the total rostral area that is Pax6+ in n =4 littermate pairs of embryos. *P<0.05, paired T-test.
Radial glial cells are elongated and highly polarized. They are densely packed in the proliferative zones where they display characteristic nuclear movements that are synchronized with the cell cycle: nuclei undergo mitosis in apical positions and S-phase in basal positions, and migrate between these positions in G1 and G2, respectively. In turn, these nuclei have an elongated shape that aligns perpendicularly to the apico-basal axis of the proliferative zone (Fig. 4A) (22–24). The Pax6 labeling highlighted this typical nuclear shape in control animals (Fig. 4B–D; dotted lines) but, in contrast, Pax6+ nuclei were less elongated and less orientated to the apico-basal axis in abnormally developing E11.5 Tor1a−/− embryos (Fig. 4B–D; solid lines). This points to a fundamental abnormality in radial glial cell behavior. We therefore further examined whether there were signs of suppressed apoptosis or elevated proliferation to explain why the Pax6+ zone was larger in E11.5 Tor1a−/− embryos. We similarly detected a few anti-cleaved caspase3 (cl-caspase3) labeled apoptotic cells in the neuroepithelium of control and Tor1a−/− embryos (Fig. 4E), and quantification of their number also failed to suggest a deficit in apoptosis (Fig. 4F). Next we examined the number of mitotic cells, which found more pHH3+ mitotic nuclei in the abnormal E11.5 Tor1a−/− proliferative zone compared to controls (Fig. 4G and H). In addition, these mitotic nuclei were more frequently localized outside the apical region: in controls ∼80% of mitotic nuclei were located immediately next to the ventricle, while this dropped to under 60% in the abnormal Tor1a−/− mice (Fig. 4G’ and I). We further examined whether this reflects a radial glial cell defect, or the abnormal presence of basal neuroprogenitor cells (Tbr2+) that inherently undergo mitosis outside the apical zone. However, we saw the abnormally localized pHH3+ nuclei were negative for Tbr2, even though we clearly detected these basal neuroprogenitor cells in other neuroepithelium regions of control and Tor1a−/− mice (Supplementary Material, Fig. S2A and B). Thus, while the proliferative zones of E11.5 Tor1a−/− embryos are appropriately comprised of different neuroprogenitor cells, our data shows that they are abnormally large, and contain cytoarchitectural defects including mislocalized and elevated numbers of mitotic nuclei.
Cellular abnormalities in the E11.5 Tor1a−/− proliferative zone. (A) Cartoon depicting radial glial cell organization in the proliferative zone. These are the predominant cell type in the embryonic proliferative zone where they assemble into a pseudostratified epithelium. Nuclei are positioned at different apico-basal heights in this epithelium and move in concert with the cell cycle: mitosis occurs when nuclei are in apical positions, while S-phase occurs when nuclei are basally located (this movement is depicted by the blue nucleus and the dark blue arrows). The apico-basal polarity and integrity of the proliferative zone depends on apically localized cytoskeletal elements and cell adhesions (red), including centrosomes (green) that are invariably located in the far apical region of the cell even as nuclei alter position. (B–D) Radial glial cell nuclei have abnormal characteristics in the E11.5 Tor1a−/− proliferative zone. (B) Anti-Pax6 and anti-nestin labeling in the rostral telencephalon (left panels) and hindbrain (right panels) of littermate control and Tor1a−/− embryos. Nuclei appear differently packed and shaped in the Tor1a−/− compared to controls. (C) Plot showing the frequency of different Pax6+ nuclear shapes. ‘Nuclear Elongation Ratio’ refers to the ratio between the long and short axis: 1 indicates that these are equal, while higher values reflect elliptical shapes (schematic in blue). Control mean=1.96, Tor1a−/− mean=1.58. Values come from the hindbrain of n=3 littermate animals of each genotype and >700 nuclei per animal. (D) Plot showing the degree to which elongated Pax6+ nuclei are aligned to the apico-basal coordinates of the proliferative zone. A 90° angle reflects that the long axis of a nucleus is parallel to the proliferative zone axis (thus perpendicular to the ventricular surface). Values come from n=3 littermate animals of each genotype and >700 nuclei per animal. (E and F) Apoptosis appears relatively normal in the hindbrain neuroepithelium of Tor1a−/− embryos. (E) Horizontal sections through the hindbrain of littermate E11.5 control and Tor1a−/− embryos where all nuclei are labeled by DAPI (blue) and apoptotic nuclei are detected by anti-cleaved (cl) caspase (green). Insets show the presence of some apoptotic cells. (F) Average number of apoptotic nuclei (per mm2 of hindbrain neuroepithelium) in littermate control and Tor1a−/− embryos. n.s.: not significant, paired T-test, n=3. (G-I) The nestin+ hindbrain proliferative zone of E11.5 Tor1a−/− embryos contains mislocalized mitotic nuclei and increased numbers of mitotic nuclei. (G) Horizontal sections of E11.5 hindbrain labeled with anti-nestin and the mitotic marker, phospho-histone H3 (pHH3). (G’) Magnification of the insets shown in (G). (H) The number of pHH3+ nuclei per 100 μm length of E11.5 hindbrain proliferative zone. Data show the average from n=3 littermate pairs of embryos. *P<0.05, paired T-test. (I) Percentage of apical pHH3+ nuclei (in terms of total pHH3+ nuclei) in the nestin+ hindbrain neuroepithelium of n=3 littermate pairs of embryos. A nucleus was considered as ‘apical’ when it was located ≤5μm from the ventricular surface. *P<0.05, paired T-test.
Excess neuronal production in Tor1a−/− embryos
Affected Tor1a−/− brains were severely disorganized by E14.5 (Fig. 1B and C) that is also the stage when neurogenesis reaches its peak in the forebrain proliferative zones that give rise to the cortex. We investigated the distribution and organization of neuronal and neural progenitor cells at this developmental stage. We again used anti-Pax6 to detect radial glial cells (red) in coronal forebrain sections matched to peripheral landmarks (Fig. 5A; Ey: eye; OE: olfactory epithelium). This clearly identified a Pax6+ proliferative zone region in all animals, although it was displaced downwards and laterally in the Tor1a−/− mutants (Fig. 5A). Given their abnormal location, we also confirmed that these Pax6+ zones were associated with the developing cortex. This indeed demonstrated a similar displacement of cells positive for the Tbr1 marker of cortical neurons (Fig. 5B). In addition, like we saw at E11.5, quantification identified that the Pax6+ region was larger in affected Tor1a−/− brains than littermate controls (Fig. 5C).
Excess neuronal tissue in E14.5 Tor1a−/− brains. (A) Images of E14.5 coronal sections aligned on the rostro-caudal axis via the position of the developing eye (Ey) and olfactory epithelium (OE). These sections are from littermate embryos. Red shows anti-Pax6 signal (radial glial cell marker) and green shows the neuronal marker, βIII-tubulin. (B) Coronal E14.5 sections where the developing cortex is detected with the cortical specific anti-Tbr1 marker (magenta), and co-labeled with βIII-tubulin (green). (C) Area that is Pax6+ in coronal sections from E14.5 littermate control and Tor1a−/− embryos, matched by landmarks as shown in (A). Data show values from n=3 littermate pairs. *P<0.05, paired T-test. (D) Area that is positive for anti-βIII-tubulin, as described above for anti-Pax6. Data show values from n =3 littermate pairs. **P<0.01, paired T-test.
We then assessed the amount of neurogenesis that had occurred in abnormal Tor1a−/− brains by labeling committed neuronal cells with anti-βIII tubulin (green). This again confirmed that neurons are produced in the abnormally developing Tor1a−/− brain. Indeed, the dorsally protruding tissue was strongly positive for this neuronal marker (Fig. 5A). Moreover, quantification of the total βIII-tubulin+ area revealed that abnormal Tor1a−/− brains had almost double the amount of neuronal tissue compared to littermate controls (Fig. 5D). We then again examined whether excess tissue was associated with lower levels of apoptosis, and thus might reflect insufficient death of newly born neurons. However, we detected minimal apoptosis in βIII-tubulin+ areas of control or abnormal Tor1a−/− brains (Supplementary Material, Fig. S3A). Thus, in addition to proliferative zone abnormalities, we find that later stage Tor1a−/− animals have excess neural tissue. This further implicates abnormal radial glial cell proliferation as the reason why brain morphogenesis is abnormal in 30% of Tor1a−/− embryos.
Tor1a deletion causes breakdown of proliferative zone cytoarchitecture
Radial glial cell proliferation is typically studied at the E14.5 stage. Interestingly, closer examination of Pax6+ cells at this time point detected that the Tor1a−/− proliferative zone was severely abnormal. The first indication of this was that we detected regions where the normal packing of Pax6+ cells was disrupted and cells appeared to surround a ‘micro’ lumen rather than orient to the apico-basal axis of the proliferative zone (Fig. 6A). These areas were reminiscent of the rosette-like structures that occur in neural tumors when proliferative cells lose polarity with respect to the apico-basal axis of the tissue, and are a hallmark of excess neural progenitor proliferation (25). We therefore examined the polarity of radial glial cells located in these abnormal clusters via staining for F-actin that strongly concentrates in the apical zone where it stabilizes adherens junctions. This confirmed that the apical domains of these radial glial cells no longer corresponded with the apico-basal axis of the proliferative zone (Fig. 6A).
Multiple cytoarchitectural defects in the forebrain proliferative zone of E14.5 Tor1a−/− embryos. (A) Abnormal rosette-like organization of Pax6+ nuclei (red) in an E14.5 Tor1a−/− proliferative zone, like that shown in Figure 5A. These structures were absent from control animals, where instead nuclear packing appears to respect the apico-basal polarity of the proliferative zone. Phalloidin staining (green) detects actin filaments. (B) Images of E14.5 control and Tor1a−/− proliferative zone co-labeled with the anti-γ-tubulin centrosome marker (green) and anti-Pax6 (red). Arrowheads point to centrosomes outside the apical region. (C) The average number of non-apical centrosomes in control and Tor1a−/− proliferative zones (per 100 µm of proliferative zone length) quantified from images like that shown in (B). Bars show the mean ± SEM from n=3 littermate pairs of embryos. *P<0.05, ratio paired T-test. (D) Apical F-actin and adherens junctions are disrupted in Tor1a−/− embryos. Images show regions of the E14.5 forebrain proliferative zone labeled with Pax6 (or DAPI) together with phalloidin to detect F-actin (top), or the adherens junction markers N-cadherin (center) and β-catenin (lower panel). Arrowheads highlight discontinuities in the band of signal that normally covers the entire ventricular surface of the proliferative zone, as well as indicating patches where these apical-domain proteins are mislocalized. (E) The percentage of the E14.5 proliferative zone that lacks an apical band of adherens junctions (AJ). Bars show the mean ± SEM from n= 3 littermate pairs of embryos. **P <0.01, T-test. (F) Confocal images of EdU+S-phase nuclei in the E14.5 proliferative zone of control and Tor1a−/− littermates. Dotted lines indicate the apical surface, defined by DAPI staining. (G) Plot showing the localization of S-phase nuclei with respect to the apical surface of the proliferative zone of E14.5 control and Tor1a−/− embryos. Measurements were made from n =3 littermate embryos of each genotype and >500 nuclei per embryo.
We therefore further analyzed whether Tor1a loss disturbed the polarity of radial glial cells. The centrosome of radial glial cells is normally permanently localized in their apical region at the surface of the ventricle (Fig. 4A). We detected centrosomes via labeling for the γ-tubulin marker. This found that the majority of centrosomal-signal in the Pax6+ zone was localized to a specific apical band in both control and abnormal Tor1a−/− embryos (Fig. 6B). However, we also detected more centrosomes outside the apical area in the abnormal Tor1a−/− proliferative zones compared to controls (Fig. 6C). Since this is further consistent with abnormal radial glial polarity, we next examined the localization of adherens junctions. These are central to proliferative zone tissue integrity, and underlie how radial glial cells detect apico-basal coordinates. We assessed three proteins associated with apical radial glial cell adherens junctions: F-actin, N-cadherin and β-catenin. As expected in controls, all three were continuously and strongly localized in a band that represents the apical domain of the E14.5 Pax6+ proliferative zone (Fig. 6D). However, we frequently detected abnormal adherens junctions in the Tor1a−/− proliferative zone. Firstly, while a ‘band’ of adherens junctions was identifiable, this did not extend along the entire length of the Pax6+ zone of abnormal Tor1a−/− embryos (Fig. 6D). We compared the total length of the Pax6+ zones that lacked a concentrated band of β-catenin or N-cadherin signal, and found that this was indeed elevated from <3% in controls to ∼15% in abnormally developing Tor1a−/− embryos (Fig. 6E). In addition, we also saw proliferative zone areas where adherens junction proteins accumulated in non-apical patches (Fig. 6D). Thus, there are multiple defects in the localization of radial glial polarity markers within the proliferative zone of abnormal Tor1a−/− embryos.
The fundamental characteristics of radial glial cells, like nuclear migration and self-renewing versus neurogenic cell divisions, are deeply intertwined with their apico-basal polarity. We therefore further examined whether radial glial cells continued to execute events related to apico-basal polarity, also noting that the presence of non-apical mitotic nuclei already suggests that this is impaired (Fig. 4G’ and I). To this aim, we examined the position where nuclei undergo S-phase by administering the thymidine-analog 5-ethynyl-2'-deoxyuridine (EdU) to pregnant mice and collecting embryos 30 min later. In controls, S-phase nuclei were located, as expected, in basal regions of the E14.5 proliferative zone (Fig. 6F). Quantification confirmed this, including that almost no nuclei (0.7%) were detected within 50 μm of the apical surface, and the majority (69%) were more than 100 μm distant (Fig. 6G). In contrast, S-phase nuclei appeared across all apico-basal levels of the abnormal Tor1a−/− proliferative zone (Fig. 6F). Again, this was confirmed by quantification, which revealed an apical shift in the overall distribution and many (33%) S-phase nuclei located close (<50 μm) to the apical surface of the proliferative zone (Fig. 6G). Thus, the abnormal Tor1a−/− proliferative zone contains cells that fail to perform a radial glial cell behavior that depends on apico-basal polarity, and this occurs alongside multiple signs of mislocalized and absent radial glial polarity markers. These data strongly implicate radial glial dysfunction in why Tor1a−/− embryos develop morphologically abnormal brains.
Excess nuclear envelope LINC complexes in the Tor1a−/− proliferative zone
We next explored how Tor1a/torsinA loss causes multiple defects in radial glial cell organization and behavior. TorsinA is a protein that resides in the lumen of the endoplasmic reticulum/nuclear envelope. Thus, torsinA cannot directly interact with the abnormally localized cellular components like actin, adherens junctions and centrosomes. However, torsinA in cultured cells is shown to affect levels and/or localization of the LINC complex (11,14,16) which, via its role coupling the nucleus to cytoskeletal networks (26,27), could potentially mediate changes in cytoskeletal and cytoskeletal-associated proteins.
The inner nuclear membrane Sun1 and Sun2 proteins anchor the LINC complex in the nuclear envelope of most cell types, including developing neurons (17). We validated that Sun-protein antibodies detect their antigens (Supplementary Material, Fig. S4A–E) and then examined their expression through neurodevelopment. This revealed that Sun2 is highest in the neurodevelopmental stages under study here, but sharply declines over embryonic development to where it is barely detectable in post-natal mice. In contrast, Sun1 is much more highly expressed in the brains of post-natal mice compared to embryos, and is not clearly detected at the developmental stages under study here (Supplementary Material, Fig. S5A). We then examined total levels of Sun1 and Sun2 in the brains of E14.5 control and affected Tor1a−/− embryos, which found no noticeable difference in anti-Sun-protein immunoreactivity (Supplementary Material, Fig. S5B and C).
Sun-proteins form LINC complexes when they localize in the nuclear envelope. However, the proliferative zone of control animals had little nuclear-associated anti-Sun1 or anti-Sun2 signal (Fig. 7A and C;Supplementary Material, Fig. S6A, C and D). This suggests that the nuclear envelope of radial glial cells contains relatively few LINC complexes. We further validated this finding by confirming that we strongly detected Sun1 and Sun2 around the nuclei of non-neural cells (in the same tissue sections; Supplementary Material, Fig. S6B and E), and seeing nuclear envelope anti-Sun2 signal associated with nuclei in the βIII-tubulin+ zone that contains committed neurons (Supplementary Material, Fig. S6D’). Furthermore, the lack of nuclear envelope Sun-protein signal was specific to the control condition. In contrast, Sun1 and Sun2 were both clearly present around nuclei in the proliferative zone of abnormally developing Tor1a−/− mice (Fig. 7A and C). The degree to which we detected this nuclear envelope localized Sun1 and Sun2 varied, but overall a significantly greater area of the abnormal Tor1a−/− proliferative zone contained cells where we clearly detected nuclear envelope Sun1 and Sun2 when compared to controls (Fig. 7B and D). We then also assessed nuclear envelope Sun1 and Sun2 in the proliferative zone of ‘asymptomatic’ Tor1a−/− mice that had morphologically normal brain development. This again detected regions of the proliferative zone where nuclei had Sun1 or Sun2 labeling, and quantification suggested that the extent of these regions was intermediate between controls and abnormally developing Tor1a−/− animals (Fig. 7B and D).
Nuclear envelope LINC complexes are detected in the Tor1a−/− proliferative zone. Confocal images of (A) Sun1, (C) Sun2 and (E) Nesprin2G immunolabeling in the forebrain proliferative zone of E14.5 control (left) and Tor1a−/− embryos with CNS abnormalities (right), together with DAPI to show nuclei. Note we confirmed that the majority of anti-Sun1 and anti-Sun2 immunoreactivity reflects Sun1 and Sun2 proteins (Supplementary Material, Fig. S4), although this also revealed the non-specific nature of some of the punctate anti-Sun1 signal. Graphs show the percentage of proliferative zones (PZ) where we observed clear nuclear envelope localized Sun1 (B), Sun2 (D) or Nesprin2G (F) in control embryos, Tor1a−/− embryos that appear normally developed (Tor1a−/− non-CNS abn) and Tor1a−/− embryos with abnormal brain morphology (Tor1a−/− CNS abn). Bars represent mean ± SEM from n=5 embryos of each genotype for anti-Sun1 labeling, and n=3 embryos for anti-Sun2 and anti-Nesprin2G. *P<0.05, **P<0.01, n.s.: non-significant; One-way ANOVA followed by Tukey test for individual group comparison.
We then turned to Nesprin2, an outer nuclear membrane LINC complex component that is thought to be active during neural development (17). Using an antibody against the ‘giant’ Nesprin2 isoform that interacts with actin, we detected minimal nuclear envelope labeling in the proliferative zone of E14.5 control animals. In contrast, Nesprin2G signal was clearly localized around nuclei in the proliferative zone of abnormally developing Tor1a−/− embryos (Fig. 7E). Quantification of the number of proliferative zone regions that contained nuclear envelope Nesprin2 again found these were significantly more frequent in abnormal Tor1a−/− embryos compared to controls, and suggested that they occur at an intermediate frequency in ‘asymptomatic’ Tor1a−/− embryos (Fig. 7F). Thus, the proliferative zone of abnormal Tor1a embryos is enriched in three different nuclear envelope LINC complex proteins that are normally rarely detected in these cells. Given previous reports that torsinA regulates the LINC complex, these data suggest that the early neurodevelopmental importance of torsinA relates to a role where it negatively regulates LINC complex levels in the nuclear envelope of proliferative neuroprogenitor cells.
Sun2 deletion prevents morphological and radial glial defects in Tor1a−/− embryos
We specifically tested whether excess LINC complexes underlie the morphologically abnormal brain development of Tor1a−/− embryos. Sun2 knock-out mice are previously characterized as viable, fertile and lacking overt neurodevelopmental abnormalities (28) (unlike double Sun1/2 knock-out mice) (17), and thus represent a genetic strategy that can reduce LINC complex levels without broadly impacting brain development. We intercrossed mice carrying Tor1a and Sun2 knock-out alleles, and then scored litters for embryos with macroscopically abnormal brain morphology. These crosses again resulted in ∼30% of Tor1a−/− embryos with grossly abnormal brains when they were wild-type for Sun2 (Fig. 8A). However, this was reduced to 13% when Tor1a−/− animals carried the heterozygous Sun2+/− genotype, and 100% of Tor1a−/− mice escaped morphological brain defects when also homozygous for Sun2−/− (Fig. 8A and B). This analysis detected all genotypes at expected Mendelian ratios (Fig. 8A) to rule out the possibility that double mutant embryos were lost to earlier appearing defects and reabsorption. We also allowed some litters to develop to birth. We saw that all Tor1a−/−; Sun2−/− still had morphologically normal brains, although these animals continued to die within 48 h as previously shown for the majority of Tor1a−/− mice (data not shown) (4).
Sun2 deletion prevents morphological defects in Tor1a−/− embryos. (A) Frequency of embryos with abnormal brain morphology. Embryos generated from Tor1a+/−; Sun2+/− intercrosses were collected, scored visually for the presence of exposed brain tissue, and genotyped. All genotypes were present at the expected Mendelian ratio, but data are not presented for all heterozygous genotypes. More than 200 embryos were assessed. (B) Lateral views of E11.5 control and Tor1a−/−; Sun2−/− embryos showing apparently normal development despite the removal of both genes. (C) Apical F-actin and adherens junctions are disrupted in Tor1a−/− embryos, but normal in a double Tor1a−/−; Sun2−/− mutant. Images show E14.5 forebrain proliferative zone labeled with DAPI and phalloidin to detect F-actin (top), or the adherens junction markers N-cadherin (center) and β-catenin (lower). Arrowheads highlight discontinuities where these proteins are absent from the apical surface of the proliferative zone, or are mislocalized within the proliferative zone. (D) Confocal imaging of the proliferative zone of E14.5 control, Tor1a−/− and Tor1a−/−; Sun2−/− embryos immunolabeled with anti-γ-tubulin (green) and anti-Pax6 (red). Arrowheads highlight centrosomes located outside the apical zone. (E) The average number of non-apical centrosomes (per 100 µm of proliferative zone length) in the E14.5 forebrain proliferative zone of control, Tor1a−/− and Tor1a−/−; Sun2−/− embryos. Bars represent mean ± SEM of n=3 embryos of each genotype. *P<0.05, n.s.: non-significant; One-way ANOVA followed by Tukey test for individual group comparison. (F) The percentage of regions within the proliferative zone (PZ) that clearly contain nuclear envelope localized Nesprin2G in control embryos, abnormally developing Tor1a−/− embryos, and Tor1a−/−; Sun2−/− embryos. Bars show the mean ± SEM from n=3 embryos of each genotype. ***P<0.001; One-way ANOVA followed by Tukey test for individual group comparison.
Finally, we tested whether Sun2 deletion rescued the radial glial cell pathology of Tor1a−/− embryos. We focused on the cytoarchitectural breakdown we detected at E14.5 since this was the most severe pathology we associated with Tor1a deletion. This identified that nuclei were normally organized in the proliferative zone of all double mutant Tor1a−/−; Sun2−/− embryos (Fig. 8C), adherens junctions markers continuously lined the apical surface of the ventricle of all animals (Fig. 8C), and centrosomes displayed their normal apical localization (Fig. 8D and E). We also further associated the restored proliferative zone structure with LINC complex levels by confirming that Sun2 loss indeed reduced the amount of nuclear envelope Nesprin2 (Fig. 8F). Thus, we conclude that Sun2 deletion produces a dose-dependent reduction in the penetrance that Tor1a−/− mice display abnormal brain morphogenesis, and does so in parallel with reducing proliferative zone LINC complex levels and several radial glial cell defects.
Discussion
Homozygous Tor1a mutations were previously shown to cause severe neurological defects and lethality in mice (4,7). These data had predicted the negative clinical consequences recently described for humans carrying homozygous TOR1A mutations (2,3). Here we now describe a previously undetected, partially penetrant component to the Tor1a loss-of-function phenotype that takes the form of abnormal brain morphogenesis. We did not detect this in mice that are homozygous for the Tor1a-Δgag mutation, suggesting that brain morphogenesis is only affected when Tor1a gene activity is fully ablated. Furthermore, we identified radial glial cell pathology as the explanation for these partially penetrant brain defects, and genetically proved that abnormal levels of LINC complexes lie in the pathway between torsinA loss and their development.
There are five definitive cases of recessive TOR1A disease described in the literature (2,3). These infants are affected from birth when they invariably suffer from severe arthrogryposis despite a structurally normal brain. Their developmental progression has been followed to ∼2 years of age, at which point they have broad neurological dysfunction including intellectual impairment. Given that they are the first reported cases, it remains unclear whether they describe the full spectrum of recessive TOR1A disease. Notably, they all carry small point mutations in the 3' coding region of TOR1A that appear to strongly suppress (but not ablate) Tor1a function in mice. To further examine this issue we turned to mice that carry a large deletion in the Tor1a gene. This bona fide complete loss-of-function allele can define the most severe impacts of homozygous Tor1a mutations (4). We now identified that ∼30% of Tor1a−/− embryos develop structural brain defects that become detectable soon after neural tube closure, while these are not seen in Tor1aΔgag/Δgag embryos. This data predicts that a) mutations that ablate TOR1A function may cause a more severe clinical presentation than described in patients surviving with TOR1A homozygous point mutations (2,3) and b) the presentation may extend to grossly abnormal brain structure that appears with incomplete penetrance. Indeed, this is supported by a conference report of an infant who was homozygous for a truncating TOR1A mutation. Unlike the fully published studies, this child failed to survive beyond a few months and imaging detected structural brain defects (8). Our data therefore predict that a spectrum of recessive TOR1A disease symptoms will emerge as more cases are reported with different types of TOR1A mutation, although we also note that all homozygous Tor1a−/− and Tor1aΔgag/Δgag mice ultimately succumb to neurological dysfunction and post-natal lethality (4).
This work also identifies radial glial cell dysfunction as the explanation for abnormal Tor1a−/− brain morphogenesis, and is the second study that identifies that Tor1a−/− causes proliferative zone defects (29). Radial glial cells are the primary embryonic neuroprogenitor cell in mice (30,31). They localize in proliferative zones that line the ventricles where they undergo self-renewing (symmetric) or neurogenic (asymmetric) divisions that lead to neurons. In turn, these expand or contract the size of the radial glial cell pool: more symmetric cell divisions are associated with a larger pool of radial glial cells and ultimately more neurons and a larger brain, while a switch to neurogenic cell divisions is associated with fewer neurons and a smaller brain (32). The evidence supporting a defect in radial glial cells in Tor1a−/− mice comes from several observations, including that the proliferative zone of this mice contains mislocalized mitotic nuclei, abnormal cytoarchitecture and excess LINC complexes. Furthermore, by E14.5 Tor1a−/− mice have 2-fold excess neuronal tissue, which is something that very strongly suggests a defect in the balance of radial glial cell divisions.
We also hypothesize that excess neuronal tissue production underlies why the brain is exposed in Tor1a−/− mice. In this model, we hypothesize that the larger brain exerts pressure that exceeds the strength of the overlying layers of mesoderm and ectoderm (developing skull and skin), which break, and thus no longer contain the developing brain. Indeed, there are other genetic mouse mutants where exencephaly develops concomitant with radial glial cells defects and excess neuron production (33,34). Additional work is now needed to define how radial glial cells are affected by Tor1a loss. Our data showing that abnormally developing Tor1a−/− mice have larger Pax6+ proliferative zones with more mitosis, but apparently normal apoptosis, point to a problem where radial glial cells undergo too many self-renewing divisions. However, more work is needed to properly test this. In addition, it will be interesting to closely examine the structurally ‘normal’ brains of human TOR1A patients (dominant and recessive disease) to detect whether there are subtle signs of hyper or hypo-production of neurons, bearing in mind that the timing and modes of cell division differ between species (35).
We also genetically establish (through Sun2 deletion) that LINC complexes lie between Tor1a/torsinA loss and radial glial cell-driven brain defects. TorsinA had been previously shown to alter LINC complexes in cultured cells, although it also still remains unclear whether they are targeted directly or indirectly by torsinA activity (11,14,16). Now, by finding that Sun2 deletion rescues the early neurodevelopmental defects, our data demonstrates that LINC complex regulation is indeed a neurodevelopmentally important event downstream of torsinA activity. Furthermore, it also shows that this is a role where torsinA reduces LINC complex levels, particularly since data comes from three different LINC complex antibodies (with different epitopes) to effectively rule out the possibility of altered epitope accessibility.
It is also surprising that radial glial cells have unusually low levels of the LINC complex given that this complex is required by most cell types. While we do not solve why, the most obvious explanation is that the balance between self-renewing/symmetric versus neurogenic/asymmetric cell divisions is impeded by strong, permanent connections between the nucleus and cytoskeleton. However, other explanations are also plausible. One is that excess LINC complexes interfere with the fundamental polarity of a radial glial cell and this, in turn, impacts their cell division. The fact that we see severe disruption of proliferative zone cytoarchitecture, including mislocalized adherens junctions, supports this possibility. Intriguingly, loss of the Caenorhabditis elegans torsin, ooc-5, also impairs cell polarity in parallel with preventing asymmetric cell division in the early embryo (18). This provides a possible model for the observations here, particularly since it is well established in mice that neural overgrowth can result when radial glial polarity is disturbed (36).
Our data strongly suggest that variability in LINC complex levels underlie why only 30% of Tor1a−/− mice have morphologically abnormal brain development, as well as arguing that there is a ‘toxic’ threshold of LINC complexes that compromises radial glial cells. This then raises the question about why LINC complex levels vary between genetically homogeneous Tor1a−/− embryos. One possibility is that all animals similarly utilize torsinA, but a second inefficient mechanism also targets the LINC complex, and that this fails to sufficiently compensate for Tor1a loss in 30% of cases for stoichiometric reasons. Candidate mechanisms include other members of the torsin family (torsinB, torsin2 and torsin3) (37). Alternatively, it is well established that embryonic development varies between genetically identical mouse embryos, including those in the same litter (38). This might in turn cause variability in LINC complex levels so that only 30% of embryos require that this is removed by torsinA to allow brain morphogenesis. While speculative, it is intriguing to consider that the mechanical forces that contribute to tissue morphogenesis will vary with factors like developmental rate (39,40), and that LINC complexes have roles managing and transducing the mechanical stresses experienced by a cell (41). Indeed, if the need for torsinA varies depending on an interaction between LINC complex levels and developmental stress (mechanical or otherwise), then this might also explain why abnormal brain morphogenesis was not reported in Tor1a−/− colonies maintained under different husbandry conditions (4,29).
The fact we find a partially penetrant Tor1a neurodevelopmental phenotype has interesting parallels with the dominant form of TOR1A disease (DYT-TOR1A dystonia). This disease is also partially penetrant, and again it is 30% of individuals that become symptomatic for currently poorly defined reasons (42). We can only speculate on whether the 30% penetrance of both situations is a coincidence, or reflects that similar mechanisms determine whether an individual with the TOR1A+/Δgag genotype develops dystonia and why brain morphogenesis fails in a Tor1a−/− mouse. The possibility they are connected is supported by a previous study that presented evidence for subtle neuroprogenitor cell abnormalities in Tor1a−/− mice that had ‘normal’ brain morphogenesis (29), and it is conceivable that small defects in embryonic neurogenesis negatively impact neurological function at the developmental stage when neurons are recruited into CNS motor circuits. Given that we define Sun2 as a genetic modifier of the penetrance of Tor1a neurodevelopmental abnormalities, it will be interesting to investigate whether Sun2 (or other LINC complex genes) are associated with DYT-TOR1A dystonia penetrance. This may also be the most straightforward route to decipher whether a common mechanism operates in both situations.
Materials and Methods
Mouse lines and tissue collection
Tor1a+/−, Tor1a +/Δgag, Sun1+/− and Sun2+/− are previously described (4,28,43). Tor1a+/− and Tor1a +/Δgag were originally produced in the SvEv129 (Taconic) background. These and the Sun2+/− line were crossed more than 10 times onto the C57Bl6 (Jackson) background (Jackson Laboratory). The days of embryonic development are defined after assigning the day of vaginal plug detection as E0.5. In addition, embryos younger than E11.5 were staged based on somite number. Days of post-natal development were defined with the birthdate as P0.
Embryos were collected from pregnant females after they were euthanized by cervical dislocation or CO2 inhalation. Embryos were dissected from the uterus in Dulbecco's modified Eagle's medium (DMEM) and in some cases imaged under brightfield illumination using a dissecting microscope. Tissues were collected from post-natal wild-type animals after they were euthanized by cervical dislocation (until P21) or CO2 inhalation. In all cases tissue destined for biochemical analysis was snap frozen in liquid nitrogen and stored at -80°C until use. Tissues destined for histological analysis were fixed overnight at 4°C in 4% paraformaldehyde in phosphate buffered saline (PBS). They were then either dehydrated and embedded in paraffin (the majority of E9.5 to E11.5 embryos) or cryoprotected in 30% sucrose, placed in embedding media, rapidly frozen on a metal plate on dry ice and stored at −80°C until required (most E14.5 tissues).
All animal procedures were approved by the Institutional Animal Care and Research Advisory Committee of the KU Leuven or of the University of Tennessee, Knoxville, and performed in accordance with the Animal Welfare guidelines of the KU Leuven, Belgium.
3D whole embryo imaging
Samples were prepared for 3D whole embryo imaging as described (44). Optical projection tomography was acquired with OPT Scanner 3001M (Bioptonics, Skyscan) using the GFP channel to record autofluorescence of formaldehyde fixed embryos. Data were analyzed using Arivis.
Immunolabeling
For analysis of paraffin embedded samples, 6 μm thick sections were dewaxed and rehydrated via a graded ethanol series. For morphological studies, sections were stained with hematoxylin and eosin (H&E). For immunofluorescent analysis, antigen retrieval was performed by boiling the sections for 20 min in sodium citrate buffer (10 mM sodium citrate, 0.05% tween-20, pH 6.0). After blocking for 1 h in 0.1% bovine serum albumin (BSA), sections were incubated overnight with primary antibodies diluted in blocking solution (0.25% Triton X-100 and 10% normal donkey serum in PBS). Then they were washed, incubated with secondary antibodies, washed, and mounted.
For cryopreserved frozen tissue, 12 μm thick sections were cut using a CryoStar NX70 cryostat. For standard immunolabeling, these were permeabilized and incubated for 1 hour in blocking solution. They were then incubated overnight with primary antibodies diluted in blocking solution, washed, incubated with secondary antibodies, washed and mounted.
Primary antibodies and fluorescent stains used in this study are: anti-cleaved caspase3 (Cell Signaling; mAb #9664), Tbr1 (Abcam; ab31940), anti-βIII-tub (TUJ-1; Abcam; ab14545), Pax6 (Biolegend; 901301), phospho-Histone H3 (pHH3; Millipore; 07–081), Tbr2 (Abcam; ab23345), γ-tubulin (Abcam; ab11317), GAPDH (Invitrogen; AM4300), β-catenin (Abcam; ab32572), N-cadherin (Abcam; ab98952), nestin (SantaCruz; sc-21248), 4',6-diamidino-2-phenylindole (DAPI) and phalloidin (Thermofisher; A12379 and A12381). Anti-Nesprin2G antibody was a gift from Didier Hodzic. A previously described anti-Sun1 (45) and commercial anti-Sun2 (Abcam; ab124916) were used for immunofluorescence, while western blotting used rabbit anti-torsinA (37), anti-Sun1 and anti-Sun2 (45). Secondary antibodies used in this study were: Alexa 488 anti-rabbit, Alexa 488 anti-mouse, Alexa 488 anti-goat, Rhodamine red anti-mouse, Rhodamine red anti-rabbit, Rhodamine red anti-goat, Cy5 anti-goat and Cy5 anti-rabbit (all from Jackson ImmunoResearch).
Imaging
Confocal images of fluorescently labeled sections were acquired with a Nikon A1R Eclipse microscope. Widefield images of H&E stained sections were acquired with a Leica SP8 microscope. At least three embryos of each genotype were processed and imaged for all experiments, from which representative images are presented.
Quantification methods
Area measurements
Pax6+ and βIII-tubulin+ zones were measured in littermate pairs of embryos to determine the size of areas occupied by proliferating radial glial or differentiating neurons, respectively. For each pair of embryos, matching sections were selected based on peripheral landmarks. A threshold signal that represented positive labeling was set manually using ImageJ. At E11.5, the number of pixels where signal exceeded this value was measured for a region of interest that covered the entire rostral brain present in a horizontal section (telencephalon to diencephalon). At E14.5, the area containing signal was assessed in coronal forebrain sections. Pixels were converted to μm values and, at E11.5, these were normalized to the total length of the embryo present on a section to offset variability coming from the angle of cutting.
Characteristics of proliferative zone nuclei
Horizontal sections from littermate E11.5 embryos were matched for dorsal-ventral position on the basis of peripheral landmarks. High power images (60× magnification) of the hindbrain neuroepithelium were then collected. These covered all areas of a proliferative zone without overlaps, although excluded those with poorly defined Pax6 labeling, or had tears, folds or other signs of damage. We ensured at least three high power images were collected from each embryo before proceeding with further quantification. The shape and angle of Pax6+ nuclei were assessed after each image was manually rotated so the ventricular surface of the hindbrain lay horizontally in the image plane (designated as 0° angle). Nuclei that were clearly Pax6+ were segmented using trainable Weka segmentation (46). Mis-segmented nuclei were discarded after manual quality control. For each nucleus, an ImageJ macro measured the fitted ellipse parameters (https://imagej.nih.gov/ij/docs/menus/analyze.html) (minor, major axis and angle) from which we calculated the nuclear elongation ratio (major axis/minor axis).
Apoptotic nuclei were defined as those positive for anti-cl-caspase3 signal. The hindbrain neuroepithelium was delineated via its morphology revealed by DAPI staining, and the area measured using ImageJ. Then the total number of apoptotic nuclei in this area were manually counted. Since apoptotic nuclei were infrequent, three pairs of matched hindbrain sections were examined for each littermate pair. These were spaced at least 50 μm apart to preclude the possibility that nuclei were double counted.
Mitotic nuclei were defined as those positive for pHH3 signal. After selecting matched pairs of horizontal sections through the E11.5 hindbrain, we imaged the Nestin+ zone at 40× ensuring no overlap between adjacent images. These images covered 150–350 μm lengths of Nestin+ hindbrain starting ∼50 μm from the central hinge point, and excluded areas that were damaged during tissue processing. Then we manually counted the number of mitotic cells within the Nestin+ neuroepithelium of these images, used ImageJ to measure the length of the neuroepithelium under assessment, and expressed the number of mitotic nuclei relative to this length. We also used ImageJ to define a 5 μm boundary above the apical surface, and manually counted the number of mitotic cells that were apical (within 5 μm of the apical surface), or non-apical (>5 μm from the apical surface). We only included embryos in this analysis when it was possible to assess >300 μm total length of hindbrain neuroepithelium.
Proliferative zone polarity
Coronal brain sections from E14.5 embryos were matched for their rostro-caudal position using peripheral landmarks. Proliferative zone centrosomes were detected using anti-γ-tubulin co-labeled with anti-Pax6. Confocal images were collected at 60× magnification that were non-overlapping, but covered all areas that were clearly Pax6+ and lacked tears, folds or other signs of processing damage. We ensured that at least two images were collected per embryo. The majority of anti-γ-tubulin signal (in Pax6+ areas) resided in an apical band. Non-apical centrosomes were manually counted as those that clearly lay outside this band, disregarding those centrosomes that still appeared to be associated with apically localized nuclei. We then used ImageJ to convert the pixel length of the apical surface to μm.
Adherens junctions were detected using anti-N-cadherin on sections co-labeled with Pax6 and DAPI. Non-overlapping adjacent images were collected at 20× to cover the entire Pax6+ proliferative zone. ImageJ was used to measure (1) the apical length of the proliferative zone (defined via the presence of Pax6+ nuclei) and (2) the length where this lacks N-cadherin staining. Only areas that were clearly damaged by tears or folding were excluded from image analysis. Data were expressed as a percentage between (1) and (2), and thus reflects the presence of adherens junctions independent to proliferative zone size.
Location of EdU+ nuclei
EdU (0.1 mg; Invitrogen) diluted in saline was administered intraperitoneally to pregnant females at E14.5 gestation, and embryos were collected after 30 min. These tissues were cryopreserved and sections (as described above), and then EdU+ nuclei detected by ‘click-chemistry’ according to manufacturers’ instructions and co-labeled with DAPI. Confocal images of the proliferative zone were collected at 60×, but excluding regions that were severely disorganized and we could not identify the apical surface, or regions that were damaged in processing. We ensured that at least three images were collected per embryo to perform quantifications. These images were then manually rotated so the proliferative zone surface lay horizontally in the image plane, and Edu+ nuclei were manually selected. For each nucleus, an ImageJ macro measured the shortest distance to the apical surface.
Nuclear envelope LINC complexes
Coronal forebrain sections from E14.5 embryos were matched for their rostro-caudal position using peripheral landmarks. These were labeled with anti-Sun1, anti-Sun2 or anti-Nesprin2G, together with DAPI and βIII-tubulin to define the proliferative zones. Confocal images of the proliferative zone were collected at 60×, and only excluded regions that were damaged in processing. We ensured that at least three images were collected per embryo to perform quantifications. We then subdivided these larger images into smaller 50 µm2 blocks to ensure they only contained the proliferative zone. This also removed information about overall brain structure to allow an observer to be blind to genotype. They were then qualitatively scored for the clear presence of nuclear envelope localized LINC protein signal.
Western blotting
Protein lysates were prepared from dissected embryonic and post-natal tissues by sonication in a solution of 100 mM Tris–HCl (pH 8.0) containing 1% SDS, 300 mM NaCl and complete protease inhibitor cocktail (Sigma). Insoluble material was removed by centrifugation at 20 000g for 10 min, and protein concentration determined using the BCA kit (Pierce). Lysates were diluted in Laemmli sample buffer and incubated for 30 min at room temperature, and then used in standard SDS-PAGE and western blotting. Immunoreactivity was detected using enhanced chemiluminescence and a Fujifilm LAS-3000 imager.
RT-PCR
Total mRNA was extracted using trizol, reverse transcription performed with the Superscript III system (Invitrogen). cDNA samples were then used in standard PCR with primers that amplify Tor1a and Gapdh and using the minimum number of cycles needed to visualize product in an ethidium bromide gel.
Statistics
Quantitative data are shown as mean ± SEM in figures. Statistical analyses were performed using GraphPad Prism v7.01, using the tests indicated in each figure legend. All T-test analysis were two-tailed. No measurements were excluded from any analysis. The n value of each experiment (individual embryos) is also indicated in the corresponding figure legends, and the minimum number of measurements made from each embryo are presented in the methods.
Supplementary Material
Supplementary Material is available at HMG online.
Acknowledgements
We particularly thank the Bioimaging Core of the VIB-KU Leuven Center for Brain & Disease Research, including Sebastian Munck and Nikky Corthout for their work. The Nikon A1R Eclipse Ti used in this work was acquired through a Hercules type I grant (AKUL/09/037) to Wim Annaert, and we also thank the Histology Core and the animal facility of KUL for their work. We thank Didier Hodzic for sharing antibodies. We also thank Andrew Copp for his advice and past members of the Goodchild lab at the University of Tennessee for effort producing Sun1 and Sun2 antibodies. Further, we appreciate the support of Phyllis I. Hanson for sharing plasmids, cell lines and helpfully commenting on our manuscript and findings, as well as that of Patrik Verstreken and other members of the Goodchild Lab.
Conflict of Interest statement. None declared.
Funding
This work was supported by the Foundation for Dystonia Research. B.D.G. is supported by the Instituut voor Innovatie door Wetenschap en Technologie (IWT) with a PhD fellowship.
