Two opposite voltage-dependent currents control the unusual early development pattern of embryonic Renshaw cell electrical activity
Abstract
Renshaw cells (V1R) are excitable as soon as they reach their final location next to the spinal motoneurons and are functionally heterogeneous. Using multiple experimental approaches, in combination with biophysical modeling and dynamical systems theory, we analyzed, for the first time, the mechanisms underlying the electrophysiological properties of V1R during early embryonic development of the mouse spinal cord locomotor networks (E11.5–E16.5). We found that these interneurons are subdivided into several functional clusters from E11.5 and then display an unexpected transitory involution process during which they lose their ability to sustain tonic firing. We demonstrated that the essential factor controlling the diversity of the discharge pattern of embryonic V1R is the ratio of a persistent sodium conductance to a delayed rectifier potassium conductance. Taken together, our results reveal how a simple mechanism, based on the synergy of two voltage-dependent conductances that are ubiquitous in neurons, can produce functional diversity in embryonic V1R and control their early developmental trajectory.
Introduction
The development of the central nervous system (CNS) follows complex steps, which depend on genetic and environmental factors and involve interactions between multiple elements of the neural tissue. Remarkably, emergent neurons begin to synchronize soon after the onset of synapse formation, generating long episodes of low-frequency (<0.01 Hz) correlated spontaneous network activity (SNA) (O'Donovan, 1999; Saint-Amant, 2010; Blankenship and Feller, 2010; Myers et al., 2005; Milner and Landmesser, 1999; Hanson and Landmesser, 2003; Momose-Sato and Sato, 2013; Khazipov and Luhmann, 2006). In the mouse embryonic spinal cord (SC), SNA is driven by an excitatory cholinergic-GABAergic loop between motoneurons (MNs) and interneurons (INs), GABA being depolarizing before embryonic day 16.5 (E16.5) (Allain et al., 2011). SNA emerges around E12.5 (Myers et al., 2005; Hanson and Landmesser, 2003; Branchereau et al., 2002; Yvert et al., 2004; Czarnecki et al., 2014), at a time when functional neuromuscular junctions are not yet established (Pun et al., 2002), and sensory and supraspinal inputs have not yet reached the spinal motor networks (Angelim et al., 2018; Marmigère and Ernfors, 2007; Ozaki and Snider, 1997; Ballion et al., 2002).
Several studies pointed out that SNA is an essential component in neuronal networks formation (Moody and Bosma, 2005; Spitzer, 2006; Katz and Shatz, 1996; Hanson et al., 2008). In the SC, pharmacologically induced disturbances of SNA between E12.5 and E14.5 induce defects in the formation of motor pools, in motor axon guidance to their target muscles, and in the development of motor networks (Myers et al., 2005; Hanson et al., 2008; Hanson and Landmesser, 2004; Hanson and Landmesser, 2006). During SNA episodes, long-lasting giant depolarization potentials (GDPs) are evoked in the SC, mainly by the massive release of GABA onto MNs (Czarnecki et al., 2014). Immature Renshaw cells (V1R) are likely the first GABAergic partners of MNs in the mouse embryo (Benito-Gonzalez and Alvarez, 2012; Boeri et al., 2018), and the massive release of GABA during SNA probably requires that many of them display repetitive action potential (AP) firing or plateau potential (PP) activity (Boeri et al., 2018).
However, little is known about the firing pattern of embryonic V1R and the maturation of their intrinsic properties. We recently found that V1R exhibit heterogeneous excitability properties when SNA emerges in the SC (Boeri et al., 2018) in contrast to adult Renshaw cells that constitute a functionally homogeneous population (Perry et al., 2015; Bikoff et al., 2016). Whether this early functional diversity really reflects distinct functional classes of V1R, how this diversity evolves during development, and what are the underlying biophysical mechanisms remain open questions. The present study addresses these issues using multiple approaches, including patch-clamp recordings, cluster analysis, biophysical modeling, and dynamical systems theory. The firing patterns of V1R and the mechanisms underlying their functional diversity are analyzed during a developmental period covering the initial phase of development of SC activity in the mouse embryo (E11.5–E14.5), when SNA is present, and during the critical period (E14.5–E16.5), when GABAergic neurotransmission gradually shifts from excitation to inhibition (Delpy et al., 2008) and locomotor-like activity emerges (Myers et al., 2005; Branchereau et al., 2002; Yvert et al., 2004).
We discover that the balance between the slowly inactivating subthreshold persistent sodium inward current (, Crill, 1996) and the delayed rectifier potassium outward current (), accounts for the heterogeneity of embryonic V1R and the changes in firing pattern during development. The heterogeneity of V1R at E12.5 arises from the existence of distinct functional groups. Surprisingly, and in opposition to the classically accepted development scheme (Sillar et al., 1992; Gao and Ziskind-Conhaim, 1998; Gao and Lu, 2008; McKay and Turner, 2005; Liu et al., 2016; Pineda and Ribera, 2010), we show that the embryonic V1R population loses its ability to support tonic firing from E13.5 to E15.5, exhibiting a transient functional involution during its development. Our experimental and theoretical results provide a global view of the developmental trajectories of embryonic V1R. They demonstrate that a simple mechanism, based on the synergy of only two major opposing voltage-dependent currents, accounts for functional diversity in these immature neurons.
Results
The delayed rectifier potassium current IKdr is a key partner of the persistent sodium current INap in controlling embryonic V1R firing patterns during development
We previously highlighted that V1R are spontaneously active at E12.5. Their response to a 2 s suprathreshold depolarizing current steps revealed four main patterns, depending on the recorded IN (Boeri et al., 2018): (1) single spiking (SS) V1R that fires only 1–3 APs at the onset of the depolarizing pulse, (2) repetitive spiking (RS) V1R, (3) mixed events (ME) V1R that show an alternation of APs and PPs, or (4) V1R that displays a long-lasting sodium-dependent PP (Figure 1A1–A4).
We also uncovered a relationship between and the ability of embryonic V1R to sustain repetitive firing (Boeri et al., 2018). However, the heterogeneous firing patterns of V1R observed at E12.5 could not be fully explained by variations in (Boeri et al., 2018), suggesting the involvement of other voltage-gated channels in the control of the firing pattern of V1R, in particular potassium channels, known to control firing and AP repolarization. Our voltage clamp protocol, performed in the presence of tetrodotoxin (TTX) (1 μM), did not disclose any inward rectifying current (hyperpolarizing voltage steps to –100 mV from = –20 mV, data not shown), but revealed two voltage-dependent outward potassium currents, a delayed rectifier current (), and a transient potassium current () in all embryonic V1R, whatever the firing pattern; Figure 1B1–B4. These currents are known to control AP duration () or firing rate (), respectively (Coetzee et al., 1999). The activation threshold of lied between –30 mV and –20 mV and the threshold of between –60 mV and –50 mV, (n = 27; N = 27 embryos) (Figure 1C1–C4). Removing external calcium had no effect on potassium current I/V curves (data not shown), suggesting that calcium-dependent potassium currents are not yet present at E12.5.
It was unlikely that the heterogeneity of V1R firing patterns resulted from variations in the intensity of . Indeed, its voltage-dependent inactivation (time constant: 23.3 ± 2.6 ms, n = 8; N = 8), which occurs during the depolarizing phase of an AP, makes it ineffective to control AP or PP durations. This was confirmed by our theoretical analysis (Figure 7—figure supplement 1). We thus focused our study on . At E12.5, PP V1R had a significantly lower (2.12 ± 0.44 nS, n = 6; N = 6) than SS V1R (5.57 ± 0.56 nS, n = 9; N = 9) and RS V1R (6.39 ± 0.83 nS, n = 7; N = 7) (Figure 1D). However, there was no significant difference in between SS V1R and RS V1R at E12.5 (Figure 1D), which indicated that variations in alone could not explain all the firing patterns observed at E12.5. Similarly, there was no significant difference in between RS V1R (0.91 ± 0.21nS, n = 8; N = 8) and PP V1R (1.24 ± 0.19 nS, n = 6; N = 6) at E12.5 (Figure 1E), indicating that variations in alone could not explain all the firing patterns of V1R at E12.5 (Boeri et al., 2018). In contrast, measured in SS V1R at E12.5 (0.21 ± 0.20 nS, n = 9; N = 9) were significantly lower compared to measured in RS V1R and in PP V1R at E12.5 (Figure 1E).
Mature neurons often display multiple stable firing patterns (O'Leary et al., 2013; Taylor et al., 2009; Alonso and Marder, 2019). This usually depends on the combination of several outward and inward voltage- or calcium-dependent conductances and on their spatial localization (O'Leary et al., 2013; Taylor et al., 2009; Alonso and Marder, 2019). In contrast, immature V1R have a limited repertoire of voltage-dependent currents ( and , and ) at E12.5, and we did not find any evidence of voltage-dependent calcium currents at this age (Boeri et al., 2018). Blocking prevented PP activity, PP-V1R becoming unexcitable, and turned RS V1R into SS V1R (Boeri et al., 2018). Therefore, we hypothesized that the different firing patterns of V1R observed at E12.5 were related to the / ratio only, with variations in the intensity of being unlikely to account for the heterogeneity of firing pattern. We found that this ratio was significantly lower for SS V1R recorded at E12.5 (/ = 0.043 ± 0.015, n = 9) compared to RS V1R (0.154 ± 0.022, n = 8) and PP V1R (0.66 ± 0.132, n = 6) (Figure 1F). We also found that the / ratio was significantly lower for RS V1R compared to PP V1R (Figure 1F).
Altogether, these results strongly suggest that, although the presence of is required for embryonic V1R to fire repetitively or to generate PPs (Boeri et al., 2018), the heterogeneity of the firing pattern observed between E12.5 is not determined by per se but likely by the balance between and .
Manipulating the balance between GNap and GKdr changes embryonic V1R firing patterns
We previously showed that blocking with riluzole converted PP V1R or RS V1R into SS V1R (Boeri et al., 2018). To confirm further that the balance between and was the key factor in the heterogeneity of V1R firing patterns, we assessed to what extent a given E12.5 SS V1R cell could change its firing pattern when was gradually blocked by 4-aminopiridine (4-AP). We found that could be blocked by micromolar concentrations of 4-AP without affecting IA (Figure 2—figure supplement 1). 4-AP, applied at concentrations ranging from 0.3 μM to 300 μM, specifically inhibited with an IC50 of 2.9 μM (Figure 2—figure supplement 1C1).
We then determined to what extent increasing the concentration of 4-AP modified the firing pattern of V1R at E12.5. Applying 4-AP at concentrations ranging from 3 μM to 300 μM changed the firing pattern of SS V1R (n = 10; N = 10) in a concentration-dependent manner (Figure 2A1–A3). In 50% of the recorded V1R, increasing 4-AP concentrations successfully transformed SS V1R into PP V1R with the following sequence: SS → RS → ME → PP (Figure 2A1). In a second group of embryonic V1R (25%), 4-AP application only evoked mixed activity, with the same sequence as aforementioned (SS → RS → ME) (data not shown). In the remaining SS V1R (25%), increasing 4-AP concentration only led to sustained AP firing (Figure 2A2). Application of 300 μM 4-AP on RS V1R at E12.5 evoked MEs or PPs (Figure 2—figure supplement 2). PPs and RS evoked in the presence of 300 μM 4-AP were fully blocked by 0.5–1 μM TTX, indicating that they were generated by voltage-gated Na+ channels (Figure 2B, C, Figure 2—figure supplement 2). It should be noted that the application of 300 μM of 4-AP induced a significant 30.5 ± 12.4% increase (p=0.0137; Wilcoxon test) of the input resistance (1.11 ± 0.08 GΩ versus 1.41 ± 0.12 GΩ; n = 11; N = 11).
These results show that, in addition to , is also a major determinant of the firing pattern of embryonic V1R. The above suggests that the firing patterns depend on a synergy between and and that the different patterns can be ordered along the following sequence SS → RS → ME → PP when the ratio is increased.
The heterogeneity of the V1R firing patterns decreases during embryonic development
It was initially unclear whether these different firing patterns corresponded to well-separated classes within the E12.5 V1R population or not. To address this question, we performed a hierarchical cluster analysis on 163 embryonic V1R based on three quantitative parameters describing the firing pattern elicited by the depolarizing pulse: the mean duration of evoked APs or PPs measured at half-amplitude (mean ½Ad), the variability of the event duration during repetitive firing (coefficient of variation of ½Ad [CV ½Ad]), and the total duration of all events, expressed in percentage of the pulse duration (depolarizing duration ratio [ddr]) (Figure 3A insets). In view of the large dispersion of mean ½Ad and ddr values, cluster analysis was performed using the (decimal) logarithm of these two quantities (Sigworth and Sine, 1987). The analysis of the distribution of log mean ½Ad, CV ½Ad, and log ddr revealed multimodal histograms that could be fitted with several Gaussians (Figure 3—figure supplement 1A1–C1). Cluster analysis based on these three parameters showed that the most likely number of clusters was 5 (Figure 3A, B), as determined by the silhouette width measurement (Figure 3B). Two clearly separated embryonic V1R groups with CV ½Ad = 0 stood out, as shown in the 3D plot in Figure 3C. The cluster with the largest ½Ad (mean ½Ad = 833.5 ± 89.99 ms) and the largest ddr (0.441 ± 0.044) contained all PP V1R (n = 35; N = 29) (Figure 3C, D, Figure 3—figure supplement 1A2, C2). Similarly, the cluster with the shortest ½Ad (9.73 ± 0.66 ms) and the lowest ddr (0.0051 ± 0.0004) contained all SS V1R (n = 46; N = 37) (Figure 3C, D, Figure 3—figure supplement 1A2, C2).
The three other clusters corresponded to V1R with nonzero values of CV ½Ad (Figure 3C). A first cluster regrouping all RS V1R (n = 69; N = 61) was characterized by smaller values of ½Ad (23.91 ± 1.43 ms), CV ½Ad (27.36 ± 1.64%), and ddr (0.11 ± 0.01) (Figure 3C–D, Figure 3—figure supplement 1A2, C2). The last two clusters corresponded to ME V1R (Figure 3C, D). The smaller cluster, characterized by a larger CV ½Ad (170.9 ± 8.9%; n = 4; N = 4), displayed a mix of APs and short PPs, while the second cluster, with smaller CV ½Ad (87.61 ± 7.37%; n = 9; N = 9), displayed a mix of APs and long-lasting PPs (Figure 3D, Figure 3—figure supplement 1B2). Their ½Ad and ddr values were not significantly different (Figure 3—figure supplement 1A2, C2).
It must be noted that three embryonic V1R (1.8%) were apparently misclassified since they were aggregated within the RS cluster although having zero CV ½Ad (Figure 3C, arrows). Examination of their firing pattern revealed that this was because they generated only two APs, although their ddr (0.16–0.2) and ½ Ad values (31.6–40.3 ms) were well in the range corresponding tto the RS cluster.
These different firing patterns of V1R might reflect different states of neuronal development (Gao and Ziskind-Conhaim, 1998; Ramoa and McCormick, 1994; Belleau and Warren, 2000; Picken Bahrey and Moody, 2003). SS and/or PPs are generally believed to be the most immature forms of firing pattern, RS constituting the most mature form (Spitzer, 2006; Tong and McDearmid, 2012). If it were so, the firing patterns of embryonic V1R would evolve during embryonic development from SS or PP to RS, this latter firing pattern becoming the only one in neonates (Perry et al., 2015) and at early postnatal stages (Bikoff et al., 2016). However, RS neurons already represent 41% of V1R at E12.5. We therefore analyzed the development of firing patterns from E11.5, when V1R terminate their migration and reach their final position (Alvarez et al., 2013), to E16.5. This developmental period covers a first phase of development (E11.5–E14.5), where lumbar spinal networks exhibit SNA, and a second phase (E14.5–E16.5), where locomotor-like activity emerges (Myers et al., 2005; Yvert et al., 2004; Allain et al., 2010; Branchereau et al., 2000). We first analyzed changes in the intrinsic properties (input capacitance , input resistance = , and spike voltage threshold) of V1R. did not change significantly from E11.5 to E13.5 (Figure 4A1), remaining of the order of 12 pF, in agreement with our previous work (Boeri et al., 2018). However, it increased significantly at the transition between the two developmental periods (E13.5–E15.5) to reach about 23.5 pF at E15.5 (Figure 4A1). A similar developmental pattern was observed for , which remained stable during the first phase from E11.5 to E14.5 ( ≈ 1–1.2 GΩ) but decreased significantly after E14.5 to reach about 0.7 GΩ at E15.5 (Figure 4A2). Spike threshold also decreased significantly between the first and the second developmental phases, dropping from about –34 mV at E12.5 to about –41 mV at E16.5 (Figure 4A3). Interestingly, this developmental transition around E14.5 corresponds to the critical stage at which SNA gives way to a locomotor-like activity (Yvert et al., 2004; Allain et al., 2010; Branchereau et al., 2000) and rhythmic activity becomes dominated by glutamate release rather than acetylcholine release (Myers et al., 2005).
This led us to hypothesize that this developmental transition could be also critical for the maturation of V1R firing patterns. The distinct firing patterns observed at E12.5 were already present at E11.5 (Figure 4B1, C), but the percentage of RS V1R strongly increased from E11.5 to E12.5, while the percentage of ME V1R decreased significantly (Figure 4C). The heterogeneity of V1R firing patterns then substantially diminished. PPs were no longer observed at E13.5 (Figure 4B2, C) and ME V1R disappeared at E14.5 (Figure 4B3, C). Interestingly, the proportion of SS V1R remained high from E13.5 to E15.5 and even slightly increased (91.23% at E14.5% and 93.33% at E15.5; Figure 4C). This trend was partially reversed at E16.5 as the percentage of RS V1R increased at the expense of SS V1R (67.86% SS V1R and 32.34% RS V1R; Figure 4B5, C). This decrease in repetitive firing capability after E13.5 was surprising in view of what is classically admitted on the developmental pattern of neuronal excitability (Moody and Bosma, 2005; Spitzer et al., 2000). Therefore, we verified that it did not reflect the death of some V1R after E13.5. Our data did not reveal any activated caspase3 (aCaspase3) staining in V1R (FoxD3 staining) at E14.5 (n = 10 SCs; N = 10) (Figure 5), in agreement with previous reports showing that developmental cell death of V1R does not occur before birth (Prasad et al., 2008).
To determine whether and also controlled the firing pattern of V1R at E14.5 (see Figure 4B3, C), we assessed the presence of and in SS V1R at this embryonic age. Both and were present in V1R at E14.5 (Figure 6—figure supplement 1, Figure 6—figure supplement 2), whereas, as in V1R at E12.5, no calcium-dependent potassium current was detected at this developmental age (not shown). In SS V1R, was significantly higher at E14.5 (11.11 ± 1.12 nS, n = 10; N = 10) than at E12.5 (Figure 1D). In contrast, was similar at E14.5 (0.13 ± 0.14 nS, n = 10; N = 10) and E12.5 (Figure 1E). We also found that the / ratio was significantly lower for SS V1R recorded at E14.5 (0.012 ± 0.004, n = 10) compared to RS V1R (0.154 ± 0.022, n = 8) and PP V1R (0.66 ± 0.132, n = 6) recorded at E12.5 (Figure 1F).
We tested the effect of 4-AP in SS V1R at E14.5. At this embryonic age, 300 μM 4-AP inhibited only 59.2% of . Increasing 4-AP concentration to 600 μM did not inhibit significantly more (60.2%) (Figure 6—figure supplement 2), indicating that inhibition of by 4-AP reached a plateau at around 300 μM. 600 μM 4-AP application had no significant effect on (Figure 6—figure supplement 2). The application of the maximal concentration of 4-AP tested (600 μM) converted SS V1R (n = 13; N = 13) to PP V1R (23.1%; Figure 6A1, B), RS V1R (38.5%; Figure 6A2, B), or ME V1R (38.4%; Figure 6B), as was observed at E12.5, thus indicating that the firing pattern of V1R depends on the balance between and also at E14.5. PP and RS recorded in the presence of 4-AP at E14.5 were fully blocked by 0.5–1 μM TTX, indicating that they were generated by voltage-gated sodium channels (Figure 6A1, A2), as observed at E12.5.
Theoretical analysis: the basic model
As shown in Figure 7A for 26 cells, in which both and were measured, the three largest clusters revealed by the hierarchical clustering analysis (SS, RS, and PP, which account together for the discharge of more than 95% of cells, see Figure 3) correspond to well-defined regions of the - plane. SS is observed only when is smaller than 0.6 nS. For larger values of , RS occurs when is larger than 3.5 nS, and V1R display PPs when is smaller than 3.5 nS. ME (4.5% of the 163 cells used in the cluster analysis), where plateaus and spiking episodes alternate, are observed at the boundary of RS and PP clusters. This suggested to us that a conductance-based model incorporating only the leak current, , , and (see Materials and methods), could account for most experimental observations, the observed zonation being explained in terms of bifurcations between the different stable states of the model. Therefore, we first investigated a simplified version of the model without and slow inactivation of .
A one-parameter bifurcation diagram of this 'basic' model is shown in Figure 7B for two values of (2.5 nS and 10 nS) and a constant injected current = 20 pA. In both cases, the steady-state membrane voltage (stable or unstable) and the peak and trough voltages of stable and unstable periodic solutions are shown as a function of the maximal conductance of the current, all other parameters being kept constant. For = 10 nS, the steady-state membrane voltage progressively increases (in gray) with , but RS (in red, see voltage trace for = 1.2 nS) is not achieved until reaches point SN1, where a saddle node (SN) bifurcation of limit cycles occurs. This fits with the experimental data, where a minimal value of is required for RS (see also Boeri et al., 2018), and is in agreement with the known role of in promoting repetitive discharge (Taddese and Bean, 2002; Kuo et al., 2006). Below SN1, the model responds to the onset of a current pulse by firing only one spike before returning to quiescence (see voltage trace for = 0.2 nS) or a few spikes when close to SN1 (not shown) before returning to quiescence. The quiescent state becomes unstable through a subcritical Hopf bifurcation (HB) at point HB1, with bistability between quiescence and spiking occurring between SN1 and HB1 points. Repetitive firing persists when is increased further and eventually disappears at point SN2. The firing rate does not increase much throughout the RS range (Figure 7—figure supplement 1C), remaining between 11.5 Hz (at SN1) and 20.1 Hz (at SN2). A stable plateau appears at point HB2 through a subcritical HB. The model is bistable between HB2 and SN2, with plateau and large amplitude APs coexisting in this range.
The model behaves very differently when is reduced to 2.5 nS (gray-blue curve in Figure 7B). It exhibits a unique stable fixed point whatever the value of is, and the transition from quiescence to plateau is gradual as is increased. No RS is ever observed. This indicates that the activity pattern is controlled not only by but also by . This is demonstrated further in Figure 7C, where was fixed at 1.2 nS while was increased from 0 to 25 nS. The model exhibits a PP until is increased past point the subcritical HB point HB2, RS sets in before at point SN2 via a SN of limit cycles bifurcation. When is further increased, repetitive firing eventually disappears through a SN bifurcation of limit cycles at point SN1, the quiescent state becomes stable through a subcritical HB at point HB1, and bistability occurs between these two points. This behavior is in agreement with Figure 7A.
Since both conductances and control the firing pattern of embryonic V1R cells, we computed a two-parameter bifurcation diagram (Figure 7D), where the stability regions of the different possible activity states and the transition lines between them are plotted in the - plane. The black curves correspond to the bifurcations HB1 and HB2 and delimit a region where only repetitive firing occurs. The red curves correspond to the SN bifurcations of periodic orbits associated with the transition from quiescence to firing (SN1) and the transition from plateau to firing (SN2). They encompass a region (shaded area) where RS can be achieved but may coexist with SS (between the HB1 and SN1 lines) or PP (in the narrow region between the HB2 and SN2 lines).
Some important features of the diagram must be emphasized: (1) minimal values of both (to ensure sufficient excitability) and (to ensure proper spike repolarization) are required for RS; (2) SS and PP can be clearly distinguished only when they are separated by a region of RS (see also Figure 7B for = 10 nS), otherwise the transition is gradual (Figure 7B for = 2.5 nS); and (3) only oblique lines with an intermediate slope cross the bifurcation curve and enter the RS region (e.g., see the red line in Figure 7D). This means that RS requires an appropriate balance between and . If the ratio / is too large (blue line) or too small (gray line), only PPs or SS will be observed at steady state. This is exactly what is observed in experiments, as shown by the cumulative distribution function of the ratio / for the different clusters of embryonic V1R in Figure 7E (same cells as in Figure 7A). The ratio increases according to the sequence SS → RS → ME → PP, with an overlap of the distributions for SS V1R and RS V1R. Note also that the ratio for ME cells (around 0.25) corresponds to the transition between RS and PP (more on this below).
Embryonic V1R cells display voltage fluctuations that may exceed 5 mV and are presumably due to channel noise. The relatively low number of sodium and potassium channels (of the order of a few thousands) led to voltage fluctuations in the stochastic version of our model comparable to those seen experimentally when the cell was quiescent (top voltage trace in Figure 7D) or when a voltage plateau occurred (bottom trace). Channel noise caused some jitter during RS (middle trace) and induced clearly visible variations in the amplitude of APs. However, RS proved to be very robust and was not disrupted by voltage fluctuations. Altogether, channel noise little alters the dynamics (compare the deterministic voltage traces in Figure 7B and the noisy traces in Figure 7D). This is likely because channel noise has a broad power spectrum and displays no resonance with the deterministic solutions of the model.
The one-parameter bifurcation diagram of the model was not substantially modified when we took into account, as shown in Figure 7—figure supplement 1. It just elicited a slight membrane hyperpolarization, an increase in the minimal value of required for firing, and a decrease of the firing frequency. The transition from repetitive firing to plateau was not affected because is then inactivated by depolarization.
The bifurcation diagram of Figure 7D accounts qualitatively for the physiological data on V1R at E12.5 presented in Figure 7A, as shown in Figure 7F where the conductance data of Figure 7A were superimposed on it. However, one must beware of making a more quantitative comparison because the theoretical bifurcation diagram was established for a constant injected current of 20 pA, whereas the current injected in experiments data varied from neuron to neuron and ranged from 10 to 30 pA in the sample shown in Figure 7A. The position of bifurcation lines in the - plane depends not only on the value of the injected current, but on the values chosen for the other parameters, which also vary from cell to cell but were kept at fixed values in the model (Ori et al., 2018). For instance, the diagrams were computed in Figure 7D, F for = 1 nS and = 13 pF, the median values of the input conductance and capacitance at E12.5, taking no account of the cell-to-cell variations of these quantities. Between E12.5 and E14.5, , which provides an estimate of the cell size, increases by 38% in average, whereas is not significantly modified (see Figure 4). As illustrated in Figure 7G, the two-parameter bifurcation diagram is then shifted upward and rightward compared to Figure 7F because larger conductances are required to obtain the same firing pattern. The observed regression of excitability from E12.5 to E14.5–E15.5 (see Figure 4C) thus comes from a decrease in density (see presumable developmental trajectories indicated by arrows in Figure 7F) together with a shift of the RS region as cell size increases. As a result, all 10 cells shown in Figure 7G are deeply inside the SS region at E14.5.
It is less straightforward to explain on the basis of our model the experiments where 4-AP changed the firing pattern of SS V1R (Figure 2). Indeed, the decrease of (Figure 7—figure supplement 2), although it may exceed 70% at the higher concentrations of 4-AP we used, is not sufficient by itself to account for the change in the firing pattern of V1R because data points in the SS cluster will not cross the bifurcation lines between SS and RS (SN1) and between RS and PP (SN2) when displaced downward in the - plane. However, 4-AP at a 300 μM concentration also decreased (by 23% in average and up to 50% in some neurons), the rheobase current with it, and the current that was injected in cells during experiments was reduced accordingly. When we take into account this reduction of both and , the two-parameter bifurcation diagram of the model remains qualitatively the same, but it is shifted leftwards and downwards in the - plane (Figure 7—figure supplement 2). As a consequence, the bifurcation lines between SS and RS (SN1) and between RS and PP (SN2) are then successively crossed when is reduced, in accordance with experimental results.
Theoretical analysis: slow inactivation of INap and bursting
Our basic model accounts for the firing pattern of 73% of the 163 cells used in the cluster analysis. However, bursting, under the form of recurring plateaus separated by brief repolarization episodes (see a typical trace in Figure 8A, left), was experimentally observed in half of PP V1R (24 out of 46), and plateaus intertwined with spiking episodes were recorded in the 13 cells of the ME cluster (8% of the total sample, see Figure 8A, right, for a typical example). Recurrent plateaus indicate membrane bistability and require that the curve be S-shaped. This occurs when is large and small (Figure 8B1, B2). However, our basic model lacks a mechanism for switching between quiescent state and plateau, even in this case. Channel noise might induce such transitions, but our numerical simulations showed that this is too infrequent to account for bursting (see voltage trace in Figure 8B1 where the plateau state is maintained despite channel noise).
To explain recurrent plateaus during a constant current pulse, we have to incorporate in our model an additional slow dynamical process. Therefore, we took into account the slow inactivation of that is observed in experiments. also inactivates slowly but over times that are much longer than the timescale of bursting, which is why we did not take its slow inactivation into account. The one-parameter bifurcation diagram of the basic model without slow inactivation of is shown in Figure 8C for = 5 nS and an injected current reduced to 10 pA (as compared to 20 pA in the previous section), so as to allow for bistability (see Figure 8B2). The curve is then S-shaped, as shown in Figure 8B1, with a bistability region for between 1.36 and 1.85 nS. This is in contrast with Figure 7B where the curve was monotonic. Adding the slow (de)inactivation of then causes periodic transitions between up (plateau) and down (quiescent) states, as illustrated by the top voltage trace on the right of Figure 8C, and the model displayed a stable limit cycle (shown in black in the bifurcation diagram on the left of Figure 8C). This mechanism is known as pseudo-plateau or plateau-like bursting (a.k.a. fold-subcritical HB bursting) (Teka et al., 2011). In contrast with square wave bursting (Bertram et al., 1995; Izhikevich, 2000a; Borisyuk and Rinzel, 2005; Rinzel, 1985), where the up state is a stable limit cycle arising from a supercritical Hopf bifurcation (Stern et al., 2008; Osinga and Tsaneva-Atanasova, 2010; Osinga and Tsaneva-Atanasova, 2010; Osinga et al., 2012), the up state here is a stable fixed point (which coexists with an unstable limit cycle). This is why one does not observe bursts of APs separated by quiescent periods as, for instance, observed in postnatal CA1 pyramidal cells (Golomb et al., 2006) and in neurons of neonatal pre-Bötzinger Complex (Del Negro et al., 2002; Rybak et al., 2004), but recurrent plateaus. The duration of the plateaus and repolarization episodes depends on the values of and . A voltage-independent time constant = 2 ms leads to up and down states of comparable durations (see top-left voltage trace in Figure 8C). In agreement with the bifurcation diagram of Figure 8C, the persistent sodium current inactivates during plateaus (phase 1, see top-right trace in Figure 8C) and de-inactivates during quiescent episodes (phase 3, see top-right trace). Transitions from the up state to the down state occurs when inactivation is maximal (phase 2) and transition from the down state to the up state when it is minimal (phase 4). Adding channel noise preserves bursting but introduces substantial randomness in the duration of plateaus and repolarization episodes (bottom-left voltage trace in Figure 8C). Moreover, it substantially decreases the duration of both plateaus and quiescent episodes by making transition between the two states easier (compare the top and bottom voltage traces on the left, both computed for ).
Increasing (or decreasing ) makes plateaus much longer than quiescent episodes (see bottom-right voltage trace in Figure 8C). This again points out to the fact that the ratio of these two conductances is an important control parameter. We also noted that adding the current lengthened the quiescent episodes (Figure 7—figure supplement 1).
Slow inactivation of also provides an explanation for mixed patterns, where plateaus alternate with spiking episodes (Figure 8A, right). They take place in our model near the transition between RS and PP, as in experiments (see Figure 8A). Slow inactivation can lead to elliptic bursting, notably when the bifurcation HB is subcritical (Izhikevich, 2000b; Su et al., 2004), which is the case here (Figure 8D). The model then displays a stable limit cycle with alternating plateaus and spiking episodes, arising from crossing the bifurcation points HB2 and SN2 back and forth (see bifurcation diagram in Figure 8D and top voltage trace). We note that sufficient de-inactivation of for triggering a new plateau (phase 1 in the bottom trace of Figure 8D) may be difficult to achieve during spiking episodes because voltage oscillates over a large range, which tends to average out the variations of the inactivation level. If de-inactivation is not sufficient, the model keeps on spiking repetitively without returning to the plateau state. This is what occurs for cells well within the RS region, far away from the RS-PP transition. It also probably explains why it was difficult in many recorded cells to elicit plateaus by increasing the injected current, activation of induced by the larger current being balanced by the increased inactivation.
Altogether, our study shows that a model incorporating the slow inactivation of accounts for all the firing patterns displayed by cells from the PP and ME clusters.
Discussion
V1R constitute a homogeneous population when referring to their transcription factor program during development (Benito-Gonzalez and Alvarez, 2012; Stam et al., 2012), their physiological function (Eccles et al., 1956), and their firing pattern at postnatal stages (Bikoff et al., 2016). Surprisingly, our electrophysiological recordings and our cluster analysis clearly indicate that distinct functional classes of V1R are transiently present during development at the onset of the SNA (E11.5–E12.5). Five different groups of embryonic V1R were defined using cluster analysis, according to their firing properties.
Development of the firing pattern of embryonic V1R during SNA
It is generally assumed that, during early development, newborn neurons cannot sustain repetitive firing (Pineda and Ribera, 2010; Spitzer et al., 2000). Later on, neurons progressively acquire the ability to fire repetitively, APs become sharper, and neurons eventually reach their mature firing pattern due to the progressive appearance of a panoply of voltage-gated channels with different kinetics (Moody and Bosma, 2005; Pineda and Ribera, 2010; Spitzer et al., 2000). Our results challenge the general view that SS is a more primitive form of excitability (Pineda and Ribera, 2010). Indeed, we show that repetitive firing and PPs dominated at early stages (E11.5–E12.5), while SS was prevailing only later (E13.5–E16.5).
The different V1R firing patterns observed at E11.5–E12.5 might reflect variability in the maturation level between V1R at a given developmental stage, as suggested for developing MNs (Vinay et al., 2000; Durand et al., 2015). However, this is unlikely since V1R transiently lose their ability to sustain tonic firing or PP after E13.5. The heterogeneous discharge patterns of V1R observed before E13.5 contrast with the unique firing pattern of V1R at postnatal age (Bikoff et al., 2016). Accordingly, the transient functional heterogeneity of V1R rather reflects an early initial developmental stage (E11.5–E13.5) of intrinsic excitability.
The physiological meaning of the transient functional involution of V1R that follows, after E12.5, is puzzling. To our knowledge, such a phenomenon was never described in vertebrates during CNS development. So far, a functional involution was described only for inner hair cells between E16 and P12 (Marcotti et al., 2003a; Marcotti et al., 2003b) and cultured oligodendrocytes (Sontheimer et al., 1989), and it was irreversible. Because most V1R cannot sustain tonic firing after E12.5, it is likely that their participation to SNA is limited to the developmental period before other GABAergic IN subtypes mature and start to produce GABA and glycine (Allain et al., 2004). Interestingly, embryonic V1R begin to recover their capability to sustain tonic firing when locomotor-like activity emerges (Myers et al., 2005; Yvert et al., 2004), a few days before they form their recurrent synaptic loop with MNs (around E18.5 in the mouse embryos; Sapir et al., 2004). One possible function of the transient involution between E12.5 and E15.5 could be to regulate the growth of V1R axons toward their targets. It is indeed known that low calcium fluctuations within growth cones are required for axon growth while high calcium fluctuations stop axon growth and promote growth cone differentiation (Henley and Poo, 2004).
Ion channels mechanisms underlying the functional heterogeneity of embryonic V1R
Blockade of leads to single spiking (Boeri et al., 2018), which emphasizes the importance of this current for the occurrence of repetitive firing and plateau potentials in V1R at early developmental stages. But these neurons can also switch from one firing pattern to another, when is decreased by 4-AP, which emphasizes the importance of . We found that the main determinant of embryonic V1R firing pattern is the balance between and .
A Hodgkin–Huxley-type model incorporating a persistent sodium current provided a parsimonious explanation of all five firing patterns recorded in the V1R population at E12.5. It provided a mathematical interpretation for the clustering of embryonic V1R shown by the hierarchical analysis and accounted for the effect of 4-AP and riluzole (Boeri et al., 2018) on the discharge. Remarkably, it highlighted how a simple mechanism involving only the two opposing currents and , could produce functional diversity in a population of developing neurons. The model explained why minimal and are required for firing, and how a synergy between and controls the firing pattern and accounts for the zonation of the plane that is observed experimentally.
Taking into account the slow inactivation of to the model allowed us to explain the bursting patterns displayed by cells of the PP and ME clusters. We showed, in particular, that smooth repetitive plateaus could be explained by a pseudo-plateau bursting mechanism (Teka et al., 2011; Osinga and Tsaneva-Atanasova, 2010). Such bursting scenario has been previously studied in models of endocrine cells (Stern et al., 2008; Tsaneva-Atanasova et al., 2010; Tagliavini et al., 2016) and adult neurons (Oster et al., 2015), but rarely observed in experiments (Chevalier et al., 2016). It contrasts with the more common square wave bursting at firing onset, that is, alternating bursts of APs and quiescent episodes, on which most studies of bursting focused (Golomb et al., 2006; Del Negro et al., 2002; Rybak et al., 2004). Our model can also display such square wave bursting, but this occurs for physiologically unrealistic parameter values, so we did not dwell on that bursting mode that we never observed in embryonic V1R. The model also provides a mathematical explanation for mixed events, where bursts of APs alternate with plateau episodes. It is due to an elliptic bursting scenario at the RS-PP transition, a firing range that the aforementioned studies did not examine. This further emphasizes the capacity of our simple model to account for a wide diversity of firing patterns.
Pseudo-plateau bursting has also been observed in the embryonic pre-Bötzinger network (Chevalier et al., 2016). However, it is produced there by the calcium-activated nonselective cationic current , while leads to square wave bursting. Pseudo-plateau bursting, displayed by half of the cells at E16.5, largely disappears at E18.5 because of the change in the balance between and during embryonic maturation (Chevalier et al., 2016). Such a scenario cannot account for the variety of discharge patterns observed in embryonic V1R at the E11.5–12.5 stage of development. Our theoretical analysis and experimental data clearly indicate that the interplay between two opposing currents is necessary to explain all the firing patterns of V1R. Our model is of course not restricted to embryonic V1R, but may also apply to any electrically compact cell, the firing activity of which is dominated by and delayed rectifier potassium currents. This is the case of many classes of embryonic cells in mammals at an early stage of their development. It can also apply to the axon initial segment, where and are known to play the major role in the occurrence of repetitive firing (Kole and Stuart, 2012).
Altogether, our experimental and theoretical results provide a global view of the developmental trajectories of embryonic V1R (see Figure 7F, G). At E12.5, conductances of embryonic V1R are widely spread in the plane, which explains the heterogeneity of their firing patterns. This likely results from the random and uncorrelated expression of sodium and potassium channels from cell to cell at this early stage. Between E12.5 and E14.5–15.5, cell size increases, and with it, while the density of sodium channels decreases (see Figures 1 and 4). The functional involution displayed by V1R between E12.5 and E15.5 thus mainly results from a decrease of coordinated with an increase of . How these synergistic processes are controlled during this developmental period remains an open issue.
It is important to note that the presence of is required for the functional diversity of V1R. Indeed, in the absence of , V1R lose their ability to generate plateau potentials or to fire repetitively. More generally, when the diversity of voltage-gated channels is limited, as observed in embryonic neurons (Moody and Bosma, 2005), changes in the balance between and non inactivating inward currents can modify the firing pattern. This can be achieved not only by , but also by other slowly or non-inactivating inward conductances, such as (Chevalier et al., 2016). Our work also clearly indicates that a change in the firing pattern can only occur if a change in inward conductances cannot be counterbalanced by a corresponding change in outward conductances. This implies that there is no homeostatic regulation of channel density to ensure the robustness of V1R excitability during its early development, contrarily to the mature CNS (O'Leary et al., 2013). In addition, the poor repertoire of voltage-gated channels at this developmental stage precludes channel degeneracy, which is also known to ensure the robustness of excitability in mature neurons (O'Leary et al., 2013).
In conclusion, our study shows that there is no universal pattern of development in embryonic neurons, and it demonstrates that a simple general mechanism involving only two slowly inactivating voltage-gated channels with opposite effects is sufficient to produce a wide variety of firing patterns in immature neurons having a limited repertoire of voltage-gated channels.
Materials and methods
Isolated SC preparation
Request a detailed protocolExperiments were performed in accordance with European Community guiding principles on the care and use of animals (86/609/CEE, CE Off J no. L358, 18 December 1986), French decree no. 97/748 of 19 October 1987 (J Off République Française, 20 October 1987, pp. 12245–12248). All procedures were carried out in accordance with the local ethics committee of local universities and recommendations from the CNRS. We used Gad1GFP knock-in mice to visualize putative GABAergic INs (Tamamaki et al., 2003), as in our previous study (Boeri et al., 2018). To obtain E12.5–E16.5 Gad1GFP embryos, 8–12-week-old wild-type Swiss female mice were crossed with Gad1GFP Swiss male mice.
Isolated mouse SCs from 420 embryos were used in this work and obtained as previously described (Delpy et al., 2008; Scain et al., 2010). Briefly, pregnant mice were anesthetized by intramuscular injection of a mix of ketamine and xylazine and sacrificed using a lethal dose of CO2 after embryos of either sex were removed. Whole SCs were isolated from eGFP-positive embryos and maintained in an artificial cerebrospinal fluid (ACSF) containing 135 mM NaCl, 25 mM NaHCO3, 1 mM NaH2PO4, 3 mM KCl, 11 mM glucose, 2 mM CaCl2, and 1 mM MgCl2 (307 mOsm/kg H2O), continuously bubbled with a 95% O2-5% CO2 gas mixture.
In the lumbar SC of Gad1GFP mouse embryos, eGFP neurons were detected using 488 nm UV light. They were localized in the ventrolateral marginal zone between the motor columns and the ventral funiculi (Stam et al., 2012). Embryonic V1R identity was confirmed by the expression of the forkhead transcription factor Foxd3 (Boeri et al., 2018).
Whole-cell recordings and analysis
Request a detailed protocolThe isolated SC was placed in a recording chamber and was continuously perfused (2 ml/min) at room temperature (RT) (22–26°C) with oxygenated ACSF. Whole-cell patch-clamp recordings of lumbar spinal embryonic V1R were carried out under direct visualization using an infrared-sensitive CCD video camera. Whole-cell patch-clamp electrodes with a resistance of 4–7 MΩ were pulled from thick-wall borosilicate glass using a P-97 horizontal puller (Sutter Instrument Co., USA). They were filled with a solution containing (in mM): 96.4 K methanesulfonate, 33.6 KCl, 4 MgCl2, 4 Na2ATP, 0.3 Na3GTP, 10 EGTA, and 10 HEPES (pH 7.2; 290 mOsm/kg-H2O). This intracellular solution led to an equilibrium potential of chloride ions, , of about –30 mV, close to the physiological values measured at E12.5 in spinal MNs (Delpy et al., 2008). The junction potential (6.6 mV) was systematically corrected offline.
Signals were recorded using Multiclamp 700B amplifiers (Molecular Devices, USA). Data were low-pass filtered (2 kHz), digitized (20 kHz) online using Digidata 1440A or 1550B interfaces, and acquired using pCLAMP 10.5 software (Molecular Devices, USA). Analyses were performed offline using pCLAMP 10.5 software packages (Molecular Devices; RRID:SCR_014284) and Axograph 1.7.2 (AxoGraph; RRID:SCR_002798).
In voltage-clamp mode, voltage-dependent K+ currents () were elicited in the presence of 1 μM TTX (Alomone Labs, Cat# T550, CAS No.: 18660-81-6) by 500 ms depolarizing voltage steps (10 mV increments, 10 s interval) after a prepulse of 300 ms at = –100 mV. To isolate , voltage steps were applied after a 300 ms prepulse at = –30 mV that inactivated the low threshold transient potassium current . was then obtained by subtracting offline from the total potassium current . Capacitance and leak current were subtracted using online P/4 protocol provided by pCLAMP 10.5.
In current-clamp mode, V1R discharge was elicited using 2 s depolarizing current steps (from 0 to ≈ 50 pA in 5–10 pA increments, depending on the input resistance of the cell) with an 8 s interval to ensure that the membrane potential returned to . When a cell generated a sustained discharge, the intensity of the depolarizing pulse was reduced to the minimal value compatible with repetitive firing.
was measured in voltage-clamp mode using a 70 mV/s depolarizing voltage ramp (Huang and Trussell, 2008). This speed was slow enough to preclude substantial contamination by the inactivating transient current and fast enough to avoid substantial inactivation of . Subtraction of the current evoked by the voltage ramp in the presence of 1 μM TTX from the control voltage ramp-evoked current revealed .
Pharmacological reagents
Request a detailed protocolDuring patch-clamp recordings, bath application of TTX (1 µM, Alomone Labs, Cat# T550, CAS No.: 18660-81-6) or 4-AP (Sigma-Aldrich Cat# T550, CAS No.: 18660-81-6) was done using 0.5 mm diameter quartz tubing positioned, under direct visual control, 50 µm away from the recording area. The quartz tubing was connected to six solenoid valves linked with six reservoirs via a manifold. Solutions were gravity-fed into the quartz tubing. Their application was controlled using a VC-8 valve controller (Warner Instruments, USA).
4-AP was used to block . To determine the concentration–response curve, curves of for different concentrations of 4-AP (0.3–300 μM) were compared to the control curve obtained in the absence of 4-AP. The percentage of inhibition for a given concentration was calculated by dividing the peak intensity of by the peak value obtained in control condition. The obtained normalized concentration–response curves were fitted using the Hill equation:
where [4-AP] is the 4-AP concentration, is the residual current (in percentage of the peak ), is the maximal inhibition achieved for saturating concentration of 4-AP, is the 4-AP concentration producing half of the maximal inhibition, and is the Hill coefficient. Curve fitting was performed using KaleidaGraph 4.5 (Synergy Software, USA).
Immunohistochemistry and confocal microscopy
Request a detailed protocolE14.5 embryos were collected from pregnant females. Once dissected out of their yolk sac, SCs were dissected and immediately immersion-fixed in phosphate buffer (PB 0.1 M) containing 4% paraformaldehyde (PFA; freshly prepared in PB, pH 7.4) for 1 hr at 4°C. Whole SCs were then rinsed out in 0.12 M PB at 4°C, thawed at RT, washed in PBS, incubated in NH4Cl (50 mM), diluted in PBS for 20 min, and then permeabilized for 30 min in a blocking solution (10% goat serum in PBS) with 0.2% Triton X-100. They were incubated for 48 hr at 4°C in the presence of the following primary antibodies: guinea pig anti-FoxD3 (1:5000, gift from Carmen Birchmeier and Thomas Müller of the Max Delbrück Center for Molecular Medicine in Berlin) and rabbit anti-cleaved Caspase-3 (1:1000, Cell Signaling Technology Cat# 9661, RRID:AB_2341188). SCs were then washed in PBS and incubated for 2 hr at RT with secondary fluorescent antibodies (goat anti-rabbit-conjugated 649; donkey anti-guinea pig-conjugated Alexa Fluor 405 [1:1000, ThermoFisher]) diluted in 0.2% Triton X-100 blocking solution. After washing in PBS, SCs were dried and mounted in Mowiol medium (Millipore, Molsheim, France). Preparations were then imaged using a Leica SP5 confocal microscope. Immunostaining was observed using a 40× oil-immersion objective with a numerical aperture of 1.25, as well as with a 63× oil-immersion objective with a numerical aperture of 1.32. Serial optical sections were obtained with a Z-step of 1 µm (40×) and 0.2–0.3 µm (63×). Images (1024 × 1024; 12-bit color scale) were stored using Leica software LAS-AF and analyzed using ImageJ 1.5 (N.I.H., USA, RRID:SCR_003070) and Adobe Photoshop CS6 (Adobe, USA, RRID:SCR_014199) software.
Cluster analysis
Request a detailed protocolTo classify the firing patterns of embryonic V1R, we performed a hierarchical cluster analysis on a population of 163 cells. Each cell was characterized by three quantitative measures of its firing pattern (see legend of Figure 5). After normalizing these quantities to zero mean and unit variance, we performed a hierarchical cluster analysis using the hclust function in R 3.3.2 software (Cran project; https://cran.r-project.org/; RRID:SCR_001905) that implements the complete linkage method. The intercluster distance was defined as the maximum Euclidean distance between the points of two clusters, and, at each step of the process, the two closest clusters were merged into a single one, thus constructing progressively a dendrogram. Clusters were then displayed in data space using the dendromat function in the R package ‘squash’ dedicated to color-based visualization of multivariate data. The best clustering was determined using the silhouette measure of clustering consistency (Rousseeuw, 1987). The silhouette of a data point, based on the comparison of its distance to other points in the same cluster and to points in the closest cluster, ranges from −1 to 1. A value near 1 indicates that the point is well assigned to its cluster, a value near 0 indicates that it is close to the decision boundary between two neighboring clusters, and negative values may indicate incorrect assignment to the cluster. This allowed us to identify an optimal number k of clusters by maximizing the overall average silhouette over a range of possible values for k (Rousseeuw, 1987) using the silhouette function in the R package ‘cluster’.
Biophysical modeling
Request a detailed protocolTo understand the relationship between the voltage-dependent membrane conductances and the firing patterns of embryonic V1R, we relied on a single-compartment conductance-based model that included the leak current, the transient and persistent components of the sodium current, and , a delayed rectifier potassium current , and the inactivating potassium current revealed by experiments. Voltage evolution then followed the equation
where is the input capacitance; the input conductance; , , , and the maximal conductances of the aforementioned currents; , , and their activation variables; the inactivation variable of ; the slow inactivation variable of ; and the inactivation variable of . is the baseline potential imposed by ad hoc current injection in current-clamp experiments; and are the Nernst potentials of sodium and potassium ions, and the injected current. All gating variables satisfied equations of the form
where the (in)activation curves were modeled by a sigmoid function of the form
with being positive for activation and negative for inactivation. The time constant was voltage-independent except for the inactivation variables and . The activation variable of was assumed to follow instantaneously voltage changes.
The effect of channel noise was investigated with a stochastic realization of the model, where channels kinetics were described by Markov-type models, assuming a unitary channel conductance of 10 pS for all channels.
Choice of model parameters
Request a detailed protocolMost model parameters were chosen on the basis of experimental measurements performed in the present study or already reported (Boeri et al., 2018). Parameters that could not be constrained from our experimental data were chosen from an experimentally realistic range of values. was set at –60 mV as in experiments (see Table 1). (average 13.15 pF, 50% between 11.9 and 15.1 pF, only 18 cells out of 246 in the first quartile below 7.2 pF or in the fourth quartile above 19 pF) and (50% of cells between 0.71 and 1.18 nS, only 7 out of 242 with input conductance above 2 nS) were not spread much in the cells recorded at E12.5, which showed that most embryonic V1R were of comparable size. Interestingly, and were not correlated, which indicated that the input conductance was determined by the density of leak channels rather than by the sheer size of the cell. Moreover, no correlation was observed between the passive properties and the firing pattern (Boeri et al., 2018). Therefore, we always set and to 1 nS and 13 pF in the model (except in Figure 6—figure supplement 2), close to the experimental medians (0.96 nS and 13.15 pF, respectively). The membrane time constant was then equal to 13 ms, which was also close to the experimental median (13.9 ms, N = 241).
was set to 60 mV (see Boeri et al., 2018). The activation curve of was obtained by fitting experimental data, leading to an average mid-activation of –36 mV and an average steepness of 9.5 mV. The experimentally measured values of were in the range 0–2.2 nS. We assumed that the activation curve of was shifted rightward by 10 mV in comparison to . No experimental data was available for the inactivation of . We chose a mid-inactivation voltage = –45 mV and a steepness = –5 mV. We also assumed that the activation time constant of both and was 1.5 ms, and that the inactivation time constant was voltage-dependent: , decreasing by one order of magnitude (from 30 ms down to 3 ms) with the voltage increase. This enabled us to account for the shape of the APs recorded in experiments, showing a slow rise time and rather long duration. The conductance was not measured experimentally. When choosing a reasonable value of 20 nS for , the model behaved very much as recorded embryonic V1R: with similar current threshold (typically 10–20 pA) and stable plateau potentials obtained for the largest values of .
When taking into account slow inactivation of (see Figure 8), we chose = –30 mV for the mid-inactivation voltage and set the steepness at –5 mV (as for the inactivation of ). For simplicity, we assumed that the inactivation time constant was voltage-independent and set at 2 s.
was set to the experimentally measured value of –96 mV (Boeri et al., 2018). The activation parameters of were obtained by fitting the experimental data: = –20 mV, = 15 mV, = 10 ms, and an activation exponent of 3. The activation and inactivation properties of were also chosen based on experimental measurements. Accordingly, = –30 mV, = –12 mV, = –70 mV, = –7 mV, and = 23 ms. When was taken into account, we assumed that , consistently with experimental data (see Figure 6—figure supplement 1).
Numerical simulations and dynamical systems analysis
Request a detailed protocolWe integrated numerically the deterministic model using the freeware XPPAUT 8.0 (University of Pittsburgh; Pennsylvania; USA; RRID:SCR_001996) (Ermentrout, 2002) and a standard fourth-order Runge–Kutta algorithm. XPPAUT was also used to compute one-parameter and two-parameter bifurcation diagrams. The stochastic version of the model was also implemented in XPPAUT and computed with a Gillespie’s algorithm (Gillespie, 1976).
To investigate the dynamics of the model with slow inactivation of , we relied on numerical simulations together with fast/slow dynamics analysis (Witelski and Bowen, 2015). In this approach, one distinguishes slow dynamical variables (here only s) and fast dynamical variables. Slow variables vary little at the time scale of fast variables and may therefore be considered as constant parameters of the fast dynamics in first approximation. In contrast, slow variables are essentially sensitive to the time average of the fast variables, much more than to their instantaneous values. This separation of time scales allows one to conduct a phase plane analysis of the full dynamics.
Statistics
Samples sizes (n) were determined based on previous experience. The number of embryos (N) is indicated in the main text and figure captions. No power analysis was employed, but sample sizes are comparable to those typically used in the field. All values were expressed as mean with standard error of mean (SEM). Statistical significance was assessed by non-parametric Kruskal–Wallis test with Dunn’s post hoc test for multiple comparisons, Mann–Whitney test for unpaired data or Wilcoxon matched pairs test for paired data using GraphPad Prism 7.0e Software (USA). Significant changes in the proportions of firing patterns with age were assessed by chi-square test for large sample and by Fisher's exact test for small sample using GraphPad Prism 7.0e Software (GraphPad Software; RRID:SCR_002798). Significance was determined as *p<0.05, **p<0.01, or ***p<0.001. The exact p values are mentioned in the result section or in the figure captions.
Data availability
All data generated or analysed during this study are included in the manuscript and supporting files. Source data files have been provided for Figures 3 (Source data files for cluster analysis).
References
-
Serotonin controls the maturation of the GABA phenotype in the ventral spinal cord via 5-HT1b receptorsAnnals of the New York Academy of Sciences 1198:208–219.https://doi.org/10.1111/j.1749-6632.2010.05433.x
-
Maturation of the GABAergic transmission in normal and pathologic motoneuronsNeural Plasticity 2011:905624.https://doi.org/10.1155/2011/905624
-
Principles of interneuron development learned from renshaw cells and the motoneuron recurrent inhibitory circuitAnnals of the New York Academy of Sciences 1279:22–31.https://doi.org/10.1111/nyas.12084
-
Postnatal development of electrophysiological properties of nucleus accumbens neuronsJournal of Neurophysiology 84:2204–2216.https://doi.org/10.1152/jn.2000.84.5.2204
-
Topological and phenomenological classification of bursting oscillationsBulletin of Mathematical Biology 57:413–439.https://doi.org/10.1016/S0092-8240(05)81776-8
-
Mechanisms underlying spontaneous patterned activity in developing neural circuitsNature Reviews Neuroscience 11:18–29.https://doi.org/10.1038/nrn2759
-
Persistent sodium current drives excitability of immature renshaw cells in early embryonic spinal networksThe Journal of Neuroscience 38:7667–7682.https://doi.org/10.1523/JNEUROSCI.3203-17.2018
-
BookUnderstanding neuronal dynamics by geometrical dissection of minimal modelsIn: Chow C, Gutkin B, Hansel D, Meunier C, Dalibard J, editors. Models and Methods in Neurophysics. Proc Les Houches Summer School 2003, (Session LXXX). Elsevier. pp. 19–72.
-
Molecular diversity of K+ channelsAnnals of the New York Academy of Sciences 868:233–255.https://doi.org/10.1111/j.1749-6632.1999.tb11293.x
-
Persistent sodium current in mammalian central neuronsAnnual Review of Physiology 58:349–362.https://doi.org/10.1146/annurev.ph.58.030196.002025
-
Developing electrical properties of postnatal mouse lumbar motoneuronsFrontiers in Cellular Neuroscience 9:349.https://doi.org/10.3389/fncel.2015.00349
-
The inhibitory pathway to motoneuronesProgress in Neurobiology 2:72–82.https://doi.org/10.1016/j.neuron.2007.02.020
-
BookSimulating, Analyzing, and Animating Dynamical Systems: A Guide to XPPAUT for Researchers and StudentsPhiladelphia: Society for Industrial and Applied Mathematics.https://doi.org/10.1137/1.9780898718195
-
Development of ionic currents underlying changes in action potential waveforms in rat spinal motoneuronsJournal of Neurophysiology 80:3047–3061.https://doi.org/10.1152/jn.1998.80.6.3047
-
A general method for numerically simulating the stochastic time evolution of coupled chemical reactionsJournal of Computational Physics 22:403–434.https://doi.org/10.1016/0021-9991(76)90041-3
-
Guiding neuronal growth cones using Ca2+ signalsTrends in Cell Biology 14:320–330.https://doi.org/10.1016/j.tcb.2004.04.006
-
Neural excitability, spiking and burstingInternational Journal of Bifurcation and Chaos 10:1171–1266.https://doi.org/10.1142/S0218127400000840
-
Subcritical elliptic bursting of bautin typeSIAM Journal on Applied Mathematics 60:503–535.https://doi.org/10.1137/S003613999833263X
-
Early patterns of electrical activity in the developing cerebral cortex of humans and rodentsTrends in Neurosciences 29:414–418.https://doi.org/10.1016/j.tins.2006.05.007
-
Sodium and calcium currents shape action potentials in immature mouse inner hair cellsThe Journal of Physiology 552:743–761.https://doi.org/10.1113/jphysiol.2003.043612
-
Specification and connectivity of neuronal subtypes in the sensory lineageNature Reviews Neuroscience 8:114–127.https://doi.org/10.1038/nrn2057
-
Physiological and morphological development of the rat cerebellar purkinje cellThe Journal of Physiology 567:829–850.https://doi.org/10.1113/jphysiol.2005.089383
-
Cholinergic and GABAergic inputs drive patterned spontaneous motoneuron activity before target contactThe Journal of Neuroscience 19:3007–3022.https://doi.org/10.1523/JNEUROSCI.19-08-03007.1999
-
Large-scale synchronized activity in the embryonic brainstem and spinal cordFrontiers in Cellular Neuroscience 7:36.https://doi.org/10.3389/fncel.2013.00036
-
The origin of spontaneous activity in developing networks of the vertebrate nervous systemCurrent Opinion in Neurobiology 9:94–104.https://doi.org/10.1016/S0959-4388(99)80012-9
-
Cross-currents between biology and mathematics: the codimension of pseudo-plateau burstingDiscrete and Continuous Dynamical Systems. Series A 32:2853–2877.https://doi.org/10.3934/dcds.2012.32.2853
-
Dynamics of plateau bursting depending on the location of its equilibriumJournal of Neuroendocrinology 22:1301–1314.https://doi.org/10.1111/j.1365-2826.2010.02083.x
-
Mechanisms for multiple activity modes of VTA dopamine neuronsFrontiers in Computational Neuroscience 9:95.https://doi.org/10.3389/fncom.2015.00095
-
Initial trajectories of sensory axons toward laminar targets in the developing mouse spinal cordThe Journal of Comparative Neurology 380:215–229.https://doi.org/10.1002/(SICI)1096-9861(19970407)380:2<215::AID-CNE5>3.0.CO;2-6
-
Firing properties of renshaw cells defined by Chrna2 are modulated by hyperpolarizing and small conductance ion currents ih and ISKThe European Journal of Neuroscience 41:889–900.https://doi.org/10.1111/ejn.12852
-
Early development of voltage-gated ion currents and firing properties in neurons of the mouse cerebral cortexJournal of Neurophysiology 89:1761–1773.https://doi.org/10.1152/jn.00972.2002
-
BookEvolution of the Action PotentialIn: Kaas JH, editors. Evolution of Nervous System. Elsevier. pp. 211–238.
-
BookBursting oscillations in an excitable membrane modelIn: Sleeman B, Jarvis R, editors. Ordinary and Partial Differential Equations Lecture Notes in Mathematics. Berlin, Heidelberg: Springer. pp. 304–316.https://doi.org/10.1007/BFb0074739
-
Silhouettes: a graphical aid to the interpretation and validation of cluster analysisJournal of Computational and Applied Mathematics 20:53–65.https://doi.org/10.1016/0377-0427(87)90125-7
-
Development of motor rhythms in zebrafish embryosProgress in Brain Research 187:47–61.https://doi.org/10.1016/B978-0-444-53613-6.00004-6
-
Pax6 and engrailed 1 regulate two distinct aspects of renshaw cell developmentJournal of Neuroscience 24:1255–1264.https://doi.org/10.1523/JNEUROSCI.3187-03.2004
-
The post-embryonic development of cell properties and synaptic drive underlying locomotor rhythm generation in Xenopus larvaeProceedings. Biological Sciences 249:65–70.https://doi.org/10.1098/rspb.1992.0084
-
Differentiation of electrical excitability in motoneuronsBrain Research Bulletin 53:547–552.https://doi.org/10.1016/S0361-9230(00)00388-9
-
Resetting behavior in a model of bursting in secretory pituitary cells: distinguishing plateaus from pseudo-plateausBulletin of Mathematical Biology 70:68–88.https://doi.org/10.1007/s11538-007-9241-x
-
Effects of noise on elliptic burstersNonlinearity 17:133–157.https://doi.org/10.1088/0951-7715/17/1/009
-
Is bursting more effective than spiking in evoking pituitary hormone secretion? A spatiotemporal simulation study of calcium and granule dynamicsAmerican Journal of Physiology-Endocrinology and Metabolism 310:E515–E525.https://doi.org/10.1152/ajpendo.00500.2015
-
Green fluorescent protein expression and colocalization with calretinin, Parvalbumin, and somatostatin in the GAD67-GFP knock-in mouseThe Journal of Comparative Neurology 467:60–79.https://doi.org/10.1002/cne.10905
-
How multiple conductances determine electrophysiological properties in a multicompartment modelThe Journal of Neuroscience 29:5573–5586.https://doi.org/10.1523/JNEUROSCI.4438-08.2009
-
From plateau to pseudo-plateau bursting: making the transitionBulletin of Mathematical Biology 73:1292–1311.https://doi.org/10.1007/s11538-010-9559-7
-
Full system bifurcation analysis of endocrine bursting modelsJournal of Theoretical Biology 264:1133–1146.https://doi.org/10.1016/j.jtbi.2010.03.030
-
Differential maturation of motoneurons innervating ankle flexor and extensor muscles in the neonatal ratEuropean Journal of Neuroscience 12:4562–4566.https://doi.org/10.1046/j.0953-816X.2000.01321.x
-
Methods of Mathematical Modelling201–213, Fast/slow Dynamical Systems, Methods of Mathematical Modelling, Cham, Springer, 10.1007/978-3-319-23042-9_10.
Article and author information
Author details
Funding
Fondation pour la Recherche Médicale (DEQ20160334891)
- Pascal Legendre
The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.
Acknowledgements
We thank Susanne Bolte, Jean-François Gilles, and France Lam for assistance with confocal imaging (IBPS imaging facility) and IBPS rodent facility team for animal care and production. We thank University Paris Descartes for hosting Yulia Timofeeva as an invited professor. This work was supported by INSERM, CNRS, Sorbonne Université (Paris), Université de Bordeaux, Université Paris Descartes, and Fondation pour la Recherche Médicale.
Ethics
Animal experimentation: Experiments were performed in accordance with European Community guiding principles on the care and use of animals (86/609/CEE, CE Off J no. L358, 18 December 1986), French decree no. 97/748 of October 19, 1987 (Journal Officiel République Française, 20 October 1987, pp. 12245-12248). All procedures were carried out in accordance with the local ethics committee of local Universities and recommendations from the CNRS. pregnant mice were anesthetized by intramuscular injection of a mix of ketamine and xylazine and sacrificed using a lethal dose of CO2 after embryos of either sex were removed. Every effort was made to minimize suffering.
Copyright
© 2021, Boeri et al.
This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.
Metrics
-
- 730
- views
-
- 136
- downloads
-
- 3
- citations
Views, downloads and citations are aggregated across all versions of this paper published by eLife.
Download links
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Further reading
-
- Neuroscience
Locomotion in mammals is directly controlled by the spinal neuronal network, operating under the control of supraspinal signals and somatosensory feedback that interact with each other. However, the functional architecture of the spinal locomotor network, its operation regimes, and the role of supraspinal and sensory feedback in different locomotor behaviors, including at different speeds, remain unclear. We developed a computational model of spinal locomotor circuits receiving supraspinal drives and limb sensory feedback that could reproduce multiple experimental data obtained in intact and spinal-transected cats during tied-belt and split-belt treadmill locomotion. We provide evidence that the spinal locomotor network operates in different regimes depending on locomotor speed. In an intact system, at slow speeds (<0.4 m/s), the spinal network operates in a non-oscillating state-machine regime and requires sensory feedback or external inputs for phase transitions. Removing sensory feedback related to limb extension prevents locomotor oscillations at slow speeds. With increasing speed and supraspinal drives, the spinal network switches to a flexor-driven oscillatory regime and then to a classical half-center regime. Following spinal transection, the model predicts that the spinal network can only operate in the state-machine regime. Our results suggest that the spinal network operates in different regimes for slow exploratory and fast escape locomotor behaviors, making use of different control mechanisms.
-
- Neuroscience
Perceiving emotions from the movements of other biological entities is critical for human survival and interpersonal interactions. Here, we report that emotional information conveyed by point-light biological motion (BM) triggered automatic physiological responses as reflected in pupil size. Specifically, happy BM evoked larger pupil size than neutral and sad BM, while sad BM induced a smaller pupil response than neutral BM. Moreover, this happy over sad pupil dilation effect is negatively correlated with individual autistic traits. Notably, emotional BM with only local motion features retained could also exert modulations on pupils. Compared with intact BM, both happy and sad local BM evoked stronger pupil responses than neutral local BM starting from an earlier time point, with no difference between the happy and sad conditions. These results revealed a fine-grained pupil-related emotional modulation induced by intact BM and a coarse but rapid modulation by local BM, demonstrating multi-level processing of emotions in life motion signals. Taken together, our findings shed new light on BM emotion processing, and highlight the potential of utilizing the emotion-modulated pupil response to facilitate the diagnosis of social cognitive disorders.