The assembly of developing motor neurons depends on an interplay between spontaneous activity, type II cadherins and gap junctions

A key organisational motif of the central nervous system is assemblies of neurons with similar synaptic inputs, axonal outputs and functions. Neurons may assemble in clusters, termed nuclei, or in laminar structures such as the cerebral cortex or the cerebellum. Neuronal nucleus formation (nucleogenesis) is required for the correct acquisition of afferent input (Sürmeli et al., 2011), implying that nuclei are a crucial feature of neuronal circuits. Whereas the development of laminar structures has been extensively studied, the question of how nuclei form is unresolved. Here, we demonstrate that cell adhesion and spontaneous activity (SA), in the form of observable calcium transients, converge to shape motor nuclei of the vertebrate brainstem (Fig. 1A).

During nucleogenesis, facial and abducens motor neurons originate from overlapping rostrocaudal domains in rhombomeres (r) 4-6, but migrate differentially to segregate into nuclei at respective lateral and medial locations (Fig. 1B) (Guthrie, 2007). We have previously shown that type II cadherins play a pivotal role in motor nucleogenesis (Astick et al., 2014; Price et al., 2002). Coinciding with the period of neuronal segregation [embryonic day (E) 4-7], type II cadherin expression becomes refined (E5-6), including a change from widespread expression of cadherin 20 (Cad20) to expression of different combinations of cadherins by specific nuclei (Astick et al., 2014). Perturbing cadherin expression disrupts the segregation of abducens and facial neurons, suggesting a role for cadherin-mediated adhesion in nucleogenesis.

We hypothesised that SA might collaborate with cadherins to regulate nucleogenesis, with an adjunct role played by gap junctions (Fig. 1A). Whereas spontaneous electrical activity is known to regulate the late maturation of neuronal circuits (Yamamoto and López-Bendito, 2012), its influence on early developmental events, and interaction with molecular mechanisms has been unclear (Benjumeda et al., 2013; Hanson et al., 2008; Spitzer, 2006). However, rhythmic SA, which occurs in the absence of synaptic input, has previously been recorded from the avian cranial motor nerves at early developmental stages, suggesting a plausible involvement in nucleogenesis (Fortin et al., 1994, 1995).

Using the genetically encoded calcium indicator GCaMP6, we have optically imaged spontaneous calcium transients in facial and abducens motor neuron populations in the chick brainstem. We find that they exhibit characteristic SA patterns at E5 and E6. Manipulation of cadherins or gap junction coupling disrupts coordinated SA during nucleogenesis, and pharmacological disruption of SA in vivo impairs nucleogenesis, demonstrating a functional role for activity. These findings suggest that a network of interactions between cadherins, gap junctions and SA regulates and stabilises the formation of neuronal assemblies during development.

We introduced the genetically encoded calcium indicator GCaMP6 (Chen et al., 2013) into chick hindbrains at E2 by electroporation (Hammond et al., 2005), and imaged activity in ‘open book’ hindbrain preparations at E5-6. The identity of GCaMP-expressing facial and abducens motor neurons was confirmed using cell position, and retrograde labelling in confocal z-stacks (Fig. 1C). Single optical planes in the r4-6 region were imaged to obtain functional time series (2-5 Hz; 3-10 min). At E5, we observed that motor neurons within a nucleus exhibited calcium transients (Movie 1), and that such activity was correlated among motor neurons within, but not between, facial and abducens populations within the same preparation (Fig. 1D), indicating that these neuronal subpopulations have independent SA patterns.

We focussed on the facial nucleus for more detailed analyses. Traces of normalised fluorescence changes (ΔF/F) from single facial motor neurons at E5 showed calcium transients of duration 4-7 s that were temporally correlated (Fig. 1E,G). Calculation of the Pearson correlation coefficient for activity in each cell showed that all activity traces were significantly correlated between cells (P<0.05 for all possible pairings). These transients occurred at low frequency, with an average of one or two active regions of interest (ROIs; groups of cells showing coordinated activity) in a typical 5 min period (Fig. 1E; Fig. 2G). The range varied from one to seven per time series, with a wide variation in size (15-160 µm²). The location and temporal distribution of ROIs was represented in phase maps that were created from the time series, representing the field of view that was imaged, in which colours were used to represent time (blue to red representing earlier to later events). Clusters of cells that showed activity were located in the more mature, lateral regions of the nucleus (Fig. 1I; Fig. 1C, asterisk) and not in medial regions containing younger, migrating motor neurons. The larger ROI sizes observed in Fig. 1I reflect the aggregation of many neighbouring cells with coherent activity, as can be seen by the quantitation of ROI size (Fig. 1K).

We found that facial motor neuron activity patterns changed at E6 (Movie 2; Fig. 1F,H). Calcium transients occurred at much higher frequency (six or seven in a 5 min period; P<0.01), but had smaller amplitude (P<0.05) (Fig. 2G,H). Activity was spread across a wider area of the nucleus (Fig. 1J), with ROIs that were significantly smaller on average than at E5 (Fig. 1L; P<0.05). Activity patterns are therefore different at E5 and E6, switching from infrequent, large coherent events involving neighbouring cells to numerous, smaller events. The correlation of activity observed within nuclei at E5 was also lost by E6, with only 34% of activity events significantly correlated (Pearson's correlation coefficient, P<0.05), coinciding with the timescale of coalescence of facial motor neurons to form a nucleus.

Pharmacological characterisation by bath-applying drugs and imaging GCaMP activity revealed that the activity depended on T type calcium channels at E5 (Fig. S1A,G), and on N and L type channels and endoplasmic reticulum-mediated calcium release at E6 (Fig. S1B,H-J). Spontaneous activity was insensitive to the sodium channel blocker tetrodotoxin (TTX) at E5 and E6 (Fig. S1D,E), but TTX abolished activity at E7 (Fig. S1F). These changes suggest that SA has a switch in calcium channel dependence between E5 and E6, with activity becoming synaptically driven from E7.

In view of the cadherin dependence of nucleogenesis in the hindbrain (Astick et al., 2014), we tested the idea that cadherin expression and SA are linked (Fig. 2A). We analysed SA patterns following disruption of cadherin signalling using a dominant-negative cadherin construct, NΔ390 (Astick et al., 2014; Kintner, 1992), or an siRNA construct, which downregulates chicken Cad20 (Fig. S2A,B). GCaMP imaging at E5 showed that, in both cases, there was a disruption of the typical SA patterns (Fig. 2B-D,G). The percentage of total events that were correlated fell to 42% for NΔ390 and 11% for Cad20 siRNA (P<0.05). For the NΔ390 embryos, the frequency of transients were significantly increased (P<0.05) and the amplitude of transients was decreased relative to controls but not significantly (Fig. 2G,H). For the Cad20 siRNA, the frequency was increased and the amplitude was reduced but neither change was significant (Fig. 2G,H). Phase maps showed ROIs to be more scattered and significantly smaller than in controls (Fig. 2E,F; Fig. S3B,C; P<0.001 and P<0.01 for NΔ390 and Cad20 siRNA, respectively). These data suggest that the global or specific de-regulation of cadherins disrupts SA patterns at E5.

In some contexts, SA events depend on gap junction coupling (Leybaert and Sanderson, 2012), and the gap junction protein connexin 43 (Cx43) has been linked to cadherin expression (Mendoza-Naranjo et al., 2012; Wei et al., 2005). In order to investigate a putative gap junction involvement, we first confirmed Cx43 expression by facial motor neurons using retrograde labelling and immunohistochemistry (Fig. 3A). To test the influence of gap junctions on SA (Fig. 3B), Cx43 was knocked down using an siRNA (Fig. S2C,D) and SA was assessed at E5 as above. Activity patterns were strongly disrupted with an increase in frequency and a decrease in correlation, with only 23% of events correlated (Fig. 3C,D; Fig. 2G). The amplitude of transients was significantly reduced relative to E5 (P<0.05) and was more similar to E6 controls (Fig. 2H). Phase maps showed regions of SA to be more dispersed, with smaller ROIs (Fig. 3F; Fig. S3D). This suggests that facial motor neuron SA patterns are mediated by Cx43-containing gap junctions.

Overall, cadherin and connexin manipulations produced activity patterns with qualitative and quantitative features that more closely resembled E6 than E5 controls.

We investigated a possible interplay between type II cadherins and Cx43 in NIH-3T3 cells, which express type II cadherins including Cad11 (also known as Cdh11) (Takagi et al., 2012), which is expressed by both facial and abducens neurons (Astick et al., 2014). We confirmed expression of Cad11 and Cx43 (also known as Gja1) at cell junctions, and quantified the density of puncta in untransfected and GFP-transfected control cells using immunohistochemistry (Fig. S4A,D). Downregulation or upregulation of cadherins using the NΔ390 dominant-negative construct or Cad20 full-length construct led to a significant decrease and increase, respectively, in the density of Cx43 puncta at cell junctions relative to controls (Fig. S4A-C; P<0.01; P<0.001). Conversely, siRNA-mediated knockdown of Cx43 expression significantly reduced Cad11 immunoreactivity at cell junctions relative to controls (Fig. S4D,E; P<0.001). These results therefore suggest that there is a reciprocal regulation of type II cadherins and Cx43 in vitro.

The results of these in vivo and in vitro experiments suggest that a dynamic interplay between cadherins and connexins regulates SA. We then asked whether, conversely, SA might regulate the topography of the nascent motor nucleus, and cadherin expression. SA was inhibited by applying a cocktail of calcium channel inhibitors found to block activity at E5-6 to chick embryos in ovo for 24 h from E5 to E6 (Fig. 4A). At E6, the distribution of facial and abducens motor neurons was analysed by immunohistochemistry and the expression levels of two relevant cadherins, Cad20 and Cad13, were analysed by in situ hybridisation (Fig. 4B) (Astick et al., 2014). A normal segregation of facial and abducens neuronal clusters and of cadherin expression was observed in vehicle-treated embryos (Fig. 4C,D). Activity blockade resulted in neuronal disaggregation, as quantified using a coalescence index for neuronal clustering (Fig. 4C-E; P<0.01), suggesting that SA might itself stabilise topography within the facial nucleus. Whereas Cad20 levels were normal in SA-inhibited embryos, Cad13 expression appeared to be reduced (Fig. 4C,D). We therefore propose the presence of a feedback loop that allows SA to regulate nucleogenesis.

This study reports a novel interplay between SA and cadherin-mediated adhesion in neuronal nucleus formation. Cadherins, gap junctions and SA form a regulatory network that underpins nucleogenesis. We propose that the initial expression of cadherins among motor neurons drives their migration and segregation into a loose, immature nucleus, whereas the onset of correlated activity regulates nuclear maturation and the emergence of functional properties. Once the nucleus has matured and cadherin expression is refined, the pattern of SA changes from spreading waves to small, dispersed cell groups.

Perturbation experiments suggested that the early (E5) phase of ensemble activity in particular is cadherin and connexin dependent. Manipulations of cadherins did not significantly decrease the amplitude of calcium transients, suggesting that the activity of individual motor neurons was largely unchanged. By contrast, Cx43 knockdown produced a significant reduction in amplitude, implying an effect on the ability of individual neurons to generate SA.

Reciprocal regulation between Cx43 and type I cadherins is well-established though a direct interaction has not been shown (Mendoza-Naranjo et al., 2012; Wei et al., 2005). Our in vitro data suggest a similar reciprocal interaction between Cx43 and type II cadherins; however, in vivo experiments failed to detect a significant change in Cx43 localisation following siRNA-mediated Cad20 downregulation (data not shown). The mechanism by which inhibition of SA results in the disruption of nucleogenesis could be via an effect on cadherin expression.

Our observations show similarities with the correlated SA that has been reported in the developing retina (Feller et al., 1996; Graeber et al., 2013), and in chick and zebrafish spinal motor neurons (Law et al., 2014; Plazas et al., 2013; Wang et al., 2009), where in one case activity was thought to be regulated by recurrent synaptic excitation.

Whereas previous studies focussed on the relationship between activity and axon guidance or topographic mapping, here we focus specifically on the relationship between SA and molecules at the level of motor neuron cell bodies. Our findings highlight the collaboration of molecular mechanisms of cell adhesion and activity in organising neuronal assemblies. Cadherin/catenin signalling has also been shown to pattern spinal motor neuron cell bodies (Demireva et al., 2011). We can speculate that events during nucleogenesis are autonomous to neuronal soma; indeed the caudal migration of mouse facial motor neurons occurs in hindbrain explants without peripheral axons (Vivancos et al., 2009).

The activity we describe might also regulate neuronal maturation and the acquisition of afferent input, which may include descending input from early longitudinal tracts and the vestibular and raphe spinal systems (Gilland and Baker, 2005; Hughes et al., 2009). The link between motor nucleogenesis and brainstem neuronal circuit formation remains to be characterised.

Fertilised Brown Bovan Gold Hen's eggs (Henry Stewart & Co, UK) were incubated at 38°C to E2-7 and staged according to Hamburger and Hamilton (1992).

Expression of constructs was achieved by in ovo electroporation at E2 as described (Ferrario et al., 2012). Retrograde labelling of the facial nerve was carried out as previously described (Simon and Lumsden, 1993). See supplementary Materials and Methods for further details.

For live imaging of GCaMP-electroporated hindbrains, avian embryos were harvested at 37°C, in oxygenated Tyrode's solution, and pinned ventral-side up in a Sylgard dish. Mesenchyme was removed from the hindbrain, and movies were acquired on an Eclipse Ni-E Upright CSU-X1 Spinning Disc confocal microscope equipped with an Andor iXon 3 EM-CCD camera and NIS elements AR software. All other images or z-stacks were acquired using a Zeiss Upright Fluorescence microscope with QImaging CCD digital camera and Volocity software, or an upright Nikon confocal microscope with EZ-C1 software.

For single motor neuron traces, regions of interest were created around identified facial and abducens motor neurons. Normalised fluorescence changes were calculated as ΔF/F=(F−F0)/(F0), where F was the background-corrected average fluorescence signal within the ROI, and F0 was the background-corrected average intensity at the resting level using Fiji software. Movies were further processed using image registration to reduce the drift. Subsequently, a voxel-wise estimation of ΔF/F was calculated based on lower percentile of baseline signal, and then an estimation of skewness was calculated to derive regions of activity quantitatively. Phase maps were also constructed for each time series. See supplementary Materials and Methods for details of ROI analysis.

GCaMP-electroporated chick embryos were dissected and hindbrains were imaged using a spinning disc confocal microscope as described. Embryos were equilibrated at 37°C in oxygenated Tyrode's solution for 30 min and pharmacological reagents were applied [0.5 µM TTX; 100 µM NiCl₂; 300 nM conotoxin; 10 µM nifedepine; 100 µM 2-aminoethoxydiphenyl borate (APB)]. Rhombomeres 4-6 of the hindbrain were then imaged for 10 min before the reagents were washed out with fresh oxygenated Tyrode's solution at 37°C, allowing activity to recover.

Eggs were windowed at E5 (HH stage 25) and either left untreated or treated with sterile Tyrode's solution or a solution containing 300 nM conotoxin (Tocris Bioscience), 10 µM nifedepine (Sigma), 100 µM NiCl₂ (Sigma) and 100 µM APB (Sigma). These reagents were infused into the egg every 4 h over 24 h and the egg then re-sealed. Embryos were fixed in 4% paraformaldehyde after 24 h, dehydrated in increasing concentrations of sucrose (up to 30%), snap-frozen, cryosectioned and processed for in situ hybridisation and immunohistochemistry (see supplementary Materials and Methods for details, including antibodies used).

NIH-3T3 cells were grown in Dulbecco's Modified Eagle Medium/Nutrient Mixture F-12 (DMEM F12, Invitrogen) supplemented with 10% foetal bovine serum (FBS; Invitrogen) and 1% penicillin-streptomycin. Cultures were grown on glass coverslips in 24-well plates to 80% confluency. Lipofectamine LTX (Invitrogen) along with DNA plus reagent (Invitrogen) was used to transfect the cells with the cadherin (NΔ390 and Cad20), and Cx43 siRNA constructs. All constructs used were co-transfected with a GFP plasmid in order to assess transfection efficiency. See supplementary Materials and Methods for details of validation of siRNA constructs and NIH-3T3 cell immunocytochemistry.

Chinese hamster ovary (CHO) cells were grown in DMEM (Invitrogen) supplemented with 10% FBS. Cells were allowed to reach at least 80% confluency before being transfected with the chicken Cad20 siRNA or chicken Cx43 siRNA constructs. Transfection was carried out as described for NIH-3T3 cells. The CHO cells were left in culture for a further 48 h before being homogenised for western blots (see supplementary Materials and Methods for details).

The significance of differences in activity frequency, amplitude and ROI size between different manipulations and stages was tested using a paired Student's t-test. Correlation of activity was assessed using Pearson's correlation coefficient. All statistical tests were performed using SPSS.

We thank Uwe Drescher, Camilla Larsen and QueeLim Ch'ng for valuable advice and discussion. We thank Aine Rubikaite for validation of siRNA constructs, and Ana Casimiro De Almeida, Chiraag Thakrar and Pranetha Baskaran for technical assistance.