Synchronized activation of striatal direct and indirect pathways underlies the behavior in unilateral dopamine‐depleted mice

Abstract For more than three decades it has been known, that striatal neurons become hyperactive after the loss of dopamine input, but the involvement of dopamine (DA) D1‐ or D2‐receptor‐expressing neurons has only been demonstrated indirectly. By recording neuronal activity using fluorescent calcium indicators in D1 or D2 eGFP‐expressing mice, we showed that following dopamine depletion, both types of striatal output neurons are involved in the large increase in neuronal activity generating a characteristic cell assembly of particular neurons that dominate the pattern. When we expressed channelrhodopsin in all the output neurons, light activation in freely moving animals, caused turning like that following dopamine loss. However, if the light stimulation was patterned in pulses the animals circled in the other direction. To explore the neuronal participation during this stimulation we infected normal mice with channelrhodopsin and calcium indicator in striatal output neurons. In slices made from these animals, continuous light stimulation for 15 s induced many cells to be active together and a particular dominant group of neurons, whereas light in patterned pulses activated fewer cells in more variable groups. These results suggest that the simultaneous activity of a large dominant group of striatal output neurons is intimately associated with parkinsonian symptoms.


| INTRODUCTION
Since pioneering work in the 1970s we know that unilateral destruction of dopamine (DA) neurons in the rodent mesencephalon induces unilateral hypokinesia, deviated posture, and spontaneous circling or turning behavior toward the lesion side (Ungerstedt & Arbuthnott, 1970). This experimental model of Parkinson's disease (PD) was initially useful in pharmacological studies and it is still useful in the characterization of function of basal ganglia outputs (Greco et al., 2008). However, one of the most striking initial findings using this model is that unilateral destruction of striatal DA input is accompanied by an increase in activity of striatal neuronal activity on the ipsilateral side (Arbuthnott, 1974;Hull, Levine, Buchwald, Heller, & Browning, 1974;Ohye, Bouchard, Boucher, & Poirier, 1970;Schultz, 1982;Schultz & Ungerstedt, 1978).
Concurrent pharmacological blockade of DA receptor 1 (D1) and DA receptor 2 (D2) SPNs produces an akinesia (often mischaracterized as catalepsy) similar to a DA lesion, but specific blockade of D1 or D2 SPNs indicates that synchronous activity is affected in greater extent by D1 compared to D2 receptor antagonists (Burkhardt, Jin, & Costa, 2009).
In order to understand the role of DA in specific subpopulations of SPNs we used calcium imaging of individual D1-or D2-expressing enhanced eGFP to visualize their activity and study their characteristic patterns of activation in cell assemblies. Moreover, we used calcium imaging and optogenetics to illustrate the specific contribution of both D1 and D2 microcircuits in the generation of the pathological circling behavior characteristic of the animal model of PD.

| Ethical standards
We used a total of 64 of 78 (14 discarded as failures in procedure or stereotaxic coordinates) Swiss Webster bacterial artificial chromosome (BAC) transgenic mice D1-enhanced fluorescent protein (eGFP) and D2-eGFP (postnatal 21-25 days, males) or C57BL/6J control mice were used. Animals were bred in the university's animal facilities. In general, we adjust the number of animals in our experiments according to the mortality rates after surgeries, the success rate of stereotaxic placement of viral injections and the general success of the experiments considering the expertise of trainees. Our experiments complied with the Guide for the care and use of laboratory animals of the US National Institutes of Health; the Society for Neuroscience Policy on the Use of Animals in Neuroscience Research and the Guiding Policies and Principles for Experimental Procedures endorsed by the government of Japan and supervised by the local Animal Care and Use Committee, approved protocol number 2018-212-2.

| Stereotaxic surgeries
Surgeries were performed under aseptic conditions in a small animal stereotaxic instrument (Leica Angle Two, Leica Biosystems, Wetzlar, Germany). Isoflorane inhalation (IsoFlo, Abbot, 1.0%-1.5% in medical O 2 ) was used as anesthetic and the respiratory rate, heart rate, and hindpaw withdrawal reflexes were monitored throughout the procedure to maintain an adequate anesthetic level as stably as possible. Animals were fixed with atraumatic earbars, the cornea protected with mycochlorin ointment 2% and the skin rubbed with anesthetic gel (10% ketoprofen and lidocaine). Post-operative care measures to control hypothermia, dehydration and pain included keeping mice warm until they came out of anesthesia (<15 min) by placing them on a disposable heating pad, administering sterile saline (0.5 ml, s.c.) immediately after surgery and leaving available, during the first 72 hr after surgery, a can of diet gel with carprofen (e.g., MediGel CPF, Portland, ME, USA). One-to-two weeks after surgery, animals were tested or the brain extracted.
Animals that received AAV1 RCaMP1h were left for at least 2 weeks to recover and allow viral expression before in vitro recording experiments. Animals that received AAV-10 ChR2-2mCherry injection had a longer surgery since a 3 mm stainless steel guide cannula (25 gauge) was inserted above the injection site and fixed to the skull with a small amount of dental cement (Super-Bond C&B, Sun Medical) followed by a 5-mm-long optic fiber (diameter: 260 μm, length 5.45 mm, Teleopto, Nagoya, Japan) placement into the injection site. A device containing the optic fiber, the high-intensity light-emitting diode (LED; 470 nm) and a connector to plug the wireless receiver was also fixed to the skull with more dental cement.

| Behavioral observations
To allow viral expression, animals were housed for at least 2 weeks before any recording or behavioral observation was performed. A couple of days before the experiments, animals were first trained for 20 min to carry a mock receiver (12 × 18 × 7 mm, 2 g) plugged to the wireless stimulation system in their home cage. On the experiment day, the infrared receiver (Teleopto, Nagoya, Japan) with the same dimensions and weight, replaced the mock receiver and animals were placed in a rectangular (45 × 20 × 15 cm) or circular (15 cm diameter × 20 cm high) testing chamber.
The stimulation device (Teleopto, Nagoya, Japan) triggered a LED of 470 nm (blue light) with intensity at the tip of 1.0 mW. Behavioral procedures consisted of a period of habituation (≈5 min) followed by a sequence of three periods repeated 10 times: 20 s pre-stimulation, 15 s of light stimulation either delivered continuously or in patterned pulses (6 ms pulses at 14 Hz) and a 30-60 s post-stimulation period. Turning behavior was measured and evaluated automatically by protocols written under Ethovision XT 11 animal tracking system (Noldus, Wageningen, The Netherlands).

| In vitro recordings of neuronal activity
For recordings of neuronal calcium activity Fluo-4 or RCaMP1h, corticostriatal slices (210-250 μm) of D1-eGFP or D2-eGFP BAC mice were visualized using an upright microscope (Olympus BX51W1F, Japan, 100W halogen lamp) fitted with brightfield (UM Plan FL 10×) and water immersion (10× Olympus) objective. The microscope was also equipped with a cooled charge-coupled device (CCD) camera (PCO.EDGE, Kelheim, Germany) with a field view of 800 × 600 μm; acquisition of images was set to 5-6 images/s. At the beginning of each experiment coordinates and landmarks of the displayed image were stored for reference and subsequent analysis. Experimental data were recorded using the CCD camera with acquisition protocols written in camware software. Short movies (180 s, 50-100 ms exposure, 250-500 ms/frames) were taken at time intervals of 5-20 min during 1-2 hr). At the end of the experiment potassium depolarization (mM): 50 KCl, 120 NaCl, 10 HEPES-Na, and 2 CaCl 2 pH7.4, was applied for 5 s to validate slice integrity, slices with small numbers of responsive cells (<10) were discarded from the analysis.
When a cell-permeant calcium indicator was used, slices were incubated in the dark at 38°C for 20-30 min with Fluo-4 (Fluo-5AM in few occasions) in 0.1% dimethyl sulfoxide diluted to a final concentration of 10 μM, pH 7.4 (95%O 2 and 5%CO 2 ). Once loaded with the calcium indicator, slices were transferred back to the microscope's recording chamber perfused with artificial cerebrospinal fluid.
The pair ChR2 and RCaMP provided independent addressable spectral channels. For neuronal activation with | 1515 JÁIDAR et Al.
ChR2 and imaging via the genetically encoded red calcium indicator RCaMP1h and RCaMP1 h activation was achieved using excitation pulses at 580 nm delivered by a xenon-arc lamp of (Lambda LS Sutter Instruments) and a bandpass filter (FB580-10, ThorLabs, Newton, NJ, USA). A combined non-overlapping calcium imaging/optogenetics was possible since light of 580 nm-the peak activation for RCaMP1hdoes not activate ChR2 (Glock, Nagpal, & Gottschalk, 2015). The emitted RCaMP1h fluorescence was passed through a long pass filter (550 nm, FEL0550, ThorLabs, Newton, NJ, USA) (Akerboom et al., 2013). Optogenetic stimulation (488 nm) was achieved using a fiber-coupled LED light source (DC2100, OGKR2 Thorlabs. Newton, NJ, USA). The same optogenetic stimulation parameters (continuous or in 14 Hz pulses) were used in slices and behaving animals.
For depolarization block assessment electrical neuronal activity of ChR2-/RCaMp1h-expressing neurons was recorded during 30 s of continuous light stimulation emitted from the fiber-coupled LED with the same intensity at the tip than the in vivo configuration (1 mW). Electrical neuronal activity was recorded using borosilicate glass micropipettes (Harvard Apparatus 30-0057, Holliston, MA, USA) heat polished to obtain direct current resistances of ∼4-6 MΩ. Micropipettes were filled with an internal solution containing in mM: 115 KH 2 PO 4 , 2 MgCl 2 , 10 HEPES, 0.5 EGTA, 0.2 Na 2 ATP, and 0.2 Na 3 GTP. The recordings were made with a microelectrode amplifier with bridge and voltage clamp modes of operation (BVC-700A, Dagan Co, Minneapolis, MN, USA). Conventional characterization of neurons was made in voltage and current clamp configurations. Access resistances were continuously monitored to be less than 20 MΩ, experiments with changes over 20% were interrupted and terminated. Software designed in LabVIEW environment (National Instruments, Austin, Texas, USA) was used for data acquisition and we performed analysis with Origin (version 8.6, Microcal, Northampton, MA, USA).
During recordings of calcium or electrical activity, slices received a constant flow of artificial cerebral spinal fluid of 2.5 ml/min.

| Processing of neuronal calcium images
For this processing we used the methods previously described Jaidar et al., 2010). Briefly, Image J (v.1.45s, National Institutes of Health, WA), MATLAB (Math-Works, Inc., Natick, MA, USA) and programs written in IDL (Exelis-Harris software version 8.2) (Cossart, Aronov, & Yuste, 2003;Mao, Hamzei-Sichani, Aronov, Froemke, & Yuste, 2001;Schwartz et al., 1998) were used to analyze the movies acquired during the recordings (see above). For each movie frame neurons were identified, their contours defined and the mean fluorescence measured as a function of time. Changes in fluorescence were computed as (Fi-Fo)/Fo, where Fi = fluorescence intensity and Fo = resting fluorescence, that is, median fluorescence of the first four frames of the movie. The calcium signals elicited by action potentials were detected based on a threshold value given by their first derivative over time (2.5× standard deviation of the noise). Images were inspected one by one to remove artifacts and slow calcium transients, most likely from glial cells Cossart et al., 2003;Jaidar et al., 2010;Sasaki, Matsuki, & Ikegaya, 2007). As previously described (Miller, Ayzenshtat, Carrillo-Reid, & Yuste, 2014) to determine whether peaks of synchrony were significant (number of coactive neurons per peak), binary activity from recorded calcium imaging (calcium spikes) were shuffled 1,000 times by randomly transposing intervals of activity within each cell. We then set a threshold corresponding to a significance level of p < 0.05 for peak detection.

| Visualization of network states
The vectorization of all recorded activity was performed to express network activity as a function of time. Active and inactive elements were therefore represented within a 250 ms time bin. With this analysis, association of bursting activity of synchronized cells occurring at a particular time can be easily identified , 2009Sasaki, Kimura, Tsukamoto, Matsuki, & Ikegaya, 2006;Sasaki et al., 2007). The inner product of all possible vector pairs (equivalent to the cosine of the angle between the vectors) was used to construct a similarity matrix of all neurons firing in synchrony. This process graphically reveals whether vectors have same or similar components (Sasaki et al., 2006(Sasaki et al., , 2007. To follow network dynamics within the neuronal microcircuits and determine how neuronal activity travels among the different network states in the field of observation, we analyzed the data using multidimensional reduction techniques previously described , 2009Jaidar et al., 2010). Briefly, (a) To reduce the dimensionality of population vectors representing network states, we applied locally linear embedding (LLE), a technique that preserves the structure of nonlinear multidimensional data (Brown, Joseph, & Stopfer, 2005;Roweis & Saul, 2000). (b) To choose the optimal number of network states, we used hard and fuzzy clustering algorithms and the Dunn's index as a validity function (Bezdek & Pal, 1998;, 2009Sasaki et al., 2007). (c) To visualize clusters of data points representing similar population vectors (i.e., network states) following activity sequences or closed cycles we projected vectors into a two-dimensional space (Jaidar et al., 2010;Liu, Khalil, & Oweiss, 2011;Sasaki et al., 2007;Schreiber, Fellous, Whitmer, Tiesinga, & Sejnowski, 2003). (d) The position of all cells involved in each state was plotted onto the original image of the slice containing the defined neuronal contours (see previous paragraph).

| Statistical analyses
Experimental and control sample size was chosen based on previous literature (Lee et al., 2016). The Shapiro-Wilk test was applied to all original data to assess normality in data distribution. Data withdrawn from a normal distribution were analyzed with the parametric statistical significance test oneway ANOVA followed by post hoc Tukey honestly significant difference (HSD). When necessary, a paired-t test was also used. Results were expressed as M = mean, SD = standard deviation and n = number of cases; statistical probability values were expressed in the results and figure legends. A simple randomization (i.e., heads-continuous; tail-pulses) was applied when the same slices were activated with pulsed or continuous light.

| Histology
Following a brief rinse in phosphate buffer 0.01 M, mice were perfused with 4% paraformaldehyde and 14% picric acid; brains were post-fixed for at least 2 hr and then cryoprotected in a 50/50 mixture of fixative and 20% sucrose in 0.01 M phosphate-buffered saline (PBS) for at least 24 hr. Sections were cut at 60 μm on a sledge microtome with a freezing stage (Yamato REM-710 electrofreeze MC-802A), washed in PBS and incubated in 20% normal goat serum. Primary antibodies to tyrosine hydroxylase (rabbit polyclonal 1:5,000, Enzo Life Sciences) or glial fibrillary acid protein (GFAP, rabbit polyclonal, Dako; 1:500) were incubated overnight at 4°C and stained with secondary goat antirabbit antibodies (Life Technologies; 1:200). At least 2 hr were allowed for binding before rinsing in PBS. Sections were mounted on slides; Vectamount AQ (Vector) or occasionally Santa Cruz mountant with DAPI was used to fix the coverslips. A spinning disk confocal microscope (Olympus BX-DSU) was used and pictures were taken using Neurolucida software and a Hammatsu (EM-CCD C91) camera.

| Spontaneous calcium activity in corticostriatal slices of intact or DA-depleted mice
The proposed involvement of DA in the generation of striatal neuronal assemblies (Burke, Rotstein, & Alvarez, 2017;Nicola, Surmeier, & Malenka, 2000;O'Donnell, 2003) became testable through calcium imaging that allows simultaneous measurement of many individual neurons Jaidar et al., 2010). With the use of population calcium imaging that allows the indirect measurement of neuronal spiking in striatal networks in vitro Jaidar et al., 2010;Lopez-Huerta et al., 2013) we characterized the population activity of SPNs in slices obtained from naïve controls and mice depleted of DA for at least 9 days. Examples of simultaneous recordings of fluo-4 calcium signals and analysis of large numbers of D1eGFP + and D2eGFP + SPNs are shown for naïve and DA-depleted mice in Figure 1 and Supporting Information Video S1. Spontaneous activity of both D1 and D2 SPNs from non-lesioned mice slices is characterized by sparse events without significant coactivity episodes (Figure 1a), and following DA depletion, a significant increase in the number of D1 and D2 active SPNs relative to controls is observed (one-way ANOVA (3,28) = 27.12, p < 0.0001), accompanied by the presence of significant coactivity peaks absent in non-lesioned controls ( responses induced by DA depletion (Figure 3) using multidimensional reduction techniques as described (see Section 2.7). A previous description of the striatal microcircuit of unidentified SPNs of rats deprived of DA, was characterized by a dominant network state (Jaidar et al., 2010). Similarly, as previously observed in rats (Jaidar et al., 2010) we observed that D1-or D2-eGFP-expressing neurons formed three sets that generated sequential activity states. Moreover, as  (Jaidar et al., 2010), one state that comprised the largest number of active neurons became dominant (refer to Figure 3a,b,d,e orange dots). Importantly, the dominant state was formed by SPNs belonging to both striatal output pathways as illustrated by the presence and absence of the plus sign in active neurons (Figure 3c,f).

| Simultaneous photoactivation of direct and indirect pathway SPNs
Consistent with the coordinated synchronous activation of indirect and direct SPNs necessary for the selection, initiation and performance of a particular movement, it has been reported that when the animal is not moving, SPNs remain inactive, but become coactive during movement initiation (Cui et al., 2013;Klaus et al., 2017). With the increase in the number of active cells and peaks of synchrony observed in DA-depleted tissue, we hypothesized that in the absence of DA, the observed increase in synchronous activation of large sets of striatal neurons could be responsible for the animal's motor impairments, such as deviated posture and persistent spontaneous turning (Costall, Marsden, Naylor, & Pycock, 1976;Dunnett & Iversen, 1982), Figure 1e.
To examine the relationship between the increment of neuronal activity and behavior, animals expressing AAV10 ChR2-mCherry (EF-1promoter) were used for the unilateral optogenetic activation through an optic probe implanted into the dorsolateral striatum (Figure 4a). We delivered to our behaving animals stimulation at the same intensity (470 nm; 1.0 mW at the tip) for 15 s in two modes as reported by others, for example, a single continuous 15 s pulse (Kravitz et al., 2010) or patterned pulses at 14 Hz (210 pulses of 6 ms each: (Jin, Tecuapetla, & Costa, 2014). When stimulation was delivered in pulses (14 Hz, 6 ms) it induced turning toward the opposite side characteristic of the activation of striatal cell assemblies (Ossowska & Wolfarth, 1995) (M 1.73 turns, SD 0.99, n = 7 mice) (Figure 4b right panel and Supporting Information Video S3). Interestingly, when light was delivered continuously in control animals, behavior changed drastically, mice turned toward the stimulated side (M 1.92 turns, SD 0.35, n = 9; Figure 4b, left panel). This turning behavior is compelling since it is similar to the one observed in DA-depleted animals (M = 1.77 turns, SD = 1.35, n = 11; Figure 1e). Optogenetically induced turning to the two types of stimulation was statistically significantly different (onetail ANOVA F(3,28) = 97.14, p < 0.001, n = 9, 7 continuous vs. patterned pulses), Figure 4b.
Our results demonstrate that in the absence of DA, spontaneous synchronous patterns of SPN activation are observed in both D1 and D2 SPNs, and that when similar large synchronous patterns of activity are induced in both D1 or D2 SPNs by continuous optogenetic stimulation, normal mice turn in circles as do animals with a DA depletion.

| Visualization in vitro of striatal dynamics under optogenetic stimulation
To further visualize how the striatal network elements behaved under optogenetic stimulations we used electrophysiology, calcium imaging, and optogenetics ( Figure 5). We combined AAV10 ChR2-mCherry (EF-1promoter) with genetically encoded calcium indicator RCaMP1h (hSyn promoter) delivered by AAV1 viral infection. It is relevant to mention that AAV10 ChR2-mCherry labels the majority of SPNs, but excludes the population of SPN contained in the striosomes (less than 20%) (Lopez-Huerta et al., 2016).
Neuronal calcium responses were recorded and their amplitude of the responsive cells was normalized by considering the brightest recorded neuronal fluorescence as 100% for each tissue slice. This analysis showed that the normalized calcium response amplitude elicited with continuous stimulation (M = 0.77, ∆F/F, SD = 0.1, n = 17 slices) was significantly larger than the response observed with stimulation in patterned pulses (M = 0.37, ∆F/F, SD = 0.1, n = 17 slices; paired t (16) = 16.22, p = 2.33504 −11 ; Figure 5 and Supporting Information Video S2). Similarly, when the same tissue slices from 11 animals were activated with stimulation delivered at random either continuously or in patterned pulses, the total number of neurons activated by continuous stimulation was significantly larger (M = 108 active neurons, SD = 32.63, n = 11) than the number of neurons activated by patterned pulses (M = 61.54 active neurons, SD = 29.65, n = 11; paired t (20) = 3.5, p = 0.005, Figure 5d).
Since prolonged electrical stimulation often induces adaptation of SPN discharge (Kita, Kita, & Kitai, 1985) and strong (10-40 mW/mm 2 ) prolonged light activation of ChR-2 can silence instead of activating neurons by depolarization block, particularly interneurons (Herman, Huang, Murphey, Garcia, & Arenkiel, 2014), we first determined in vitro, the firing characteristics of our ChR2-RCaMP1h-expressing neurons by stimulating with 470 nm (blue light), 1.0 mW at the tip, for 30 s (see Section 3). A total of 14 SPNs (n = 7 for each D1 or D2 neurons) responded in similar proportions of activation/ adaptation. During depolarization block assessment neurons were exposed to a continuous light pulse for 30 s, the duration of response of activated neurons before undergoing to depolarization block was: D1: M = 17.5 s, SD = 6.38, n = 3 and D2: M = 18.25s, SD = 11.02, n = 4. The other neurons continued firing for the whole exposure time. To our behaving animals stimulation at the same intensity (470 nm; 1.0 mW at the tip) for 15 s in two modes as reported by others, for example, (Jin et al., 2014;Kravitz et al., 2010): a single continuous 15s pulse or patterned pulses at 14 Hz (210 pulses of 6 ms each), all the experimental stimulations were therefore shorter than the average time to cease firing in both groups of F I G U R E 3 D1 or D2eGFP expressing neurons formed three sets that generated sequential activity states with one predominant set. (a and d) Analyses of coactive cells with significant peaks of synchrony (above the dotted line) also seen in Figure 1c were used for multidimensional reduction (see 2.7). (b and e). The network states reconstructed from the raster plots show that D1 or D2SPNs generate three sequential activity states (one color per state). The numbers indicate the percentage of transitions between states in the direction of the arrows. The "orange" state clearly dominates. (c and f). Superimposed onto the ROI or map of all observed active neurons, the spatial location of active SPNs can be observed for each one of the states. Colored circles represent active neurons that belong to the same state, circles with no color represent active neurons that do not belong to that particular state, and the plus sign inside the colored circles represents positive eGFPs, that is, represented as red green dots in the raster plots of Figure 1 cells. Although depolarization block is likely induced in up to half the stimulated cells, the obtained pattern of activation of larger calcium transients was indication that most of the striatal cells were not blocked from firing impulses. The activation was clearly sufficient to induce behavioral changes, in vivo, suggesting that the cells excited did influence behavior and so a large recruitment of output neurons is the likeliest explanation of the outcome of the optogenetic stimulation.
We studied striatal responses from both direct and indirect pathway (D1eGFP and D2eGFP mice) from the same slices ( Figure 6). Light stimulation induced synchronous activation in similar proportions, in both positive and negative eGFP neurons regardless of stimulation delivered continuously or in patterned pulses (D1-D2: range 47-54 synchronous neurons; one-way ANOVA F(7,36) = 1.2, p = 0.32, n = 5,6; Figure 5a To visualize the differences in population activity evoked by continuous or patterned optogenetic stimulation we created similarity maps (Carrillo-Reid, Hernandez-Lopez, Tapia, , 2009Jaidar et al., 2010) (Figure 7a,b). We found that similarity index between synchronized neurons was significantly higher with continuous stimulation compared to responses following patterned pulses, in other words, continuous stimulation not only activates a larger population of neurons, but it has the tendency of recruit the same neurons across stimuli, whereas patterned pulses recruits fewer and more alternated neurons. Network dynamic between DA-depleted and pharmacological-activated striatal cell assemblies, were mimicked by the two forms of light stimulation. That is, pathologically engaged under continuous stimulation and flexible or alternating under patterned stimulation (continuous: M = 0.52, SD = 0.19 n = 12 mice; pulsed: M = 0.4, SD = 0.23 n = 12; paired t (11) = 4.24, p < 0.001; Figure 7c).

| DISCUSSION
Early studies of striatal cells after DA damage were unanimous in finding the striatal cells more active in anaesthetized animals for example (Arbuthnott, 1974;Hull et al., 1974;Ohye et al., 1970;Schultz, 1982;Schultz & Ungerstedt, 1978). Then came the formulation that made a clear prediction that in PD the direct pathway SPNs should be less active, whereas the indirect pathway SPNs were more active (Albin, Young, & Penney, 1989). This formulation had its origin in changes in genetic activity in the two groups of cells (Gerfen & Young, 1988;Gerfen et al., 1990). The idea was very powerful and led to new therapies including globus pallidus lesions (Laitinen, Bergenheim, & Hariz, 1992), then to subthalamic nucleus deep brain stimulation (Bergman, Wichmann, & DeLong, 1990) and thence to deep brain stimulation (Polack et al., 2007). However, there were difficulties in relating this success to the electrophysiology of individual cells of the two types once they became identifiable by transgenesis. Some laboratories did find differences in the expected direction (Flores-Barrera, Vizcarra-Chacon, behavior similar to 6-OHDA treated animals, that is, toward the side stimulated. With stimulation in patterned pulses (14 Hz) animals displayed turning behavior opposite to the stimulated side (right). Behavioral sessions lasted approximately 10-15 min divided as follows: a period of habituation (≈5 min) followed by three sequences repeated 10 times: 1-20 s pre-stimulation, 2-15 s of light stimulation either delivered continuously or in pulses and 3-30-60 s post-stimulation. As described in Section 2.3, a tracking system was used to automatically measure turning. A contour detection snapshot of a recorded mouse illustrates the deviated posture typical of an ipsilateral DA depletion, but in this case in an intact mouse during optogenetic stimulation | 1521 JÁIDAR et Al. Shen et al., 2008;Warre et al., 2011) but as the methodology became more sophisticated, the obvious interactions of both cell types with cortical activity and the beta power increase in the electrocorticogram, common in PD, has changed the emphasis from "faster or slower firing" to changes in the intrastriatal dynamics which result from the interactions of the many effects of DA removal on functional relationships within the striatal network (Cui et al., 2013;Klaus et al., 2017;Lindahl & Hellgren Kotaleski, 2016;Parker, Kim, Alberico, Emmons, & Narayanan, 2016;Parker et al., 2018;Perez-Ortega et al., 2016; da Silva, Tecuapetla, Paixao, & Costa, 2018).
Analyses of many cells at a time have revealed that such neuronal activation is characterized by an increased number of coactive neuronal discharges that involve both types of SPNs (Jaidar, Carrillo-Reid, & Bargas, 2012;Jaidar et al., 2010). Using calcium imaging in eGFP transgenic animals, we first confirmed the characteristics of SPN activity in the absence of DA, and revealed that both populations of SPNs equally participate in this abnormal microcircuit activity. In our in vitro recordings of spontaneous activity in DA deprived animals, both populations of D1 and D2 SPNs to some extent increased their activity. Therefore, although electrophysiological and anatomical evidence indicates that D2, but not  (Day et al., 2006) but see (Suarez, Solis, Aguado, Lujan, & Moratalla, 2016) and an increase in excitability (Day, Wokosin, Plotkin, Tian, & Surmeier, 2008), in the absence of DA, perhaps the increase in excitability could somewhat compensate the loss of dendritic spines. A recent electrophysiological study in anesthetized rats by Sharott and collaborators claimed that only D2 cells were changed in firing rate but their laborious methodology meant a very small fraction of identified D1 cells in the lesioned animals (n = 5 in Figure 3 and n = 8 in figure 8) (Sharott, Vinciati, Nakamura, & Magill, 2017). Our method, although in vitro, samples at least 50 cells per slice of both genotypes, the calcium signals are not single action potentials like those of the electrophysiological studies but the likeliest  Figure 4a); colors representing individual active neurons and coactivity are, black: all eGFP negative; green: D1eGFP + ; red D2eGFP + . (e) In spite of a differential neuronal activation induced by stimulation delivered continuously or in patterned pulses, the percentage of activated SPNs within groups of positive or negative D1 or D2eGFP neurons remains around 50%, as expected. No preferential neuronal activation was observed. (f) Total number of coactivated neurons from both D1 or D2 neurons with positive eGFP expression. Note that both cell types showed higher synchronization to optogenetic stimulation delivered continuously than in patterned pulses (one-way ANOVA F(3,92) = 43.56, p ≤ 0.0001) explanation for the transients we measure is a burst of at least three spikes . The synchronization of striatal cells is common to both studies, with the Sharott et al. (2017) having correlated the synchronous discharges of D2 cells with beta frequencies in the EEG in their anaesthetized animals; an effect we could not test in vitro. Parker et al. (2018) reported on calcium transients in freely moving animals with a miniature camera and implanted GRIN lens. Although they too see reductions in D1 cell activity, they also see a dislocation of the response of (particularly D2) cells to the initiation of movements after 6-OHDA injections. During normal movement both D1 and D2 cells are tuned to movements but the rate changes obliterate the movement relations in D2 cells. These different results reflect the within animal control that we lose in our ex vivo models, but we do find a similar disruption of patterned activity in "clusters" of cells that they report.
The effect of bilateral SPN activation using optogenetic techniques has been studied (Parker et al., 2016), but only a few have explored unilateral effects of SPN manipulations (Cui et al., 2013;Jin et al., 2014;Lee et al., 2016). We observed similar results following stimulation in patterned pulses. This suggests that stimulation in patterned pulses is sufficient to stimulate striatal SPNs of either type. Consistently, network analyses revealed that stimulation delivered in patterned pulses (14 Hz/15 s), compared to continuous stimulation, was less efficient in recruiting neurons, but obviously sufficient to induce contralateral turning. The smaller number of coactivated neurons by pulses (14 Hz) induced contralateral turning characteristic of tonic activation of striatal cell assemblies Ossowska & Wolfarth, 1995). Reminiscent of the patterning of responses in striatal cells (independent of output class) observed in the awake mice (Parker et al., 2018). In contrast, from the analyses of network dynamics, it is clear that continuous stimulation induces a significantly higher similarity index compared to patterned stimulation, as also seen in dopamine-depleted rats (Jaidar et al., 2010).
Therefore, it seems plausible that continuous stimulation induces a dominant recurrent pattern of network activity sufficiently pronounced to induce ipsilateral turning like the spontaneous turning observed in DA-depleted animals. Such largely synchronous activity patterns would perhaps look "unstructured" if viewed with the small field of view in the in vivo camera, but in our study is obviously a very dispersed concurrent activation. Pulsed patterned stimulation could induce a restorative neuronal activity by recruiting fewer coactive neurons. This is a speculative functional relationship obtained from the observed in vitro neuronal network activation by the two stimulation parameters, that remains to be proven with in vivo recordings of neural networks, once technical advances allow such striatal recordings of hundreds of cells, without removal of the overlying cortical tissue.
The reasons for the differences in neuronal recruitment between patterned and continuous stimulation, although interesting, are not the focus of this paper; methodological aspects perhaps involving a ChR2-activated calcium conductance; reciprocal multisynaptic influences between direct and indirect SPNs and motor cortex (Burke et al., 2017;Freeze, Kravitz, Hammack, Berke, & Kreitzer, 2013;O'Hare et al., 2016;Oldenburg & Sabatini, 2015;Shipp, 2017), intrastriatal F I G U R E 7 Analysis of network dynamics of D1 and D2 SPNs. The similarity index matrices of the stimulation induced by continuous (a) and stimulation in patterned pulses (b) show that continuous delivery of light activated sets of neurons with higher similarity index across stimulations, than stimulation delivered in patterned pulses. (c) Summary box plots to illustrate the significantly higher mean similarity index of vectorized activity observed with continuous versus patterned stimulation (paired t (11) = 4.24, p < 0.001, n = 12 mice). Black numbers: vector numbers; blue numbers: responses from one of the four stimulations influences on SPNs (Planert, Szydlowski, Hjorth, Grillner, & Silberberg, 2010;Taverna, Ilijic, & Surmeier, 2008) and or the participation of interneurons (Damodaran, Evans, & Blackwell, 2014;Dehorter et al., 2009) should be considered in further analyses. Modulation of neuronal assemblies by specific patterns of neuronal stimulation may represent a modified animal model to study the pathogenesis in basal ganglia.
Our purposes were to optogenetically activate selective groups containing both D1-and D2-eGFP-expressing neurons enough to induce the striatal network activity that presumably supports the spontaneous turning in dopamine-depleted animals. Therefore, it may be possible that the larger synchronization of the striatal circuit (via continuous stimulation) more closely reproduced the effects of DA depletion. Consequently, the abnormal turning observed in the 6-hydroxydopamine animal model of PD may result from the pathological striatal synchronous activity that follows DA depletion.
It may be meaningful that to reproduce the pathological state, strong neuronal activation is necessary. It is clear that the stronger continuous stimulation of both D1 and D2 SPNs was required to induce the ipsilateral turning, as it is also the case for the bilateral excitation of D2 SPNs to elicit parkinsonian bradykinesia (Kravitz et al., 2010). A robust inhibition and excitation of different subsets of substantia nigra reticulata neurons has been reported following bilateral striatal optogenetic stimulation of D1 or D2 SPNs, for 1,000 ms but not for 100 ms (Freeze et al., 2013).
Recently, in freely moving animals, it was reported that both SPN populations become active in localized compact groups of synchronized neurons, that alternate their activity over time (Barbera et al., 2016;Klaus et al., 2017;London et al., 2018), these results are similar to previous slice recordings treated with NMDA . In contrast, slices from DA-depleted animals report neuronal activity locked into a dominant network state with reduced alternations between the assemblies (Jaidar et al., 2010;Lopez-Huerta et al., 2013). Normal striatal motor control requires activation of both direct and indirect SPNs population in a "complementary" and "coordinate" network dynamic (Jin et al., 2014;Klaus et al., 2017;Tecuapetla, Jin, Lima, & Costa, 2016;Vicente, Galvao-Ferreira, Tecuapetla, & Costa, 2016). Very recent results from a modeling study suggest that for planned movements both output pathways need to be engaged (Bahuguna, Weidel, & Morrison, 2018).
We conclude that the concept of neural networks has provided neuroscientists with an important tool not only to study their properties and capabilities (Blaeser, Connors, & Nurmikko, 2017;Carrillo-Reid, Lopez-Huerta, Garcia-Munoz, Theiss, & Arbuthnott, 2015;Carrillo-Reid et al., 2009Liu et al., 2011), but to activate different functional processes and the related behavior. Fear memories and freezing behavior have been induced by optically stimulating dentate gyrus neurons whose activity during passive avoidance was previously identified (Liu et al., 2012;Ramirez et al., 2013); for review see (Carrillo-Reid, Yang, Kang Miller, Peterka, & Yuste, 2017).
Here we have used neural network manipulation to demonstrate that: (a) DA depletion caused simultaneous activation of both types of identified D1 and D2 SPNs, (b) the contralateral turning behavior, characteristic of unilateral activated striatal neurons, was induced by patterned pulsed optogenetic stimulation of both types of SPNs, (c) ipsilateral turning behavior, as observed in DA-depleted animals, was induced in normal animals by continuous optogenetic stimulation that mimics the pathological microcircuit dynamics that accompany this animal model of PD. Our experiments confirm that large, unilateral and synchronous discharges of SPNs can simulate the ipsilateral turning model of PD in intact animals, and suggest that dopamine degeneration causes circling due to a similar large increase in unilateral SPNs discharge. The clinical relevance of this finding is supported by past studies using turning behavior as a model for PD. Resent results implicating glutamate receptors in the hyperexcitability of SPNs in monkeys (Singh et al., 2018) only add to the idea that this striatal pathological activity is close to the symptom development in PD.