SUMMARY
Dopamine (DA) is a critical modulator of brain circuits that control voluntary movements, but our understanding of its influence on the activity of target neurons in vivo remains limited. Here, we use two-photon Ca2+ imaging to simultaneously monitor the activity of direct and indirect-pathway spiny projection neurons (SPNs) in the striatum of behaving mice during acute and prolonged manipulations of DA signaling. We find that, contrary to prevailing models, DA does not modulate activity rates in either pathway strongly or differentially. Instead, DA exerts a prominent influence on the overall number of direct and indirect pathway SPNs recruited during behavior. Chronic loss of midbrain DA neurons in a model of Parkinson’s disease selectively impacts direct pathway ensembles and profoundly alters how they respond to DA elevation. Our results indicate that DA regulates striatal output by dynamically reconfiguring its sparse ensemble code and provide novel insights into the pathophysiology of Parkinson’s disease.
INTRODUCTION
The neuromodulator dopamine (DA) is an essential component of basal ganglia circuits that control goal-directed behaviors. Considerable evidence in humans, primates and rodents implicates DA in the selection and execution of vigorous movement and in supporting motor learning via its actions in striatum, the largest input nucleus of the basal ganglia (Graybiel 2005; Turner & Desmurget 2010; Dudman & Krakauer 2016; Klaus et al., 2019). In humans for instance, loss of striatum-projecting midbrain DA neurons in Parkinson’s disease is associated with bradykinesia, while drugs that elevate DA act as stimulants and are behaviorally reinforcing.
Striatal output is mediated by two populations of inhibitory spiny projection neurons (SPNs) belonging to the direct and indirect pathways (dSPNs and iSPNs). Although long believed to exert opposite effects on movement, dSPNs and iSPNs are now known to be concurrently active and to work in concert to produce coherent sequences of voluntary movements (Cui et al., 2013; Barbera et al., 2016; Klaus et al., 2017; Markowitz et al., 2018; Meng et al., 2018; Parker et al., 2018; Sheng et al., 2019). Both receive excitatory inputs from overlapping cortical and thalamic areas and modulatory afferents from midbrain DA neurons (Kress et al., 2013; Wall et al., 2013; Guo et al., 2015). Because dSPNs express Gαs-coupled D1-type DA receptors and iSPNs Gαi-coupled D2-type DA receptors, DA is thought to modulate both pathways in opposite ways and to promote imbalances that impact how the basal ganglia orchestrate and modify behavior (Gerfen & Surmeier 2011; Klaus et al., 2019). However, the cellular and circuit mechanisms of DA action in vivo remain poorly understood.
Electrophysiological studies in vitro have identified several ion channels and synaptic properties in SPNs susceptible to differential modulation by DA (Tritsch & Sabatini 2012; Zhai et al., 2019). A widely-held view is that DA increases the firing rate of dSPNs and depresses that of iSPNs (Nelson & Kreitzer 2014). Although supported by some chronic DA depletion studies modeling Parkinson’s disease (Mallet et al., 2006; Parker et al., 2018; Ryan et al., 2018), others report no or even opposite changes in discharge rate under similar conditions (Liang et al., 2008; Ellens & Leventhal 2013; Willard et al., 2019). Additional aspects of striatal physiology susceptible to acute pathway-specific DA modulation, like bursting and synaptic efficacy have been comparatively less explored in vivo.
Recent imaging studies show that distinct motor actions are represented in striatum by separate groups – or ensembles – of SPNs, and that individual renditions of a given action recruit only a subset of SPNs from the larger action-specific ensemble (Barbera et al., 2016; Klaus et al., 2017; Markowitz et al., 2018). It is presently not clear how plastic these neural representations are, and if DA plays a role in sculpting them. Here, we test this possibility using two-photon Ca2+ imaging to monitor dSPN and iSPN ensembles simultaneously in the striatum of behaving mice. We report that acute DA manipulations strongly modify the overall size of movement-related dSPN and iSPN ensembles, while exerting comparatively little influence on the rate of Ca2+ events. In addition, we show that chronic loss of midbrain DA neurons in a model of Parkinson’s disease is associated with a specific decrease in the number of active dSPNs and with significant changes in how dSPN ensembles respond to DA replacement.
RESULTS
Baseline activity in dSPNs and iSPNs is concurrent and balanced
To resolve the activity of individual striatal neurons in vivo, we performed two-photon microscopy through a chronically implanted window (Howe & Dombeck 2016; Bloem et al., 2017) in head-fixed mice (N = 18) on a freely rotating circular treadmill (Fig. 1a). We expressed the Ca2+ indicator GCaMP6f (Chen et al., 2013) virally in the dorsolateral striatum under control of the synapsin promoter to label all neurons. Experiments were conducted using Drd1atdTomato transgenic mice, in which dSPNs are selectively labeled red (Ade et al., 2011). This approach allows differentiation of Ca2+ signals arising in dSPNs from those in tdTomato-negative neurons, which are overwhelmingly iSPNs, and which together with dSPNs account for over 90% of all striatal neurons (Gerfen & Surmeier 2011). We obtained large fields of view (FOV; 500 x 500 μm, n = 1–3 per mouse) with clearly identifiable dSPNs interspersed among neurons containing GCaMP6f only (Fig. 1b and Supplementary Video).
a. Schematic of experimental setup.
b. Top: representative two-photon maximum projection image of dorsolateral striatum. Red: td-tomato labelled dSPNs, green: striatal neurons virally transduced to express GCaMP6f (scale bar: 50 μm). Inset shown at bottom in green (left) and red (middle) channels only, and composite (right; scale: 30 μm).
c. Top: example Ca2+ fluorescence traces from dSPNs (red) and putative iSPNs (green). Middle: Mean ΔF/F across all active dSPNs and iSPNs. Bottom: treadmill velocity highlighting two self-initiated locomotor bouts.
d. Proportion of all imaged neurons assigned to dSPNs, iSPNs and other neurons (n = 30 FOVs in 18 mice).
e. Mean Ca2+ transient waveform (± SEM) imaged from active dSPNs and iSPNs.
f. Comparison of Ca2+ transient amplitude in dSPNs versus iSPNs. Correlation coefficient r is indicated in plot and mean ± SEM in blue (p = 0.004; Wilcoxon signed-ranks; n = 30 FOVs).
g. Comparable percentages of dSPNs exhibit Ca2+ transients in the same FOV on imaging sessions separated by d days (p = 0.76; Wilcoxon signed-ranks; n = 30 FOVs).
h. Same as g for iSPNs (p = 0.78; Wilcoxon signed-ranks).
i. Percentage of imaged dSPNs and iSPNs exhibiting Ca2+ transients per FOV (n = 30). Mean ±SEM shown in blue.
j. Percentage of all imaged dSPNs and iSPNs that display Ca2+ transients at different treadmill velocities (n = 30 FOVs).
k. Percentage of all imaged dSPNs and iSPNs that display Ca2+ transients aligned to locomotion onset (left) and offset (right) overlaid with treadmill velocity (black).
l. Mean Ca2+ transient frequency in all active dSPNs vs. iSPNs (n = 30 FOVs) during spontaneous treadmill running (black) and while immobile (gray). No significant difference was observed between dSPNs and iSPNs while moving (blue; p = 0.8, Wilcoxon signed-ranks) or immobile (yellow; p = 0.4, Wilcoxon signed-ranks).
m. Same as j for Ca2+ transient frequency in dSPNs and iSPNs.
n. Same as k for Ca2+ transient frequency in dSPNs and iSPNs.
Data in j, k, m and n represent mean ± SEM. dSPNs are shown in red, iSPNs in green.
Self-paced forward locomotion was associated with prominent increases in fluorescence in groups of neurons both positive and negative for tdTomato (Fig. 1c). Population Ca2+ signals in tdTomato-positive dSPNs resembled data obtained using photometry in mice expressing GCaMP6f selectively in dSPNs (Supplementary Fig. 1a). Amongst tdTomato-negative neurons, the vast majority resembled dSPNs in morphology and, when active, showed Ca2+ transients comparable to those observed in dSPNs (Fig. 1e,f) and activity consistent with signals obtained specifically from iSPNs (Supplementary Fig. 1b–d). These cells were consequently classified as iSPNs. A small fraction of tdTomato-negative neurons consistently exhibited distinct Ca2+ dynamics, suggesting that they represent striatal interneurons (Gritton et al., 2019; Howe et al., 2019). Indeed, the proportions of imaged neurons assigned to dSPNs, iSPNs and interneurons agree with anatomical estimates (Fig. 1d). This imaging approach therefore offers the ability to simultaneously monitor and compare the activity of hundreds of striatal neurons (mean ± SEM): 327 ± 13 per FOV, range: 131–442) belonging to both direct and indirect pathways with high spatial resolution during a simple behavior.
The fraction of all imaged dSPNs and iSPNs displaying Ca2+ transients in each FOV across all forward locomotor bouts were comparable across imaging sessions on separate days (Fig. 1g,h), with a slightly greater proportion of active iSPNs as compared to dSPNs (active iSPNs: 20.5 ± 2.2 % of all imaged iSPNs; active dSPNs: 18.0 ± 2.2 % of all imaged dSPNs; p = 0.022, Wilcoxon signed-rank; Fig. 1i). Forward locomotion was associated with the activation of a sparse subset of SPNs from the larger ensembles defined across several locomotor bouts. The number of SPNs contributing to each subset grew larger as treadmill speed increased without notable temporal offset between pathways (Fig. 1j,k). The end of movement was associated with a progressive decline in the fraction of active SPNs (Fig. 1k). The frequency of Ca2+ transients in individual SPNs was also comparable across pathways and similarly reflected treadmill velocity (Fig. 1l–n). Thus, consistent with previous reports (Barbera et al., 2016; Klaus et al., 2017; Markowitz et al., 2018), self-paced forward locomotion is associated with sparse activation of subsets of SPNs that over time define larger, action-related ensembles, with few notable differences distinguishing dSPNs from iSPNs at either low (< 0.2 cm/s; i.e. immobile) or high (> 0.4 cm/s) treadmill speeds.
DA neuron lesions impact population-level Ca2+ event rates
To estimate the influence of DA on striatal activity, we first injected the neurotoxin 6-hydroxydopamine (6-OHDA) into the substantia nigra pars compacta (SNc) ipsilateral to the striatal hemisphere under investigation in a subset of mice (N = 12). This manipulation causes DA neurons to degenerate within hours, resulting in near complete and irreversible loss of tyrosine hydroxylase (TH)-positive cell bodies in SNc and commensurate reductions in DA transporter (DAT) fluorescence in dorsal striatum (Fig. 2a,b and Supplementary Fig. 2a,b). Consistent with DA’s established role in producing vigorous movement (Panigrahi et al., 2015), unilateral SNc lesions mildly reduced the mean velocity of self-paced locomotor bouts compared to sham-treated mice (N = 6; Two-way ANOVA lesion x time: F2,30 = 5.56, p = 0.009) as early as 24 hours and for up to a month after treatment (Supplementary Fig. 2c), but not their duration (F2,30 = 0.99, p = 0.38; Supplementary Fig. 2d) or frequency (F2,30 = 0.15, p = 0.86; Supplementary Fig. 2e). Lesioned mice also showed a strong and persistent ipsiversive turning bias in a closed-arm plus-maze (F2,27 = 6.0, p = 0.007; Fig. 2c), further providing evidence of the immediate and long-lasting motor impairments that result from unilateral loss of SNc neurons.
a. Representative coronal forebrain (top) and ventral midbrain (bottom) sections from a mouse injected with 6-OHDA into the right SNc and stained for DAT (top) and TH (bottom). DS, dorsal striatum; VS, ventral striatum; VTA, ventral tegmental area. Scale bars: 1 mm (top), 0.5 mm (bottom).
b. Quantification of immunofluorescence signal for DAT in dorsal and ventral striatum (left) and TH in SNc and VTA (right) in sham (gray; N = 6) and 6-OHDA lesioned (blue; N = 12) mice on treated vs. intact hemispheres (mean ± SEM overlaid). Asterisks depict significant difference from sham (DS: p = 3.2 × 10−4; VS: p = 0.001; SNc: p = 1.1 × 10−4; VTA: p = 1.1 × 10−4; Mann-Whitney).
c. Turning bias measured in a closed-arm plus maze from a subset of vehicle (N = 5, gray) and 6-OHDA injected (N = 6, blue) mice 1 day before, 1 day after and 30 days after intracranial injection surgery. Values approaching 1 and −1 reflect strong turning bias toward (ipsiversive) or away from (contraversive) the lesioned hemisphere, respectively. Within-group post-hoc statistic (Bonferroni, vs. pre-surgery): Sham, day 1: p = 0.98; day 30: p = 0.28; 6-OHDA, day 1: p = 0.003; day 30: p = 0.015.
d. Mean (± SEM) Ca2+ event rate across the entire population of dSPNs per FOV during self-initiated movement before, 1 day after or 30 days after ipsilateral sham (n = 9 FOVs in 6 mice) or 6-OHDA lesion (n = 21 from 12 mice) of SNc neurons. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p > 0.99; day 30: p > 0.99. 6-OHDA, day 1: p = 0.011; day 30: p = 0.008.
e. Same as d for iSPNs. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p > 0.99; day 30: p > 0.99. 6-OHDA, day 1: p = 1.9 × 10−7; day 30: p = 0.46.
f. Same as d for Ca2+ event rate bias between pathways. Values approaching 1 and −1 reflect bias towards dSPNs and iSPNs, respectively. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p = 0.67; day 30: p = 0.59. 6-OHDA, day 1: p = 8.1 × 10−12; day 30: p = 8.4 × 10−5.
g. Mean population-level Ca2+ event rate in dSPNs vs. treadmill velocity before, 1 day after or 30 days after SNc lesion (n = 21 FOVs). Shaded regions indicate SEM.
h. Same as g for iSPNs.
i. Same as g for dSPNs (top) and iSPNs (middle) vs. treadmill velocity (bottom) aligned to the onset (left) and offset (right) of locomotor bouts.
A recent study using one-photon microendoscopy revealed that DA depletion results in large and opposite changes in the frequency of Ca2+ transients imaged from dSPNs and iSPNs in separate cohorts of freely behaving mice (Parker et al., 2018). We therefore attempted to replicate their findings under our experimental conditions by monitoring Ca2+ signals in dSPNs and iSPNs concurrently within the same FOV. We first confirmed that DA depletion (n = 21 FOVs in 12 mice) did not compromise our ability to detect Ca2+ transients, as their amplitude in dSPNs and iSPNs did not significantly vary over imaging sessions compared to sham-treated mice (n = 9 FOVs in 6 mice; Two-way ANOVA lesion x time: dSPN: F2,45 = 1.31, p = 0.3; iSPN: F2,45 = 1.50, p = 0.2; Supplementary Fig. 3a–c), excluding shifts in signal-to-noise as primary contributors to 6-OHDA-mediated phenotypes. In agreement with Parker and colleagues (2018), we observed a significant effect of 6-OHDA on the overall rate of Ca2+ events imaged across the entire FOV (i.e. population-level Ca2+ event rate) in both dSPNs (Two-way ANOVA lesion x time: F2,45 = 3.95, p = 0.026) and iSPNs (F2,45 = 14.84, p = 1.12 × 10−5) compared to vehicle-treated mice. Striatal activity did not appreciably change in sham-treated mice, establishing a stable baseline from which to identify differences specifically related to the loss of DA neurons. By contrast, population-level rates in dSPNs decreased by half during locomotion immediately after 6-OHDA infusion, while they more than doubled in iSPNs at all recorded velocities (Fig. 2d–i). Population-level Ca2+ event rates in chronically lesioned mice remained depressed in dSPNs but returned to baseline in iSPNs during movement (Fig. 2d–i), consistent with previous observations (Parker et al., 2018). These results therefore indicate that DA depletion strongly biases striatum-wide activity towards the indirect pathway both acutely and chronically in head-fixed and freely-moving mice. However, they do not clearly establish how these population-level effects arise. We investigate this in the following section.
Lesioning DA neurons differentially alters dSPN and iSPN ensemble size
The population-level Ca2+ event rate changes reported in Fig. 2 may result from changes in the frequency of Ca2+ transients experienced by individual SPNs, differences in the overall size of SPN ensembles recruited by a given behavior, or both (Fig. 3a). To distinguish between these possibilities, we took advantage of the high spatial resolution and signal-to-noise of two-photon microscopy (Helmchen & Denk 2005). Given our reported effects on frequency (Fig. 2) and the prevalence of rate-based models of basal ganglia function and DA modulation (reviewed in Calabresi et al., 2014; Nelson & Kreitzer 2014), we first asked whether acute and chronic DA depletion alters the frequency of Ca2+ transients in individual SPNs. Surprisingly, we failed to detect a significant effect of 6-OHDA compared to sham in either dSPNs (Two-way ANOVA lesion x time: F2,45 = 1.05, p = 0.36; Fig. 3b) or iSPNs (F2,45 = 1.64, p = 0.21; Fig. 3c) across recorded treadmill velocities (Fig. 3e). Because the mean Ca2+ transient frequency per active iSPN trended up immediately after 6-OHDA infusion and trended down in dSPNs in chronically lesioned mice, DA depletion did result in a slight imbalance in activity rates towards the indirect pathway (F2,45 = 7.98, p = 0.001; Fig. 3d). However, this effect alone cannot account for the strongly dichotomous behavior of dSPNs and iSPNs reported above (Fig. 2) and before (Parker et al., 2018).
a. Schematic of possible circuit mechanisms accounting for population-level Ca2+ event rate changes following DA depletion. Left: locomotion is associated with the activation of sparse subsets of SPNs that, over time, define action-specific dSPN and iSPN ensembles (filled neuronal profiles and their Ca2+ fluorescence signals. After DA depletion, population-level changes in Ca2+ event rate in all dSPNs and iSPNs per FOV may arise from modifying the frequency of Ca2+ transients in active SPNs (middle) or from altering the overall size of SPN ensembles associated with locomotion (right).
b. Mean (± SEM) Ca2+ transient frequency per dSPN active during self-initiated movement before, 1 day after, or 30 days after sham (n = 9 FOVs in 6 mice) or 6-OHDA-mediated (n = 21 FOVs from 12 mice) SNc ablation.
c. Same as b for iSPNs.
d. Same as b for Ca2+ transient frequency bias between pathways. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p > 0.99; day 30: p = 0.23. 6-OHDA, day 1: p = 0.008; day 30: 5.0 × 10−4.
e. Mean Ca2+ transient frequency per dSPN (left) and iSPN (right) vs. treadmill velocity before, 1 day after or 30 days after SNc lesion (n = 21 FOVs). Shaded regions indicate SEM.
f. Percentage of all imaged dSPNs that exhibit Ca2+ transients per FOV before, 1 day after, or 30 days after sham (n = 9 FOVs in 6 mice) or 6-OHDA-mediated (n = 21 FOVs from 12 mice) SNc lesion. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p = 0.42; day 30: p > 0.99. 6-OHDA, day 1: p = 1.8 × 10−4; day 30: p = 0.001.
g. Same as f for iSPNs. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p = 0.78; day 30: p > 0.99. 6-OHDA, day 1: p = 8.0 × 10−7; day 30: p > 0.99.
h. Same as f for bias in number of active cells per FOV between pathways. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery): Sham, day 1: p > 0.99; day 30: p > 0.99. 6-OHDA, day 1: p = 3.1 × 10−14; day 30: p = 4.2 × 10−5.
i. Same as e for percentage of all imaged dSPNs (left) and iSPNs (right) that show Ca2+ transients at different treadmill velocities. Note similarities with Fig. 2g, h.
j. Percentage of dSPN and iSPN activated per self-initiated locomotor bout aligned to onset and offset before, 1 day after, or 30 days after 6-OHDA lesion. Note similarities with Fig. 2i.
We therefore considered the alternative possibility that lesioning DA neurons alters the size of SPN ensembles associated with forward locomotion. Indeed, we observed a strong effect of 6-OHDA on ensemble size compared to vehicle-injected mice in dSPNs (Two-way ANOVA lesion x time: F2,53 = 3.66, p = 0.03; Fig. 3f), iSPNs (F2,53 = 11.92, p = 5.3 × 10−5; Fig. 3g) and in pathway bias (F2,53 = 19.0, p = 6.1 × 10−7; Fig. 3h). Specifically, the total number of dSPNs displaying Ca2+ transients over the course of an imaging session was halved immediately after 6-OHDA infusion and remained low thereafter (Fig. 3f). This is unlikely to result from diminished locomotor speed because within the same FOV, the total number of active iSPNs and the subset of iSPNs recruited per locomotor bout doubled in acutely lesioned mice (Fig. 3g) across recorded velocities (Fig. 3i) and movement bouts (Fig. 3j). In addition, imaging sessions on motorized treadmills with constant velocity and distance traveled yielded similar results (Supplementary Fig. 3j–l). Acute loss of SNc neurons therefore alters the extent of dSPN and iSPN ensembles in opposite ways, generating a strong imbalance in favor of the indirect pathway (Fig. 3h). Interestingly, the expansion of iSPN ensembles was not sustained in time and returned to baseline in chronically lesioned animals (Fig. 3f–j). Because compensatory changes were not observed in dSPNs, the overall size of SPN ensembles recruited over the course of imaging sessions remained significantly biased towards the indirect pathway in chronically lesioned mice (Fig. 3h).
DA receptor signaling rapidly reconfigure SPN ensembles
Neurotoxic lesions cannot reveal DA’s dynamic modulatory actions and only implicate DA indirectly, as SNc neurons release multiple transmitters (Trudeau et al., 2014) and 6-OHDA-mediated cell death triggers inflammatory responses (Cicchetti et al., 2002). To more directly assess DA’s influence on striatal activity on the time scale of minutes, we compared SPN Ca2+ signals before and after systemic administrations of a cocktail of DA receptor antagonists in DA-intact mice. Under our experimental conditions, this manipulations did not affect mean velocity or bout duration, but significantly curtailed the number of spontaneously-initiated locomotor bouts compared to vehicle-treated animals (Supplementary Fig. 4a–c). We therefore limited our analyses to imaging sessions on motorized treadmills to keep velocity and distance traveled constant.
Consistent with our earlier findings, reducing DA signaling did not change the frequency of Ca2+ transients in individual dSPNs recruited by forward locomotion (p = 0.2, Mann-Whitney; vehicle: n = 10 FOVs in 5 mice; DA antagonist: n = 11 FOVs in 5 mice; Fig. 4a). The frequency of Ca2+ transients in iSPNs increased (p = 0.007, Mann-Whitney; Fig. 4b), but not enough to bias striatal activity towards the indirect pathway (p = 0.22, Mann-Whitney; Fig. 4c). In stark contrast, blocking DA receptors significantly affected SPN ensemble size: it reduced the total number of dSPNs recruited by forward locomotion by half (p = 0.006, Mann-Whitney; Fig. 4d) and expanded iSPNs more than three-fold (p = 5.7 × 10−6, Mann-Whitney; Fig. 4e), leading to a strong imbalance in favor of the indirect pathway (p = 5.7 × 10−6, Mann-Whitney; Fig. 4f). These data indicate that loss of DA minimally impacts the frequency of Ca2+ transients displayed by SPNs, yet it evokes substantial reconfiguration of the overall number of dSPNs and iSPNs associated with a given behavior.
a. Mean Ca2+ transient frequency per dSPN active during locomotion after systemic administration of vehicle (left; n = 10 FOVs in 5 mice), a cocktail of D1 and D2 receptor antagonists (middle; n = 11 FOVs in 5 mice, p = 0.20, Mann-Whitney) and a cocktail of D1 and D2 receptor agonists (right; n = 11 FOVs in 7 mice, p = 0.88, Mann-Whitney) normalized to pre-drug baseline.
b. Same as a for iSPNs (antagonists: p = 0.007; agonists: p = 0.02; Mann-Whitney).
c. Same as a for bias in frequency between pathways (antagonists: p = 0.22; agonists: p = 0.36; Mann-Whitney).
d. Fraction of all imaged dSPNs active per FOV after systemic administration of vehicle (left; n = 10 FOVs in 5 mice), D1 and D2 receptor antagonists (middle; n = 11 FOVs in 5 mice, p = 0.006, Mann-Whitney) or D1 and D2 receptor agonists (right; n = 11 FOVs in 7 mice, p = 0.015, Mann-Whitney) normalized to pre-drug baseline.
e. Same as d for iSPNs (antagonists: p = 5.7 × 10−6; agonists: p = 2.3 × 10−5; Mann-Whitney).
f. Same as d for bias in ensemble size between pathways (antagonists: p = 5.7 × 10−6; agonists: p = 2.0 × 10−4; Mann-Whitney).
Mean ± SEM are overlaid for each group. Asterisks depict significant difference from vehicle.
We next investigated whether elevating DA signaling would exert the opposite effect. To do so, we imaged striatal activity upon pharmacological stimulation of D1 and D2 receptors (n = 11 FOVs in 7 mice) while mice locomoted on motorized treadmills to mitigate treatment effects on spontaneous behavior (Supplementary Fig. 4a–c). Here too, we failed to detect changes in the frequency of Ca2+ transients during locomotion in individual dSPNs (p = 0.9, Mann-Whitney; Fig. 4a) or between pathways (p = 0.36, Mann-Whitney; Fig. 4c), despite a small but statistically significant decrease in Ca2+ transient frequency in iSPNs compared to vehicle (p = 0.021, Mann-Whitney; Fig. 4b). By contrast, stimulating DA receptors markedly altered the size of SPN ensembles associated with forward locomotion, yielding a strong imbalance in favor of the direct pathway (p = 2.0 × 10−4, Mann-Whitney; Fig. 4f). The latter was mediated by a near complete decline in the prevalence of active iSPNs (p = 2.3 × 10−5, Mann-Whitney; Fig. 4e) since, contrary to DA’s presumed excitatory influence on dSPNs, activation of DA receptors also diminished the number of active dSPNs (p = 0.015, Mann-Whitney; Fig. 4d). Importantly, mild elevation of endogenous DA using the presynaptic DA reuptake antagonist nomifensine similarly diminished the number of active dSPNs without altering Ca2+ transient frequency (Supplementary Fig. 4d, e). Together, these data indicate that DA does not strongly modulate the frequency of Ca2+ transients in individual dSPNs and iSPNs that comprise action-specific ensembles. Instead, DA controls the number of SPNs recruited with each motor action as well as the extent of the larger SPN ensembles that are defined over multiple rendition of such action. While the overall number of iSPNs recruited in this task is inversely related to extracellular DA, dSPNs follow an ‘inverted-U’ shaped function, where too little or too much DA negatively impacts the size of their ensemble.
Chronic DA depletion alters how SPN ensembles respond to DA
Motor impairments in Parkinson’s disease are typically managed by elevating striatal DA with L-DOPA, the metabolic precursor for DA. We therefore wondered whether acutely restoring DA with L-DOPA would affect SPN ensemble size, and if so how. To address this, we treated a subset of the mice (N = 6) that had received unilateral vehicle or 6-OHDA infusions in SNc a month or more earlier with a single dose of L-DOPA. As expected, this treatment had a significant effect on the behavior of 6-OHDA vs. sham-lesioned mice (N = 5) in the plus-maze (Two-way ANOVA lesion x levodopa: F1,18 = 94.7, p = 1.4 × 10−8), reversing the turning bias of lesioned mice within minutes (p = 6.2 × 10−10, post-hoc Bonferroni; Fig. 5a). In the striatum of DA-depleted mice locomoting on motorized treadmills (n = 12 FOVs), L-DOPA elevated the frequency of Ca2+ transients in individual dSPNs and reversed the mild frequency imbalance caused by DA depletion (Fig. 5c), in agreement with previous work (Parker et al., 2018). In addition and most notably, L-DOPA significantly altered the size of SPN ensembles (Fig. 5b): it halved the overall number of iSPNs exhibiting Ca2+ transients and expanded the number of active dSPNs 10-fold, reaching upwards of 70% of all imaged dSPNs in some FOVs (Fig. 5d). This resulted in a reversal of ensemble size imbalance in DA-depleted mice towards the direct pathway (Fig. 5d). These data therefore indicate that DA favors the recruitment of dSPNs relative to iSPNs in both DA-intact and DA-depleted mice, but that the underlying mechanisms profoundly differ (Fig. 5e).
a. Turning bias measured in closed-arm plus maze before and after L-DOPA treatment in a subset of DA-intact (N = 5, gray; p = 0.15 vs. baseline, post-hoc Bonferroni) and chronically DA-depleted mice (N = 6, blue; p = 6.24 × 10−10 vs. baseline, post-hoc Bonferroni).
b. Example maximum projection image of dorsal striatum (left) and spatial distribution of dSPNs (red) and iSPNs (green) displaying Ca2+ transients during imaging sessions immediately preceding (middle) and following (right) systemic L-DOPA administration. Scale bar: 50 μm.
c. Mean Ca2+ transient frequency per dSPN (left) and iSPN (middle) active in chronically-lesioned mice before and after elevating striatal DA with L-DOPA (n= 12 FOVs in 6 mice). Right: frequency bias index. Mean ± SEM are overlaid for each group (dSPN: p = 0.023; iSPN: p = 0.052; pathway bias: p = 0.016; all vs. baseline, Wilcoxon signed rank).
d. Same as c for active SPN ensemble size (dSPN: p = 0.001; iSPN: p = 0.016; pathway bias: p = 4.9 × 10−4; all vs. baseline, Wilcoxon signed rank).
e. Summary diagrams depicting observed changes in SPN ensemble size in response to changing striatal DA levels in intact and DA-depleted mice.
DISCUSSION
In this study, we sought to reveal how acute and prolonged manipulations of DA signaling impact striatal output. To do so, we simultaneously monitored intracellular Ca2+ signals in dSPNs and iSPNs using two-photon microscopy to permit the identification of subtle imbalances between SPN populations long proposed to lie at the heart of basal ganglia function and dysfunction (Nelson & Kreitzer 2014; Klaus et al., 2019). We specifically examined changes in the frequency and amplitude of Ca2+ transients exhibited by individual neurons, in addition to changes in the overall size of SPN ensembles, all during a low-dimensional locomotor behavior to facilitate comparisons across imaging sessions.
In agreement with prior imaging studies (Cui et al., 2013; Barbera et al., 2016; Klaus et al., 2017; Markowitz et al., 2018; Meng et al., 2018; Parker et al., 2018), we found that the activity of dSPNs and iSPNs is balanced and concurrently elevated during movement under baseline conditions. Individual forward locomotion bouts recruited a subset of the overall ensemble of SPNs associated with that behavior in proportion to treadmill velocity. Individual dSPNs and iSPNs were not systematically activated across different locomotor bouts, at particular phases within movement bouts or at specific treadmill velocities, arguing against a specific role for individual SPNs in the initiation, maintenance or cessation of self-paced locomotion (Sales-Carbonell et al., 2018). Instead, these observations are consistent with dorsolateral striatum relying on a sparse ensemble code to guide behavior (Barbera et al., 2016; Klaus et al., 2017; Markowitz et al., 2018). The latter may result from a combination of anatomical and functional features proper to striatal neurons and circuits (Wilson & Kawaguchi 1996; Kincaid et al., 1998; Tepper et al., 2008; Ponzi & Wickens 2013; Reig & Silberberg 2014), and from the dynamic and heterogeneous activity of their excitatory (Churchland & Shenoy 2007; Robbe 2018; Sauerbrei et al., 2018) and modulatory inputs (Howe & Dombeck 2016; da Silva et al., 2018).
Changing DA levels or DA receptor signaling appreciably altered striatal activity, but not in a manner predicted by rate-based models of dopaminergic action (Calabresi et al., 2014; Nelson & Kreitzer 2014), as any effect on Ca2+ transient frequency in individual SPNs was generally mild and comparable across pathways. Indeed, acute DA manipulations did not modify the frequency of Ca2+ transients in dSPNs or produce large rate imbalances between pathways despite slight but significant frequency changes in iSPNs. We also failed to detect consistent effects on Ca2+ transient amplitude. Our results stand in contrast to a recent imaging study reporting changes in the rate of Ca2+ events in SPNs upon acute and chronic DA manipulations in freely behaving mice (Parker et al., 2018). This inconsistency may result from technical differences: we carried out experiments on motorized treadmills to promote movement at defined speeds and mitigate behavioral confounds. In addition, our imaging approach clearly distinguishes changes in the frequency and amplitude of Ca2+ transients per active neuron from changes in the overall number of active cells (Fig. 3a), which can more easily be conflated with one-photon endoscopic techniques (Helmchen & Denk 2005; Ziv & Ghosh 2015; Zhou et al., 2018). Accordingly, we ascribe most of the changes in Ca2+ event rate observed at the population level (Fig. 2) to modifications in the overall number of SPNs recruited during behavior.
Still, our observations may be surprising in light of previous reports of DA-dependent changes in firing rates measured electrophysiologically. Because many such studies do not distinguish dSPNs from iSPNs, it is conceivable that both populations comprise neurons that elevate and depress their firing rates, resulting in little to no net change at the population level. Indeed, Ryan and colleagues (2018) did not detect mean firing rate differences in iSPNs during movement in chronically lesioned mice, in line with our findings (Fig. 3c). Alternatively, because somatodendritic Ca2+ elevations in SPNs are tightly correlated with bursts of action potentials (Kerr & Plenz 2002) and modulated by local interneurons independently of mean firing rate (Owen et al., 2018), GCaMP6f fluorescence may not be sensitive to moderate changes in firing rate and is more likely to reflect bursting activity patterns. As such, Ca2+ signals constitute a measure of neural activity in their own right, highlighting bursting events likely to impact synaptic plasticity in SPNs (Carter & Sabatini 2004; Jedrzejewska-Szmek et al., 2017) and discharge patterns in downstream nuclei (Ellens & Leventhal 2013; Willard et al., 2019). Importantly, our data indicate that the factors that control the rate of Ca2+ transients are neither strongly nor differentially modulated by DA, pointing to a common synaptic drive to dSPNs and iSPNs.
By far the largest and most consistent effects of DA manipulations were changes in the overall size of SPN ensembles associated with forward locomotion (Fig. 5e). Specifically, reducing DA signaling rapidly depressed the total number of dSPNs exhibiting Ca2+ transients and simultaneously doubled that of iSPNs, causing a strong imbalance in favor of the indirect pathway. This supports the notion that DA normally acts to promote the recruitment of dSPNs while limiting that of iSPNs. Interestingly, elevating DA signaling beyond physiological levels using DA receptor agonists or a DA transporter inhibitor did not promote the opposite. Instead, both dSPN and iSPN ensembles shrank. This observation challenges the implicit notion that high DA levels invariably promote activation of the direct pathway. Thus, iSPN ensemble size is inversely related to extracellular DA, whereas dSPN ensembles follows an “inverted-U” shaped response. Interestingly, a similar relationship has been extensively described in prefrontal cortex with DA acting on D1 receptors to regulate cognitive functions such as working memory (Vijayraghavan et al., 2007), pointing to a common mechanism of dopaminergic modulation across brain areas.
It is increasingly recognized that motor impairments in Parkinson’s disease reflect aberrant activity patterns within the basal ganglia that arise not only from striatal DA loss, but also from subsequent homeostatic circuit adaptations (Zhai et al., 2019). The latter are also believed to underlie the dyskinetic side effects that result from chronic DA replacement therapy (Cenci 2014; Picconi et al., 2018). Consistent with this, we observed stark differences in striatal activity between acute and persistent loss of DA. In iSPNs, ensemble size returned to pre-lesion values in chronically lesioned mice. These and other recent data (Ketzef et al., 2017; Parker et al., 2018; Ryan et al., 2018; Willard et al., 2019) do not support the long- and widely-held notion that persistent DA depletion in Parkinson’s disease results in uncontrolled disinhibition of the indirect pathway. Instead, they reveal a remarkable capacity for homeostatic plasticity in iSPNs reflected in diminished intrinsic excitability, dendritic atrophy, spine loss and elevated inhibition from surrounding interneurons (Gittis & Kreitzer 2012; Zhai et al., 2019).
By contrast, dSPNs failed to show functional recovery from acute DA loss, as few dSPNs displayed Ca2+ signals up to a month after 6-OHDA injection, despite an increase in intrinsic excitability (Fieblinger et al., 2014; Ketzef et al., 2017). This echoes previous observations that dSPNs are more difficult to sample electrophysiologically and to evoke activity from in chronically lesioned animals (Mallet et al., 2006), possibly because the number and strength of excitatory inputs onto these cells declines (Fieblinger et al., 2014; Parker et al., 2016; Ketzef et al., 2017) and GABAergic transmission rises (Lemos et al., 2016). This diminished ability to recruit dSPNs may underlie the bradykinesia that defines Parkinson’s disease. Interestingly, homeostatic adaptions in dSPNs appear to rely more heavily on DA receptor signaling than in iSPNs, as chronic DA depletion drastically altered how dSPNs respond to DA receptor stimulation, leading to excessive recruitment of dSPNs not normally associated with forward locomotion. The resulting degradation of action-selective dSPN ensembles may help explain why long-term L-DOPA treatment eventually induces uncontrolled involuntary movements (Cenci 2014; Picconi et al., 2018).
Together, our data indicate that the size of SPN ensembles encoding motor actions is not fixed, but is rather dynamically modulated by DA. Thus, in addition to its established role in shaping plasticity at individual synapses (Shen et al., 2008), DA also regulates the number of neurons eligible to participate in such plasticity. In doing so, DA may directly contribute to the production of vigorous motor actions (Panigrahi et al., 2015; da Silva et al., 2018; Yttri & Dudman 2018) and the refinement of motor skills (Yin et al., 2009; Sheng et al., 2019). Interestingly, DA does this without strongly or differentially modulating the frequency of Ca2+ transients in dSPNs or iSPNs. Instead, DA may regulate the number of SPNs that ‘tune in’ to a common input driving somatodendritic Ca2+ spikes, a key element governing plasticity across brain areas in health and disease (Lerner & Kreitzer 2011; Zhuang et al., 2013; Bittner et al., 2017; Nanou & Catterall 2018). Future investigations will determine the cellular mechanisms that underlie this ability, the time course of such changes in response to synaptic release of DA, and the impact of varying ensemble size on the production and learning of motor actions by the basal ganglia.
AUTHOR CONTRIBUTIONS
N.X.T. conceived of the project. M.M., J.R.M. and N.X.T. designed and performed experiments, analyzed and interpreted the data, and wrote the manuscript. A.G.B. helped with behavioral training and analysis.
DECLARATION OF INTERESTS
The authors declare no competing interests.
METHODS
Animals
All procedures were performed in accordance with protocols approved by the NYU Langone Health Institutional Animal Care and Use Committee. Mice were housed in group before surgery and singly after surgery under a reverse 12-hour light-dark cycle (dark from 10 a.m. to 10 p.m.) with ad libitum access to food and water. Drd1atdTomato transgenic mice were purchased from The Jackson Laboratory (stock #: 016204) and bred with C57BL/6J wild type mice (stock #: 000664). Transgenic mice expressing Cre selectively in iSPNs (Adora2aCre_KG139) or dSPNs (Drd1aCre_EY217) were generously provided by Chip Gerfen (NIH) and maintained on a C57BL/6J background. Control experiments in Supplementary Fig. 1 were performed using offspring of Adora2aCre_KG139 mice bred to a tdTomato reporter (Ai14; The Jackson Laboratory, Stock #: 007908). Experiments were carried out using both male and female mice heterozygous for all transgenes at 8–24 weeks of age.
Surgery
Mice were administered Dexamethasone (4 mg/kg, intraperitoneal) 1–2 hours prior to surgery. They were then anaesthetized with isoflurane, placed in a stereotaxic apparatus (Kopf Instruments) on a heating blanket (Harvard Apparatus) and administered Ketoprofen (10 mg/kg, subcutaneous). The scalp was shaved and cleaned with ethanol and iodine solutions before exposing the skull. A custom titanium headpost was implanted over lambda using C&B metabond (Parkell) to allow head fixation. To achieve widespread viral expression of GCaMP6f in striatum, 100 nl of AAV1-Syn-GCaMP6f-WPRE-SV40 (University of Pennsylvania Vector Core) was injected 1.7 mm below dura at a rate of 100 nl/min (KD Scientific) into the right dorsolateral striatum at four locations (anterior/lateral from bregma, in mm): 0.7/1.7; 0.7/2.3; 1.3/1.7 and 1.3/2.3. Injection pipettes were left in place for 5 min before removal. A 3 mm craniotomy was then drilled (centered at 1.0 mm anterior and 2.0 mm lateral from bregma) and cortical tissue was aspirated until the corpus callosum lying above the striatum was exposed, as described previously (Howe & Dombeck 2016). A custom 9 gauge thin-walled stainless-steel cannula (Microgroup; 2.3 mm in height) sealed at one end with a 3 mm glass coverslip (Warner Instruments) using optical glue (Norland #71) was placed above the striatum and cemented to the skull using C&B metabond. Mice were allowed to recover in their cage for 2 weeks before head-fixation habituation, treadmill training and imaging.
6-OHDA lesions
For experiments characterizing the effects of SNc neuron loss on striatal activity, a second stereotaxic surgery was performed under isoflurane anesthesia after baseline imaging sessions. Desipramine (25 mg/kg) and pargyline (5 mg/kg) were administered intraperitoneally prior to surgery to increase the selectivity and efficacy of 6-OHDA lesions (Thiele et al., 2012). A small craniotomy was performed above the SNc ipsilateral to the imaged striatum (−3.1 mm posterior from bregma, 1.3 mm lateral) and 3 μg of 6-OHDA (total volume: 200 nl) was injected in SNc (4.2 mm below dura) at a rate ~100 nl/min. Sham-lesioned mice were treated identically except that 6-OHDA was omitted from the injected solution (0.2% ascorbic acid in 0.9% sterile NaCl solution). To evaluate motor impairments, mice were placed in a plus maze consisting of 4 identical arms (45 cm × 7 cm × 15 cm) at 90° to each other and monitored from above with a near infrared camera (Basler; acA2000-165um) for 10 min. The number of ipsiversive, contraversive and straight choices upon reaching the center of the maze were automatically quantified in MATLAB (Mathworks) and verified visually. Data were expressed as a turning bias index, defined as the difference between ipsiversive and contraversive turns, divided by the total number of turns.
Immunohistochemistry
Mice were deeply anesthetized with isoflurane and perfused transcardially with 4 % paraformaldehyde in 0.1 M sodium phosphate buffer. Brains were post-fixed for 1–3 days, sectioned coronally (50–100 μm in thickness) using a vibratome (Leica; VT1000S) and processed for immunofluorescence staining for tyrosine hydroxylase (Millipore; AB152, 1:1000) and dopamine transporter (Millipore; MAB369, 1:1000) using standard methods. Brain sections were mounted on superfrost slides and coverslipped with ProLong antifade reagent with DAPI (Molecular Probes). Endogenous tdTomato and GCaMP6 fluorescence were not immuno-enhanced. Whole sections were imaged with an Olympus VS120 slide scanning microscope and high resolution images of regions of interest were subsequently acquired with a Zeiss LSM 800 confocal microscope. TH and DAT immunofluorescence were quantified in ImageJ (NIH) by measuring mean pixel intensity in similarly-sized regions of interest in VTA, SNc, dorsal striatum and ventral striatum in both intact and lesioned hemispheres. After subtracting background signal from adjacent unstained brain regions, fluorescence intensity values in the lesioned hemisphere were expressed as a percentage of values measured on the intact side. Three measurements were obtained from each region of interest along the anterior-posterior axis and averaged together for each mouse.
Reagents
Drugs (all from Tocris, unless specified otherwise) were reconstituted and stored according to the manufacturers’ recommendations. 6-hydroxydopamine (6-OHDA; Sigma-Aldrich) was dissolved freshly into sterile 0.9% NaCl and 0.2% ascorbic acid immediately prior to administration to minimize oxidation. Working concentration aliquots of desipramine (25 mg/kg), pargyline (Sigma-Aldrich; 5 mg/kg), S(-)raclopride (Sigma-Aldrich; 1 mg/kg), SCH23390 (Fisher Scientific; 0.2 mg/kg), nomifensine (10 mg/kg), (-)quinpirole (3 mg/kg), SKF81297 (3 mg/kg) and L-DOPA (10 mg/kg) were prepared daily in sterile physiological saline and administered intraperitoneally at minimum 20 min prior to experimentation. L-DOPA was co-administered along with the peripheral DOPA decarboxylase inhibitor benserazide hydrochloride (Sigma-Aldrich, 12 mg/kg).
Two-Photon Imaging
Imaging was performed through a 20X long working-distance air objective (Edmund Optics; #58373) on a galvo-resonant scanning microscope (Thorlabs; Bergamo-II) equipped with GaAsP photomultiplier tubes and under the control of ScanImage 5 software (Vidrio Technologies). GCaMP6f and tdTomato were excited a using a pulsed dispersion-compensated Ti:Sapphire laser (Coherent; Chameleon Vision II) tuned to 940 nm (40–80 mW at the sample). Mice were head-fixed above a freely rotating circular treadmill consisting of a 6” plastic saucer (Ware Manufacturing) mounted on a rotary encoder (US Digital, Serial #: MA3-A10-125-B). Motorized, fixed-distance and velocity trials were conducted by connecting a integrated servo motor (Teknic Clearpath: CPM-MCVC-2310S-RQN) to the treadmill to impose 10 s-long running bouts every 20 s at 3 different speeds (6, 9 and 12 cm/s). Mouse posture was simultaneously monitored using a near infrared camera (Basler; acA2000-165um), with each video frame triggered by the two-photon imaging frame clock. The latter was recorded in Wavesurfer (https://www.janelia.org/open-science/wavesurfer) to synchronize imaging and behavioral data. Striatal fields of view (500 x 500 μm; 1–3 per mouse) near the center of the imaging window were selected based on the expression of GCaMP6f and tdTomato, and were continuously imaged with a resolution of 512 x 512 pixels at 30 Hz frame rate for a minimum of 20 min each. Bleaching of GCaMP6f was negligible over this time frame. The same fields of view were imaged multiple times on non-consecutive days using alignment based on blood vessels and a reference image.
Image Processing
Time series of two-photon images were concatenated and motion corrected, and Ca2+ fluorescence traces were extracted from individual neurons with minimal neuropil contamination using a custom MATLAB (Mathworks) pipeline described in detail in Driscoll et al. (2017) and kindly made available by Chris Harvey (Harvard Medical School; https://github.com/HarveyLab/Acquisition2P_class.git). Briefly, putative cell bodies were selected manually in the mean intensity image. Highly correlated and spatially-contiguous fluorescence sources within 30 μm of the selected pixel were then identified automatically based on the correlation structure of the pixel time series to delineate the spatial footprint of active neuronal processes. Fluorescence time series were computed by averaging across all pixels within that footprint and were manually classified into cell bodies, processes (not analyzed), and background (neuropil). This approach is advantageous over manually defined regions of interest as it segregates dendritic processes that overlap with somata if they exhibit distinct fluorescence time series. Each putative cell was paired with a neighboring background source, which was subtracted from the cell body’s fluorescence time series. Segmentation and neuropil subtraction were manually verified for each cell and adjusted when necessary using a graphical user interface to obtain clean neuropil-subtracted fluorescence traces without apparent negative-going transients. GCamp6f-labelled cell bodies were manually labeled as a dSPNs based on tdTomato fluorescence in the mean intensity image, and tdTomato-negative cells were labeled as putative iSPNs or ‘other’ (not analyzed) based on cell morphology, baseline Ca2+ fluorescence intensity and Ca2+ transient kinetics (Fig. 1 and Supplementary Fig. 1) by an experimenter blind to experimental conditions.
Image and behavior analyses
Quantification of imaging and behavioral data was carried out in MATLAB using custom-code available online (https://github.com/TritschLab). Ca2+ transients were detected using the built-in MATLAB findpeaks() function on each cell’s neuropil-subtracted fluorescence trace smoothed with a 150 ms window if they exhibited a minimum peak height and minimum peak prominence of 5 standard deviations greater than baseline, and a minimum width of 140 ms at half peak prominence. These criteria are intentionally stringent so as to exclude events not clearly resolved from baseline fluorescence at the cost of underestimating Ca2+ transient frequency per SPN. Ca2+ transient rise time is defined as the time required for fluorescence to rise from 20 to 80 % of peak height, and decay time as the time required for fluorescence to decay to one third of its peak. All imaged neurons displaying at minimum one Ca2+ transient were deemed active. Individual fields of view were only selected for longitudinal imaging and quantification if at least five active dSPNs and five active iSPNs were observed at baseline in self-initiated and motorized locomotor trials, and if mice traveled at minimum five meters in self-initiated trials to ensure adequate estimation of SPN ensemble size and mean SPN Ca2+ transient frequency. For each imaging session and field of view, we report the fraction of all dSPNs and iSPNs that display Ca2+ transients (i.e. active dSPN and iSPN ensemble size), the frequency of Ca2+ transients recorded per active dSPNs or iSPNs while moving or immobile averaged across each FOV, and the mean amplitude of Ca2+ transients recorded in all active dSPNs or iSPNs averaged for each FOV. Population-level Ca2+ event rates in Fig. 2 were calculated by summing all Ca2+ transients imaged across the entire population of dSPNs and iSPNs in each FOV, divided by the amount of time spent moving or remaining immobile. Activity in dSPNs and iSPNs was compared for each imaging session and field of view by computing a bias index, consisting of the difference in ensemble size, mean Ca2+ transient frequency or amplitude between dSPNs and iSPNs, divided by their sum. Treadmill velocity was extracted from positional information provided by the rotary encoder, down-sampled to 30 Hz and aligned to the two-photon imaging frame rate. Immobility is defined as any period of time beginning at least 0.5 s after treadmill velocity decreases below 0.2 cm/s, lasting at minimum 4 s, and ending 0.5 s before treadmill velocity exceeds 0.2 cm/s again and for which no postural movement is detected using the infrared camera. Movement bouts are defined as any period of time when absolute treadmill velocity exceeds 0.4 cm/s for a minimum of 4 s, and is preceded and followed by 4 s of immobility. Movement onsets and offsets are the first and last time points, respectively, at least two standard deviations away from treadmill velocity measured while immobile. To relate Ca2+ transient frequency to treadmill velocity, we determined treadmill speed for each recorded Ca2+ transient and divided the total number of Ca2+ transients per velocity (binned to 0.4 cm/s) per SPN by the amount of time spent by a mouse at that velocity over the course of an imaging session. To relate the fraction of active SPNs to velocity, we averaged treadmill velocity in 200 ms bins, determined the fraction of all imaged dSPNs or iSPNs active within each bin and calculated the mean percentage of active dSPNs and iSPNs at different treadmill speeds (binned to 0.4 cm/s). To align the fraction of active cells to movement onset/offset, we calculated the mean percentage of SPNs exhibiting Ca2+ transients in 200 ms-long bins around all locomotor bout onsets/offsets. For frequency of Ca2+ transients per movement onset/offset, we performed a similar procedure except that the mean number of Ca2+ transients imaged per FOV per bin was further divided by the total number of active SPNs and bin duration.
Statistical analyses
Data (reported in text and Figures as mean ± SEM) were compared in Prism 8 (Graphpad) using Mann-Whitney U test for independent groups, Wilcoxon signed-rank test for matched samples and Two-way ANOVAs (Mixed-effect model) for all multiple comparison, as indicated in text and figure legends. Significant interactions were followed-up with within-group pairwise t-tests with baseline or prelesion data corrected for multiple comparisons (Bonferroni). P values smaller than 0.05 were considered statistically significant and assigned the following nomenclature in Figures: * p < 0.05, ** p < 0.01, *** p < 0.005 and **** p < 0.001. Reported n values represent the number of fields of view imaged from N mice.
a. Fiber-photometry recording (top) and treadmill velocity (bottom) recorded from a Drd1Cre mouse virally expressing Cre-dependent GCaMP6f in dorsolateral striatum. Population Ca2+ fluorescence is most prominent during bouts of forward locomotion.
b. Same as a for an Adora2aCre mouse expressing GCaMP6f in iSPNs.
c. Representative in vivo two-photon maximum projection image of dorsolateral striatum in an Adora2aCre mouse expressing a red (tdTomato) Cre reporter in iSPNs (Ai14 line) and Cre-independent GCaMP6f virally (AAV-Syn-GCaMP6f) in all neurons. Inset: Detail of boxed area in green (top), red (middle) and merged (bottom) channels. Scale bar: 50 μm.
d. Example Ca2+ fluorescence traces from five active iSPNs (red) and five putative dSPNs (green) in FOV shown in c. Mean fluorescence signal in both pathways is highest during locomotor bouts on motorized treadmill (black).
a. Representative confocal image of a coronal midbrain section stained for TH 24 h after unilateral lesion of SNc (dashed outline) with 6-OHDA. Scale bar: 300 μm.
b. Number of TH-positive neurons identified in SNc and VTA on injection side relative to contralateral side 24 h after intracranial injection (VTA, p = 0.008; SNc, p = 0.004, n = 9 sections from 3 mice, Wilcoxon signed rank).
c. Quantification of mean self-initiated treadmill velocity before, 1 day after and 30 days after unilateral injection of vehicle (N = 6, gray) or 6-OHDA (N = 12, blue) in SNc. Within-group post-hoc statistics (Bonferroni, vs. pre-surgery) for velocity: sham, day 1: p > 0.99; day 30: p = 0.14. 6-OHDA, day 1: p = 0.035; day 30: p = 0.033.
d. Same as c for locomotor bout duration.
e. Same as c for bout frequency.
a. Mean Ca2+ transient amplitude in dSPNs active during self-initiated movement immediately before, 1 day after and 30 days after vehicle (gray, n = 9 FOVs in 6 mice) or 6-OHDA (n = 21 FOVs in 12 mice) infusion in SNc.
b. Same as a for iSPNs.
c. Same as a for bias in amplitude between pathways.
d. Mean Ca2+ transient frequency per dSPN active while immobile immediately before, 1 day after and 30 days after vehicle (gray, n = 9 FOVs in 6 mice) or 6-OHDA (n = 21 FOVs in 12 mice) infusion in SNc. No significant interaction was detected between sham and 6-OHDA lesioned mice over time (Two-way ANOVA lesion x time: F2,71 = 0.63, p = 0.54).
e. Same as d for iSPNs. A significant interaction was detected between sham and 6-OHDA lesioned mice over time (Two-way ANOVA lesion x time: F2,71 = 3.19, p = 0.0473). Within-group post-hoc analyses (Bonferroni) revealed no significant changes in Ca2+ transient frequency in iSPNs sham mice (p > 0.99 vs. prelesion on day 1 and p = 0.47 vs. prelesion on day 30), but a significant increase in 6-OHDA injected mice on day 1 (p = 0.006 vs. pre-lesion) but not on day 30 (p = 0.066 vs. pre-lesion).
f. Same as d for Ca2+ transient frequency bias between pathways. No significant interaction was detected between sham and 6-OHDA lesioned mice over time (Two-way ANOVA lesion x time: F2,67 = 1.12, p = 0.33).
g. Mean ± SEM Ca2+ transient frequency per dSPN (left) and iSPN (right) versus treadmill velocity before, 1 day after (acute) or 30 days after (chronic) unilateral vehicle injection in SNc (n = 9 FOVs in 6 mice).
h. Same as g for the percentage of all imaged dSPNs or iSPNs that show Ca2+ transients at different treadmill velocities.
i. Mean Ca2+ transient frequency recorded in dSPNs (left) and iSPNs (middle) active during fixed-velocity locomotion on a motorized treadmill and their relative bias (right) immediately before, 1 day after or 30 days after unilateral 6-OHDA injection in SNc (dSPNs: day 1 p = 0.70 and day 30 p = 0.56; iSPNs: day 1 p = 0.002 and day 30 p = 0.19; pathway bias: day 1 p = 0.065 and day 30 p = 0.084; all vs. prelesion, Wilcoxon signed rank, n = 14 FOVs in 7 mice).
j. Same as i for the percentage of all imaged dSPNs or iSPNs exhibiting Ca2+ transients over the course of imaging sessions (dSPNs: day 1 p = 0.003 and day 30 p = 3.7 × 10−4; iSPNs: day 1 p = 1.2 × 10−4 and day 30 p = 0.88; pathway bias: day 1 p = 1.2 × 10−4 and day 30 p = 1.2 × 10−4; all vs. prelesion, Wilcoxon signed rank).
k. Cumulative distribution of the proportion of active dSPNs (left) and iSPNs (right) recruited vs. total distance travelled before (prelesion), 1 day after (acute) or 30 days after (chronic) unilateral 6-OHDA injection in SNc. More than 75% of ensemble size is captured within the first five meters travelled under all experimental conditions.
a. Mean self-initiated treadmill velocity after systemic administration of vehicle (N = 5 mice), D1 and D2 receptor antagonists (N = 5 mice, p = 0.84 vs. vehicle, Mann-Whitney) or D1 and D2 receptor agonists (N = 7 mice, p = 0.88 vs. vehicle, Mann-Whitney) normalized to pre-drug baseline.
b. Same as a for locomotor bout duration (D1 and D2 receptor antagonists: p = 0.15; D1 and D2 receptor agonists: p = 0.01; all vs. vehicle, Mann-Whitney).
c. Same as a for bout frequency (D1 and D2 receptor antagonists: p = 0.008; D1 and D2 receptor agonists: p = 0.88; all vs. vehicle, Mann-Whitney).
d. Mean Ca2+ transient frequency per dSPN (left) and iSPN (middle) active during locomotion on a motorized treadmill before and after pharmacological DA elevation with nomifensine. Right: frequency bias between pathways (dSPN: p = 0.24; iSPN: p = 0.58; pathway bias: p = 0.88; all vs. prelesion, Wilcoxon signed rank, n = 14 FOVs in 6 mice).
e. Same as a for active dSPNs and iSPNs ensemble size (dSPN: p = 0.005; iSPN: p = 2.4 × 10−4; pathway bias index: p = 0.68; all vs. prelesion, Wilcoxon signed rank).
Supplementary Video
Two-photon time lapse series (30 Hz frame rate) of Gcamp6f fluorescence in dorsolateral striatum as mice engage in bouts of self-paced forward locomotion. Replay speed: 1x.
ACKNOWLEDGEMENTS
We thank Adam Carter, Un Kang, Michael Long, Adam Mar, Margaret Rice, Dick Tsien and members of the Tritsch laboratory for comments on the manuscript. We thank the GENIE Program and the Janelia Research Campus, specifically V. Jayaraman, R. Kerr, D. Kim, L. Looger, and K. Svoboda for making GCaMP6f available. M.M. is supported by a Marlene and Paolo Fresco Postdoctoral Fellowship. N.X.T. is an Alfred P. Sloan Research Fellow in Neuroscience, and is supported by grants from the Dana, Leon Levy and Whitehall Foundations, as well as from the National Institutes of Health (R00NS087098 and DP2NS105553).