1,25-Dihydroxyvitamin D modulates L-type voltage-gated calcium channels in a subset of neurons in the developing mouse prefrontal cortex

Schizophrenia has been associated with a range of genetic and environmental risk factors. Here we explored a link between two risk factors that converge on a shared neurobiological pathway. Recent genome-wide association studies (GWAS) have identified risk variants in genes that code for L-type voltage-gated calcium channels (L-VGCCs), while epidemiological studies have found an increased risk of schizophrenia in those with neonatal vitamin D deficiency. The active form of vitamin D (1,25(OH)2D) is a secosteroid that rapidly modulates L-VGCCs via non-genomic mechanisms in a range of peripheral tissues, though its non-genomic effects within the brain remain largely unexplored. Here we used calcium imaging, electrophysiology and molecular biology to determine whether 1,25(OH)2D non-genomically modulated L-VGCCs in the developing prefrontal cortex, a region widely implicated in schizophrenia pathophysiology. Wide-field Ca2+ imaging revealed that physiological concentrations of 1,25(OH)2D rapidly enhanced activity-dependent somatic Ca2+ levels in a small subset of neurons in the developing PFC, termed vitamin D-responsive neurons (VDRNs). Somatic nucleated patch recordings revealed a rapid, 1,25(OH)2D-evoked increase in high-voltage-activated (HVA) Ca2+ currents. Enhanced activity-dependent Ca2+ levels were mediated by L-VGCC but not associated with any changes to Cacna1c (L-VGCC pore-forming subunit) mRNA expression. Since L-VGCC activity is critical to healthy neurodevelopment, these data suggest that suboptimal concentrations of 1,25(OH)2D could alter brain maturation through modulation of L-VGCC signalling and as such may provide a parsimonious link between epidemiologic and genetic risk factors for schizophrenia.


Introduction
Schizophrenia is a poorly understood group of mental disorders with a lifetime prevalence of 0.7% 1 . The disorder is associated with both common and rare genetic variants, as well as a range of environmental exposures. Recent genome-wide association studies have identified genetic variants of L-type voltage-gated calcium channel (L-VGCC) subunits that are associated with an increased risk for schizophrenia (CACNA1C, CACNB2, CACNA2D1) 2 . Evidence from whole-exome sequencing 3 and copy number variant studies 4 have also implicated L-VGCCrelated genes as risk factors for schizophrenia. While these risk variants do not result in gross dysfunction in L-VGCCs, it is thought that they may contribute to subtle changes in the developmental expression or activity of L-VGCC subunits 5 , potentially leading to neurodevelopmental abnormalities. Indeed, modulation of L-VGCC activity during brain development has been shown to alter the extent and complexity of dendritic arborisation, as well as the developmental maturation and migration of neurons in both the hippocampus 6,7 and cortex 8 . Moreover, deletion of Cacna1c during embryonic development, but not in adulthood, resulted in altered adult behavioural phenotypes relevant to schizophrenia 9 . Taken together, these findings support the hypothesis that factors that alter L-VGCC activity in the developing brain may contribute to risk of schizophrenia.
There is also a growing body of epidemiological evidence implicating developmental vitamin D deficiency as a risk factor for schizophrenia. Epidemiological studies have demonstrated that winter/spring season of birth is associated with increased risk for schizophrenia-vitamin D deficiency is most prevalent during these seasons 10 . There is also an increased risk of schizophrenia in darkskinned migrants to Nordic and northern European countries, groups known to be at increased risk of vitamin D deficiency 11,12 . Neonatal vitamin D status has also been directly linked to an increased risk for schizophrenia. A case-control study (n = 848) found that neonates with vitamin D deficiency had a twofold increased risk of being diagnosed with schizophrenia in later life (Incidence Rate Ratio = 2.1; 95% confidence intervals (CIs) 1.3-3.5) 13 . This finding was recently replicated using a larger sample (n = 2602; Incidence Rate Ratio = 1.44; 95% CIs 1.12-1.85) 14 . The biological plausibility of vitamin D deficiency as a risk factor for schizophrenia is supported by the expression of the vitamin D receptor (VDR) and 25-hydroxyvitamin D-1α-hydroxylase (an enzyme required for 1,25-dihydroxyvitamin D (1,25(OH) 2 D) production) in the human brain 15 . Further, rodent experiments have demonstrated that transient developmental vitamin D deficiency is associated with persistent neurochemical and behavioural changes that involve biological pathways of interest to schizophrenia, including the dopamine system 16,17 .
To date, this research has been based on the assumption that the biological consequences of developmental vitamin D deficiency operated via the classical genomic pathways. For example, vitamin D deficiency impacts on neuronal proliferation and differentiation through transcriptional activity, which is mediated by VDR forming a heterodimer with the retinoic acid X receptor and then binding to response elements in the genome [18][19][20][21][22] . Longterm 1,25(OH) 2 D treatment has also been shown to confer neuroprotective effects following insult, which correlated with the downregulation of both L-VGCC mRNA expression and L-VGCC surface expression 23,24 . However, despite the documented non-genomic effects of 1,25(OH) 2 D on L-VGCC-dependent calcium influx in several tissues including bone 25 , muscle 26 and pancreas 27,28 , its non-genomic effects within the brain remain largely unexplored. To our knowledge, only a single study has implicated 1,25(OH) 2 D with rapid changes to Ca 2+ signalling in the brain, using liquid scintillation spectrometry to reveal 1,25(OH) 2 D-induced slice uptake of 45 Ca 2+ that was dependent on L-type, but not N-type, VGCC activity 29 . Based on the convergence of these genetic and environmental risk factors for schizophrenia upon L-VGCC-related mechanisms, we investigated the non-genomic effects of 1,25(OH) 2 D on neuronal L-VGCC activity in the developing brain using both functional and molecular approaches.

Material and methods
All experimental and animal care procedures were in accordance with the Australian Code of Practice for the Care and Use of Animals for Scientific Purposes and approved by the University of Queensland Animal Ethics Committee.

Calcium imaging
Following 25-min recovery at 34°C, slices were loaded with Texas Red Hydrazide (TxRed; 1.8 μM; ThermoFisher Scientific; Ex/Em 582/602) for 5 min, before being transferred into standard ACSF and equilibrated to room temperature for at least 30 min. Slices were then transferred to a small volume (3 mL) incubation chamber made in-house for Cal-520 AM loading (AAT Bioquest; Ex/Em 492/514) at either room temperature or 34°C (Physio-Suite, Kent Scientific) for 2-3 h, containing (in mM): 118 NaCl, 2.5 KCl, 25 NaHCO 3 , 10 glucose, 1.3 MgCl 2 , 2.5 CaCl 2 , and 1.2 NaH 2 PO 4 , 0.01 Cal-520, 0.03% pluronic F-127, pH 7.4 (95% O 2 and 5% CO 2 ). Slices were incubated in L-VGCC blocker nifedipine (nif; 10 μM) for >30 min prior to imaging where indicated and were imaged in alternation with bath application of 1,25(OH) 2 D alone. Following loading, slices were transferred to a submerged imaging chamber perfused with oxygenated ACSF maintained at 25 ± 1°C (TC-324B, Warner) and secured with a platinum harp strung with parallel nylon threads. Cell depolarisation was evoked using electric field stimulation (efs) generated by two parallel platinum electrodes, separated by a distance of 5 mm, with an isolated stimulator (DS2A, Digitimer Ltd; square pulse, 1 ms, 0.05 Hz, 20-40 V). Both the imaging and stimulation systems were controlled by Metafluor (version 7.10, Molecular Devices) through a 74HC08 AND gate box (Scitech). Fluorescence imaging of the developing PFC was performed using a Polychrome V monochromator (TILL Photonics) combined with an sCMOS camera (pco.edge 5.5, PCO) and ×40 water-immersion objective, which captured a visual field of approximately 411 μm × 338 μm. Wide-field Cal-520 (50 ms) and TxRed (30 ms) fluorescent images were acquired at 0.1 Hz (2 × 2 binning), with every second of Cal-520 exposure time-locked to efs (0.05 Hz). Cal-520 and TxRed fluorescent images were collected using a multi-band dichroic and filter set (FITC/TxRed-A, Semrock; Supplementary Fig. 1a-c). TxRed images were collected to monitor for changes in z-depth via in-plane astrocytic processes (<3 μm), and manual depth corrections were made as required. Imaging sessions were excluded from the data set when changes in z-depth were observed (movement of the imaged z-plane towards or away from the objective lens). Instantaneous ΔF/F (0.05 Hz) was calculated for individual somatic regions of interest (ROIs) as fluorescence during field stimulation (F2) minus the immediately preceding fluorescence at rest (unstimulated, F1), divided by the fluorescence at rest (F1, Fig. 1a). Single-cell ROI fluorescence levels were extracted offline from these acquired temporal z-stacks using Fiji. Background subtraction (rolling ball, radius 50 pixels) and XY image stabilisation were applied using built-in plugins. Instantaneous ΔF/F analysis of wide-field images was automated using custom MATLAB scripts and converted to txt files that were baselined (flat or extrapolated sloping function) and quantified using the Axograph software (Axograph X, version 1.7.2). ROI data met signal-to-noise inclusion criteria if bath application of 1,25(OH) 2 D induced a percentage change in the instantaneous ΔF/F that exceeded three standard deviations of >5 min pre-1,25(OH) 2 D baseline ΔF/F amplitude. 1,25(OH) 2 D was bath applied at the physiological concentration of 0.1 nM 30 from a 1-mM stock solution (dimethyl sulfoxide (DMSO)). Where specified, imaging sessions were conducted in the presence of synaptic blockers (synblock), which included 6-cyano-7-nitroquinoxaline-2-3-dione (10 μM; Tocris), D-2-amino-5-phosphonopentanoic acid (50 μM; Tocris) and picrotoxin (100 μM; Sigma-Aldrich); the L-VGCC agonist Bay K8644 (BAYK; 2.5 μM; Sigma-Aldrich); or the L-VGCC antagonist nif (10 μM; Sigma-Aldrich). The duration of an imaging session baseline was dependent on the applied pharmacology: ACSF 10-20 min, synblock 20 min, nif 20 min 31,32 (Supplementary Fig. 1d). Neurons that ceased to respond to electrical stimulation following bath application of synblock were excluded from the total cell count imaged per slice. Only electrically responsive, TxRed-negative cells were included in vitamin D-responsive neuron (VDRN) subsets. VDRN slice Responding % was calculated from the total number of Cal-520-positive cells in focus per imaged area (TxRed-positive and TxRed-negative) to avoid errors stemming from inaccurate TxRed staining. As such, VDRN Proportion Responsive (%) statistics may underestimate the percentage of imaged neurons that were VDRNs.

Nucleated patch recordings
High-voltage-activated (HVA) Ca 2+ channel currents were recorded in nucleated patch configuration 33 using barium as the charge carrier. Nucleated patches were extracted from the soma of visually identified L2/3 PFC neurons (series resistance <15 MΩ). Briefly, negative pressure (10-15 kPa; Series 475 MARK III Digital Manometer, Dwyer Instruments, IN, USA) 34 was applied to the patch pipette in whole-cell configuration, drawing the nucleus towards the pipette tip as it was slowly withdrawn from the soma, to produce an outside-out patch of somatic membrane internally scaffolded by the nucleus. Following isolation of the nucleated patch, negative pipette pressure was reduced to 3-5 kPa for the remainder of the experiment, and extracellular Ca 2+ was exchanged for Ba 2+ (5-10 mM), containing (in mM): 115.5/110.5 NaCl, 2.5 KCl, 25 NaHCO 3 , 10 glucose, 1.3 MgCl 2 , 5/10 BaCl 2 , and 1.2 NaH 2 PO 4 , pH 7.4 (95% O 2 and 5% CO 2 ). VGCC currents were isolated pharmacologically by bath application of tetradotoxin (1 μM) and 4-aminopyridine (5 mM) to block voltage-gated sodium and voltage-gated potassium channels, respectively, which were only applied once the nucleated patch was isolated from the remnant (anucleated) neuron and situated in the external ACSF (input resistance range typically 1-5 GΩ). Recording pipettes were fabricated from borosilicate glass that was pulled and fire-polished to a tip resistance between 3 and 6 MΩ (GC150F, 1.5 mm, Harvard Apparatus, UK) when filled with caesium-based internal solution, containing (in mM): 120 CsMeSO 4 , 10 TEACl, 10 HEPES, 4 Mg 2 ATP, 0.3 Na 3 GTP, 20 phosphocreatine, and 0.3 EGTA (pH 7.3 with CsOH; osmolarity~295-300 mOsm/kg). VGCC currents were evoked with a 50-ms depolarising pulse from −80 to 0 mV every 20 s (0.05 Hz). Current-voltage (I-V) relationships were investigated by 50-ms step depolarisations from −90 to +40 mV in 10-mV increments from a holding potential of −80 mV. Linear leak and capacitive currents were subtracted online by a P/N protocol (15 repetitions of voltage step −80 to −100 mV). 1,25(OH) 2 D was applied at the physiological concentration of 0.1 nM 30 from a 1-mM stock solution (DMSO). Recordings were made using a MultiClamp 700B

Statistics
Calcium imaging statistics were evaluated using unpaired two-tailed t test with Welch's correction (unequal variances t test). All data were analysed using Prism 7 (GraphPad Software) and are reported as mean ± SEM. Additional details and experimental design are included in Supplementary Methods.

Simultaneous measurement of single-cell VGCC activity from a population of neurons
Conventional methods for the functional investigation of VGCC activity have low spatial resolution, in that they involve single-cell recording configurations in order to provide sufficient voltage control. To locate and quantify the effects of 1,25(OH) 2 D on VGCCs in the brain, we developed a high-throughput imaging method that allowed the simultaneous measure of single-cell activitydependent cytosolic Ca 2+ from a population of neurons ( Fig. 1a-d). This method combines wide-field Ca 2+ fluorescence imaging with field stimulation in brain slices, thereby retaining cortical layer organisation (Fig. 1a). Cytosolic Ca 2+ levels were visualised using the fluorescence indicator Cal-520 (50-ms exposure; 0.1 Hz; Fig. 1b), with every second exposure time-locked to field stimulation (F2; 0.05 Hz; Fig. 1c, d). This approach evoked stable activity-dependent increases in cytosolic Ca 2+ in postnatal PFC neurons (P8-P12) throughout the imaged area, for extended imaging durations (>1 h). Bath application of dihydropyridine (DHP) L-VGCC agonist BAYK (2.5 μM) confirmed sufficient sensitivity to detect changes in L-VGCC activity (Fig. 1e), resolving changes to ΔF/F as low as 7% (59 ± 6%, n = 38; Fig. 1f). BAYK increased Ca 2+ ΔF/F within 3 min of application (2.3 ± 0.05 min), which was consistent with the effect of BAYK on single-channel L-VGCC recordings 35 . Further, BAYK evoked an increased Ca 2+ ΔF/F in 47.5% of the imaged neurons (Fig. 1f), consistent with L-VGCC protein expression in the PFC 9 . Application of equimolar DMSO (10 7 -fold dilution, equivalent to dilution of 1,25(OH) 2 D stock) evoked negligible effect on Ca 2+ ΔF/F measured from imaged neurons (1/144 DMSO-responsive neuron, 0.7%; Fig. 1f).

1,25(OH) 2 D enhanced activity-dependent cytosolic Ca 2+ levels in a subset of neurons
Using the approach described above, 1,25(OH) 2 D (0.1 nM) was bath applied to brain slices containing PFC.

Discussion
Vitamin D deficiency, a prevalent exposure in many parts of the world 40 , reduces the expected (basal) concentration of 1,25(OH) 2 D 41 . While developmental vitamin D deficiency is associated with a range of altered brain outcomes 19 , the assumption to date has been that these properties were mediated via classical genomic pathways (i.e., involving the nuclear receptor VDR). Here we provide evidence that 1,25(OH) 2 D also affects brain function via rapid, non-genomic mechanisms. Physiological levels of 1,25(OH) 2 D rapidly enhanced Ca 2+ influx through L-VGCCs in a small subset of neurons in the PFC, leading to elevated intracellular calcium levels during neural activity (Fig. 4). We propose that suboptimal concentrations of this secosteroid during critical periods of brain development may result in altered L-VGCC function in VDRNs. Since L-VGCC activity is critical to healthy neurodevelopment 7 , vitamin D deficiency may disrupt the excitability and maturation of VDRNs, with possible consequences for local circuit integration, function, and information processing. This non-genomic mechanism may represent a link between two independent genetic and epidemiologic risk factors for schizophrenia.

1,25(OH) 2 D non-genomically modulated L-VGCC activity in a subset of PFC neurons
We show with single-cell resolution that 1,25(OH) 2 D rapidly enhanced activity-dependent Ca 2+ ΔF/F through L-VGCCs in a subset of developing PFC neurons. Changes to L-VGCC-mediated Ca 2+ influx can be evoked through (1) modulation of single-channel conductance or gating, such as through DHP agonists and antagonists 38,42 and/or (2) regulation of the number of channels present in the plasma membrane via protein trafficking. Interestingly, the kinetics of this effect were different to those evoked by L-VGCC agonist BAYK, where 1,25(OH) 2 D induced a smaller mean change to Ca 2+ ΔF/F, with a slower mean response time, and was not observed in all neurons that expressed L-VGCCs. This difference likely reflects an alternative mechanism of action compared to BAYK, which binds directly to the pore-forming α1 subunit of L-VGCCs to enhance channel gating. Further, this finding indicates that only a subset of L-VGCCexpressing neurons contained the machinery required for initiation of 1,25(OH) 2 D-evoked effects, which may suggest the involvement of specific auxiliary L-VGCC subunits 43,44 or second messenger signalling pathways. Indeed, a previous study showed that non-genomic effects of 1,25(OH) 2 D in neurons are dependent on Ca 2+ /calmodulin-dependent protein kinase II (CaMKII) and increased VDRN Ca 2+ ΔF/F in a manner comparable to Bay K8644 alone (Fig. 1i) and was attenuated by preincubation in L-VGCC antagonist nifedipine (p < 0.0001). Pre-incubation of imaged slices in nifedipine significantly reduced both the proportion of detected VDRN (c p = 0.01) and magnitude of detected Ca 2+ ΔF/F changes in response to 1,25(OH) 2 D (d p < 0.0001). Data = Mean ± SEM (Unpaired two-tailed t test with Welch's correction, *p < 0.05, ****p < 0.0001) phosphoinositide 3-kinase (PI3K) 29 , where both CaMKII and P13K have previously been shown to promote trafficking of L-VGCC neurons through the β 2 subunit 45,46 .
The propensity of these second messengers to facilitate L-VGCC activity through the β 2 subunit is noteworthy here, since variants of the β 2 gene (CACNB2) have also been associated with an increased risk for schizophrenia 2,3 . In order to determine whether 1,25(OH) 2 D enhances L-type VGCC Ca 2+ influx in VDRNs by increasing Ca V 1.2 surface expression levels, future studies will require live imaging of Ca V 1. L-VGCCs, brain development and schizophrenia L-VGCCs are highly expressed in the developing brain and are widely known to play a critical role in neuronal development 7 , neuronal excitability, synaptic plasticity, homoeostatic plasticity and the transduction of neuronal activity into gene expression 47 . L-VGCC activity plays a critical role in mediating cortical neurite extension and radial migration 8 , and both ionic and conformational signalling in L-VGCC are required to drive neuronal gene expression 48 . Indeed, disruption to L-VGCC function in hippocampal parvalbumin interneurons during development significantly reduced cell number and dendritic arbor complexity 7 . L-VGCCs are also critical in the regulation of basal and burst firing activity in dopaminergic neurons of the ventral tegmental area 49 , which are considered responsible for the control of basal dopamine levels in downstream areas 50 , and are also the target of significant PFC projections 51 . Notably, both parvalbumin interneurons and the dopamine system are implicated in current hypotheses in schizophrenia research 52 . Critically, disruption to L-VGCC activity has also been implicated in behavioural phenotypes of interest to psychiatry. A recent transgenic animal study suggested that deletion of Cac-na1c in forebrain glutamatergic neurons during embryonic development, but not adulthood, was associated with hyperactivity, impaired hippocampal synaptic plasticity, cognitive impairment and reduced sociability 9 . These findings support the hypothesis that factors influencing L-VGCC activity in the developing brain may contribute to risk of schizophrenia.
In addition to schizophrenia 3 , variants in L-VGCCrelated genes such as CACNA1C and CACNB2 have also been linked to a range of other mental disorders 53 . In particular, variants in L-VGCC-related genes have been linked to the risk of autism-related phenotypes 54 . Recent studies based on the Generation R birth cohort have found that gestational vitamin D deficiency is also associated with an increased risk of both autism-related traits and autism-spectrum disorder 55,56 . These findings are in keeping with the expectation that genetic and environmental risk factors are often shared across a range of psychiatric phenotypes.
Genetic and epidemiological studies have provided important insights into the aetiology of schizophrenia; however, it is rare that discoveries from these fields converge on a shared neurobiological pathway. Indeed, studies that can combine genetics, epidemiology and basic neuroscience are considered better able to triangulate on causal mechanisms 57 . This approach led us to investigate the links between two established risk factors for schizophrenia: risk variants in L-VGCC genes and developmental vitamin D deficiency. Using functional approaches, we have demonstrated that vitamin D rapidly enhanced L-VGCC activity in a subset of PFC neurons during brain development. We propose that optimal concentrations of 1,25(OH) 2 D contribute to the normal (expected) development and maturation of VDRNs through the non-genomic modulation of L-VGCC activity. Conversely, we propose that suboptimal concentrations of 1,25(OH) 2 D may disrupt developmentally critical L-VGCC-dependent processes within the immature cortical circuit. In other words, vitamin D deficiency may Fig. 4 A non-genomic role for 1,25(OH) 2 D in the developing brain. VDRNs responded rapidly to 1,25(OH) 2 D treatment with increased Ca 2+ influx mediated by L-VGCCs. Through their modulation of L-VGCC, optimal levels of 1,25(OH) 2 D may contribute to the healthy development of VDRNs within the maturing cortical circuit, such as through neurite extension, neuronal excitability, and gene expression produce a transient channelopathy-like state, in that the activity of L-VGCCs is altered during critical periods of neurodevelopment. However, unlike genetic diseases that affect L-VGCC activity, vitamin D deficiency is preventable with supplementation. Unravelling the underlying mechanism, as well as the identity and role these neurons play within the circuit, may provide further insight into the aetiology and pathophysiology of schizophrenia.