DIFFERENTIAL ROBUSTNESS TO SPECIFIC POTASSIUM CHANNEL DELETIONS IN MIDBRAIN DOPAMINERGIC NEURONS Alexis Haddjeri-Hopkins, Béatrice Marqueze-Pouey, Monica Tapia, Fabien Tell, Marianne Amalric, Jean-Marc Goaillard

To cite this version:

Alexis Haddjeri-Hopkins, Béatrice Marqueze-Pouey, Monica Tapia, Fabien Tell, Marianne Amalric, et al.. DIFFERENTIAL ROBUSTNESS TO SPECIFIC POTASSIUM CHANNEL DELETIONS IN MIDBRAIN DOPAMINERGIC NEURONS. 2020. ￿hal-03026595￿

HAL Id: hal-03026595 https://hal.archives-ouvertes.fr/hal-03026595 Preprint submitted on 26 Nov 2020

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. DIFFERENTIAL ROBUSTNESS TO SPECIFIC POTASSIUM CHANNEL

DELETIONS IN MIDBRAIN DOPAMINERGIC NEURONS

Alexis HADDJERI-HOPKINS1, Béatrice MARQUEZE-POUEY1, Monica TAPIA1,

Fabien TELL1, Marianne AMALRIC2 and Jean-Marc GOAILLARD1, 3

1 UMR_S 1072, Aix Marseille Université, INSERM, Faculté de Médecine Secteur

Nord, Marseille, FRANCE

2 Aix Marseille Université, CNRS, LNC, FR3C, Marseille, FRANCE

3 Corresponding author

Author contributions : M.A. and J-M.G. designed research. A.H-H., B.M-P. and

M.T. performed research. A.H-H., B.M-P., M.T. and F.T. analyzed data. A.H-H.,

F.T., M.A. and J-M.G. wrote the manuscript.

Corresponding author: Jean-Marc GOAILLARD.

UMR_S 1072, Aix Marseille Université, INSERM, Faculté de Médecine Secteur

Nord, Marseille, FRANCE.

Email: [email protected]

Conflict of interest

The authors declare no competing interests.

1 Abstract

Quantifying the level of robustness of neurons in ion channel knock-out (KO) mice depends on how exhaustively electrical phenotype is assessed. We characterized the variations in behavior and electrical phenotype of substantia nigra pars compacta

(SNc) dopaminergic neurons in SK3 and Kv4.3 potassium channel KOs. SK3 and

Kv4.3 KO mice exhibited a slight increase in exploratory behavior and impaired motor learning, respectively. Combining current-clamp characterization of 16 electrophysiological parameters and multivariate analysis, we found that the electrical phenotype of SK3 KO neurons was not different from wild-type neurons, while that of Kv4.3 KO neurons was significantly altered. Consistently, voltage-clamp recordings of the underlying currents demonstrated that the SK current charge was unchanged in SK3 KO neurons while the Kv4-mediated A-type current was virtually abolished in Kv4.3 KO neurons. We conclude that the robustness of SNc dopaminergic neurons to potassium channel deletions is highly variable, due to channel-specific compensatory mechanisms.

2

INTRODUCTION

Is the electrical phenotype of neurons robust to ion channel deletion, and what are the rules that define the level of robustness ? Until recently, the absence of a clear phenotype when artificially mutating or deleting a given gene was often taken as evidence for lack of functional relevance, and many studies based on genetically- modified animals with "negative phenotypes" have probably never been published

(Barbaric et al, 2007; Edelman & Gally, 2001). However, it is now accepted that biological systems have ways of coping with gene mutation or deletion, especially through functional redundancy and pleiotropy of the genes (Kirschner & Gerhart,

1998; Kitano, 2004; Klassen et al, 2011). Studying transgenic animals with negative phenotypes is therefore a unique way of gaining insights into the molecular mechanisms underlying the astonishing robustness of biological systems.

Ion channels, in particular voltage-dependent ion channels, constitute a striking example of functional redundancy (Amendola et al, 2012; Marder & Goaillard, 2006;

Taylor et al, 2009). Indeed, in most neuronal types, many different subtypes of voltage-dependent ion channels are expressed, far more than the theoretical minimum needed to generate the appropriate pattern of activity (Podlaski et al, 2017; Prinz et al,

2004). Consistently, computational studies have demonstrated that ion channel redundancy underlies the multiplicity of biophysical solutions that can confer a given electrical phenotype (Drion et al, 2015; O'Leary et al, 2014; Taylor et al, 2009). As an unsurprising consequence, a number of studies have now demonstrated that the phenotype of ion channel knock-outs (KOs) is often far from what can be expected based on acute blockade of ion channels (Carrasquillo et al, 2012; Kulik et al, 2019;

Nerbonne et al, 2008; Swensen & Bean, 2005), suggesting that compensatory

3 mechanisms are at play. However, the depth of phenotype characterization determines how well its variations or stability are assessed, such that negative phenotypes might sometimes be the result of an under-assessment of neuronal function (Barbaric et al,

2007).

The Kv4.3 potassium channel and the small-conductance calcium-activated potassium channel 3 (SK3) are widely expressed in several brain areas (Serodio & Rudy, 1998;

Vacher et al, 2006) and in several cardiac tissues, including the pacemaker node

(Kv4.3) (Serodio et al, 1996). In spite of this, previous reports suggested that the SK3 and Kv4.3 KO mice do not display an overt phenotype (Carrasquillo et al, 2012;

Jacobsen et al, 2008; Nerbonne et al, 2008). Interestingly, both of these ion channels are strongly expressed in SNc DA neurons and play important roles in the control of their pattern of activity (Amendola et al, 2012; de Vrind et al, 2016; Deignan et al,

2012; Hahn et al, 2003; Liss et al, 2001; Serodio & Rudy, 1998; Seutin et al, 1993;

Vandecasteele et al, 2011; Wolfart et al, 2001; Wolfart & Roeper, 2002). Specifically, the Kv4.3 ion channel is responsible for the A-type potassium current controlling spontaneous firing frequency and post-inhibitory rebound (Amendola et al, 2012;

Hahn et al, 2003; Liss et al, 2001), while the SK3 channel is involved in the control of firing regularity and excitability (Deignan et al, 2012; Wolfart et al, 2001). SNc DA neurons project onto the dorsal striatum (forming the nigrostriatal pathway) where they release DA. As a consequence, their activity has been demonstrated to critically influence motor learning, habitual and goal-directed actions (Balleine, 2019; Wise,

2004; Yin & Knowlton, 2006).

We used behavioral tests to address alterations in motor activity and motor learning, and current-clamp and voltage-clamp electrophysiology to evaluate the robustness of the SNc DA neuron phenotype to the deletion of SK3 and Kv4.3 channels. In

4 addition, following the approach developed in a previous study (Dufour et al, 2014a), we used multivariate analysis of current clamp-recorded parameters to evaluate global alterations in electrical phenotype. While the loss of Kv4.3 ion channel does not seem to be compensated at all, genetic deletion of SK3 is associated with very slight variations in electrical phenotype of these neurons. The comparison of phenotype variations after chronic deletion or acute pharmacological blockade of these channels allowed us to quantify the precise level of robustness of SNc DA neurons to the deletion of Kv4.3 and SK3, respectively. Voltage-clamp experiments suggest that

Kv4.3 deletion is not compensated by Kv4.2 while SK3 loss is partly compensated by

SK2 expression in SNc DA neurons. This study demonstrates that SNc DA neurons are differentially robust to potassium channel deletions, depending on the strength of the compensatory mechanisms engaged. It also provides a general framework for assessing the degree of robustness of electrical phenotype in response to ion channel deletion.

RESULTS

We first sought to determine whether subtle alterations in motor behavior might have been overlooked in previous studies using the SK3 and Kv4.3 KO mice (Carrasquillo et al, 2012; Jacobsen et al, 2008; Nerbonne et al, 2008). We focused on simple motor tasks known to be modulated by the activity of the nigrostriatal DA pathway, such as spontaneous exploration (Kravitz et al, 2010) and learning of a new motor skill

(Beeler et al, 2012; Durieux et al, 2012; Giordano et al, 2018; Yin et al, 2009). In particular, motor learning on the accelerating rotarod is sensitive to DA receptor antagonists and to targeted lesions of the dorsal striatum (Beeler et al, 2012; Durieux et al, 2012; Giordano et al, 2018; Yin et al, 2009).

5 Assessment of motor behavior in SK3 and Kv4.3 KO mice

Behaviors of SK3 and Kv4.3 KO mice were compared to wild-type (WT) littermates in order to assess any significant alteration in motor function. Spontaneous locomotor activity was first measured using actimetry chambers, while motor learning abilities were evaluated using the accelerating rotarod test (Figure 1A). SK3 and Kv4.3 KO mice displayed a global level of locomotor activity similar to WT littermates: SK3

KO vs WT, 202.00 ± 18.63, n=12 vs 185.08 ± 19.06, n=12, p=0.532, unpaired t-test;

Kv4.3 KO vs WT, 127.69 ± 11.46, n=13 vs 117.40 ± 14.87, n=10; p=0.583, unpaired t-test; Figure 1B). However, the analysis of more specific locomotor features revealed a significant increase in the number of rearing events in SK3 KO mice compared to WT (254.2 ± 31.8, n=12 vs 178.5 ± 15.2, n=12, p=0.043, unpaired t-test) while no difference was found between Kv4.3 KO mice and WT (238.6 ± 37.5, n=10 vs 178.9 ± 29.4, n=13, p=0.246, unpaired t-test; Figure 1B). We then wondered whether motor learning could be affected by the chronic deletion of SK3 or Kv4.3.

Mice were submitted to 10 consecutive trials on the accelerating rotarod, and latencies before falling were measured. While SK3 KO mice showed similar learning curves to their WT littermates (performance index 164.5 ± 21.4, n=12 vs 157.5 ± 12.7, n=12, p=0.782, unpaired t-test; Figure 1C), Kv4.3 KO mice displayed a deficit in motor learning, as indicated by the decreased value of the performance index compared to

WT littermates (176.9 ± 21.4, n=11 vs 111.2 ± 8.3, n=13, p=0.006, unpaired t-test;

Figure 1C). In summary, the SK3 KO mice showed a slightly increased number of rearings while Kv4.3 displayed a slight alteration in motor learning. Overall, the modifications in motor behavior observed in SK3 or Kv4.3 KO mice, although rather weak, are consistent with alterations in the function of the nigrostriatal pathway

(Balleine, 2019; Yin & Knowlton, 2006). Therefore, we next investigated whether the

6 electrophysiological phenotype of SNc DA neurons displayed significant variations in both KO mice.

SNc DA neuron pacemaking activity in SK3 and Kv4.3 KO mice

The first distinctive electrophysiological feature of SNc DA neurons is their ability to generate, both in vivo and in vitro, a tonic regular pattern of activity known as pacemaking activity (Kitai et al, 1999; Paladini & Tepper, 2016). This pattern of activity is characterized by its frequency and its regularity, estimated by the coefficient of variation of the interspike interval (CVISI, Figure 2). As mentioned before, Kv4.3 channels have been demonstrated to control firing frequency

(Amendola et al, 2012; Hahn et al, 2003; Liss et al, 2001) while SK3 channels have been shown to increase firing regularity (decrease CVISI) (Deignan et al, 2012;

Wolfart et al, 2001). Consistent with these documented roles, firing frequency was strongly increased specifically in Kv4.3 KO mice (table 1, Figure 2B) while regularity was significantly decreased (CVISI increased) specifically in SK3 KO mice

(table 1, Figure 2B).

Action potential shape in SK3 and Kv4.3 KO mice

We next analyzed AP shape, which can be extracted from recordings of spontaneous activity (Figure 3). SNc DA neurons display a characteristic slow (half-width > 1.0 ms) biphasic action potential (AP) followed by a prominent after-hyperpolarization

(AHP) (Dufour et al, 2014a; Kimm et al, 2015; Vandecasteele et al, 2011). In mouse

SNc DA neurons, the AHP often displays two components, a fast AHP (fAHP) with a hyperpolarizing trough ~10-30ms after the AP and a medium AHP (mAHP) with a trough ~50-100ms after the AP (Figure 3A). However, these components have no clear-cut correspondence with identified subtypes of calcium-activated potassium channels. In fact, apamin-sensitive SK channels seem to be involved in both

7 components. Depending on the SNc DA neuron, the fAHP or mAHP dominates, dictating different repolarization kinetics, and we therefore distinguished these two components in our analysis of AP shape. The following AP parameters were measured: AP threshold, AP amplitude, AP half-width; AHP, fAHP and mAHP trough voltages; AHP, fAHP and mAHP latencies. In spite of the reported role of Kv4 channels in the control of AP shape in other neuronal types (Carrasquillo et al, 2012;

Nerbonne et al, 2008), most of the AP parameters remained unchanged in Kv4.3 KO

SNc DA neurons compared to WT (Figure 3B-D). Only the latency of the mAHP was found to be significantly shorter in Kv4.3 KO mice compared to WT (Figure 3D). In contrast, AP threshold and AHP trough were found to be different in SK3 KO mice compared to WT. Specifically, AP threshold was significantly hyperpolarized (table

1, Figure 3B), AHP trough voltage was significantly depolarized (table 1, Figure

3C). Surprisingly, the mAHP was not significantly modified in the SK3 KO mice

(table 1; Figure 3C,D).

Post-inhibitory rebound in SK3 and Kv4.3 KO mice

SNc DA neurons display a peculiar response to hyperpolarizing current steps (Figure

4A). During hyperpolarization, these neurons display a strong voltage rectification (or sag) due to the IH current (Mercuri et al, 1995; Neuhoff et al, 2002). Then, upon release of the hyperpolarization, the repolarization is biphasic, with a rapid phase followed by a slow linear repolarizing phase leading to AP firing (Amendola et al,

2012; Kita et al, 1986; Nedergaard, 1999). This post-inhibitory rebound profile has been shown to be controlled by the IH and IA currents (Amendola et al, 2012; Neuhoff et al, 2002; Tarfa et al, 2017), this latter being supposedly carried exclusively by the

Kv4.3 potassium channels in this cell type (Hahn et al, 2003; Liss et al, 2001). In particular, the IA current is responsible for the slow repolarizing phase and the

8 subsequent delay in AP firing (Amendola et al, 2012; Tarfa et al, 2017). While the amplitude of the voltage sag during hyperpolarization was found to be similar in SK3

KO, Kv4.3 KO and WT mice, the rebound delay was almost completely abolished in

Kv4.3 KO mice (table 1, Figure 4B) while it was unaltered in SK3 KO mice (table

1).

Excitability in SK3 and Kv4.3 KO mice

SNc DA neurons respond to depolarizing current steps with a characteristic adapting firing pattern (Figure 5A): the frequency at the start of the pulse is higher than the frequency at the end of the pulse, and the highest frequencies reached by SNc DA neurons usually do not exceed 50 Hz (Blythe et al, 2009; Vandecasteele et al, 2011).

This excitability profile can be described by the following measures: the gain of the response at the start (GS, gain start) and at the end (GE, gain end) of the pulse, and the spike frequency adaptation (SFA) index (GS/GE). Several studies have demonstrated that the adapting firing profile of SNc DA neurons is strongly related to the activity of apamin-sensitive SK channels (SK2 and SK3) (Dufour et al, 2014a;

Grace & Bunney, 1984; Vandecasteele et al, 2011). We analyzed the three aforementioned excitability parameters in Kv4.3 KO mice and found no significant changes compared to WT (table 1, Figure 5B). In contrast, consistent with previous results (Vandecasteele et al, 2011), GS was significantly increased in SK3 KO mice compared to WT while both GE and SFA index were unchanged (table 1, Figure

5B).

Estimating the robustness of electrical phenotype in SK3 and Kv4.3 mice

So far, we showed that the electrophysiological phenotype of SNc DA neurons is significantly modified in the SK3 and Kv4.3 KO mice, although the animals show only slight alterations in motor behavior (Figure 1). The electrophysiological changes

9 observed in each transgenic line are consistent with the documented roles of the corresponding ion channels, albeit with substantial variations in the size of the changes between the two KO lines. Specifically, in the SK3 KO mice, the AHP is slightly decreased, its latency increased, and the regularity of spontaneous firing is slightly decreased (Figures 2, 3). In the Kv4.3 mice, spontaneous firing frequency is strongly increased and rebound delay is almost completely abolished (Figures 2, 5).

These changes suggest that, in spite of an overall "negative" phenotype of the SK3 and Kv4.3 animals, SNc DA neurons are not totally robust to SK3 or Kv4.3 deletion.

In order to precisely quantify the extent of robustness of these neurons to SK3 and

Kv4.3 deletion, we next compared the electrophysiological effect of these chronic genetic manipulations with the effect of acutely blocking the ion channels with pharmacological agents.

Partial robustness of electrical phenotype in SK3 KO mice

Both the SK2 and SK3 subunits of the SK family are expressed in SNc DA neurons, although SK3 seems to be more strongly expressed than SK2 (Wolfart et al, 2001).

These two subunits are sensitive to the bee-venom toxin apamin in the nanomolar range (Fakler & Adelman, 2008; Weatherall et al, 2010). Based on the reported stronger expression of SK3 and in the absence of an SK3-specific blocker, we compared the effect of acute application of a saturating concentration of apamin with the effect of SK3 deletion. We first compared the effect of these two treatments on the properties of the AHP. As described before, we analyzed the changes in amplitude and kinetics of both the fAHP and the mAHP. As expected, apamin induced a significant decrease in AHP amplitude, affecting both the fAHP and the mAHP, although the decrease in mAHP was larger, contrasting with the lack of difference of amplitude of both components in the SK3 KO mice (Figure 6A, B). Consistent with

10 its predominant effect on the mAHP, AHP latency was strongly decreased in the presence of apamin, in contrast with the lack of significant change in the SK3 KO mice (Figure 6C). The effect of apamin on firing regularity was consistent with previous studies (de Vrind et al, 2016; Soden et al, 2013; Wolfart et al, 2001; Wolfart

& Roeper, 2002), with an increase in CVISI much larger than the one observed in the

SK3 KO mice (Figure 7A, B). We next measured the effect of apamin on excitability: apamin induced a strong increase in both gain start and gain end, while only gain start was modified in SK3 KO mice (Figure 7C, D). Overall, these experiments show that, while SK3 is accepted to be the dominant SK subunit expressed in SNc DA neurons, its chronic deletion leads to electrophysiological modifications much milder than the acute blockade of SK channels using apamin. Thus SNc DA neuron electrophysiological phenotype appears to be partially robust to SK3 chronic deletion.

Absence of robustness of electrical phenotype in Kv4.3 KO mice

Although two studies suggested that Kv4.2 might be expressed in SNc DA neurons

(Ding et al, 2011; Tapia et al, 2018), the published literature strongly suggests that

Kv4.3 is fully responsible for the IA current in SNc DA neurons (Hahn et al, 2003;

Liss et al, 2001; Serodio & Rudy, 1998). The scorpion toxin AmmTX3 has been shown to selectively block Kv4 channels at nanomolar concentrations (Vacher et al,

2001). We used saturating concentrations of AmmTX3 and compared the changes in electrophysiological parameters with those observed in the Kv4.3 KO mice (Figure

8). Consistent with data obtained in rat SNc DA neurons, AmmTX3 significantly increased spontaneous firing frequency (Figure 8A, B). Interestingly, the percentage of increase in frequency induced by AmmTX3 was almost identical to the change in frequency observed in Kv4.3 KO mice (table 2, Figure 8B). We then tested the effect of AmmTX3 on rebound delay, the other electrophysiological parameter strongly

11 controlled by Kv4 ion channels (Amendola et al, 2012; Tarfa et al, 2017). AmmTX3 induced a strong reduction in duration of rebound delay, which was almost identical to the reduction in delay observed in Kv4.3 mice (table 2, Figure 8C,D). The quantitative similarity of the effects of acute blockade and chronic deletion of Kv4.3 indicates that SNc DA neurons are not robust to Kv4.3 loss, and that compensatory mechanisms are absent.

Voltage-clamp analysis of the ISK and IA currents

To better understand the differences in robustness of electrophysiological phenotype between the SK3 and the Kv4.3 KO mice, we quantified the changes in the properties of the ion currents carried by the SK and Kv4 ion channels (Figure 9). We first characterized the SK current (ISK) in WT and SK3 KO SNc DA neurons (Figure 9A): while the amplitude of ISK was significantly reduced in amplitude in SK3 KO mice, its time constant of decay was significantly longer, leading to an overall constancy of the charge carried by the SK channels (table 1, Figure 9A). We then characterized the A-type current (IA) carried by Kv4 channels in WT and Kv4.3 KO SNc DA neurons (Figure 9B): in contrast with ISK, both the amplitude and the inactivation time constant of IA were strongly reduced in Kv4.3 KO mice, leading to an 80% reduction in the charge carried by Kv4 channels (table 1, Figure 9B). Interestingly, the small residual IA observed in Kv4.3 KO mice was blocked by AmmTX3, suggesting that it is carried by Kv4.2 ion channels (Ding et al, 2011; Tapia et al,

2018). As shown in the current-clamp and voltage-clamp recordings though, this residual current does not play a significant functional role. Although our data strongly suggest that Kv4.3 loss is not compensated in SNc DA neurons, previous studies

(Carrasquillo et al, 2012; Nerbonne et al, 2008) have shown that Kv4.2 or Kv4.3 loss in cortical pyramidal neurons is compensated by increases in delayed rectifier currents

12 (IKDR). We measured IKDR in WT and Kv4.3 KO SNc DA neurons (Figure 9C): both peak and steady-state IKDR amplitudes were analyzed, and no change was observed between WT and Kv4.3 KO mice. Therefore, Kv4.3 loss does not seem to be compensated, neither by Kv4.2, nor by delayed rectifier potassium channels.

Stability of ISK influence in SK3 KO mice

Our voltage-clamp analysis of ISK suggests that SK3 loss is compensated by an increase in SK2, as ISK charge is kept constant in spite of the decrease in amplitude of the current. If ISK charge is maintained, then its functional incidence on firing should also be maintained in SK3 KO mice, meaning that its acute blockade should produce changes similar to the ones observed in WT mice. Thus we next compared the effect of saturating concentrations of apamin in the WT and in the SK3 KO animals (Figure

10, table 2). Apamin application in SK3 KO mice induced strong changes in AHP amplitude and latency (Figure 10A, B), in CVISI (Figure 10C, D) and in excitability that were not significantly different from the changes induced by the toxin in WT animals (table 2). These results confirm that ISK functional impact is maintained in

SK3 KO mice, extending the results already described in the previous current-clamp analysis and the voltage-clamp measurements.

Multivariate analysis of robustness of electrical phenotype

The data obtained so far show that the phenotypic changes induced by knocking-out the SK3 or Kv4.3 potassium channels are both qualitatively and quantitatively different. However, it is difficult to perform a global evaluation of phenotype variation from the sum of the univariate analyses. In order to compare the global effect of the two genetic manipulations on the electrical phenotype of SNc DA neurons, we first used linear discriminant analysis (LDA), a supervised dimensionality-reduction analysis allowing a high-dimensional discrimination of

13 groups of observation. As balanced group sizes are required, LDA was run only on

WT (n=22), SK3 KO (n=26) and Kv4.3 KO (n=28) groups. The 8 most discriminant electrophysiological variables (spontaneous firing frequency, CVISI, AP threshold, AP amplitude, AP half-width, AHP trough voltage, AHP latency and rebound delay,

Figure 11A) were used to perform LDA and obtain a 2-dimensional representation of the phenotype in F1/F2 space (Figure 11B). The mean vector difference between groups was found to be highly significant (Wilk's lambda = 0.22; F=9.3, p=5.10-15), providing a statistical criterion ascertaining that groups are significantly discriminated. Specifically, as demonstrated by the cross-validation, most Kv4.3 KO neurons (82%) were well classified by the discriminant function, while this percentage was lower for SK3 KO (72%) and WT neurons (62%), suggesting that

LDA mainly discriminates Kv4.3 KO neurons from the two other classes (see Figure

11B). These results are explained by the fact that rebound delay and firing frequency

(strongly modified in Kv4.3 KO mice) are major contributors to F1 (85.67%), while

CVISI and AP threshold (modified in SK3 KO mice) contribute mainly to F2

(14.33%). These results confirm that the changes in phenotype in SK3 KO mice are milder than the ones observed in Kv4.3 KO mice. In order to confirm and extend the multi-dimensional analysis, we then used an unsupervised non-linear dimensionality- reduction technique for high-dimensional data set visualization (t-distributed

Stochastic Neighbor Embedding or t-SNE) (Van der Maaten & Hinton, 2008) that allowed us to identify clusters of neurons analyzed in current-clamp recordings (WT,

KOs and pharmacological treatments) based on the 8 electrophysiological parameters used for LDA. Consistent with the LDA results, WT and SK3 KO observations were found to cluster in the same region of the phenotypic space, while Kv4.3 KO, apamin- treated and AmmTX3-treated observations were located in a separate region of

14 phenotypic space (Figure 11C). In other words, while all pharmacological treatments and Kv4.3 deletion induce significant variations in phenotype, deleting SK3 does not induce a significant variation in phenotype. It must be noted however, that 4 Kv4.3

KO neurons (4/28=14%) are not well classified, neither using LDA nor using t-SNE

(see Figure 11B and 11C left, see discussion).

DISCUSSION

In the current study, we determined the level of robustness of SNc DA neurons to the deletion of either SK3 or Kv4.3 potassium channel and also assessed the robustness of nigrostriatal-related motor functions in SK3 and Kv4.3 KO mice. Using multivariate analyses, we demonstrated that SK3 KO SNc DA neurons display an electrical phenotype not significantly different from WT SNc DA neurons while Kv4.3 SNc DA neurons display a significantly altered electrical phenotype. The results from acute pharmacological blockade of SK channels or Kv4 channels are consistent with these conclusions: acute blockade of SK channels leads to a phenotype strikingly different from chronic deletion of SK3 while acute blockade of Kv4 channels and Kv4.3 chronic deletion lead to quantitatively "similar" phenotypes. Voltage-clamp recordings also support these conclusions as the charge carried by ISK is maintained in the SK3 KO mouse while the charge carried by IA is very significantly decreased (-

80%) in Kv4.3 KO mice. The acute blockade of SK channels in SK3 KO neurons suggests that the constancy of ISK charge is due to a compensatory increase in SK2 expression in SK3 KO neurons. On the other hand, Kv4.3 loss is not associated with a compensatory increase in Kv4.2-carried IA current or in IKDR, as shown in other neuronal types (Carrasquillo et al, 2012; Nerbonne et al, 2008). In summary, we demonstrate that SNc DA neurons are robust to SK3 deletion, due to a compensation

15 of SK3 loss by SK2. In spite of the expression of the functionally similar Kv4.2 channel, SNc DA neurons are not robust to Kv4.3 deletion because no compensatory increase in Kv4.2 is triggered in Kv4.3 KO neurons. Intriguingly though, both transgenic mice appear rather robust to either potassium channel deletion, displaying only subtle alterations in spontaneous locomotion and rotarod motor learning.

Comparison with previous studies

To our knowledge the Kv4.3 KO mice were used only in one study focused on the robustness of firing of cortical pyramidal neurons (Carrasquillo et al, 2012) and no behavioral data were available. Our rotarod results suggesting that motor learning might be slightly impaired in Kv4.3 KO mice constitute the first behavioral characterization of this transgenic line. In contrast, the SK3 KO mice were already the object of a study centered on dopamine-related behaviors (Jacobsen et al, 2008) and another study focused on SNc DA neurons (Deignan et al, 2012). Our behavioral results showing a mild increase in spontaneous locomotor events are consistent with the increase of dopamine tone in the striatum reported by Jacobsen and colleagues

(2008). Concerning the electrical phenotype of SK3 KO SNc DA neurons, Deignan and colleagues (2012) described minor modifications of the extracellularly recorded spontaneous activity: the ISI (the inverse of firing frequency) does not appear to be significantly different from WT, while CVISI seems to be slightly increased (Deignan et al, 2012). These results are consistent with our current-clamp analysis, which revealed a weak but significant increase in CVISI in SK3 KO mice without a significant change in spontaneous firing frequency (Figure 1). The voltage-clamp results however differed from the ones presented here. Deignan and colleagues reported a 90% reduction in ISK amplitude in SK3 KO mice, while we observed a

~40% decrease in peak amplitude. Moreover, we found that ISK decay was

16 significantly slower, leading to an overall constancy of the charge carried by the current. This maintenance of the charge in SK3 KO mice is consistent with the global robustness of spontaneous firing (Figure 2), and also supported by the observation that SK channel acute blockade in SK3 KO neurons leads to the same changes in firing pattern (Figure 10). The discrepancies in voltage-clamp recordings might be explained by the number of neurons recorded: ISK measurements were performed in 5

SK3 KO SNc DA neurons in Deignan's study, while ISK was recorded in 26 neurons in the current study.

Kv4.2 expression in SNc DA neurons

The data we obtained from the Kv4.3 KO shed light on the potassium channels carrying the IA current in SNc DA neurons. So far, Kv4.3 has been considered to be the main channel responsible for IA (Amendola et al, 2012; Ding et al, 2011; Khaliq &

Bean, 2008; Liss et al, 2001), as Kv4.1 mRNA has never been detected in SNc DA neurons (Ding et al, 2011; Liss et al, 2001) and the expression of Kv4.2 is a subject of debate: while immunohistochemical stainings in standard conditions have failed to reveal Kv4.2 protein expression in SNc DA neurons (Dufour et al, 2014b; Liss et al,

2001), its mRNA was detected by single-cell PCR in ~50% of SNc DA neurons (Ding et al, 2011; Tapia et al, 2018). In that respect, the current-clamp and voltage-clamp results we obtained from Kv4.3 KO neurons are very interesting. First the electrical phenotype of Kv4.3 KO neurons is quantitatively "identical" to AmmTX3-treated WT neurons, suggesting that most if not all of IA is due to Kv4.3 in SNc DA neurons.

However, the voltage-clamp recordings reveal the presence of a small and fast

AmmTX3-sensitive residual current in the KO neurons. Altogether, these findings suggest that Kv4.2 is indeed expressed in SNc DA neurons, even though it does not have a significant functional influence on firing in most neurons. At this point

17 however, it is impossible to determine whether the small expression found in Kv4.3

KO SNc DA neurons corresponds to an up-regulation of expression compared to WT.

As mentioned before though, 4 Kv4.3 SNc DA neurons out of 28 seem to behave differently, and are clustered with WT neurons when performing LDA or t-SNE multivariate analysis (see Figure 11C). These neurons have a long rebound delay, a slower spontaneous firing frequency, and also display a significantly larger and slower residual IA current. Thus, we cannot rule out the possibility that Kv4.2 is significantly expressed (or up-regulated in Kv4.3 KO mice) in a subpopulation of SNc

DA neurons (~15% here), providing a compensatory mechanism for the loss of

Kv4.3.

Multivariate analysis and definition of phenotype robustness

One of the interests of the current study is to provide a detailed characterization of the variations of electrical phenotype in two different transgenic mice. Most studies on ion channel KO mice were often restricted to the analysis of specific electrophysiological parameters, such as action potential shape and excitability

(Carrasquillo et al, 2012; Nerbonne et al, 2008), bursting patterns of activity

(Swensen & Bean, 2005), regularity of firing (Deignan et al, 2012), etc. Following the framework that we applied to the characterization of post-natal development of electrical phenotype (Dufour et al, 2014a), we combined an exhaustive univariate analysis of 16 electrophysiological parameters and the use of multivariate analyses to provide a global evaluation of phenotype variation in ion channel KO mice. This approach allowed us to provide not only a fine evaluation of the variations of each firing feature but also a statistical argument about the variation of "global" electrical phenotype of SNc DA neurons. Even though some electrophysiological parameters may be missing from our analysis (in particular concerning synaptic integration), the

18 final multivariate picture we provide (Figure 11) is consistent with both the current- clamp analysis of individual parameters and the voltage-clamp recordings of ISK, IA and IKDR: i) the SK3 KO neurons display a multivariate phenotype and an ISK charge not significantly different from WT neurons while ii) the Kv4.3 KO neurons display a strongly reduced IA current (-80%) not compensated by IKDR, and associated with a significant alteration in multivariate phenotype. As mentioned in the introduction, quantifying robustness strongly depends on the accuracy and exhaustiveness of phenotype definition (Barbaric et al, 2007), and the approach we are presenting here seems to provide a way to measure fine variations in electrical phenotype in the face of various perturbations.

Functional redundancy does not systematically imply compensation

One particularly surprising result of the current study is the difference in robustness of the two potassium channel KOs. In the case of SK3 deletion, the loss seems to be roughly compensated by an increase in SK2 expression. Both channels are known to be expressed in WT neurons, even though SK3 is more strongly expressed than SK2

(Deignan et al, 2012; Tapia et al, 2018; Wolfart et al, 2001). These two calcium- activated potassium channels are extremely similar in their functional properties, and are both highly sensitive to apamin (Weatherall et al, 2011). Moreover, the expression of the two subunits suggests that SK3-SK2 heterotetramers could be present in SNc

DA neurons (Tapia et al, 2018; Wolfart et al, 2001). As discussed in the previous section, the situation could be considered to be quite similar for Kv4.3 and Kv4.2.

Both channels display very similar biophysical properties (Birnbaum et al, 2004), and can associate with the same auxiliary subunits, giving rise to IA current in different cell types (Birnbaum et al, 2004; Cai et al, 2004; Carrasquillo et al, 2012; Nerbonne et al, 2008). Even though the presence of Kv4.2 protein in SNc DA neurons has been

19 debated, our results suggest that functional Kv4.2 channels are present at least in

Kv4.3 KO SNc DA neurons. However, in contrast with the results obtained in SK3

KO neurons, no compensatory increase in Kv4.2 is observed in most Kv4.3 KO neurons. To summarize, in both cases two functionally redundant subunits are expressed in SNc DA neurons: SK3/SK2 for the calcium-activated potassium current and Kv4.3/Kv4.2 for IA. Surprisingly however, compensation and subsequent robustness is observed only in SK3 KO neurons.

How can we explain such a difference? Several hypotheses can be proposed. First, SK channels are known to critically control the firing pattern of SNc DA neurons, as they determine the regularity of firing, and by extension the transition between regular firing and bursting patterns of activity (de Vrind et al, 2016; Deignan et al, 2012;

Dufour et al, 2014a; Seutin et al, 1993; Soden et al, 2013; Vandecasteele et al, 2011;

Wolfart et al, 2001; Wolfart & Roeper, 2002). Thus, a substantial decrease in SK channel expression should lead to dramatic changes in firing, with the occurrence of bursts of activity. As a consequence, calcium dynamics might be significantly altered.

On the other hand, Kv4.3 channels "only" control firing frequency, and their absence does not compromise pacemaking regularity. We may hypothesize that changes in firing frequency are not associated with changes in calcium dynamics dramatic enough to trigger compensatory changes in functionally redundant channels (Kv4.2), while changes in firing pattern (pacemaking to bursting) may trigger the up-regulation of SK2 when SK3 is lost. In addition, the SK2 gene might be under the regulatory control of activity-dependent factors while Kv4.2 gene promoter sequences might not be sensitive to changes in activity. Studies in invertebrates have also suggested that ion channel expression levels might be under the control of activity-independent regulatory mechanisms, either at the mRNA (MacLean et al, 2003) or at the protein

20 level (Kulik et al, 2019). Strong correlations in the levels of expression of different ion channel genes, suggesting activity-dependent or -independent co-regulation, have been observed in SNc DA neurons (Tapia et al, 2018). One may postulate that activity-independent regulation selectively affects SK2 expression, not Kv4.2. At this point, it is impossible to determine which hypothesis is true, but our results strongly suggest that compensation of SK3 loss occurs in most SNc DA neurons, while the most optimistic view of Kv4.3 KO neurons suggests that only 15% of them are insensitive to Kv4.3 loss. This finding is particularly interesting when compared to the conclusions drawn from theoretical studies. When databases of models of a given neuronal type are built to study the relationship between phenotype and conductance parameter space (Drion et al, 2011; Prinz et al, 2004; Taylor et al, 2009), each ion conductance is allowed to vary over a given range, independent of the variations of other conductances. In such models, as every conductance is allowed to freely vary

(increase or decrease in expression), functional redundancy implies compensation.

Our results suggest that biological neurons do not use all the available theoretical solutions, and that purely biological constraints (such as genetic mechanisms of regulation or sensitivity to calcium changes) play a major role in determining which ion channel loss might be compensated, and which one might not be. In theory, Kv4.2 should be perfectly able to compensate for Kv4.3 loss, but in SNc DA neurons, it does not. In SNc DA neurons, even though Kv4.2 is expressed and gives rise to a measurable current, this current is very small in most cases, and does not significantly influence activity nor compensate the loss of Kv4.3. Understanding why SNc DA neurons are robust to SK3 but not to Kv4.3 deletion will require an investigation of the activity-dependent mechanisms regulating the expression of the functionally redundant SK2 and Kv4.2 channels. In addition, as suggested by previous studies

21 (Schulz et al, 2007; Tapia et al, 2018), the regulatory rules controlling the expression of functionally redundant ion channels might differ between cell types. For instance, while Kv4.2 (and the associated IA current) loss is compensated by an increase in IKDR in cortical pyramidal neurons (Carrasquillo et al, 2012; Nerbonne et al, 2008), our study shows that this type of compensatory link between IA and IKDR does not exist in

SNc DA neurons. As suggested by previous findings (Schulz et al, 2007; Tapia et al,

2018), the activity-dependent regulatory rules determining the homeostatic relationships in ion channel expression might be specific to each cell type, because they are genetically coupled to cell identity genes (Tapia et al, 2018).

MATERIAL AND METHODS

Animals

Female & male P16-P80 WT (n=39 animals), transgenic SK3 KO (n=21, Jackson laboratory) and Kv4.3KO (n=24, Deltagen) mice from C57BL6/J genetic background were housed with free access to food and water in a temperature-controlled room

(24°C) on a 12:12 h dark–light cycle (lights on at 07:00 h). All efforts were made to minimize the number of animals used and to maintain them in good general health, according to the European (Council Directive 86/609/EEC) and institutional guidelines for the care and use of laboratory animals (French National Research

Council).

Behavioral Experiments

Female and male WT (n=22), transgenic SK3 KO (n=12) and Kv4.3 KO (n=13) mice aged P56-P63 at the start of the behavioral testing were used to evaluate changes in motor function.

Locomotor and exploratory activities

22 Actimetry was monitored in individual activity chambers (20 cm × 11.2 cm × 20.7 cm) housed within a sound-attenuating cubicle and under homogeneous light illumination (Imetronic, Pessac, France). Each chamber was equipped with two pairs of infrared photobeams located 1.5 and 7.5cm above the floor level of the chamber.

The number of back-and-forth movements (animals breaking the lower photobeams) as well as the number of vertical movements (animals breaking the upper photobeams) were recorded in 5-min bins over 90 min. Numbers of back-and-forth movements (locomotion) and vertical movements (rearing) are shown as mean ± SEM for each time bin over the whole period of recording time. Locomotion and rearing activities over the whole 90-min period were normalized and displayed as a mean percentage of control littermates.

Motor learning

Motor learning was evaluated on the accelerating rotarod test (5-40 RPM in 5 min) that consisted of a 10cm-diameter rod. On the first day, mice were allowed to freely explore the non-rotating apparatus for 60 seconds and subsequently trained, as many times as necessary to hold on the rotating rod (5 RPM) for at least two 60-second trials, each trial being separated by 10-minute breaks. Mice were allowed to recover one hour before the first test. Testing phase consisted in 10 consecutive trials on the accelerating rod separated by 15-minute breaks that allowed consolidation of performance. Results are shown as the average latency to fall off the rod (mean ±

SEM) at each trial. A performance index was calculated for each individual and consisted in the average latency of the last 3 trials divided by the average latency of the first 3 trials multiplied by 100 (!"#$!%# !"#$%&' !!!" ×100). !"#$!%# !"#$%&' !!!

Electrophysiology

23 51 neurons from 17 WT mice, 28 neurons from 9 SK3 KO mice and 33 neurons from

11 Kv4.3 KO mice were recorded (current-clamp and voltage-clamp).

Acute midbrain slice preparation

Acute slices were prepared from P16-P25 animals of either sex. All experiments were performed according to the European and institutional guidelines for the care and use of laboratory animals (Council Directive 86/609/EEC and French National Research

Council). Mice were anesthetized with isoflurane (CSP) in an oxygenated chamber

(TEM SEGA) before decapitation. After decapitation the brain was immersed briefly in oxygenated ice-cold low-calcium aCSF containing the following (in mM): 125

NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 0.5 CaCl2, 4 MgCl2, and 25 D-glucose, pH 7.4, oxygenated with 95% O2/5% CO2 gas. The cortices were removed and then coronal midbrain slices (250 µm) were cut in ice-cold oxygenated low-calcium aCSF on a vibratome (vibrating microtome 7000smz, Camden Instruments, UK). Following

20-30 min incubation in oxygenated low-calcium aCSF at 35°C, the acute slices were then incubated for a minimum of 30 min in oxygenated aCSF (containing in mM: 125

NaCl, 25 NaHCO3, 2.5 KCl, 1.25 NaH2PO4, 2 CaCl2, 2 MgCl2, and 25 glucose, pH

7.4, oxygenated with 95% O2/5% CO2 gas) at room temperature before electrophysiological recordings.

Drugs

Kynurenate (2 mM, Sigma-Aldrich) and picrotoxin (100 µM, Sigma-Aldrich) to block excitatory and inhibitory synaptic activity, respectively. Apamin (300 nM, Alomone) was used to identify and measure calcium-activated potassium currents carried by SK channels. AmmTX3 (1µM, Alomone) was used to measure the transient potassium current (IA) carried by Kv4 channels. Drugs were bath applied via continuous perfusion in aCSF.

24 Electrophysiology recordings and analysis

All recordings (112 neurons from 37 mice) were performed on midbrain slices continuously superfused with oxygenated aCSF at 30–32°C. Picrotoxin and kynurenate were systematically added to the aCSF for all recordings to prevent contamination of the intrinsically generated activity by glutamatergic and GABAergic spontaneous synaptic activity. Patch pipettes (3.2–4.0 MOhm) were pulled from borosilicate glass (GC150TF-10, Harvard Apparatus) on a DMZ-Universal Puller

(Zeitz Instruments) and filled with a patch solution containing the following (in mM):

20 KCl, 10 HEPES, 0.5 EGTA, 2 MgCl2, 2 Na-ATP, and 120 K-gluconate, pH 7.4,

290–300 mOsm. Whole-cell recordings were made from SNc dopaminergic neurons visualized using infrared differential interference contrast videomicroscopy

(QImaging Retiga camera; Olympus BX51WI microscope), and were identified based on their location, large soma size (>30µm), and electrophysiological profile (regular slow pacemaking activity, large spike half-width, large sag in response to hyperpolarizing current steps). For voltage-clamp experiments, only whole-cell recordings with an uncompensated series resistance <7 MOhm (compensated 85–

90%) were included in the analysis. For current-clamp pharmacology experiments, higher series resistances were tolerated as long as the bridge compensation was properly adjusted to 100%. Liquid junction potential (-13.2 mV) and capacitive currents were compensated on-line. Recordings were acquired at 50kHz and were filtered with a low-pass filter (Butterworth characteristic 8.4 kHz cutoff frequency for current-clamp recordings and Bessel characteristic 5kHz cutoff frequency for voltage- clamp recordings). For current-clamp recordings, 1s hyperpolarizing current steps were injected to elicit a hyperpolarization-induced sag (due to IH activation).

Current-clamp recordings and protocols

25 The spontaneous firing frequency was calculated from a minimum of 30 seconds of stable current-clamp recording (with no injected current) within the first 5 minutes of obtaining the whole-cell configuration. The coefficient of variation of the interspike interval (CVISI) was extracted from the same recording. Action potentials (APs) generated during this period of spontaneous activity were averaged and several parameters were extracted: AP threshold, AP amplitude, AP duration at half of its maximal height (AP half-width), AHP trough voltage, AHP latency. Hyperpolarizing current steps and depolarizing current steps were used to characterize the post- inhibitory rebound (Amendola et al, 2012) and the excitability properties (Dufour et al, 2014a; Vandecasteele et al, 2011). The gain start, gain end and spike frequency adaptation (SFA) index used to define excitability were calculated as described before

(Dufour et al, 2014a).

Voltage-clamp recordings

For voltage-clamp recordings of ISK, a hybrid-clamp protocol was used to elicit maximal calcium-activated potassium current. This protocol consisted of a depolarization step (0 mV; 500 ms) from a holding potential of -70 mV. During this strong depolarization step, unclamped action currents were triggered that allowed large calcium entries necessary for SK channels to activate. Returning to holding potential (-70 mV) for 2500 ms allowed calcium-activated potassium channels to fully deactivate. The apamin-sensitive current was then isolated by repeating this stimulation protocol during apamin superfusion: the apamin-sensitive potassium current elicited by SK channels was obtained by substraction of the potassium current after apamin from the total potassium current recorded before apamin application.

Measurements of ISK properties (amplitude, time-to-peak, charge) were extracted from subtracted traces and the decay time constant was obtained from a mono-

26 exponential fit. For voltage-clamp recordings of IA and the delayed rectifier current

(IKDR), tetrodotoxin (1 µM, Alomone), nickel (200 µM, Sigma-Aldrich) and cadmium

(400 µM, Sigma-Aldrich) were also added to the aCSF. The IA current was elicited by a protocol consisting in a 500 ms prestep at -100 mV (to fully de-inactivate IA) followed by a 500 ms voltage step to -40 mV (to activate IA without eliciting delayed rectifier potassium currents). The current generated by the same protocol using a prestep at -40mV (to fully inactivate IA) was subtracted to isolate IA. IA properties

(peak amplitude and total charge) were measured after subtracting the baseline at -

40mV. The inactivation time constant was extracted from a mono-exponential fit of the decay of the current. IKDR was elicited by using a protocol consisting of a prestep at -30mV (to fully inactivate IA) followed by incremental depolarizing voltage steps up to +40mV.

Data acquisition

Data were acquired using an EPC 10 USB patch-clamp amplifier (HEKA) and the

Patchmaster software acquisition interface (HEKA). Analysis was performed using

FitMaster v2x73 (Heka).

Statistics

Behavior

The statistical analysis of actimetry and rotarod behaviors consisted in unpaired t-tests between SK3 KO or Kv4.3 KO mice and their respective WT littermates. For actimetry assessment, the numbers of horizontal (locomotion) and vertical (rearing) photobeam breaks were measured and compared between genotypes. Data are represented as line and scatter plots for the number of horizontal and vertical photobeam breaks per 5-min bin. The number of total movements normalized to control littermate was then used to determine significant changes in spontaneous

27 locomotion. For motor learning on the accelerating rotarod, the average latency to fall off the rod was measured for each trial. Statistical differences in motor learning were assessed by comparing the performance index. Data are represented as line and scatter plots for the average latency to fall off the rod.

Electrophysiology

The univariate statistical analysis of electrophysiological data, performed according to the distribution properties of the data, included standard sigmoidal fitting procedure, paired t-test or Wilcoxon signed rank test, one–way ANOVA and two-way repeated measures ANOVA with genotype and pharmacological condition as the independent factors (all conducted using SigmaPlot 11.0, Jandel Scientific), with p < 0.05 considered to be statistically significant. One-way ANOVA tests were followed by adapted post-hoc tests between groups (Holm-Sidak or Bonferroni’s test), with p<0.05 considered as statistically significant (SigmaPlot 11.0, GraphPad Prism 6). In most figures, data are represented as scatter plots or box and whisker plots, with all individual points appearing on the graphs. For rotarod experiments and pharmacological experiments, data are represented as mean ± SEM (scatter or bar plots). Multivariate analysis included Linear Discriminant Analysis (LDA) and t- distributed Stochastic Neighbor Embedding (t-SNE). LDA and t-SNE were run in R using the MASS and the Rtsne packages, respectively. Before performing LDA, 11 variables were selected among the 16 original electrophysiological parameters by removing redundant parameters. Then the 8 most relevant electrophysiological variables (Spontaneous firing frequency, CVISI, AP amplitude, AP threshold, AP half- width, AHP trough voltage, AHP latency and rebound delay) were selected among 11 using a stepwise forward/backward model using the Wilk's lambda criterion (klaR package in R, p threshold=0.05). For LDA, the mean vector difference between

28 groups was tested using the Rao's approximation of the Wilk's lambda test. Cross- validation of the confusion matrix was calculated with 15 randomly selected neurons.

LDA was run on WT, SK3 KO and Kv4.3 KO groups only, as equal group sizes are required for efficient analysis. t-SNE was run on the same data set (same 8 electrophysiological variables), to which we added measurements from acute blockade experiments (apamin and AmmTX3 application). Euclidean distances were used as a measure of similarity. Perplexity was varied from 10 to 30 with a number of iterations set to 1000 to check for stability, and otherwise default option values (Rtsne function, Rtsne package in R) were used. Figures were prepared using SigmaPlot

11.0, GraphPad Prism 6 and Adobe Illustrator CS5.

ACKNOWLEDGEMENTS

This work was supported by the European Research Council (Consolidator grant

616827 CanaloHmics to J.M.G., supporting A.H-H. and M.T.) and the Fondation de

France (grant 00076344 to J-M.G. and M.A., supporting A.H-H.). We thank O.

Toutendji and C. Iborra-Bonnaure for technical assistance.

REFERENCES

Amendola J, Woodhouse A, Martin-Eauclaire MF, Goaillard JM (2012) Ca2+/cAMP- sensitive covariation of IA and IH voltage dependences tunes rebound firing in dopaminergic neurons. The Journal of Neuroscience 32: 2166-2181

Balleine BW (2019) The meaning of behavior: discriminating reflex and volition in the brain. Neuron 104: 47-62

Barbaric I, Miller G, Dear TN (2007) Appearances can be deceiving: phenotypes of knockout mice. Briefings in functional genomics & proteomics 6: 91-103

Beeler J, Frank MJ, McDaid J AE, Turkson S, Bernardez Sarria MS, McGehee DS, Zhuang X (2012) A role for dopamine-mediated learning in the pathophysiology and treatment of Parkinson’s disease. Cell Reports 2: 1747-1761

29 Birnbaum SG, Varga AW, Yuan LL, Anderson AE, Sweatt JD, Schrader LA (2004) Structure and function of Kv4-family transient potassium channels. Physiological reviews 84: 803-833

Blythe SN, Wokosin D, Atherton JF, Bevan MD (2009) Cellular mechanisms underlying burst firing in substantia nigra dopamine neurons. The Journal of neuroscience 29: 15531-15541

Cai X, Liang C, Muralidharan S, Kao J, Tang C, Thompson S (2004) Unique roles of SK and Kv4.2 potassium channels in dendritic integration. Neuron 44: 351-364

Carrasquillo Y, Burkhalter A, Nerbonne J (2012) A-type K+channels encoded by Kv4.2, Kv4.3 and Kv1.4 differentially regulate intrinsic excitability of cortical pyramidal neurons. The Journal of Physiology 590: 3877-3890 de Vrind V, Scuvée-Moreau J, Drion G, Hmaied C, Philippart F, Engel D, Seutin V (2016) Interactions between calcium channels and SK channels in midbrain dopamine neurons and their impact on pacemaker regularity: Contrasting roles of N- and L-type channels. European journal of pharmacology 788: 274-279

Deignan J, Lujan R, Bond C, Riegel A, Watanabe M, Williams JT, Maylie J, Adelman JP (2012) SK2 and SK3 expression differentially affect firing frequency and precision in dopamine neurons. Neuroscience 217: 67-76

Ding S, Matta SG, Zhou FM (2011) Kv3-like potassium channels are required for sustained high-frequency firing in basal ganglia output neurons. Journal of neurophysiology 105: 554-570

Drion G, Massotte L, Sepulchre R, Seutin V (2011) How modeling man reconcile apparently discrepant experimental results: the case of pacemaking in dopaminergic neurons. Plos Computational Biology 7: e1002050

Drion G, O'Leary T, Marder E (2015) Ion channel degeneracy enables robust and tunable neuronal firing rates. Proceedings of the National Academy of Sciences of the United States of America 112: E5361-5370

Dufour MA, Woodhouse A, Amendola J, Goaillard JM (2014a) Non-linear developmental trajectory of electrical phenotype in rat substantia nigra pars compacta dopaminergic neurons. Elife 3: e04059

Dufour MA, Woodhouse A, Goaillard JM (2014b) Somatodendritic ion channel expression in substantia nigra pars compacta dopaminergic neurons across postnatal development. Journal of neuroscience research 92: 981-999

Durieux PF, Schiffmann SN, de Kerchove d'Exaerde A (2012) Differential regulation of motor control and response to dopaminergic drugs by D1R and D2R neurons in distinct dorsal striatum subregions. The EMBO journal 31: 640-653

30 Edelman GM, Gally JA (2001) Degeneracy and complexity in biological systems. Proceedings of the National Academy of Sciences of the United States of America 98: 13763-13768

Fakler B, Adelman JP (2008) Control of K(Ca) channels by calcium nano/microdomains. Neuron 59: 873-881

Giordano N, Iemolo A, Mancini M, Cacace F, De Risi M, Latagliata EC, Ghiglieri V, Bellenchi GC, Puglisi-Allegra S, Calabresi P, Picconi B, De Leonibus E (2018) Motor learning and metaplasticity in striatal neurons: relevance for Parkinson's disease. Brain 141: 505-520

Grace AA, Bunney BS (1984) The control of firing pattern in nigral dopamine neurons: burst firing. The Journal of Neuroscience 4: 2877-2890

Hahn J, Tse TE, Levitan ES (2003) Long-term K+ channel-mediated dampening of dopamine neuron excitability by the antipsychotic drug haloperidol. The Journal of Neuroscience 23: 10859-10866

Jacobsen JP, Weikop P, Hansen HH, Mikkelsen JD, Redrobe JP, Holst D, Bond CT, Adelman JP, Christophersen P, Mirza NR (2008) SK3 K+ channel-deficient mice have enhanced dopamine and serotonin release and altered emotional behaviors. Genes, brain, and behavior 7: 836-848

Khaliq ZM, Bean BP (2008) Dynamic, nonlinear feedback regulation of slow pacemaking by A-type potassium current in ventral tegmental area neurons. The Journal of Neuroscience 28: 10905-10917

Kimm T, Khaliq ZM, Bean BP (2015) Differential regulation of action potential shape and burst-frequency firing by BK and Kv2 channels in substantia nigra dopaminergic neurons. The Journal of Neuroscience 35: 16404-16417

Kirschner M, Gerhart J (1998) Evolvability. Proceedings of the National Academy of Sciences of the United States of America 95: 8420-8427

Kita T, Kita H, Kitai ST (1986) Electrical membrane properties of rat substantia nigra compacta neurons in an in vitro slice preparation. Brain Research 372: 21-30

Kitai ST, Shepard PD, Callaway JC, Scroggs R (1999) Afferent modulation of dopamine neuron firing patterns. Current Opinion in Neurobiology 9: 690-697

Kitano H (2004) Biological robustness. Nature Reviews Genetics 5: 826-837

Klassen T, Davis C, Goldman A, Burgess D, Chen T, Wheeler D, McPherson J, Bourquin T, Lewis L, Villasana D, Morgan M, Muzny D, Gibbs R, Noebels J (2011) Exome sequencing of ion channel genes reveals complex profiles confounding personal risk assessment in epilepsy. Cell 145: 1036-1048

31 Kravitz AV, Freeze BS, Parker PR, Kay K, Thwin MT, Deisseroth K, Kreitzer AC (2010) Regulation of parkinsonian motor behaviours by optogenetic control of basal ganglia circuitry. Nature 466: 622-626

Kulik Y, Jones R, Moughamian AJ, Whippen J, Davis GW (2019) Dual separable feedback systems govern firing rate homeostasis. Elife 8: e45717

Liss B, Franz O, Sewing S, Bruns R, Neuhoff H, Roeper J (2001) Tuning pacemaker frequency of individual dopaminergic neurons by Kv4.3L and KChip3.1 transcription. The EMBO journal 20: 5715-5724

MacLean JN, Zhang Y, Johnson BR, Harris-Warrick RM (2003) Activity- independent homeostasis in rhythmically active neurons. Neuron 37: 109-120

Marder E, Goaillard J-M (2006) Variability, compensation and homeostasis in neuron and network function. Nature Review Neuroscience 7: 563-574

Mercuri NB, Bonci A, Calabresi P, Stefani A, Bernardi G (1995) Properties of the hyperpolarization-activated cation current Ih in rat midbrain dopaminergic neurons. The European journal of neuroscience 7: 462-469

Nedergaard S (1999) Regulation of action potential size and excitability in substantia nigra compacta neurons: sensitivity to 4-aminopyridine. Journal of neurophysiology 82: 2903-2913

Nerbonne J, Gerber B, Norris A, Burkhalter A (2008) Electrical remodelling maintains firing properties in cortical pyramidal neurons lacking KCND2-encoded A- type K+currents. The Journal of Physiology 586: 1565-1579

Neuhoff H, Neu A, Liss B, Roeper J (2002) I(h) channels contribute to the different functional properties of identified dopaminergic subpopulations in the midbrain. The Journal of Neuroscience 22: 1290-1302

O'Leary T, Williams AH, Franci A, Marder E (2014) Cell types, network homeostasis, and pathological compensation from a biologically plausible ion channel expression model. Neuron 88: 1308

Paladini CA, Tepper JM (2016) Neurophysiology of substantia nigra dopamine neurons: modulation by GABA and glutamate. Handbook of Behavioral Neuroscience 24: 335-360

Podlaski WF, Seeholzer A, Groschner LN, Miesenbock G, Ranjan R, Vogels TP (2017) Mapping the function of neuronal ion channels in model and experiment. Elife 6: e22152

Prinz AA, Bucher D, Marder E (2004) Similar network activity from disparate circuit parameters. Nature Neuroscience 7: 1345-1352

Schulz DJ, Goaillard JM, Marder EE (2007) Quantitative expression profiling of identified neurons reveals cell-specific constraints on highly variable levels of gene

32 expression. Proceedings of the National Academy of Sciences of the United States of America 104: 13187-13191

Serodio P, Rudy B (1998) Differential expression of Kv4 K+ channel subunits mediating subthreshold transient K+ (A-type) currents in rat brain. Journal of neurophysiology 79: 1081-1091

Serodio P, Vega-Saenz de Miera E, Rudy B (1996) Cloning of a novel component of A-type K+ channels operating at subthreshold potentials with unique expression in heart and brain. Journal of neurophysiology 75: 2174-2179

Seutin V, Johnson SW, North RA (1993) Apamin increases NMDA-induced burst- firing of rat mesencephalic dopamine neurons. Brain Research 630: 341-344

Soden M, Jones G, Sanford C, Chung A, Güler A, Chavkin C, Luján R, Zweifel L (2013) Disruption of dopamine neuron activity pattern regulation through selective expression of a human KCNN3 mutation. Neuron 80: 997-1009

Swensen A, Bean B (2005) Robustness of burst firing in dissociated Purkinje neurons with acute or long-term reductions in sodium conductance. The Journal of Neuroscience 25: 3509-3520

Tapia M, Baudot P, Formisano-Treziny C, Dufour MA, Temporal S, Lasserre M, Marqueze-Pouey B, Gabert J, Kobayashi K, Goaillard JM (2018) Neurotransmitter identity and electrophysiological phenotype are genetically coupled in midbrain dopaminergic neurons. Scientific Reports 8: 13637

Tarfa RA, Evans RC, Khaliq ZM (2017) Enhanced sensitivity to hyperpolarizing inhibition in mesoaccumbal relative to nigrostriatal dopamine neuron subpopulations. The Journal of Neuroscience 37: 3311-3330

Taylor AL, Goaillard JM, Marder E (2009) How multiple conductances determine electrophysiological properties in a multicompartment model. The Journal of Neuroscience 29: 5573-5586

Vacher H, Diochot S, Bougis P, Martin-Eauclaire M-F, Mourre C (2006) Kv4 channels sensitive to BmTX3 in rat nervous system: autoradiographic analysis of their distribution during brain ontogenesis. European Journal of Neuroscience 24: 1325- 1340

Vacher H, Romi-Lebrun R, Mourre C, Lebrun B, Kourrich S, Masméjean F, Nakajima T, Legros C, Crest M, Bougis P, Martin-Eauclaire M (2001) A new class of scorpion toxin binding sites related to an A-type K+ channel: pharmacological characterization and localization in rat brain. FEBS letters 501: 31-36

Van der Maaten L, Hinton G (2008) Visualizing data using t-sne. Journal of Machine Learning Research 9: 2579-2605

33 Vandecasteele M, Deniau JM, Venance L (2011) Spike frequency adaptation is developmentally regulated in substantia nigra pars compacta dopaminergic neurons. Neuroscience 192: 1-10

Weatherall KL, Goodchild SJ, Jane DE, Marrion NV (2010) Small conductance calcium-activated potassium channels: from structure to function. Progress in neurobiology 91: 242-255

Weatherall KL, Seutin V, Liegeois JF, Marrion NV (2011) Crucial role of a shared extracellular loop in apamin sensitivity and maintenance of pore shape of small- conductance calcium-activated potassium (SK) channels. Proceedings of the National Academy of Sciences of the United States of America 108: 18494-18499

Wise R (2004) Dopamine, learning and motivation. Nature Review Neuroscience 5: 483-494

Wolfart J, Neuhoff H, Franz O, Roeper J (2001) Differential expression of the small- conductance, calcium-activated potassium channel SK3 is critical for pacemaker control in dopaminergic midbrain neurons. The Journal of Neuroscience 21: 3443- 3456

Wolfart J, Roeper J (2002) Selective coupling of T-type calcium channels to SK potassium channels prevents intrinsic bursting in dopaminergic midbrain neurons. The Journal of Neuroscience 22: 3404-3413

Yin HH, Knowlton BJ (2006) The role of the basal ganglia in habit formation. Nature Reviews Neuroscience 7: 464-476

Yin HH, Mulcare SP, Hilário MRF, Clouse E, Holloway T, Davis MI, Hansson AC, Lovinger DM, Costa RM (2009) Dynamic reorganization of striatal circuits during the acquisition and consolidation of a skill. Nature Neuroscience 12: 333-341

34 A day 0 day 5 day 10 day 15

Basal actimetry Motor learning (Rotarod) B SK3-/- (n=12) SK3 WT littermate (n=12) 35 35 Kv4.3-/- (n=13) 30 Kv4.3 WT littermate (n=10) 30 25 25 20 20 15 15 Rearing

Locomotion 10 10 5 5

0 (# of vertical beambreaks) 0

(# of horizontal beambreaks) 0 10 20 30 40 50 60 70 80 90 0 10 20 30 40 50 60 70 80 90 Time (min) Time (min)

n.s. n.s. n.s. 200 300 *

150 200 % of control) (

100 % of control) (

100 50

(n=12) (n=12) (n=10) (n=13) Rearing Locomotion 0 0 -/- -/- -/- -/- 3 .3 3 .3 K 4 K 4 SK3 S Kv4.3 v SK3 S Kv4.3 v K K

WT littermate WT littermate WT littermate WT littermate C n.s. 100 400 100 400 ** x (%) 80 80 x (%) 300 300 nd e nd e i i

60 60 e e c 200 c 200 n n a 40 40 a Latency (s) Latency (s) m m r 100 r 100 o o f 20 WT (n=12) 20 WT (n=11) f r r SK3-/- n=12) Kv4.3-/- (n=13) P e

( P e 0 0 0 0 3 4 5 6 7 8 1 2 3 1 2 9 10 4 5 6 7 8 9 10 Trial Trial

Figure 1. Assessment of motor behavior in SK3 and Kv4.3 KO mice. A, schematic representing the timeline of behavioral experiments performed on SK3, Kv4.3 KO mice and WT littermates. Mice were habituated to their new environment on day 0, actimetry was tested on day 5 and rotarod learning assessed on day 15. B, changes in basal locomotion measured in actimetry chambers. Top left, line and scatter plot showing the mean number of horizontal movements per 5-min bin in SK3 and Kv4.3 mice and their WT littermates. Bottom left, scatter plot showing the total number of horizontal movements per 90-min session in SK3 and Kv4.3 KO mice normalized to their WT littermate average. Top right, line and scatter plot showing the mean number of rearing events per 5-min bin in SK3 and Kv4.3 mice and their WT littermates. Bottom right, scatter plot showing the total number of rearing events per 90-min session in SK3 and Kv4.3 KO mice normalized to their WT littermate average. C, changes in motor learning measured using a rotarod assay. The latency to falling o the rotating rod (with increasing rotating speed) was assessed over 10 consecutive trials (shown on the line and scatter plots). The performance index ((average latency 8-10)/(average latency 1-3)×100) was then used to evaluate the learning ability of SK3 (left scatter plot) and Kv4.3 KO mice (right scatter plot). ** p<0.01, * p<0.05, n.s. non-signicant. A WT SK3-/- Kv4.3-/-

-60 mV -60 mV

B 6 *** 35

) 30

z 5 H (

25 y

4 ) * %

( 20

u e n c 3 ISI e q

V 15 r f 2 C 10

Firing 1 5

0 0 WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- (n=26) (n=33) (n=29) (n=26) (n=33) (n=29)

Figure 2. Changes in pacemaking in SNc DA neurons in SK3 and Kv4.3 KO mice. A, current-clamp recordings showing the spontaneous pattern of activity in WT (left, black trace), SK3 KO (middle, green trace) and Kv4.3 KO mice (right, red trace). B, left, box and whisker plot showing the distribution of values for spontaneous ring frequency in WT, SK3 and Kv4.3 KO mice. Right, box and whisker plot showing the distribution of values for CVISI in WT, SK3 and Kv4.3 KO mice. * p<0.05, *** p<0.001. Scale bars: A, horizontal 1s, vertical 20mV. Horizontal gray dotted lines indicate -60mV. A WT SK3-/- Kv4.3-/-

fAHP mAHP -40 mV

B -25 * 90 3.0 -30 80 2.5 -35 70 2.0 -40 60 1.5 -45 50 1.0 -50 AP half-width (ms) AP threshold (mV) AP amplitude (mV) -55 40 0.5

-60 30 0.0 WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- (n=25) (n=33) (n=29) (n=25) (n=33) (n=29) (n=25) (n=33) (n=29) C -50 -50 -50 * n.s. (p=0.079)

-60 -60 -60

-70 -70 -70 AHP trough (mV) fAHP trough (mV) mAHP trough (mV) -80 -80 -80

WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- (n=25) (n=33) (n=29) (n=18) (n=25) (n=21) (n=16) (n=14) (n=11) D 200 50 200 * 40 150 150

30 100 100 20

50 50 AHP latency (ms) fAHP latency (ms)

10 mAHP latency (ms)

0 0 0 WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- (n=25) (n=33) (n=29) (n=18) (n=25) (n=21) (n=16) (n=14) (n=11)

Figure 3. Changes in action potential shape in SK3 and Kv4.3 KO mice. A, current-clamp recordings showing the shape of the action potential in WT (black traces), SK3 KO mice (green traces) and Kv4.3 KO mice (red traces) on a slow (left) and fast time-scale (right). The inset shows the dominance of the fAHP (black trace) or the mAHP (gray trace) in two di erent neurons at a higher magnication (action potentials were truncated for illustration purposes). B, box and whisker plots showing the distribution of values for the AP threshold (left), AP amplitude (center) and AP half-width (right) in WT, SK3 and Kv4.3 KO mice. C, box and whisker plots representing the distribution of values for AHP (left), fast AHP (fAHP, center) and medium AHP (mAHP, right) trough voltages in WT, SK3 and Kv4.3 KO mice. D, box and whisker plots representing the distribution of values for AHP (left), fAHP (center) and mAHP latencies in WT, SK3 and Kv4.3 KO mice. * p<0.05, n.s. non-signicant. Scale bars: A, left, horizontal 50ms, vertical 20mV. Inset, horizontal 50 ms, vertical 10mV. Right, horizontal 2ms, vertical 20mV. Horizontal gray dotted lines indicate -40mV. A WT SK3-/- Kv4.3-/-

B 45 1400 *** 40 1200

35 ) s m 1000 30 25 800 20 600 15 400

Voltage sag (mV) 10 ( y a l e d d n u o b e R 5 200 0 0 WT SK3-/- Kv4.3-/- WT SK3-/- Kv4.3-/- (n=28) (n=32) (n=33) (n=25) (n=28) (n=32)

Figure 4. Changes in post-inhibitory rebound properties in SK3 and Kv4.3 KO mice. A, current-clamp recordings showing the voltage response of SNc DA neurons to a current step (bottom gray traces) hyperpolarizing membrane voltage to ~-120mV in WT (black trace), SK3 (green trace) and Kv4.3 KO mice (red trace). B, box and whisker plots representing the distribution of values for voltage sag amplitude (left) and post-inhibitory rebound delay (right) in WT, SK3 and Kv4.3 KO mice. *** p<0.001. Scale bars: A, horizontal 500ms, vertical 20mV. 200ms, vertical 20mV.200ms, vertical gainend(center) (left), andSFAfor gainstart index (right)in WT, SK3andKv4.3 KO mice. bars: A,horizontal *p<0.05. Scale calculated for theseneurons isindicated above theplot. B,bow andwhiskerplots representing ofvalues thedistribution corresponding potentials to thetrains shown ofaction onthetop row. adaptationThe (SFA) spike frequency index potentials.each train of action Bottom, line and scatter plots representing the instantaneous frequency ring mice (red trace). Gray dotted indicate rectangles and gain end on the interspikeused to calculate intervals the gain start response ofSNcDA neurons to current adepolarizing step (gray traces) (blacktrace), in SK3(green trace)WT andKv4.3 KO Figure 5.Changes inexcitability inSK3andKv4.3 KO mice. A,top, recordings current-clamp showing thevoltage B A

Gain start (Hz/100pA) Frequency (Hz) 1 1 2 2 3 3 10 15 20 25 0 5 0 5 0 5 0 5 0 5 Gain start 0 (n=14) (SFA index=2.18) WT 0.2 Time (s) * 0.4 SK3 (n=22) WT 0.6 Gain end -/- 0.8 Kv4.3 (n=28) 1 -/-

Gain end (Hz/100pA) 10 15 20 25 0 5 1 1 1 0 2 4 6 8 0 2 4 0 (SFA index=2.52) (n=14) 0.2 WT Time (s) SK3 0.4 SK3 (n=22) 0.6 -/- -/- 0.8 Kv4.3 (n=28) 1 -/-

SFA index 10 15 20 25 0 5 0 2 4 6 8 0 (SFA index=1.69) (n=14) WT 0.2 Kv4.3 Time (s) 0.4 SK3 (n=22) 0.6 -/- -/- 0.8 Kv4.3 (n=28) 1 -/- A WT apamin

-40 mV Figure 6. Comparing the changes in AHP properties during acute blockade or chronic deletion of SK channels. A, current-clamp recordings showing the shape of the action potential in SNc DA neurons in WT mice in control condition (black traces) and after apamin B application (green traces) on a slow (left) and fast -50 -60 time-scale (right). B, top left, line and scatter plot -62 -55 *** *** representing the e ect of apamin on fAHP trough -64 n.s. voltage in individual WT SNc DA neurons. Top -60 -66 right, line and scatter plot comparing the average -65 -68 -70 change in fAHP trough after apamin application -70 -72 (left) or SK3 channel deletion (right, same data as -75 -74 Figure 3C, center, represented in a di erent fAHP trough (mV) fAHP trough (mV) -76 -80 -78 manner). Bottom left, line and scatter plot (n=8) -85 -80 representing the e ect of apamin on mAHP WT apamin WT apamin WT SK3-/- trough voltage in individual WT SNc DA neurons. Bottom right, line and scatter plot comparing the average change in mAHP trough after apamin -50 *** -60 *** -62 application (left) or SK3 channel deletion (right, -55 -64 n.s. same data as Figure 3C, right, represented in a -60 -66 (p=0.079) di erent manner). C, left, line and scatter plot -65 -68 representing the e ect of apamin on AHP latency -70 in individual WT SNc DA neurons. Right, bar plot -70 -72 -75 -74 comparing the average change in AHP latency

mAHP trough (mV) mAHP trough (mV) -76 after apamin application (left) or SK3 channel -80 -78 (n=9) deletion (right, same data as Figure 3D, left, -85 -80 WT apamin WT apamin WT SK3-/- represented in a di erent manner). *** p<0.001, n.s. non-signicant. Scale bars: A, left, horizontal C 50ms, vertical 20mV. A, right, horizontal 2ms, n.s. vertical 20mV. Horizontal gray dotted lines 100 *** 120 *** indicate -40mV. 100 80

80 60 60 40 40

AHP latency (ms) 20 20

(n=13) AHP latency (% of control) 0 0 WT apamin WT apamin WT SK3-/- A WT apamin

-60 mV Figure 7. Comparing the changes in ring regularity and excitability during acute blockade or chronic deletion of SK channels. A, B current-clamp recordings showing the 140 (n=12) 1000 *** spontaneous pattern of activity of a WT SNc DA 120 *** 750 neuron in control condition (black trace, left) and 100 after apamin application (green trace, right). B,

(%) 80 500 left, line and scatter plot showing the change in ISI 60

(% of control) CVISI induced by apamin application in individual CV 40 ISI * WT SNc DA neurons. Right, bar plot comparing

CV 250

20 the average change in CVISI after apamin 0 0 application (left) or SK3 channel deletion (right, WT apamin WT apamin WT SK3-/- same data as Figure 2B, right, represented in a di erent manner). C, current-clamp recordings

C WT apamin showing the train of action potentials induced by Gain start Gain end a 140pA depolarizing current step (gray trace) in a WT SNc DA neuron in control condition (black trace, left) or in the presence of apamin (green trace, right). Gray dotted rectangles indicate the interspike intervals used to calculate the gain start and gain end on each train of action potentials. D, top left, line and scatter plot showing the change in gain start induced by apamin application in individual WT SNc DA neurons. Top right, bar plot showing the average D change in gain start after apamin application (left) 50 *** *** 250 or SK3 channel deletion (right, same data as 40 * Figure 5B, left, represented in a di erent 200 manner). Bottom left, line and scatter plot 30 150 showing the change in gain end induced by 20 apamin application in individual WT SNc DA 100 neurons. Bottom right, bar plot showing the 10 Gain start (Hz/100pA) 50 average change in gain end after apamin Gain start (% of control) (n=14) application (left) or SK3 channel deletion (right, 0 0 WT apamin WT apamin WT SK3-/- same data as Figure 5B, center, represented in a di erent manner). * p<0.05, ** p<0.01, *** 18 ** ** p<0.001, n.s. non-signicant. Scale bars: A, 16 250 horizontal 1s, vertical 20mV. C, horizontal 200ms, 14 200 vertical 20mV. Horizontal gray dotted lines in A 12 n.s. indicate -60mV. 10 150 8 6 100 4

Gain end (Hz/100pA) 50 2 Gain end (% of control) (n=14) 0 0 WT apamin WT apamin WT SK3-/- A WT AmmTX3

-60 mV

B ** ** *** 5 200

4 150

3 100 2 50 1 Firing frequency (Hz) (n=9)

0 Firing frequency (% of control) 0 WT AmmTX3 WT AmmTX3 WT Kv4.3-/-

C WT AmmTX3

D

600 *** 150 (n=18) *** *** 500 125

400 100

300 75

200 50

Rebound delay (ms) 100 25 Rebound delay (% of control) 0 0 WT AmmTX3 WT AmmTX3 WT Kv4.3-/-

Figure 8. Comparing the changes in spontaneous activity and post-inhibitory rebound during acute blockade or chronic deletion of Kv4 channels. A, current-clamp recordings showing the spontaneous pattern of activity of a WT SNc DA neuron in control condition (black trace, left) and after AmmTX3 application (red trace, right). B, left, line and scatter plot showing the change in spontaneous ring frequency induced by AmmTX3 application in individual WT SNc DA neurons. Right, bar plot comparing the average change in spontaneous ring frequency after AmmTX3 application (left) or Kv4.3 channel deletion (right, same data as Figure 2B, left, represented in a dierent manner). C, current-clamp recordings showing the voltage response of a WT SNc DA neuron to a hyperpolarizing current step (bottom gray traces) in control condition (left, black trace) and after AmmTX3 application (right, red trace). D, left, line and scatter plot showing the change in rebound delay induced by AmmTX3 application in individual WT SNc DA neurons. Right, bar plot showing the average change in rebound delay after AmmTX3 application (left) or Kv4.3 channel deletion (right, same data as Figure 4B, right, represented in a dierent manner). ** p<0.01, *** p<0.001. Scale bars: A, horizontal 1s, vertical 20mV. C, horizontal 500ms, vertical 20mV. Horizontal gray dotted lines in A indicate -60mV. A 300 700 *** 100 WT 250 600 80 500 200 ** n.s. SK3-/- 400 60 150 300 40

100 charge (pA.s) amplitude (pA) 200

0 mV SK SK I (500 ms) I 50 20

decay time constant (ms) 100 SK 0 I 0 0 -70 mV -70 mV WT SK3-/- WT SK3-/- WT SK3-/- (n=15) (n=26) (n=15) (n=25) (n=14) (n=17) B Kv4.3-/- )

1000 s 500 140 m AmmTX3 *** 120 WT 800 400 ) ** )

s ***

. 100 p A

Kv4.3-/- 600 300 p A 80

400 200 60 charge (

amplitude ( 40 A A I -40 mV I 200 100 -60 mV tivation time constant ( 20

0 i n a c 0 0 A -100 mV WT Kv4.3-/- I WT Kv4.3-/- WT Kv4.3-/- (500 ms) (n=14) (n=28) (n=14) (n=28) (n=14) (n=28)

C steady peak state

8 8

6 6

4 4

2 2 Peak current (nA) +20mV WT (n=9) WT (n=9) Kv4.3-/- (n=11) Steady state current (nA) Kv4.3-/- (n=11) -30mV 0 0 -40 -30 -20 -10 0 10 20 30 -40 -30 -20 -10 0 10 20 30 -60mV -60mV Holding voltage (mV) Holding voltage (mV)

Figure 9. Voltage-clamp analysis of IA and ISK in SK3 and Kv4.3 KO mice. A, properties of ISK in SNc DA neurons in WT and SK3 KO mice. Left, voltage-clamp recordings of ISK obtained from SNc DA neurons in a WT (black trace) and an SK3 KO mouse (green trace) in response to a 'hybrid-clamp' holding voltage protocol (gray trace). Right, box and whisker plots showing the distribution of values for ISK amplitude (left), time constant of decay (center) and charge (right) in WT and SK3 KO mice. B, properties of IA in SNc DA neurons in WT and Kv4.3 KO mice. Left, voltage-clamp recordings of IA obtained from SNc DA neurons in a WT (black trace) and a Kv4.3 KO mouse (red trace) in response to a voltage step to -40mV (gray trace). The small residual IA current present in the Kv4.3 KO mice is blocked by AmmTX3 (inset, light red trace). Right, box and whisker plots showing the distribution of values for IA amplitude (left), time constant of inactivation (center) and charge (right) in WT and Kv4.3 KO mice. C, properties of the delayed recti er potassium current (IKDR) in SNc DA neurons in WT and Kv4.3 KO mice. Left, voltage-clamp recordings of IKDR obtained in a WT (black trace) and a Kv4.3 KO mouse (red trace) in response to a voltage step to +20mV (gray trace). The peak and steady-state components of IKDR are indicated by arrowheads. Right, line and scatter plots representing the average current-voltage relationships of the peak (left) and steady-state IKDR (right) obtained from 9 WT (gray circles) and 11 Kv4.3 SNc DA neurons (red circles). Data are represented as mean ± SEM. ** p<0.01, *** p<0.001, n.s. non-signi cant. Scale bars: A, horizontal 500ms, vertical 50pA. B, horizontal 200ms, vertical 200pA. C, horizontal 200ms, vertical 1nA. horizontal 20mV. 1s, vertical Horizontal gray dotted linesinAand C indicate -40mVand-60mV, respectively. horizontal 20mV. 50ms, bars:A,left, (right). *p<0.05,***p<0.001. Scale vertical horizontal Right, 20mV. 2ms, vertical C, SK3 KO mice barplotshowing (right).Right, theaverage induced andSK3KO by changein CVISI apaminin (left) mice WT induced bychange inCVISI apamin application inindividualSNcDA neurons samedata asFigure (left, in and 7B, left) WT blacktrace)condition apaminapplication (left, (right, andafter green trace). lineandscatter D,left, plotshowing the recordingscurrent-clamp inSNcDA showing thespontaneous neurons pattern ofactivity inSK3KO mice in control application inindividualSNcDA neurons in samedata asFigure (left, andSK3KO 6C , left) miceWT (right).C, andSK3KOin (left) mice lineandscatter (right).Right, induced plotshowing by the changeinAHPlatency apamin WT and scatter plot showing the change in AHP trough voltage induced by apamin application in individual SNc DA neurons condition line (blacktraces) (right). B,left, apaminapplication (green andafter traces) andfasttime-scale onaslow (left) recordingsA, current-clamp potential showing inSNcDA theshapeofaction neurons inSK3KO mice incontrol Figure in potential 10.Comparing andspontaneous activity andSK3KO ofapaminonaction thee ect WT mice. D C B A 1 1 CVISI (%) 1 AHP trough (mV) 2 4 6 8 0 2 4 ------8 8 7 7 6 6 5 5 0 0 0 0 0 0 0 0 5 0 5 0 5 0 5 0 W W T T *** *** (n=13) apamin apamin (n=12) SK3 SK3 (n=15) *** *** -/- -/- (n=13) apamin apamin a a -60 mV -40 mV SK3 SK3 p p a a m m -/- -/- i i n n

CVISI (% of control) 1 AHP latency (ms) 2 4 6 8 0 1 1 1 1 0 0 0 0 0 20 40 60 80 20 40 60 0 0 0 0 0 0 0 0 0 W W T (n=13) T *** apamin apamin SK3 SK3 (n=15) -/- -/- * apamin apamin A B 4 1 F2

3 0.75 CVISI AP half-width AP 0.5 AHP amplitude 2 trough

) 0.25 % F1 3 0 1 4 .

( -6 -4 -2 4 6 Firing 2 -0.25 Rebound F frequency delay AHP -0.5 latency -2

-0.75 AP threshold -3 -1 WT (n=22)

-1 -0.75 -0.5 -0.25 0 0.25 0.5 0.75 1 -4 Kv4.3-/- (n=28) SK3-/- (n=26) F1 (85.67 %) C 20 WT (n=22) Kv4.3-/- (n=28) 15 SK3-/- (n=26) WT + apamin (n=6) 10 SK3-/- + apamin (n=6) Kv4.3 WT + AmmTX3 (n=9) chronic 5 deletion Kv4

tSNE2 0 acute blockade SK acute -5 blockade (WT & SK3-/-) SK3 WT -10 chronic deletion -15 -20 -15 -10 -5 0 5 10 15 20 -20 -15 -10 -5 0 5 10 15 20 tSNE1 tSNE1

Figure 11. Multivariate analysis of electrical phenotype. A, polar plot representing the respective contribution of each of the 8 electrophysiological parameters chosen for LDA to the two factors (F1, X axis; F2, Y axis) retained from the LDA. Spontaneous ring frequency (Firing frequency), CVISI, AP amplitude, AP threshold, AP half-width, AHP trough voltage (AHP trough) and AHP latency were extracted from spontaneous activity recordings. Rebound delay was extracted from hyperpolarizing current step protocols (see Figure 4). F1 and F2 accounted for 85.67 and 14.33% of the variance of the measurements, respectively. B, scatter plot representing the factor loadings of the WT (gray circles), SK3 KO (green circles) and Kv4.3 KO (red circles) in the F2 vs F1 space. All individual points are represented for each genotype, and the ellipses of corresponding colors indicate the 95% con dence interval of each distribution. The centroids of each group also appear as diamonds at the center of the ellipses. C, left, t-SNE representation of the distribution of the electrophysiological phenotype of individual SNc DA neurons in the WT, SK3 KO, Kv4.3 KO, WT+apamin, SK3 KO+apamin and WT+AmmTX3 conditions. Right, same representation using the average t-SNE1 and t-SNE2 values for each condition. The small ellipses around each point indicate the con dence interval of the mean. Colored arrows schematically indicate the trajectories from the WT group to the dierent treated groups. WT SK3 KO Kv4.3 KO ANOVA Electrophysiological parameters mean SD median IQR n mean SD median IQR n p mean SD median IQR n p F p Frequency (Hz) 1.80 0.57 1.70 0.74 25 1.85 1.02 1.57 1.04 32 b 3.03 0.88 3.11 0.82 28 b 28.554 p<0.001 Pacemaking n.s. p<0.001 b b CVISI (%) 6.00 2.00 5.68 2.41 25 9.08 0.71 7.82 6.38 32 p<0.05 7.24 5.83 5.60 4.12 28 n.s. 11.228 p=0.004 AP threshold (mV) -42.1 5.55 -43.3 5.51 25 -45.9 4.53 -46.1 4.43 33 p<0.05b -43.0 5.44 -44.0 6.20 29 n.s.b 10.93 p=0.004 AP amplitude (mV) 64.5 7.86 67.3 11.4 25 65.0 10.6 66.4 15.8 33 N/Aa 67.0 8.70 67.3 10.3 29 N/Aa 0.561 p=0.573 AP half-width (ms) 1.47 0.25 1.44 0.31 25 1.58 0.34 1.47 0.30 33 N/Ab 1.64 0.25 1.60 0.29 29 N/Ab 5.651 p=0.059 AHP trough voltage (mV) -73.9 5.03 -74.4 5.47 25 -70.3 4.94 -71.2 4.76 33 p<0.05b -71.1 5.38 -72.0 6.21 29 n.s.b 7.241 p=0.027 Action Potential fAHP trough voltage (mV) -72.0 4.17 -73.0 4.28 18 -69.6 5.06 -69.9 5.17 25 N/Ab -71.6 4.71 -72.7 5.29 21 N/Ab 3.662 p=0.160 mAHP trough voltage (mV) -74.6 5.54 -75.4 6.61 16 -71.4 3.98 -71.8 4.37 14 N/Aa -70.1 6.04 -71.4 8.29 11 N/Aa 2.717 p=0.079 AHP latency (ms) 50.7 29.8 66.6 55.6 25 39.2 35.7 20.4 40.9 33 N/Ab 33.5 19.1 24.0 32.0 29 N/Ab 3.375 p=0.085 fAHP latency (ms) 17.3 6.23 15.5 6.48 18 18.6 6.40 16.3 8.01 25 N/Ab 21.2 6.26 20.8 8.08 21 N/Ab 5.948 p=0.051 mAHP latency (ms) 72.0 12.1 70.7 9.88 16 93.6 41.5 92.4 57.8 14 n.s.b 57.1 9.05 54.5 14.7 11 p<0.05b 10.501 p=0.005 Post-inhibitory Sag ampitude (mV) 33.8 4.14 34.2 2.97 28 33.3 3.43 33.7 3.42 32 N/Ab 34.7 3.20 34.9 4.04 33 N/Ab 2.263 p=0.323 rebound b b Rebound delay (ms) 277 198 248 92 25 301 216 218 222 28 n.s. 51.5 29.9 64.9 23.7 32 p<0.001 49.446 p<0.001 Gain start (Hz/100pA) 10.1 4.36 9.06 7.32 14 15.6 6.35 14.2 10.3 22 p<0.05b 13.6 4.93 12.4 6.34 28 n.s.b 6.179 p=0.046 Excitability Gain end (Hz/100pA) 4.34 2.47 3.89 1.74 14 5.53 2.80 4.95 3.07 22 N/Ab 4.60 1.88 4.62 3.08 28 N/Ab 2.564 p=0.427 SFA index 2.62 1.27 2.34 0.93 14 3.19 1.47 2.76 1.84 22 N/Ab 3.29 1.35 3.02 2.17 28 N/Ab 4.211 p=0.122 c Amplitude (pA) 107.7 43.20 118.6 75.83 15 68.91 27.35 71.75 46.76 26 p<0.01 c ISK Deactivation tau (ms) 177.9 44.12 176.3 36.58 15 311.6 114.2 305.7 150.3 25 p<0.001 c Charge (pA.s) 19.0 6.88 17.8 9.14 14 19.3 10.2 15.4 13.1 17 n.s. c Amplitude (pA) 357.4 156.7 306.3 80.2 14 245.3 95.3 231.0 91.0 28 p<0.01 c IA Inactivation tau (ms) 166.4 86.0 143.4 63.1 14 47.7 59.0 25.3 12.9 28 p<0.001 c Charge (pA.s) 39.8 21.2 36.3 27.0 14 8.3 12.6 3.8 5.7 28 p<0.001

Table 1. Current-clamp and voltage-clamp parameters measured in WT, SK3 KO and Kv4.3 KO SNc DA neurons. Statistically significant differences are indicated by the gray cell shading and bold font. The type of statistical test applied is indicated by the superscript letters next to the p values: a, 1-way ANOVA followed by Holm-Sidak's multiple comparison tests versus control group; b, 1-way ANOVA on rank followed by Dunn's multiple comparison tests versus control group; c, unpaired t-test. 2-way repeated measures WT control SK3 KO control Electrophysiological parameters ANOVA mean SD median IQR n mean SD median IQR n F1 (strain) p1 (strain)

Pacemaking CVISI (%) 6.23 2.36 5.75 3.17 12 8.77 7.46 4.07 3.71 13 3.882 0.052 AHP trough voltage (mV) -73.9 5.68 -74.4 5.69 13 -70.4 4.86 -70.5 3.74 15 0.406 0.532 fAHP trough voltage (mV) -70.9 5.02 -72.9 6.45 8 -69.1 4.41 -71.8 3.02 11 mAHP trough voltage (mV) -74.7 5.73 -75.4 6.22 9 -72.1 3.61 -71.8 3.82 11 Action potential AHP latency (ms) 57.5 28.4 67.0 51.3 13 46.3 20.4 41.8 76.8 15 1.706 0.207 Control fAHP latency (ms) 17.3 15.5 6.23 6.48 8 18.6 6.40 16.3 8.01 11 mAHP latency (ms) 72.4 70.7 12.2 10.6 9 88.1 28.8 90.6 52.8 11 Gain start (Hz/100pA) 10.1 4.36 9.06 7.32 14 15.6 14.2 6.35 10.3 22 0.517 0.475 Excitability Gain end (Hz/100pA) 4.3 2.47 3.89 1.74 14 5.53 4.95 2.80 3.07 22 0.505 0.480 SFA index 2.62 1.27 2.34 0.93 14 3.19 2.76 1.47 1.84 22 1.242 0.294 mean SD median IQR n mean SD median IQR n F2 (drug) p2 (drug) F1xF2 p(F1xF2)

Pacemaking CVISI (%) 44.17 17.02 44.46 20.90 12 54.44 43.23 33.22 37.01 13 131.560 <0.001 0.772 0.382 AHP trough voltage (mV) -64.5 4.60 -65.0 6.58 13 -65.8 3.82 -65.7 5.59 15 39.126 <0.001 0.834 0.373 fAHP trough voltage (mV) -64.5 4.60 -65.0 6.58 13 -65.8 3.82 -65.7 5.59 15 mAHP trough voltage (mV) Action potential Apamin AHP latency (ms) 16.1 4.80 15.0 7.61 13 15.5 4.21 14.9 3.65 15 194.970 <0.001 3.544 0.075 (300nM) fAHP latency (ms) 16.1 4.80 15.0 7.61 13 15.5 4.21 14.9 3.65 15 mAHP latency (ms) Gain start (Hz/100pA) 21.4 10.6 19.4 18.3 14 20.4 20.6 7.82 7.73 22 59.013 <0.001 9.312 0.004 Excitability Gain end (Hz/100pA) 8.1 3.7 7.2 4.5 14 6.51 5.93 3.05 3.38 22 14.321 <0,001 4.899 0.034 SFA index 3.70 3.94 2.14 3.30 14 4.05 3.63 2.80 4.05 22 0.444 0.509 1.134 0.332

WT control Electrophysiological parameters mean SD median IQR n Pacemaking Frequency (Hz) 1.41 0.63 1.33 0.93 9 Control Post-inhibitory Rebound delay (ms) 207 119 176 130 18 rebound mean SD median IQR n p Pacemaking Frequency (Hz) 2.36 1.15 2.20 1.40 9 p<0.001 d AmmTX3 Post-inhibitory (1µM) Rebound delay (ms) 28.0 16.4 20.5 12.9 18 e rebound p<0.001

Table 2. Effect of acute pharmacological blockade on current-clamp parameters. Upper table, effect of the SK blocker apamin on current-clamp parameters in WT and SK3 KO SNc DA neurons. Lower table, effect of the Kv4 blocker AmmTX3 on pacemaking and rebound delay in WT SNc DA neurons. Statistically significant differences are indicated by the gray cell shading and bold font. The type of statistical test applied is indicated by the superscript letters next to the p values: d, paired t-test; e, Wilcoxon signed rank test; f, 2-way repeated- measure ANOVA followed by a Holm-Sidak's multiple comparison test.