Prenatal alcohol exposure disrupts Sonic hedgehog pathway and primary cilia genes in the mouse neural tube

2.1. Animals

Male and female adult C57BL/6J mice (mus musculus) were obtained from Jackson Laboratories (Stock No: 000664; Bar Harbor, ME). Males were housed singly and females were housed in groups of up to five per standard polycarbonate cage with cob bedding, a shelter, and nesting material. All mice had ad libitum access to food (Prolab Isopro RMH 3000, LabDiet, St. Louis, MO) and water, and were maintained on a 12:12 hr light/dark cycle. For mating, 1–2 female mice were placed into the cage of a male for 1–2 h and were checked for a vaginal plug to confirm copulation. E0.0 was defined as the beginning of the mating session when the plug was found. Mated females were weighed and housed in a clean cage with up to 5 mice. All experimental procedures and euthanasia protocols were approved by the Institutional Care and Use Committee (IACUC) at University of North Carolina (approval #18–203) and accordance with the National Institutes of Health guide for the care and use of laboratory animals (NIH Publications No. 8023, revised 1978).

For behavior studies, male Kif3a^+/− (B6.129-Kif3a^tm1Gsn/J, Stock No: 003537; Jackson Laboratories, Bar Harbor, ME) and female C57BL/6J mice were obtained, housed, and mated as described above to produce Kif3a^+/+ and Kif3a^+/− offspring. Generation of Kif3a^−/− mice was avoided as this mutation is embryonically lethal. All dams were singly housed in clean cages on E15. Alcohol- and vehicle-treated litters (see below) were housed with their dams, culled to a maximum of 8 pups/litter at postnatal day (PD) 3, weighed, ear marked, and tail-clipped for genotyping on PD14, and left undisturbed until weaning at PD28 when they were housed in same sex groups with their littermates. All the mice from each litter were tested on behavioral experiments conducted between PD28–33 by an experimenter who was unaware of the prenatal treatments and Kif3a genotype.

2.2. Neurulation-stage alcohol exposure (NAE)

On E9.0, pregnant dams were administered two doses of ethanol (25% vol/vol ethyl alcohol, Pharmaco-Aaper, Brookfield, CT, at a dose of 2.9 g/kg) in Lactated Ringer’s solution 4 hr apart via intraperitoneal (i.p.) injection. These mice were designated as the NAE group. A separate group of mice were administered an equal volume of vehicle (1.5 ml/100 g body weight). This model of alcohol exposure has been previously shown to result in maternal blood alcohol concentrations of ~400 mg/dl [27]. For molecular studies, dams were humanely sacrificed via CO₂ followed by cervical dislocation either 6, 12, or 24 hr after the first ethanol injection. Embryos were dissected out and placed into cold RNase-free Dulbecco’s saline solution. For all assays, embryos were stage-matched based on somite number (E9.25: 21–22 somites, E9.5: 24–25 somites, E10: 30–31 somites). A maximum of two embryos per litter were used in each assay to minimize litter effects.

2.3. Embryo immunohistochemistry

The chorion and amnion were removed from all embryos and placed into 4% paraformaldehyde for ~72 h. Stage-matched embryos were then processed for paraffin-embedding on a Leica tissue processing station and sectioned at 7 μm on a microtome. A total of 6–10 sections per embryo, encompassing the rostroventral neural tube (RVNT), were processed for immunohistochemistry. Briefly, paraffin residue was dissolved in xylenes and sections were rehydrated in ethanol washes. Slides were quenched in 10%H₂O₂/90% methanol for 10 min and incubated for 30 min in 10% Citra Plus Buffer (BioGeneX, Fremont, CA) in a steam chamber. The slides were then blocked in normal goat serum/Triton-X/bovine serum albumin blocking solution for 1 hr and incubated overnight with primary antibody (Arl13b, 1:500, NeuroMabs, University of California-Davis) at 4°C. Slides were then incubated with secondary antibody (anti-mouse Alexafluor 488, 1:1000, Thermofisher, Waltham, MA) for 2 hr at room temperature and cover slipped with Vectashield HardSet Antifade Mounting Medium with DAPI (Vector, Burlingame, CA).

2.4. Confocal imaging and image quantification

Cilia were imaged on a Zeiss 880 confocal microscope with a 40x oil lens. Image stacks were obtained at a step of 0.46 μm between images. Stacks were then compressed into a single image containing a depth color code with Fiji (ImageJ) software [28]. Cilia within the known volume of the RVNT were counted with the Cell Counter plug-in and expressed as number of cilia per 100 μm³. RVNT volume and cell layer width were determined using images of the same sections taken with a 10x lens. The DAPI-labeled cell layer was traced to obtain volumetric measurements and the cell layer width was measured at multiple points within each RVNT (2 measurements per slide) using Fiji software.

2.5. Gene expression assays

The RVNT of stage-matched embryos were dissected and placed into lysis buffer and stored at −80°C until processing. RNA was isolated from the supernatant using the RNeasy Plus Micro Kit (Qiagen, Valencia, CA). Nucleic acid concentration and quality were determined using the Qubit 3.0 fluorometer and Nanodrop 2000 (Thermofisher, Waltham, MA). Generation of cDNA used a consistent starting amount of RNA across all embryos (100 ng). Multiplex qRT-PCR was used to determine alcohol-induced gene expression changes (n = 6–10 embryos/treatment). Taqman probes (Invitrogen) and Taqman Multiplex PCR mix (Applied Biosystems) were used to determine expression changes in the following genes: Shh (Mm00436528), Gli1 (Mm00494654), Gli2 (Mm01293117), Ccdn1 (Mm00432359), Ccdn2 (Mm00438071), Fgf15 (Mm00433278), Cep41 (Mm00473478), Hap1 (Mm00468825), Rilpl2 (Mm01199587), Evc (Mm00469587), Nek4 (Mm00478688), and Dpcd (Mm00620237). For all assays, 18s (Mm03928990) or Pgk1 (Mm000435617) were used as reference genes. Reference genes were confirmed to be unaffected by prenatal treatment in separate embryos prior to use in these experiments. However, 18s expression showed high baseline variability in the E9.25 embryos, leading to the use of Pgk1 as a reference gene for this time point. All reactions were run in triplicate and amplicon specificity was confirmed with gel electrophoresis.

2.6. Western Blot

RVNT’s of each litter (n = 5–10 samples per tube) were pooled and lysed in RIPA buffer with 1X Halt protease and phosphatase inhibitors (Thermofisher, Waltham, MA) in order to obtain sufficient protein for western blot analyses. For E9.25, 5 vehicle and 6 NAE litters were used, and for E9.5, 6 vehicle and 7 NAE litters were used. Protein concentrations were determined with the Micro BCA kit (Thermofisher, Waltham, MA) and colorimetrics measured at 540 nm on a spectrophotometer. 15 μg of sample with 4X Laemmli sample buffer were added to each lane of a tris-glycine gel in 1X TGX running buffer and run at 150V for 45 min. Protein weights were determined with the Dual Color protein ladder (Bio-Rad, Hercules, CA). Proteins were transferred to a membrane using the iBlot2 system and incubated overnight at 4°C with primary antibody (anti-Gli3, 1:500, AF3690, R&D Systems, Minneapolis, MN). Membranes were then incubated in secondary antibody (anti-goat Alexafluor 680, 1:10,000, Thermofisher, Waltham, MA) for 1 hr at room temperature and imaged on a Licor Odyssey scanner. GAPDH (1:3000, 14C10, Cell Signaling, Danvers, MA) was used as an internal loading control. The relative intensities of the Gli3 protein bands (Gli3^FL [~190 kDa] and Gli3^Rep [~83 kDa]) were analyzed with ImageJ Gel Analyzer software and normalized to internal control bands. The ratio of Gli3^FL: Gli3^Rep forms was calculated and expressed as a percentage of total Gli3.

2.7. Adolescent behavioral procedures

All the mice from each litter generated from the mating of a Kif3a^+/− male and C57BL/6J female (n = 16 litters vehicle-exposed and 19 litters NAE) were tested on behavioral experiments conducted between PD28–33 by an experimenter unaware of the prenatal treatments and Kif3a genotype. The experiments were conducted in the UNC Behavioral Phenotyping Core during the light portion of the 12:12hr light:dark schedule. All mice were tested in the following testing order: rotarod trials 1–3; elevated plus maze (EPM); open-field; rotarod trials 4 and 5. The EPM was tested before the open field to minimize potential carry-over effects, as the EPM is more sensitive to the effects of prior testing history [29, 30]. Technical issues during testing resulted in two male animals being excluded from the EPM and one male from open field analyses.

The rotarod (Ugo-Basile, Stoelting Co., Wood Dale, IL) measured the latency to fall off or rotate around the top of a dowel which progressively accelerated from 3 rpm to 30 rpm during the maximum of a 5-min test, conducted during three repeated trials on the first day of testing and two repeated trials on the second day of testing. Each trial was separated by ~45 sec. The EPM was 50 cm above the floor and contained two open arms (30 cm length, 220 lux) and two closed arms (20 cm high walls, 120 lux). During the 5-min test, an observer recorded the number of entries and time spent in each of the arms. These data were used to calculate the percent of open arm time [(open arm time/total arm time) x 100], the total arm entries, and the percent of open arm entries [(open arm entries/total arm entries) x 100]. The open field (41 × 41 × 30 cm) was illuminated (120 lux at the edges, 150 lux in the center), housed within a sound-attenuated chamber and equipped with upper and lower grids of photobeams for the detection of horizontal and vertical activity and position of the mouse (Versamax System: Accuscan Instruments, Columbus, OH). The open field was conducted over 60 min and the primary measures were horizontal activity, total distance traveled, time spent in the center of the chamber, and distance traveled in the center. Number of rears and rotations were also recorded. Open field data were analyzed as four separate 15-min epochs (i.e. min 0–15, min 16–30, min 31–45, and min 46–60), based on prior findings that the initial minutes of the open field test are most sensitive to developmental drug exposure [11, 31].

2.8. Adolescent brain measurements

Following completion of behavioral tasks on PD37, mice were deeply anesthetized with tribromoethanol anesthesia and transcardially perfused with 1X PBS followed by 10% formalin. Brains were removed and stored in formalin for at least one week. Prior to paraffin embedding, fixed whole brains were cleared by rocking at room temperature in 1X PBS for one week and 70% ethanol for one week (each solution changed daily). Brains were then cut into two sections, the frontal lobes and cerebellum, and paraffin-embedded on a Leica tissue processing station. Tissue was then embedded in paraffin block and sectioned at 10 μm on a microtome. Every 5^th slide (containing 3–4 sections per slide) throughout the entirety of the cortex was stained using cresyl violet. One section per slide was imaged and the ventricles traced using Fiji/ImageJ [28]. Total area of the ventricles was summed per section and averaged across the entire brain. For midline brain width and medial height, lines were drawn across images of one section of each slide throughout the cortex and measured in Fiji. Midline brain width was measured at the widest part of the cortex (dorsal/ventral bregma ~3.0) and height was measured as close to the midline as possible (medial/lateral bregma 0.25); measurements were averaged across the entire brain.

2.9. Statistical analyses

Cilia number, RVNT volume, and Western blot band intensity were analyzed using unpaired t-tests comparing prenatal treatment (NAE vs. vehicle) at each time point with Welch’s correction when necessary. The Benjamini-Hochberg correction was used if multiple unpaired t-tests were run within an experiment (false discovery rate = 0.10) [32]. Raw p-values shown in tables and in the text remain statistically significant following correction unless otherwise noted. Multiplex qRT-PCR gene expression data are expressed as log2 fold change calculated using the 2^(−ΔΔCT) method [33]. Data were analyzed using either unpaired t-tests with corrected p-values (cilia-related genes) or two-way ANOVAs (Treatment x Time Point) with corrected t-tests run as post hocs for analyses with significant interactions (Shh path genes, Gli3, cell cycle genes). For behavior assays, litter means were analyzed for each treatment and genotype to control for between litter effects [34]. Males and females were analyzed separately, because our previous studies demonstrated sex differences on these measures [12, 35]. EPM data were analyzed using two-way ANOVAs (treatment x genotype) and open field and rotarod data were analyzed with three-way ANOVAs (treatment x genotype x time bin or trial). For the behavioral analysis, we predicted, a priori, that any effects of NAE or Kif3a^+/− would be exaggerated when these treatments were combined, in other words, that the NAE Kif3a^+/− mice would have the largest difference from the vehicle-treated Kif3a^+/+ mice. Ventricle area measurements were analyzed using a three-way ANOVA (treatment x genotype x bregma), while width and height measurements were analyzed with two-way ANOVAs (treatment x genotype). Post hoc analyses were conducted when appropriate. Statistical differences were considered significant at an adjusted p-value threshold of 0.05.

3.1. NAE downregulates expression of the Shh pathway in the RVNT

Previous work has demonstrated that the shape and size of the ventricles, pituitary, and hypothalamus are altered following NAE [11–14], suggesting that during neurulation the RVNT (from which these structures are derived) is particularly vulnerable to alcohol. Dysregulation of Shh signaling within the RVNT is likely a mechanism for structural changes to midline brain regions following NAE, however, this hypothesis has never been tested [36]. To determine whether NAE alters the Shh pathway in the RVNT, we administered two doses of 2.9 g/kg alcohol during mid-neurulation (E9.0) to model binge-like alcohol exposure. We collected the RVNT of stage-matched embryos 6, 12, and 24 hr later (E9.25, E9.5, and E10, see Section 2.2 for somite numbers) and analyzed Shh, Gli1, and Gli2 gene expression. For each gene, two-way ANOVAs were run using the factors of Treatment x Time point (statistics in Table 1), with post hoc t-tests between NAE and Veh groups performed for each time point in the event of a significant interaction. First, we found main effects of Treatment and Time Point and a significant interaction for Shh gene expression. Post hocs revealed that NAE significantly downregulated Shh at both the 6 and 12 hr time points (t(14) = 2.741, p = 0.0159 and t(14) = 3.608, p = 0.003, respectively; Fig 1A; Table 1), extending previously reported reductions in Shh following gastrulation-stage alcohol [9, 10] and demonstrating that NAE also impairs Shh levels in the RVNT, but is independent of concurrent apoptosis in this region [5]. Following Shh activation, Gli1 is rapidly upregulated, whereas Gli2 is constitutively expressed. However, without Shh present, Gli2 is cleaved and tagged within the primary cilium for degradation. Activation of Smoothened (Smo) by Shh deactivates this cilia-mediated cleavage of Gli2 and allows it to be shuttled outside of the cilia to activate downstream genes [37]. A main effect of treatment was also found for Gli1, with NAE embryos showing reduced Gli1 expression compared to controls (Table 1, Fig 1A). No effects of NAE were found for Gli2 expression. In addition, we measured the relative amounts of the two forms of Gli3 protein 6, 12, and 24 hr following the beginning of NAE. In the absence of Shh and with normal cilia functioning, Gli3 is found predominantly in the cleaved repressor form (Gli3^Rep). Upon Shh pathway activation, the balance of Gli3^Rep to the full-length activator form (Gli3^FL) is tipped slightly to the activator form to allow Gli1 and Gli2 to proceed with normal gene transcription, including cell cycle genes such as the cyclin family, as well as having a positive feedback effect on Shh itself [38]. A two-way ANOVA revealed a significant Treatment x Time interaction (statistics in Table 1). Post hoc t-tests found that NAE led to a significantly shifted ratio at the 12 hr time point, with a higher percentage of Gli3^Rep compared to vehicle treatment (65.27% Gli3^Rep in NAE vs. 54.7% Gli3^Rep in vehicle-treated; Fig 1B; t₍₁₁₎ = 2.67, p = 0.022). The shift in Gli3 to favor the repressor form at this time point coincides with significant reductions in Shh and Gli1 gene expression, as well as smaller RVNT volumes in NAE embryos (Fig 1). The ratio of Gli3^FL:Gli3^Rep in the RVNT did not significantly differ from control levels 6 or 24 hr post-NAE (Fig 1B). No difference in total amount of Gli3 was observed between treatment groups at any time point and the amount of Gli3^Rep did not significantly differ between the E9.25, E9.5, or E10 vehicle-treated groups (F_(2,14) = 0.967, p = 0.401). Changes to Shh pathway transduction could alter the developmental trajectory of craniofacial and CNS regions derived from the RVNT by reducing normal cell proliferation, resulting in the midline defects observed in mice following NAE.

3.2. NAE downregulates cell cycle gene expression and decreases RVNT volume

One of the outputs of the Shh signaling pathway is the expression of cell cycle genes. Based on the observed decreases in Shh and Gli1 gene expression and altered Gli3^FL:Gli3^Rep following NAE, we hypothesized that NAE would also lead to altered expression of genes regulating cell proliferation. To assess changes in Shh-mediated gene expression, we analyzed three cell proliferation genes downstream of Shh, Cyclin D1 (Ccnd1), Cyclin D2 (Ccnd2), and Fibroblast growth factor 15 (encoded by Fgf15; homologous to Fgf19 in humans), in the RVNT 6, 12, and 24 hr following NAE. Ccnd1 and Ccnd2 regulate progression of the G1 phase of the cell cycle [39] and are regulated by the Shh pathway Gli family of transcription factors [40, 41]. Furthermore, Ccnd1 is important for expanding ventral neural precursors in the early mouse diencephalon [42]. In accordance with the alcohol-induced reductions in Shh, significant main effects of treatment were found for Ccnd1 and Ccnd2 (Fig 2A, Table 1), with NAE embryos displaying reduced expression compared to vehicle-treated embryos. Fgf15 depends on Shh signaling and largely overlaps with Gli1 expression patterns early in development [42, 43]. Functionally, Fgf15 promotes cell cycle exit and differentiation of neuronal precursors [44]. Main effects of both treatment and time point were seen for Fgf15 expression (Fig 2A; Table 1), however the NAE-induced reduction in Fgf15 was most pronounced on E9.25.

Given the disruption of cell cycle-related gene expression, we next analyzed the volume of the RVNT 6, 12, and 24 hr after NAE to examine how the size of this region changed over time and in response to alcohol (Fig 2B). Images taken from the RVNT were obtained at 40x for volumetric analyses; section thickness was also measured at 40x. A significant time (hours post-exposure) x treatment interaction was observed (F_(2,39) = 6.164, p = 0.0047), as well as main effects of both time (F_(2,39) = 22.08, p < 0.0001) and treatment (F_(2,39) = 5.018, p = 0.031). As expected, the volume of the RVNT increased across time, with E10 embryos having significantly larger RVNTs compared to both E9.25 (p < 0.0001) and E9.5 (p = 0.0002), demonstrating the exceptional growth of the neural tube over this period. There was a trend towards the RVNT of E9.5 embryos being larger than E9.25 (p = 0.065), however this relationship was not statistically significant. Importantly, RVNT volume was significantly smaller in NAE compared to vehicle-treated embryos (0.197±0.01 mm³ vs. 0.3224±0.04 mm³) on E9.5 (p = 0.0085) and tended to be smaller on E10 as well, though this difference did not reach statistical significance (p = 0.1). The thickness of the RVNT cell layer was also assessed, however while the cell layer did grow thicker across time (F_(2,39) = 9.142, p = 0.0006), there was no significant effect of treatment (Fig 1C). E10 embryos had increased widths compared to both E9.25 (p = 0.0049) and E9.5 (p = 0.0012). The increase in cell layer width corresponds to the larger overall volume at E10. The rapid changes in RVNT volume across the first 24 hr post-alcohol exposure indicate that significant structural changes are taking place and encourages further investigation into all factors by which alcohol could affect growth, including alterations to cell proliferation. Additionally, while NAE-induced dysmorphologies are evident later in development [11–14], no overt gross anatomical differences were observed between vehicle and NAE-treated embryos at E9.5 (S2 Fig). Thus, alcohol-related changes in tissue volume and shape are likely subtle at this time point and region- and timing-specific. Even though the volumetric changes induced by NAE were transient, a brief period of growth inhibition within this region of the neural tube during such an important and dynamic period of early development could have long-lasting consequences on face and brain structure.

3.3. NAE alters expression of key cilia-related genes, but not cilia density in the RVNT

Shh transduction occurs within the axoneme of primary cilia, nearly ubiquitous structures of mammalian cells that play a particularly important role during embryonic development. Dysregulation of RVNT primary cilia poses a risk to the normal development of brain regions that arise from this area of the neural tube, namely midline structures such as the hypothalamus, septum, pituitary, and preoptic area [45], many of which have been shown to be altered by NAE. Additionally, midline anomalies such as hypertelorism are hallmarks of genetic ciliopathies and have also been reported in a subset of patients with heavy prenatal alcohol exposure [4], suggesting an overlapping etiology between ciliopathies and prenatal alcohol.

Based on the reduction in Shh and downstream cell cycle gene expression, we hypothesized that we might find a coinciding NAE-induced change in either 1) the number of primary cilia or 2) cilia-related genes in the RVNT. First, we examined the direct impact of NAE on primary cilia in the RVNT by analyzing cilia density. Primary cilia were labeled for the cilia-specific small GTPase Arl13b at 6, 12, or 24 hr following NAE (Fig 3A) and the number of cilia within the known volume of the RVNT were analyzed using 3D image stacks from confocal images. A significant time (hours post-exposure) x treatment interaction was found (F_(2,39) = 3.642, p = 0.036), as well as a main effect of time (F_(2,39) = 18.93, p < 0.0001) (Fig 3B). No main effect of treatment was observed. No significant differences were seen between vehicle and NAE embryos at any time point, though there was a trend towards an increase in cilia density at the E9.5 (12 hr post-exposure) time point (p = 0.087). The density of cilia in the RVNT significantly decreased over time, as there was a higher density of cilia at E9.25 compared to both E9.5 and E10 (p < 0.0001 for both time points). This reduced density is likely due to the substantial growth of the neural tube during neurulation (Fig 2B–C), however the lack of a treatment effect precludes alcohol-induced effects on cilia number as the mechanism of action of reduced cell cycle gene expression.

We further investigated primary cilia homeostasis and stability by analyzing cilia-related genes 6 or 12 hr following NAE. Alcohol-induced dysregulation of cilia-related genes has been reported in two previous studies from our lab [46, 47]. RT-PCR was performed on six genes of interest (Hap1, Rilpl2, Cep41, Nek4, Evc, and Dpcd) to investigate whether cilia-related genes changes persist to the 12 hr time point (Table 2, Fig 3C). These genes were chosen due to their known roles in important and varied ciliary processes including ciliogenesis, ciliary protein trafficking, or cilia stability through maintenance of post-translational modification and/or genes that have mutations associated with human genetic ciliopathies. Based on our a priori hypothesis that cilia-related genes would be most affected at the 6 hr time point (E9.25), each time point was analyzed separately. At E9.25, NAE upregulated two of the genes, Hap1 and Rilpl2, in the RVNT (Table 2; Fig 3C). However, expression levels of these genes returned to baseline by E9.5 (Table 2, Fig 3C). Three other genes (Cep41, Evc, and Dpcd) showed marked downregulation 6 hr following NAE (E9.25) (Table 2, Fig 3C). Dpcd remained significantly downregulated 12 hr after NAE; however, expression of Cep41 and Evc had returned to control levels. These genes have prominent roles related to cilia function, stability, and ciliogenesis. The NAE-induced dysregulation of these genes precedes and coincides with the downregulation of important morphogenic and cell proliferation pathways as well as reductions in RVNT volume.

3.4. Partial knockdown of cilia gene Kif3a interacts with NAE to affect adolescent behavior

To further test the hypothesis that disrupted cilia function is a mechanism for the consequences of NAE, we utilized a transgenic mouse line with a partial deletion of Kif3a, a gene that codes for an intraflagellar transport protein. Full deletion of Kif3a, is a well-characterized ciliopathy model [24, 26, 48, 49], and if NAE interacts with cilia function, then Kif3a heterozygosity would both phenocopy and exacerbate NAE. We focused on a NAE effect that we have previously shown to be robust and reproducible, behavioral change during adolescence [11]. Adolescent male and female mice exposed to E9.0 alcohol were tested on a battery of behavioral tasks, but since the largest behavioral changes occurred in males, we discuss behavior from male, rather than female, mice in detail (for female data refer to Fig S3B, S4A–D; S1 Table). Sex-specific effects of early gestational alcohol have been previously reported to emerge in adolescence and adulthood and likely result from postnatal sex differences, as opposed to embryonic effects of alcohol.

We first examined motor coordination, using an accelerating rotarod performance across five repeated trials. Although rotarod performance for all mice improved from the first to the fifth trial (F_(4,192) = 28.4, p < 0.0001), NAE persistently impaired motor coordination, as revealed by a main effect of NAE, regardless of Kif3a genotype (F_(1,48) = 5.1; p = 0.028; Fig S3). Previous research has suggested that the etiology of cerebellar deficits caused by mid-stage NAE is likely alcohol-induced apoptosis in the rostral rhombic lip, from which the cerebellum arises [5]. The lack of an NAE x genotype interaction indicates that primary cilia are not critical for these motor incoordination effects. While it is possible that alcohol earlier or later in development could have more direct effects on primary cilia-mediated mechanisms in the cerebellum, the middle of neurulation appears to be a critically sensitive period for cerebellar defects, as our previous research has suggested that alcohol administered on E7 (O’Leary-Moore & Sulik, unpublished observations) or E8 [11] has a less pronounced impact on motor coordination.

However, when we examined elevated plus maze (EPM) exploration, a two-way ANOVA found a significant genotype x treatment interaction (F_(1,49) = 6.9; p = 0.01) on total arm entries. Vehicle-treated Kif3a^+/− and NAE Kif3a^+/+ mice were more active than were vehicle-treated Kif3a^+/+ mice (p = 0.006 and p = 0.04, respectively) (Fig 4A), while NAE Kif3a^+/− had comparably high levels of activity but were not statistically different from vehicle-treated Kif3a^+/+ mice (p = 0.061). This result replicates our previous findings on early NAE [11] and indicates that Kif3a heterozygosity phenocopies a locomotor hyperactivity caused by NAE. Supporting the hypothesis that NAE and Kif3a^+/− also affect anxiety-like behavior, there was a significant genotype x treatment interaction on the percentage of open arm time (F_(1,49) = 3.9; p = 0.05) (Fig 4B), but not open arm entries (Fig 4C). Post hoc analysis revealed that both vehicle-treated and NAE Kif3a^+/− mice spent a greater percentage of time on the open arms than did the vehicle-treated Kif3a^+/+ mice (p < 0.001 and p = 0.01, respectively). Although NAE in Kif3a^+/+ tended to phenocopy Kif3a^+/− on the open arm time, this effect was not significant in the post hoc analyses. To test whether NAE alone affects open arm time in the Kif3a^+/+ mice and confirm our previous results using a two-sample t-test [11, 12], we ran this same statistical analysis (t₍₂₀₎ = 3.0; p = 0.007; Fig 4B) and found that that early NAE increases the percent of time on the open arms. Overall, these results indicate that NAE and Kif3a heterozygosity phenocopy each other’s effects on the EPM.

Since the EPM is a relatively brief exposure to a complex novel environment and our procedure did not assess habituation to the maze, we further measured exploratory behavior and activity in an open field test across a 1-hr period. A three-way mixed ANOVA revealed a significant genotype x treatment x time bin interaction on overall horizontal activity (F_(3,141) = 4.3; p = 0.006) (Fig 4D). Post hoc analysis revealed that NAE Kif3a^+/− mice were more active than vehicle-treated Kif3a^+/+ mice during each of the 15-min time bins (p = 0.002, 0.001, 0.032, and 0.046, respectively), and were more active than NAE Kif3a^+/+ mice during the 2^nd and the final 15-min time bins (p = 0.018 and 0.006, respectively). NAE Kif3a^+/− mice were also more active than vehicle-treated Kif3a^+/− mice during the final 15-min time bin (p = 0.013). Since the persistently high activity of the NAE Kif3a^+/− mice suggests impaired habituation, we calculated a habituation index, expressing data from each 15-min time bin as a percentage of the initial time bin (i.e. min 0–15). A significant 3-way interaction (genotype x treatment x time bin; F_(3,141) = 2.0; p = 0.039) revealed that while all groups reduced their activity over the hour, in the final time bin, NAE Kif3a^+/− mice were significantly less habituated than were the vehicle-treated Kif3a^+/− mice and the NAE Kif3a^+/+ mice (p = 0.02 and 0.03, respectively). These groups had an initial bout of heightened activity, relative to the vehicle-treated Kif3a^+/+ mice, but NAE Kif3a^+/− mice maintained a higher level of activity throughout the test.

There were significant main effects of NAE on several specific measures of activity, including repeated beam breaks (F_(1,47) = 4.1; p = 0.05, S2 Table), center distance (F_(1,47) = 9.4; p = 0.004, Fig 4E), and center time (F_(1,47) = 12.5; p = 0.001, Fig 4F). Kif3a^+/− genotype also had significant effects upon repeated beam breaks (F_(1,47) = 5.2; p = 0.03, S2 Table), center distance (F_(1,47) = 10.6; p = 0.002, Fig 4E), and center time (F_(1,47) = 7.4; p = 0.009, Fig 4F) Both NAE and Kif3a^+/− genotype significantly affected the number of rears, rotations, and total distance (statistics shown in S2 Table). On each of these measures, NAE Kif3a^+/− mice were the most active treatment group indicating an additive effect of Kif3a genotype and NAE.

To further illustrate the heightened activity of the NAE Kif3a^+/− mice relative to the other groups, as predicted by our hypothesis, we analyzed overall horizontal activity (Fig 4D), center distance (Fig 4E), and center time (Fig 4F) as 60-min totals and used Bonferroni post hoc tests to compare NAE Kif3a^+/− to NAE Kif3a^+/+ mice (effect of genotype within NAE mice) or vehicle Kif3a^+/− mice (effect of NAE within the heterozygotes), vehicle Kif3a^+/− to vehicle Kif3a^+/+ (effect of genotype alone), and NAE Kif3a^+/+ to vehicle Kif3a^+/+ mice (effect of NAE within the wild-types). The NAE Kif3a^+/− mice were more active than their NAE Kif3a^+/+ littermates (p = 0.001 for horizontal activity, 0.002 for center distance, and 0.006 for center time), and the vehicle Kif3a^+/− mice (p = 0.003 for horizontal activity, 0.015 for center distance, and 0.003 for center time). There were no significant effects of NAE or Kif3a^+/− alone. The heightened overall activity levels, as well as activity specific to the center of the open field which may reflect altered anxiety-like behavior, confirm our prior results from early NAE that open field activity is a sensitive index for the effects of early gestational alcohol exposure [11, 12]. The greater sensitivity of the Kif3a^+/− in the open field versus the EPM is likely due to differences in the test duration because the differences between treatment and genotype groups appear to become greater as exploration patterns adapt to the environment.

Brain width and ventricle area were also analyzed in a subset of adolescent male mice (n = 4–5 per group) chosen based on performance closest to the group means on open field horizontal activity. Neither genotype nor prenatal treatment significantly affected ventricle area, though Kif3a^+/− mice tended to have larger ventricles (F_(1,13) = 3.767, p = 0.0743; Fig S5A). A genotype-treatment interaction was found for midline brain width (F_(1,13) = 9.206, p = 0.0096; Fig S5B), as well as a main effect of treatment (F_(1,13) = 27.55, p = 0.0002). Dunnett’s multiple comparison test was used to compare each group to vehicle-treated Kif3a^+/+ animals, and it was determined that the brains in all three other groups were narrower (vs. Kif3a^+/+ NAE: p = 0.0001; vs. Kif3a^+/− Veh: p = 0.0308; vs. Kif3a^+/− NAE: p = 0.0018). In addition, a main effect of genotype was found for medial brain height, with Kif3a^+/− mice having shorter medial height compared to Kif3a^+/+ mice (F_(1,13) = 6.332, p = 0.0258; Fig S5C). Thus, the partial deletion of Kif3a resulted in smaller brains in adolescence that coincided with behavioral abnormalities.

4.1. NAE disrupts the Shh pathway and cilia function in the neural tube

These studies confirm that NAE downregulated Shh pathway expression, similar to other FASD models targeting early points of gestation [6–10]. We observed decreased Shh and Gli1 expression in the RVNT of NAE embryos during the first 24 hr post-exposure. These reductions coincided with increased levels of the Gli3^Rep form 12 hr following NAE. These data are the first to report altered Shh signaling following alcohol exposure during neurulation. While Gli2 gene expression was not affected by NAE at either time point, it is possible that the ratio of the different forms of Gli2 was altered by NAE while not impacting levels of mRNA, similar to what was found for Gli3. Like Gli3, Gli2 is found in full-length activator and cleaved repressor forms, though studies have suggested that Gli2 acts primarily as a transcriptional activator in the mouse neural tube [52–54], meaning the majority of Gli2 will be found in the full-length form. However, currently available antibodies cannot reliably identify the various forms of Gli2, in part due to the dearth of antibodies that bind to the correct epitope of the protein to allow for detection of the cleaved repressor (N-terminus). It will be possible to address this question once validated antibodies targeting different amino acid sequences of mouse Gli2 become available.

The molecular mechanisms of action of alcohol on the Shh pathway remain unclear. Alcohol could act directly on the Shh pathway through interference with upstream regulator proteins, such as Hoxd’s [55], Sox2/3 [56], or Hand2 (best described in the embryonic limb bud) [57], or the Shh co-receptor Cdon [58], disruption of cholesterol [59] or other proteins necessary for Shh modulation, or activation of pathway inhibitors such as PKA [8] or Tbx2 [60]. Our data suggest that alcohol could indirectly disrupt the Shh pathway as a downstream consequence of the observed dysregulation of primary cilia genes and their related functions. Downregulation of Gli proteins can cause a negative feedback loop in Shh expression as the Gli proteins target multiple members of the Shh path, including Ptc [61]. The possibility that alcohol initiates a negative feedback loop of the Shh pathway is supported by the concomitant reduction in expression of pro-proliferative genes known to be downstream of Gli’s, Ccnd1, Ccnd2, and Fgf15. Expression of other targets of Gli-mediated transcription may have been impacted as well. The unique expression profiles of these genes during the 24 hr post-NAE indicate possible differences between these molecules in alcohol sensitivity or the presence of compensatory mechanisms in the neural tube. The impact of reduced pro-proliferative genes was observed as a smaller RVNT volume at E9.5. Since previous work from our laboratory has shown little to no excessive apoptosis in the rostral basal or floor plates of the neural tube following E9.0 alcohol exposure [5], the most likely explanation for the reduced RVNT volume is decreased cell proliferation. The lack of increased apoptosis in the rostroventral portion of the neural tube following NAE stands in stark contrast to the pronounced apoptosis throughout the neuroectoderm following exposure at earlier time points and suggests region- and timing-specific mechanisms of alcohol-induced tissue damage. While not quantitatively measured in this study, the RVNT’s of NAE embryos displayed some shape differences compared to controls, as can be seen in Fig 2. Since neural patterning is a primary function of Shh, altered cell migration in the RVNT could be another consequence of disrupted Shh signaling. It is also possible that reduced activation of the Shh pathway in the ventral neural tube allows for redistribution of morphogens and increased expression of Wnt within the dorsal neural tube. Since the dorsal neural tube also contributes to craniofacial development through Wnt and Bmp signaling, an imbalance in the morphogenic gradients would likely result in abnormal growth trajectories of regions arising from these portions of the neural tube.

The Shh pathway requires properly functioning primary cilia and our data demonstrating NAE-induced disruptions to cilia gene expression support that NAE impairs primary cilia function. While cilia density was not affected in the current study, NAE resulted in changes to cilia-related genes in the RVNT that emerge within the first 6 hr following alcohol exposure and extend to the Gli family of transcription factors after 6–12 hr. These data demonstrate that NAE rapidly dysregulates primary cilia gene expression and identifies ciliogenesis, protein trafficking, and maintenance of cilia stability as targets of NAE. At E9.25, NAE upregulated Hap1 and Rilpl2 in the RVNT (Table 2; Fig 3C), two genes associated with ciliogenesis [62, 63]. Hap1 is named for its role in Huntington’s disease [62] and has been shown to mediate ciliogenesis through interaction with the Huntingtin protein (Htt). While little is known regarding ciliopathies in patients with Huntington’s disease, patients with Huntington’s, and mice with introduction of mutated Htt, exhibit abnormally increased cilia number and length [62, 64]. Furthermore, mutated Htt disrupts the normal binding of Hap1 to dynein, affecting protein transport [65]. In addition to its role in ciliogenesis, Rilpl2 is involved in ciliary protein transport [63]. Increased expression of ciliogenesis-related genes could be a result of alcohol-induced disruptions of normal cell cycle progression in the RVNT, as cilia are retracted and extended during mitosis.

In contrast, Cep41, Evc, and Dpcd were downregulated in NAE embryos at E9.25 (Table 2, Fig 3C). Cep41 is mutated in patients with the genetic ciliopathy Joubert syndrome [66]. Patients with Joubert syndrome display many symptoms similar to other ciliopathies, including renal, ocular, digital, and neurological abnormalities. Notably, patients with Joubert syndrome also show clinical symptoms that overlap with some features of FASD, including cerebellar dysplasia, impaired motor function, abnormal eye development, polydactyly, and craniofacial dysmorphologies. In mouse and zebrafish models, loss of Cep41 causes ciliopathy-like phenotypes, including brain malformations [66]. Thus, even transient downregulation of this gene could impact normal CNS development. In addition, Cep41 is implicated in polyglutamylation of ciliary α-tubulin; loss of tubulin glutamylation could result in structural instability and affect ciliary assembly and transport. Further studies are needed to determine if changes to Cep41 expression are directly linked to glutamylation status of ciliary tubulin in the RVNT following NAE. The gene Dpcd is associated with another ciliopathy, Primary Ciliary Dyskinesia, caused by dysfunction of motile cilia in the respiratory tract and reproductive systems, resulting in respiratory difficulties and infertility [67]. Dpcd has been implicated in the formation and maintenance of cilia and is upregulated during cell division. Finally, downregulation of Evc could directly impact trafficking of the Shh pathway proteins Smo and the Gli family [68, 69]. Ellis-van Creveld syndrome, caused by a mutation in Evc, is a genetic ciliopathy presenting with polydactyly and other digital anomalies, congenital heart defects, and other skeletal abnormalities. Similar skeletal malformations have been noted in humans and rodent models of FASD [70–72]. The decrease in Cep41 was particularly interesting, as polyglutamylation of tubulin is important for protein trafficking within the cilia, including Shh pathway proteins [73]. In addition, Evc is known to play a role in Gli protein trafficking [68]. Together, the downregulation of these two genes suggests disrupted Gli signaling within the cilia as a likely mechanism for the later reduction in Ccnd1 expression. The observed downregulation of Dpcd, a gene normally upregulated during cilia formation and cell division [67], further supports the conclusion that the cell cycle has been affected by NAE as early as 6 hr post-exposure. Transient changes in cell proliferation in this region of the neural tube would impact the development of midline brain and craniofacial tissue [4, 11, 13, 14].

4.2. NAE interacts with cilia gene knockdown to alter exploratory behavior during adolescence

Mice lacking one copy of the Kif3a gene were used to demonstrate that prenatal alcohol interacts with cilia function. The Kif3a mutant mouse is a well-characterized model of genetic ciliopathy [24, 48, 74] as Kif3a mutant mice phenocopy many human ciliopathies, including hypertelorism, facial clefting, and brain abnormalities [48, 75]. Similar to the defects seen in conditional Kif3a mutant mice, NAE mice can exhibit abnormal cortical hemispheres and ventral midline brain structures [13, 14] and hypertelorism has been reported in some patients with heavy prenatal alcohol exposure who lack the classic FAS facial features [4, 76]. Kif3a heterozygous mice have been shown to display similar physical abnormalities, such as situs inversus, to the full knockout but at a much lower rate [77], as Kif3a^−/− die by E10. Kif3a^+/− mice also have reported reduced Kif3a gene expression in the bone and osteoblasts at 6 weeks old [78]. These mice also display abnormal Gli2 expression [78], supporting that Kif3a interacts with Shh signaling and that Kif3a^+/− mice have abnormal cilia function, as Gli2 processing is a critical function of primary cilia during development. In the current study, exploratory behavior in two novel environments (EPM and open field), which relies heavily on the cortico-limbic and hypothalamic circuitry derived from the RVNT, was sensitive to both NAE and Kif3a genotype effects, as well as NAE-genotype interactions. NAE and Kif3a^+/− similarly heightened exploration of the EPM open arms and the center of the open field and increased total activity on the EPM and in the open field. Increased exploratory behavior suggests an impaired stress and or anxiety-related response, as previously observed following gestational alcohol exposure [35]. While the current evidence cannot definitively show that the perturbed behaviors observed in NAE and Kif3a^+/− mice are due to a common mechanism, the persistence of hyperactivity and impaired open field habituation observed in the NAE Kif3a^+/− mice lends support to the hypothesis that variants of a key primary cilia gene might act as a risk factor for certain behavioral effects of NAE.