Highly efficient manipulation of nervous system gene expression with NEPTUNE

Summary Genetic loss and gain of function in mice have typically been studied by using knockout or knockin mice that take months to years to generate. To address this problem for the nervous system, we developed NEPTUNE (NEural Plate Targeting by in Utero NanoinjEction) to rapidly and flexibly transduce the neural plate with virus prior to neurulation, and thus manipulate the future nervous system. Stable integration in >95% of cells in the brain enabled long-term overexpression, and conditional expression was achieved by using cell-type-specific MiniPromoters. Knockdown of Olig2 by using NEPTUNE recapitulated the phenotype of Olig2−/− embryos. We used NEPTUNE to investigate Sptbn2, mutations in which cause spinocerebellar ataxia type 5. Sptbn2 knockdown induced dose-dependent defects in the neural tube, embryonic turning, and abdominal wall closure, previously unreported functions for Sptbn2. NEPTUNE thus offers a rapid and cost-effective technique to test gene function in the nervous system and can reveal phenotypes incompatible with life.


INTRODUCTION
Recent progress in genomics, metagenomics, and transcriptomics has accelerated the identification of gene variants associated with human disease (Chiu and Miller, 2019;Taylor et al., 2015). In vivo validation in mammalian systems has, however, lagged behind, given that generation of mouse models is limited by the long reproduction time of mice. Ultrasound-guided in utero transduction with fluorescently traceable lentiviruses, carrying RNAi or Cre recombinase, into mouse embryos demonstrated noninvasive and highly efficient transduction of surface epithelium (Beronja et al., 2010(Beronja et al., , 2013 allowing rapid deciphering of gene networks in skin (Adam et al., 2015;Schramek et al., 2014). In contrast, genetic manipulation of embryonic mouse brain has largely relied on in utero electroporation (Saito and Nakatsuji, 2001;Tabata and Nakajima, 2001) or viral infection (Artegiani and Calegari, 2013) post neurulation, which targets fewer cells and cannot achieve long-term genetic manipulation, nor systemic effects.
The neural plate, prior to the onset of neurulation, is a single layer of columnar cells exposed to the amniotic fluid. We hypothesized that in utero injection at this stage, embryonic day 7.5 (E7.5) in mice, should achieve widespread transduction of the future brain and spinal cord. Injection of fluorescent beads into amniotic fluid at E7.5 is well-tolerated by mouse embryos (Slevin et al., 2006) and injection at E8.5 can target the future brain (Gaiano et al., 1999), suggesting that viral transduction at E7.5 could offer a significant improvement in targeting efficiency. Injections of minute virus of mice (MVM) at E9.5-E12.5 (Itah et al., 2004), or recombinant adeno-associated virus 2 (rAAV) at E15 (Lipshutz et al., 2001), were well-tolerated. However, only 41% of embryos injected with retrovirus at E8.5 survived, and of these 41% displayed exencephaly (Gaiano et al., 1999), indicating that injections at E8.5 with virus MOTIVATION Current methods used to alter gene expression in mouse embryos to study brain development lack efficacy and are time consuming. NEPTUNE provides an addition to the molecular toolbox, allowing rapid and widespread manipulation of gene expression before neurulation in vivo. This technique can be adapted to specifically target central nervous system (CNS) cell subtypes such as neurons or glia. Functional analysis can be performed within days after injection. NEPTUNE can thus reduce the financial and ethical costs of animal research, while accelerating the process from hypothesis to results in vivo.
either is detrimental to development or requires further technical development.
Embryonic virus injection can achieve widespread transduction of most organs (Lipshutz et al., 2001); therefore, a method to achieve cell-type-specific expression would be desirable for targeting the nervous system exclusively. Conditional CRE expression to excise LoxP-flanked genetic sequences is a gold standard for spatiotemporal cell-type-specific expression but is associated with adverse effects related to random Cre integration, unpredictable genome editing, and toxic effects of CRE itself (Cain-Hom et al., 2017;Loonstra et al., 2001;Pomplun et al., 2007). An approach avoiding the use of dedicated Cre mouse strains is the use of MiniPromoters (Portales-Casamar et al., 2010) to drive cell-type-specific expression in brain, including in neuronal progenitors, astrocytes, and oligodendrocytes. MiniPromoters are promoters of less than 4 kb, based on cell-type-specific gene expression in human brain, and verified as faithful and specific in the mouse brain (Portales-Casamar et al., 2010).
Variable penetrance of some genetic mutations, the unexpected behavior of gene-targeted loci, and genetic compensation can confound gene analyses. In mice, genetic background can alter phenotype penetrance, but even in a defined genetic background penetrance can be variable (Dickinson et al., 2016). Additionally, genetic ''knockouts'' generated by using different strategies (deletion of start exons, all exons, or CRISPR/Cas9 deletion of a few exons) can yield different phenotypes (Hosur et al., 2020). CRISPR/Cas9-mediated gene editing can lead to exon skipping and hypomorphism, neomorphism, or gain of function of the targeted gene, leading to unexpected phenotypes that do not reflect loss of function (Anderson et al., 2017;Chen et al., 2018;Uddin et al., 2015). Phenotypic differences between knockdown and knockout models, including mice, zebrafish, and Arabidopsis, could be explained by genetic compensation, by which related genes are upregulated in response to gene knockout but not knockdown (El-Brolosy and Stainier, 2017). Therefore, the development of widespread and long-lasting short hairpin RNA (shRNA) knockdown in mouse embryonic nervous system would be beneficial.
In this study, we developed a method to manipulate gene expression in the developing murine nervous system termed NEural Plate Targeting by in Utero NanoinjEction (NEPTUNE). In utero transduction of the neural plate at E7.5 achieved near 100% targeting of the future brain and ca. 80% of the spinal cord. To achieve conditional expression while avoiding the use of Cre mice, we used MiniPromoters (Portales-Casamar et al., 2010) to drive cell-type-specific expression in neuronal progenitors as well as glia. As proof of principle, we knocked down Olig2, which recapitulated the phenotypes of the published knockout (Takebayashi et al., 2002). Finally, using NEPTUNE, we knocked down Sptbn2 and discovered early functions for Sptbn2 in neural tube development, embryonic turning, and abdominal wall closure. Our results show that NEPTUNE is a powerful method for manipulation of gene expression in the developing nervous system that can reveal functions for human disease-associated genes.

RESULTS
Optimal parameters for maximum survival and transduction lead to widespread and stable targeting of developing brain Prior to neural tube closure in mice, around E7.5, the neural plate is exposed to the amniotic fluid and should be accessible to injected virus, allowing targeting of the future nervous system (Figure 1A). Neurulation stages can be discerned by ultrasound until neural fold fusion of brain at E9.5 ( Figure S1A). Injections at E8.5 Figure 1. Ultrasound-guided in utero nanoinjection of lentivirus at E7.5 results in widespread and stable transduction of the murine CNS (A) The amniotic cavity was injected at E7.5 and embryos were collected at E13.5 or as adults at 6 months. E13.5 brains were divided in half and assessed for GFP positivity by using flow cytometry of dissociated cells or CellProfiler analysis of sectioned brain halves. (B-D) (B) Optimal injection volume was assessed by injecting 23-483 nL of viral resuspension buffer (VRB) at E7.5. Red dotted line denotes drop in survival at volumes >207 nL. Each data point represents survival percentage in one set/litter of injected embryos; for litter details see Figure S1C. SO, surgery only. (C) Distension of the amniotic cavity (double-headed arrows) is visible on ultrasound (scale bars represent 1 mm) and calculating volume after injection showed that (D) survival drops if the total volume increases more than 90%. Each datapoint represents the percent total volume increase in one embryo/amniotic cavity. Red datapoints reflect embryos that were aborted. (E) Ultrasound images of externalized uterus and four embryos (Emb) at E7.5. Theiler stage (TS) 11b was most suitable for high efficiency and high survival. Emb 1* is likely resorbing or severely underdeveloped. Scale bars represent 1 mm. (F and G) (F) GFP + heads in whole embryos at E13.5 and (G) GFP signal in brain at 6 months. (H) Targeting efficacy is dependent on viral titer. injections of lentivirus (207 nL) with four titers show dose-dependent transduction, with a titer of 2 3 10 10 ifu/mL resulting in transduction of 85%-99% of cells in brain at E13.5, assessed by flow cytometry. Each datapoint represents flow-cytometry quantification from one half brain; the other half was used for quantification with CellProfiler. (I and K) Representative images of E13.5 half brains from embryos transduced at E7.5 with VRB, 69 or 207 nL of 2 3 10 10 ifu/mL hPGK-H2B-GFP virus, with corresponding flow-cytometry data below (n = 3 for each condition). Top row is forebrain, middle row is diencephalon, and bottom row is hindbrain. Scale bars represent 200 mm (half brain) and 50 mm (boxed regions). AC, amniotic cavity; HF, head fold; NI, noninjected; ifu, infectious units. were too late to achieve consistently high levels of transduction, presumably because neural fold fusion is ongoing ( Figure S2). We therefore focused on E7.5 for in utero transduction, before neural fold fusion, when the amniotic cavity is discernible by ultrasound and the microcapillary needle tip can be accommodated in the cavity (Figures S1A and S1B). To determine the optimal parameters for embryo survival and maximum transduction, we injected a range of volumes of viral resuspension buffer into the amniotic cavity (23-483 nL) at E7.5 and recorded the percentage of injected embryos per litter surviving at E13.5 ( Figures  1B and S1C). Sham surgery identified a baseline resorption rate of 0%-25%. Therefore, we drew the cutoff for acceptable survival rates at 70% and identified 207 nL as the highest permissible volume for normal survival ( Figure 1B). At E7.5, embryonic development and amniotic cavity size varies even within litters. Therefore, the best predictor of survival was not volume but the relative increase in amniotic volume, whereby an increase >90% resulted in embryo resorption ( Figures 1C and 1D). We therefore selected 207 nL (62%-162% volume increase, mean = 95% ± 38% SD) and 69 nL (9%-44% volume increase, mean = 28% ± 14% SD) for further study. A volume of 345 nL consistently resulted in resorption (325%-1,227% volume increase, mean = 615% ± 421% SD) and was excluded from further experiments. We compared appearance on ultrasound with the Theiler stage (TS) of embryos dissected out immediately after imaging and identified TS 11b as the optimal injection stage for 207 nL. This late neural plate stage, characterized by an elongated allantoic bud, precedes the appearance of the cranial neural (head) folds (HF), which arise at TS 11c ( Figure 1E). To optimize selection of females for surgery and take into account the 3Rs (replacement, reduction, refinement), ultrasound confirmation of pregnancy before surgery can be used for staging to ensure ideal conditions ( Figure S1D).

Lentiviral transduction is even across CNS regions and cell types
The neural tube closes at specific neural tube closure points (Nikolopoulou et al., 2017), making these regions potentially less amenable to lentiviral manipulation due to decreased accessibility and subsequent uneven viral transduction. In mice, the initial neural tube closure point 1 is at the hindbrain/cervical boundary, closure point 2 is at the forebrain/midbrain boundary, and closure point 3 is at the most rostral end of the forebrain. Closure/zippering proceeds rostrally and caudally from closure points 1 and 2, whereas closure proceeds caudally from closure point 3 toward closure point 2 (Nikolopoulou et al., 2017). Injections at E8.25-E8.5 ( Figure S2A) reflected partial neural tube closure with variable efficiency of transduction of forebrain, midbrain, hindbrain, and spinal cord at E11.5 (Figures S2B-S2E). Transduction was highest in forebrain, the region which is open longest (mean = 24% ± 11.1% SD) and lowest in hindbrain (mean = 5.5% ± 2.7% SD), the region which closes first (Figures S2B and S2C). Transduction was around 8% in thoracic and lumbar spinal cord (thoracic mean = 6.2% ± 3.8% SD; lumbar mean = 10.3% ± 6.4% SD) ( Figures S2D and S2E).
We therefore focused on E7.5 injections and assessed transduction efficiency throughout the brain and spinal cord, in neural precursors or neurons, to determine whether the entire nervous system was evenly targeted by injection at this stage (Figures 2 and 3, split channels for Figures 3B-3E and 3G-3J in Figure S3). H2B-GFP was highly and evenly expressed from forebrain to hindbrain, in both SOX2 + neural precursors and post-mitotic NeuN + neurons (Figures 2A-2G). H2B-GFP was also widely expressed in the cerebellum at 6 months ( Figure 2H), which develops from a cerebellar primordium first evident at E12.5. H2B-GFP was stably expressed in both GFP + calbindin-positive (CALB1 + ) Purkinje neurons and SOX2 + glia ( Figure 2H).
A similarly even transduction pattern was present in the spinal cord at thoracic levels, which form via primary neurulation (Figures 3 and S3). Virus injection at E7.5 efficiently transduced the neural plate contributing to SOX2 + progenitor cells and NeuN + mature neurons in the spinal cord ventricular zone and parenchyma ( Figures 3A-3C), as well as delaminating neural crest cells ( Figures 3A and 3D) and dorsal root ganglia ( Figures 3A and 3E). Spinal cord transduction efficiency was slightly lower than the corresponding brain transduction ( Figure S3B, brain control data from brain halves analyzed by CellProfiler and also presented in Figure S1E, black data points).
Lumbar and sacral spinal cord is formed via secondary neurulation from the caudolateral epiblast, which contains neuromesodermal progenitors (NMPs) that are internalized around E8.5-E9.0 (Wymeersch et al., 2021). In utero nanoinjection at E7.5 results in widespread H2B-GFP positivity in caudal embryo in lumbar and sacral levels ( Figures 3F-3K); there was H2B-GFP positivity not only in skin and spinal cord but also in mesenchyme, suggesting targeting of NMPs contributing to both neural and mesodermal tissues (Figures 3L and 3M-note the H2B-GFP positivity in mesenchyme in Figure 3M, compared with absence of H2B-GFP in mesenchyme in head or thoracic spinal cord sections in Figures 3L and 2A-2C). In conclusion, in utero nanoinjection at E7.5 targets brain and spinal cord formed by primary or secondary neurulation.
To comprehensively assess which cells are transduced at E7.5 and express the DCX-H2B-GFP construct, we compared expression of H2B-GFP in hPGK-H2B-GFP-and DCX-H2B-GFP-injected whole embryos. At E18.5, GFP expression was prominent in skin of hPGK-H2B-GFP-injected embryos (Figure 5A,left) whereas GFP expression in DCX-H2B-GFP-injected embryos was most prominent in cortex ( Figure 5A, right), and skin positivity was comparable with an uninjected control. qPCR of GFP + cells sorted from whole bodies at E13.5 confirmed 7.1-fold enrichment of skin targeting by hPGK-H2B-GFP (±1.3 SD, Figure 5B) but no targeting by DCX-H2B-GFP ( Figure 5C). In hPGK-H2B-GFP-injected embryos, the GFP + signal was clear and abundant in epithelial cells in skin, lung, and stomach, and scattered GFP + cells could be found in the liver and heart ( Figures 5D-5I). qPCR analysis of sorted GFP + cells from E13.5 hPGK-H2B-GFP whole bodies ( Figure S5) further confirmed widespread transduction predominantly in cells either in contact with amniotic fluid or which are expected to be derived from the neural crest, including cells expressing Dcx (neuroblasts, Figures S5A and S5B), Sox2 (neural stem cells and epithelial cells, for example in the lung [Que et al., 2009], Figure S5D), and Cdh1 (also known as epithelial cadherin, in epithelial cells, Figure S5E). Similar to E18.5 brain analysis, hPGK-H2B-GFP GFP + cells at E13.5 were not significantly enriched for Olig1 ( Figure S5C) and were depleted for CD31/Pecam-1 and CD34 (markers of vascular and hematopoietic cells, Figures S5F and  5G). In sum, these data suggest that NEPTUNE preferentially targets epithelial cells and neural plate.
In conclusion, NEPTUNE achieves >95% transduction of brain with the hPGK promoter, and expression can be directed to specific cell types by using MiniPromoters, with an efficiency dependent on the promoter used.
NEPTUNE knockdown of Olig2 recapitulates knockout phenotypes As proof of principle, we next aimed to test whether NEPTUNE could recapitulate established nervous system phenotypes.
Olig2 is a transcription factor required for development of motoneurons and oligodendrocytes, and Olig2 À/À embryos display a striking phenotype with a shortened crown-rump length in newborn pups and a reduction in HB9/Islet1 + motoneurons and platelet-derived growth factor receptor a-positive (PDGFRa + ) oligodendrocytes (Takebayashi et al., 2002). We expect that crown-rump length shortening is a strong phenotype that can only be recapitulated with robust knockdown throughout the spinal cord. Knockdown of Olig2, using either of two different shRNA constructs, with NEPTUNE resulted in a shortened crown-rump length at E14.5 ( Figure 6A), and qPCR for Olig2 in whole spinal cord at E18.5 confirmed RNA silencing by both shRNAs ( Figure 6B). Survival of injected embryos was similar to that under control conditions at E10.5, E14.5, and E18.5 ( Figure 6C). OLIG2 protein expression in the spinal cord motor neuron progenitor domain (pMN) was nearly absent or absent in shOlig2-H2B-GFP embryos at E10.5 and E14.5 ( Figures  6D and 6E). Olig2 knockdown resulted in a decrease in HB9 + and Islet1 + motoneurons at E10.5 ( Figures 6F and 6G) and a loss of PDGFRa + oligodendrocytes at E18.5 ( Figure 6H, boxed regions 1 and 3). Importantly, PDGFRa expression was still obvious in vasculature surrounding the spinal cord ( Figure 6H, boxed regions 2 and 4). Thus, Olig2 knockdown with NEPTUNE recapitulated the Olig2 À/À phenotype (Takebayashi et al., 2002).
NEPTUNE reveals a role for Spbtn2 in neurulation and abdominal wall closure Finally, to test whether NEPTUNE can provide insights into gene function or human disease, we focused on a gene associated with neurodevelopmental defects. Spectrin Beta, Non-Erythrocytic 2 (Sptbn2) is highly expressed in the CNS and mediates intracellular vesicle transport and cytoskeleton dynamics (Liem, 2016). Mutations in SPTBN2 are associated with SCA5 (OMIM: 600224) (Cho and Fogel, 2013;Ikeda et al., 2006;Nicita et al., 2019;Ranum et al., 1994;Wang et al., 2014) and SCAR14 (OMIM: 615386) (Elsayed et al., 2014;Lise et al., 2012;Yıldız Bö l€ ukbas xı et al., 2017). Mouse models for Sptbn2 mutation display adult-onset defects in motor function, but these mice express shorter SPTBN2 isoforms (Stankewich et al., 2010) or splice variants (Perkins et al., 2010), suggesting that these models might not reflect complete loss of function. We therefore decided to test whether knockdown of Sptbn2 at E7.5 induced an ataxic phenotype or a more severe phenotype, as would be expected if the splice variants or shorter variants can compensate for loss of full-length SPTBN2.
Sorting of shSptbn2-GFP + cells from E9.5 whole embryos, injected at E7.5, confirmed silencing of Sptbn2 mRNA by 80% in targeted cells ( Figure 7A). Of 36 embryos injected at E7.5 and analyzed at E9.5, six (17%) were resorbed (similar to baseline resorption rates, Figure 1B), and the remaining 83% could be divided into three classes by phenotype ( Figure 7B). We defined Class 1 as the mildest phenotype, with a subtle straightening of the body axis (eight embryos, 22%). Class 2 was characterized by a failure to undergo embryonic turning and a skewed body axis (ten embryos, 28%), and Class 3 was developmentally delayed and had not undergone turning (12 embryos, 33%, Figure 7B). Dorsal views of these three classes revealed a kinked and asymmetric neural tube in all three classes ( Figure 7C).
To determine whether Sptbn2 knockdown resulted in reduced SPTBN2 protein levels, we first sorted GFP + cells from shSptbn2-H2B-GFP-and hPGK-H2B-GFP-injected embryos for western blotting. However, this yielded highly variable results, which might reflect degradation of SPTBN2 in dissociated cells ( Figure S6F-note distinct bands in crushed positive control whole cerebellum versus smearing and extra bands in dissociated cerebellum). We therefore injected embryos with shSptbn2-H2B-GFP or hPGK-H2B-GFP and collected embryos at E9.5, pooling three embryos per wild-type pool and four embryos by class severity per pool, and adjusted samples for equal total protein concentration. There were reduced levels of SPTBN2 in all three classes, with the greatest reduction in Class 3, the most severely affected embryos ( Figures 7D and 7E).
Axial skewing and abdominal wall defects have previously been described for compound mutants with disruption in planar cell polarity components Scrib, Celsr1, or Vangl2 (Murdoch et al., 2014), which govern the process of convergent extension and bending along the mediolateral axis of the neural tube (reviewed in Nikolopoulou et al., 2017). To test whether Sptbn2 knockdown affects neurulation via related mechanisms, we further analyzed the Foxa2 staining at E9.5 in Sptbn2 shRNA virus-transduced embryos ( Figure 7I). Sections from thoracic to sacral levels revealed a wider floorplate region (red arrows in Figure 7I, boxed regions) which was more pronounced in lumbar sections, similar to the floorplate widening seen in compound planar cell polarity (PCP) mutants (Murdoch et al., 2014). In Class 3 embryos, the sacral spinal cord was open. In sum, NEPTUNE achieved strong downregulation of Sptbn2 mRNA and a dose-dependent phenotype severity correlating with the efficiency of SPTBN2 protein downregulation. Sptbn2 silencing induced defects in neural tube, embryonic turning, and abdominal wall closure, which mimicked the phenotype of PCP mutants (Murdoch et al., 2014). Furthermore, the data show that Sptbn2 loss of function is incompatible with life and suggest that previous models of Sptbn2 mutations might reflect hypomorphic models.

DISCUSSION
Ultrasound-guided in utero nanoinjection has proved to be a powerful tool to unravel genetic networks in skin (Beronja et al., 2010(Beronja et al., , 2013Schramek et al., 2014). Here, we show that the developing mouse nervous system can be targeted in a highly efficient and reproducible manner, achieving over 95% transduction efficiency throughout the brain and spinal cord (Figures 1, 2, and 3). Transduction is stable to adulthood (Figure 1). Using MiniPromoter sequences enabled expression in defined cell types without the need for transgenic Cre mouse lines (Figures 4, 5, S4, and S5). Proof-of-principle experiments demonstrated that NEPTUNE could recapitulate the published Olig2 À/À phenotype (Takebayashi et al., 2002) (Figure 6). In addition, we demonstrated that knockdown of Sptbn2 in vivo results in dose-dependent defects including embryonic lethality by E18.5, abdominal wall defects and a skewed body axis at E13.5, and embryonic turning defects with a widened floorplate at E9.5 (Figure 7). Our data thus reveal a function for this gene in neurulation (Figure 7), likely via PCP-related neurulation mechanisms, which warrants further investigation. NEPTUNE allows high efficiency and widespread transduction of the nervous system. We identified volumes injected, viral titer, and freshness of the virus as crucial determinants of transduction and survival efficiency (Figures 1 and S1), and determined the optimal parameters to achieve >95% transduction in brain and spinal cord. To achieve conditional expression, we replaced the PGK promoter with MiniPromoters (Portales-Casamar et al., 2010) for DCX, GFAP, and OLIG1. Conditional gene expression is typically achieved by using Cre mice, but these mice are often generated by random Cre integration and less frequently by knocking Cre into the locus of a cell-type-specific gene. Importantly, random Cre integration is associated with DNA rearrangements around the integration site, inclusion of unreported genetic sequences, and multiple integrations (Cain-Hom et al., 2017). Furthermore, CRE expression itself can be toxic in vivo (Pomplun et al., 2007) and can induce DNA damage via endogenous cryptic LoxP sites (Loonstra et al., 2001), confounding interpretation of phenotypes obtained with Cre mice. Finally, CRE-mediated regulation of gene expression also entails a delay between conditional transcription of Cre, translation of CRE, and LoxP site editing, leading ultimately to expression of the conditional sequence of interest. This delay can be important when studying early developmental phenotypes. The MiniPromoters displayed robust expression, with no leakiness in SOX2 + precursor cells ( Figures 4I-4K), but had variable efficacy in labeling neurons, astrocytes, and oligodendrocytes. Whereas the DCX-H2B-GFP construct was specifically active in neurons ( Figure 4I, 4L-4N, S4A, and S5) and GFAP-H2B-GFP was expressed specifically in astrocytes ( Figures 4J and S4B), the  Figure S4C). This might reflect transient Olig1 expression in glial precursors contributing to astrocytes and oligodendrocytes, in which expression is downregulated during differentiation into astrocytes, whereas the GFP reporter is stabilized via fusion to H2B and might be expressed longer than OLIG1. Conditional expression with NEPTUNE with MiniPromoters is thus possible, but each promoter construct should be tested and validated. Future work to combine NEPTUNE with, for example, DIO systems (Double-floxed Inverted Open reading frame, also known as Flex switches [Schn€ utgen et al., 2003]) could further extend the toolbox for achieving conditional expression by using Cre mice of interest.
Previous work established that in utero transduction could target skin and has been used to elucidate mechanisms of development and disease in skin (Beronja et al., 2010(Beronja et al., , 2013Schramek et al., 2014). These injections generally utilize injection of 1 mL at E9.5, a stage at which the amniotic cavity is larger and the embryo is easy to identify on ultrasound ( Figure S1A). Gaiano et al. (1999) showed, over 20 years ago, that E8.5 amniotic cavity injection could target the nervous system, but transduction efficiency was uneven and E8.5 injection yielded 41% survival, of which 41% of embryos exhibited exencephaly, likely due to the high volumes injected. Here, we adapted the technique to injection at E7.5 to target the entire neural plate and NMPs contributing to the future nervous system. Survival of E7.5 virus-injected embryos, with optimal conditions, ranged from 40% to 90% (e.g. Figure 7G) and did not lead to brain malformations under these conditions. Future modifications to NEPTUNE could increase its versatility. The rise of AAV evolution (Chan et al., 2017;Davidsson et al., 2019) and, thus, engineering viral tropism, might allow targeted transduction of specific cells during embryonic development. The insertion of a large promoter sequence reduced viral titer obtained from virus production by about 50%, as expected (Kumar et al., 2001) (viral titers described in STAR Methods); therefore, improved systems for virus production or packaging would also further strengthen the power and versatility of this technique.
The development of NEPTUNE can reduce the number of mice used in research. It has been estimated that of 3,872 targeted genes in the mouse, 45% are not essential for viability or fertility (Liao and Zhang, 2007). Conversely, 14% of mutated genes lead to embryonic lethality, usually at or before mid-gestation (Liao and Zhang, 2007;Wilson et al., 2005). Combining these two, the probability of knocking out a gene, and obtaining knockout mice to study that exhibit an interesting and important phenotype during embryogenesis, is around 40%. The development of NEPTUNE might allow screening for relevant neural phenotypes and circumvents the risk of severe embryonic lethality in heterozygous mice. Furthermore, NEPTUNE could be used to investigate genetic redundancy and dissect signaling networks. As an example, breeding triple heterozygous mice to obtain triple knockouts would generate 64 pups to obtain one wild type and one triple knockout. With NEPTUNE, it would thus be possible to directly generate the genotypes/knockdowns of interest, avoiding waste. Furthermore, validation of gene function across different strains of mice is facilitated, given that back-crossing would not be required, further improving and facilitating reproducibility.
The Sptbn2 phenotype (Figure 7) is more severe than expected, based on our current knowledge of SPTBN2 in the human ataxic syndromes SCA5 and SCAR14. However, given its role as a cytoskeletal component linking actin and the cell membrane (Liem, 2016), it is perhaps less surprising that disruption of Sptbn2 would lead to neural tube defects and defects in embryonic turning. Importantly, in most patients, the described mutations are missense or in-frame deletions, suggesting that remnant protein might be sufficient to execute some SPTBN2 functions (Cho and Fogel, 2013;Elsayed et al., 2014;Ikeda et al., 2006;Lise et al., 2012;Nicita et al., 2019;Ranum et al., 1994;Wang et al., 2014;Yıldız Bö l€ ukbas xı et al., 2017). Likewise, the Sptbn2 mutant mice still express slightly shorter forms of SPTBN2, suggesting some functional rescue (Perkins et al., 2010;Stankewich et al., 2010). Our data, using two different shRNAs, suggest that Sptbn2 also has a crucial role in neural development and embryonic turning. With respect to the skewed body axis and abdominal wall closure defects, the Sptbn2knockdown phenotype mimics PCP mutants with single or compound mutations in Scrib, Celsr1, or Vangl2 (Murdoch et al., 2014), suggesting that Spectrin might play a key role in mediating PCP programs during neurulation and turning. However, PCP mutants generally have an extensively open neural tube, which was not present in any of the Sptbn2-knockdown embryos, highlighting a key difference.
In sum, NEPTUNE is a powerful technique to modulate gene expression during embryonic development. It can achieve widespread, stable, and conditional expression in the brain and spinal cord, and can be used to reveal roles for genes in crucial embryonic processes.
Cell Reports Methods 1, 100043, August 23, 2021 13 Article ll OPEN ACCESS Figure 7. NEPTUNE-mediated shRNA knockdown of Sptbn2 impairs survival and causes dose-dependent defects in neural tube and in embryonic turning NEPTUNE with virus encoding shSptbn2-H2B-GFP or Scr-H2B-GFP, and embryos collected at E9.5, E13.5, or E18.5 (A) qPCR for Sptbn2 in GFP + cells sorted from whole embryos at E9.5. Two Scr-H2B-GFP embryos were pooled per ''Scr'' sample (each dot = 2 embryos). Three to four shSptbn2-H2B-GFP embryos were pooled per sample (each dot = 3 or 4 embryos). (B) At E9.5, shSptbn2-H2B-GFP embryos presented with three phenotypic severities; 17% are resorbed. Class 1 phenotype (22%) was defined as a straightened body axis, Class 2 phenotype (28%) as a body with a rightward-skewed lumbar and sacral region, and Class 3 phenotype (33%) as developmentally delayed and failure to undergo embryonic turning. (C) Dorsal views of neural tube of control Scr-H2B-GFP-and shSptbn2-H2B-GFP-injected embryos from the three classes. In control, arrowheads denote a straight spinal cord with even closure. In Class 1-3 shSptbn2-H2B-GFP embryos, arrowheads denote kinks in the spinal cord with uneven closure. and died prior to E13.5. Future work comparing shRNA knockdown and CRISPR editing of the corresponding gene should be undertaken to compare phenotypes upon knockdown and genetic perturbation. Genetic compensation in genetically perturbed organisms might mask critical functions (Rossi et al., 2015), and it will thus be of great interest to compare traditional knockouts, NEPTUNE-mediated CRISPR gene editing, and NEPTUNE-mediated shRNA knockdown.
Although it is possible to achieve widespread (up to 99%) and stable (tested up to 6 months) transduction of the developing mouse brain with NEPTUNE, we could not reach 100% of the cells in the future brain, and similarly current efficiency for spinal cord was a maximum 79%. Lentiviral packaging capacity also limits the production of high-titer lentivirus, as exemplified by a 50% reduction in titer when using MiniPromoters to drive GFP expression. Although NEPTUNE can achieve cell-type-specific effects without dedicated Cre mice, combination of NEPTUNE with Cre or Cas9 mice would further improve conditionality or capacity to knock genes out while allowing for smaller lentiviral loads.

STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:

RESOURCE AVAILABILITY
Lead contact Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Dr. Emma R. Andersson (emma.andersson@ki.se).

Materials availability
All newly generated items used in this study are available upon request and following standard Material Transfer Agreement (MTA), due to institutional recommendations.

Data and code availability
This study did not generate computer algorithms or codes. All data available in the main text, supplementary materials, and raw data are deposited in Mendeley Data at https://doi.org/10. 17632/rr6c34zgsw.1.

EXPERIMENTAL MODELS AND SUBJECT DETAILS
Animals CD1 wild type mice were obtained from Charles River Laboratories (Germany). Animals were housed according to European regulations, with a standard day and night cycle with food and water ad libitum. From the age of 8 weeks, females were checked for estrus and plugged overnight. Gestation was defined as embryonic day (E) 0.5 at noon of the same day of vaginal plug. Ethical approval for all experiments described here was granted by the Swedish Board of Agriculture (Jordbruksverket) with permit numbers N59/14, 8188-2017 and 2987-2020.

Cell lines and culture conditions
Lenti-XTM 293T cells (Clontech) were grown at 37 C/ 5% CO 2 in DMEM complete medium (DMEM, high glucose, GlutaMAX, pyruvate supplemented with 10% fetal bovine serum, 1% penicillin/streptomycin and 1% geneticin). For virus production, low passage Article ll OPEN ACCESS cells (p4-p6) were thawed and seeded into a 75cm 2 cell culture flask one week prior to virus production. Upon transfection with plasmids for virus production, DMEM complete medium was replaced with DMEM virus medium for the rest of the process (DMEM, high glucose, GlutaMAX, pyruvate Supplement with 10% fetal bovine serum, 1% penicillin/streptomycin). NE4C cells (ATCC) were grown until passage 15 at 37 C/ 5% CO 2 in MEM complete medium (MEM, GlutaMAX supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin). Upon transduction with lentivirus for titration experiments, MEM complete medium was replaced by MEM infect medium (MEM, GlutaMAX supplemented with 10% fetal bovine serum and 0.1 mg/mL Polybrene).
Both Lenti-XTM and NE4C cell lines were free from mycoplasma contamination and underwent regular screening for mycoplasma.

METHOD DETAILS
Cloning of shRNA constructs 5 mg of plasmid, containing shRNA (TRC, Sigma Aldrich) or scrambled control shRNA (gift from David Sabatini) or hPGK-H2B-GFP reporter (gift from Elaine Fuchs, also known as LV-H2B-GFP) were cut with SacII and SphI for 15 min to 1 h at 37 C in CutSmart buffer.
(Vector backbone for all plasmids is pLKO.1). SphI cuts within the 5' LTR and SacII cuts just upstream of the hPGK promoter. Digests were run on a 1% Agarose gel at 100 V for 1 h. For shRNA or scrambled control shRNA digests (insert), the band at 2243 bp (containing the U6 promoter, shRNA insert and 5'LTR), and for GFP reporter digests (vector) the band at 5263 bp (containing the hPGK promoter, the H2B-GFP reporter and the 3'LTR) were excised and DNA was extracted using the Qiagen QuickGel extraction kit. Ligations were set up with a molar ratio 3:1 of insert:vector. Using NEB Quick ligation kit, ligation mixture was incubated at room temperature for 5 min. The reaction was put on ice and One Shot Stbl3 chemically competent cells were transformed following the manufacturers' protocol. 100 mL transformation mix was spread on an Agar plate containing 100 mg/mL Ampicillin, and the plate was incubated overnight at 37 C. The next day, 3-5 colonies were picked for screening and incubated on a shaker in 2 mL Terrific Broth containing 100 mg/mL Ampicillin for 8 h. Plasmid DNA was extracted using PureLink Quick Plasmid Miniprep Kit. A diagnostic digest was performed using NdeI and SacII. Vectors were validated by sequencing using the Eurofins barcode sequencing service.

Endotoxin-free plasmid prep
Bacteria containing the plasmids of interest were grown in 100 mL (for high copy plasmids) and 200 mL (for low copy plasmids) of Terrific Broth containing 100 mg/mL Carbenicillin, on a horizontal shaker for 20-22 h. Bacteria were centrifuged to pellet at 4000 g for 20 min at 4 C. Medium was discarded and bacterial DNA was extracted using PureLink TM Endotoxin-Free Maxi Plasmid Purification Kit, following the manufacturer's instructions. Concentration was measured with a nanodrop.
Transfection 5*10 4 NE4C cells/well were seeded on 12-well plate 16-20 hours prior to transfection in MEM complete medium. Lipofectamine2000 and 1mg plasmid DNA were diluted in Opti-MEM to a final DNA:Lipofectamine ratio of 1:3 and incubated at room temperature for 5 min. Transfection Mix was added and incubated for 8 hours, before being replaced by fresh MEM complete medium. 24 or 48 hours aftertransfection, cells were collected for RNA or protein analysis (described further down under ''RNA extraction and qPCR'' or ''Western Blot'').

Virus production
One 75 cm 2 flask of LentiX-293T cells, at 80% confluency, was split into four 225 cm 2 flasks. 24 hours prior to transfection, when cells were at 80%-90% confluency, one 225 cm 2 flask was seeded 1:1 on one 500 cm 2 plate (2 plates per virus were used, typically two viruses were produced at a time Article ll OPEN ACCESS vector plasmid, 275 mg of psPAX2 (packaging plasmid) and 180 mg of pMD2.G (VSV-G plasmid) were mixed in a 50 mL conical tube (pMD2.G and psPAX2 were a gift from Didier Trono). 2.28 mL of 2 M CaCl2 in MQ water were added to a final volume of 9.5 mL. Next, 9.5 mL 2xHBS (50 mM HEPES, 1.5 mM Na2HPO4, 280 mM NaCl, pH 7.07) was added and the tube was inverted 4 times. After incubation at room temperature for 60 seconds, the mixture was added to 165mL of pre-warmed DMEM virus medium (see above). 14-16 h after transfection, medium was replaced with fresh DMEM virus medium. Viral supernatant was collected at 46 h after transfection, and at 65 h. Upon collection, supernatant was filtered through a 0.45 mM Millipore low-protein binding filter units and kept at 4 C. Viral supernatant was concentrated by first using low-speed centrifugation through 100 kDa MW cutoff Millipore Centricon 70 Plus cartridges to condense the volume to < 4 mL, followed by ultracentrifugation through a 20% sucrose cushion at 45,000 rpm (MLS 50 Rotor) for 2 h, to pellet the lentiviral particles. That pellet was resuspended in 25-30 mL viral resuspension buffer (20mM Tris pH 8.0, 250 mM NaCl, 10 mM MgCl2 and 5% sorbitol) and 5 mL aliquots were stored at -80 C.
Virus titration Viral titers for constructs containing the hPGK-H2B-GFP reporter were determined via spinfection of NE4C cells. 24 h prior to transduction, cells were seeded in MEM complete medium in a 6-well plate at a density of 5*10 8 cells/well. For calculation of the viral titer, the total number of cells per well, available for spinfection, must be determined. Therefore, on Day 0, one well was trypsinized and the cells were resuspended in 1.5 mL MEM complete medium. Cell number per mL volume was assessed using a Countess 3 Automated Cell Counter and adjusted for the total volume of 1.5 mL to obtain the total number of cells in one well. Medium was replaced with MEM infect medium in the remaining five wells. 5, 50, and 500 mL of virus diluted 1:2,000 and 1 mL of concentrated virus were added to four individual wells, one well was kept as non-transduced control. Plates were centrifuged for 30 min at 1100 g and 37 C. MEM infect medium was replaced with MEM complete medium and cells were incubated for 48 h at 37 C. The percentage of GFP+ cells was recorded using a Canto LSR II flow cytometer. In brief, cells were washed with PBS and trypsinized. Cell suspensions were collected in 1.5 mL tubes and spun for 5 min, 200 g. The cell pellet was washed and resuspended in PBS twice, before fixation in 4% formalin for 15min at room temperature. After two additional washing steps, the cell suspension was filtered through a 25mm cell strainer and kept on ice until analysis. This titration yields ifu/mL. For MiniPromoter constructs, NucleoSpin RNA virus kit and LentiX qRT-PCR Titration Kit were used. The manufacturer's instructions were followed. qRT-PCR Titration yields viral particles/mL. Particles/mL was converted to ifu/mL based on comparisons of titers obtained with the two methods for the same viral preps, yielding a correction factor of 159 (159-fold higher particle titer compared to actual ifu/mL titer).

Pregnancy verification by ultrasound and embryo staging
Pregnancy in plug-positive females was confirmed via ultrasound (US) the day before injection (verified at E6.5 for injections at E7.5, or at E7.5 for injections at E8.5). The pregnant female was placed in an induction box and anesthetized with an initial dose of 3%-4% Isoflurane. Once anesthetized, the female was moved to a pre-warmed (37 C) surgical table. To maintain anaesthesia, the snout was placed into a nose cone and the Isoflurane dose was lowered to 1.5%-2%. The female was placed on its back and all four paws were gently fixed to the table with surgical tape. The fur on the lower abdomen was removed by applying commercial Veet Hair Removal cream with a cotton swab. The abdomen was wiped clean with water to remove fur, and dried, before US-gel was applied. The USprobe was lowered into the gel and remained static, while the position of the female and table could be adjusted up/down, left/right using two wheels steering the surgical table. Once pregnancy was verified or refuted, US-gel was wiped off the abdomen, and the abdomen was cleaned with water. Confirmed pregnant females were placed into a separate, new cage and awakening was monitored over the first few minutes with a final check 15 minutes after removal from anesthesia. Embryo number and amniotic cavity size for E7.5 embryos was assessed on the day of injection. The mouse was anesthetized following the same procedure as described above. Using ultrasound, the left and right uterine horns were completely scanned. The individual amniotic cavities were assessed in higher magnification images and recorded. Individual cavity sizes were not measured, and assessment was rapid and based on visual appearance only. The female that had been assessed was labelled Viral Titers
Needle loading and petri dish preparation Glass capillaries were pulled and ground to a bevelled tip in-house, following the protocol of (Beronja et al., 2013). Needles were back-filled with mineral oil using a 1 mL syringe attached to a 25G needle. This also flushes the ground needle tip from any residual debris caused by the grinding. Prior to injection, needles were attached to the nanoinjector (Harvard Apparatus). The metal plunger was pushed out entirely, followed by loading a small amount of air into the needle, creating a small barrier in the form of a bubble between the mineral oil and the virus which was loaded thereafter. Virus (stored at -80 C) was brought to the injection room and kept on dry ice until loading. One vial was placed on wet ice, allowing the aliquot to thaw. The tube was centrifuged briefly in a table-top centrifuge and the entire volume (5mL) was transferred onto a piece of parafilm. The needle was lowered into the virus drop and was filled. Once the entire drop was loaded into the needle, an additional air bubble was created at the tip of the needle to prevent the fine tip from clogging. The nanoinjector with attached and loaded needle was always turned towards the back of the ventilated hood, away from the experimenter, when any other steps than injections themselves were performed. Petri dishes with a 2x2 cm round hole in the bottom were purchased together with pre-cut pieces of silicon membrane (both Visual Sonics). The elastic membrane piece was glued to the exterior of the hole and a ca.1cm long incision was made in the membrane. During surgery, the exposed uterus is pulled through this elastic opening, so the size of the incision should be adjusted depending on the embryonic stage to be injected.

Surgery and ultrasound-guided nanoinjections
Anaesthesia induction and maintenance as well as animal placement as described above. Both eyes were covered with eye gel to prevent drying and 0.1 mg/kg Buprenorphine pain killer was injected subcutaneously. With a pair of surgical scissors, a 1-2 cm vertical midline incision was made in the lower abdomen. Two cuts were made, one to open the skin and one to open the muscle layer underneath. With two pairs of surgical forceps, both uterine horns were carefully exposed, and the total number of embryos was recorded. Embryos were counted from the ovary to the cervix. For the injections, only the 3-4 embryos nearest the ovary of the left or right uterine horn were left exposed, while the remaining embryos were carefully pushed back into the abdominal cavity with a sterile cotton swab. The exposed embryos were pulled through the elastic membrane of the modified petri dish. Commercial play dough was used to fashion four feet to secure the petri dish on the surgical table above the female. The dish was then filled with sterile PBS until the embryos were immersed. To prevent leakage, the elastic bottom was pushed down with a cotton tip, so that it adhered to the wet, surrounding skin. An additional piece of play dough was placed on the needle-averted side of the embryos, in order to immobilize them and to prevent any unwanted movement during injections. The US-probe and needle were lowered into the PBS and aligned with the first embryo, by moving the table via the wheels. An image of the amniotic cavity was taken, and the diameter was measured, which allowed for calculation of the maximum volume that could be injected without causing resorption. The needle should be inserted through the uterine wall into the amniotic cavity in one single movement. After injection of the desired volume, the needle was kept inside the amniotic cavity for an additional five seconds to allow the entire volume to exit the needle and avoid spilling of virus out of the amniotic cavity upon withdrawal of the needle. Then the needle was removed, again in one single movement. Once all embryos were injected, the US-probe and needle were lifted out of the PBS and the stabilizing playdough was removed. Once the last embryo was injected and put back into the abdomen, the muscular layer was sutured using 6-0 prolene and the skin was closed with EZ-clips. The female was placed into a cage on a 37 C heating mat and recovery was carefully monitored, with an additional check 15-30 min after surgery. All work was carried out in a laminar airflow hood. In between mice, all surgical instruments (forceps, scissors etc.) were sterilized in a 200 C glass bead sterilizer. Disposable products (cotton tips, tissues) were replaced with fresh ones.

Tissue collection and fixation
Pregnant female mice were sacrificed in a CO 2 chamber. Whole uterine horns were exposed, and embryos were dissected out under a microscope in sterile, room-temperature PBS. Tissues were fixed in 4% formalin in PBS and dehydrated with a 30% sucrose buffer. Samples were embedded in OCT freezing medium on dry ice and stored at À80 C. 12 mm cryosections were prepared at À20 C on SuperFrost+ glass slides. Sections were kept at À20 C for short term storage, or À80 C for long-term storage. Values were normalized to the housekeeping gene transcript beta actin, by dividing the relative amount of the target by the relative amount of beta actin per sample. In experiments involving GFP sorted and unsorted conditions, the values were additionally normalized to the average of the GFPÀ condition. Primer sequences were determined using Primer3 (Table S1) and checked for amplification across exon borders. In silico PCR confirmed specific target amplification.

Cell Profiler
Separate images of DAPI and GFP (488) from specific brain regions were exported to jpeg file format and loaded into Cell Profiler (see Key resources table for details). A ready-made pipeline from Cell Profiler was used (link in Key resources table), in which the cell nuclei were identified as primary objects in both images separately. The diameter range for the size of nuclei was set by measuring the size of several nuclei in one picture (output in pixel) and for calculating the threshold, Otsu and a two-class thresholding method was used. Finally, the number of nuclei in the GFP picture was divided by the number of nuclei in the DAPI picture to obtain the percentage of GFP+ nuclei per section.
Western blotting E9.5 whole embryos or p10 cerebella were snap frozen and crushed with a mortar and pestle on dry ice. P9 cerebella were dissociated as described under ''Tissue Dissociation''. The cell pellet or tissue powder was reconstituted in RIPA lysis buffer with Complete Protease inhibitor cocktail. NE4C cells were washed with ice-cold PBS and complete lysis buffer was added directly to the wells. Protein concentration was determined by Bradford assay. 25mg of sample was mixed with Laemli buffer and boiled at 95 C for 8 min. Samples were separated on a 4-20% MiniProtean precast gel and wet-transferred overnight at 30V to an Amersham Hybond 0.45 PVDF blotting membrane. Membranes were blocked in 5% non-fat milk in TBST (Tris-buffered saline + 0.1% Tween20) for at least 1h at room temperature and incubated in primary antibody overnight at 4 C. The following day, membranes were washed 3x15min in TBST at room temperature. Secondary antibody was added for 1 hour at room temperature, followed by 3x15 min washing. Membranes were developed with SuperSignal West Femto Substrate. The relative protein concentration was calculated in Image Lab using the individual band volumes. Membranes were rinsed in H 2 O to remove chemiluminescent substrate. For stripping and re-probing, the membrane was incubated in Restore Western Blot Stripping Buffer for 15 minutes at room temperature, followed by 3x15 min washing in TBST. Antibody incubation (primary and secondary) as well as blot development were performed as described above. Primary antibodies used were: Beta-actin (1:1,000); Gapdh (1:1,000) and Sptbn2 (1:200). Secondary antibodies were used 1:5,000.

QUANTIFICATION AND STATISTICAL ANALYSIS
Quantification of GFP + nuclei with Cell Profiler was performed on 20x tile images of brain or spinal cord sections. One section every 134mm was used for acquisition and quantification, resulting in 8-10 sections in total per embryo (covering fore-to hindbrain region or thoracic to sacral region in spinal cord). Figures 2G and S2C show the mean number of GFP + nuclei in 2-4 sections per brain region from three different embryos, while in Figure S1E each data point represents the total mean of GFP + nuclei per brain per embryo. In Figure S2 and 3A, every data point represents the number of GFP + nuclei on one section. Data were analysed using GraphPad Prism 9 and are presented as mean ± SD. When more than two groups were compared, a one-way ANOVA test was performed with the following parameters: No matching or pairing of replicates across groups, assuming normal distribution (Gaussian distribution) and assuming that groups do not have equal variances (Brown-Forsythe and Welch ANOVA test). In addition, multiple comparisons were done to assess differences between groups. To compare two groups only, a two-tailed Student's t test was performed. Statistical method used is indicated in the individual figure legends. Statistical significance was defined as *p < 0.05, not significant was indicated as ns.