The Functional Organization of Cortical and Thalamic Inputs onto Five Types of Striatal Neurons Is Determined by Source and Target Cell Identities.

To understand striatal function, it is essential to know the functional organization of the numerous inputs targeting the diverse population of striatal neurons. Using optogenetics, we activated terminals from ipsi- or contralateral primary somatosensory cortex (S1) or primary motor cortex (M1), or thalamus while obtaining simultaneous whole-cell recordings from pairs or triplets of striatal medium spiny neurons (MSNs) and adjacent interneurons. Ipsilateral corticostriatal projections provided stronger excitation to fast-spiking interneurons (FSIs) than to MSNs and only sparse and weak excitation to low threshold-spiking interneurons (LTSIs) and cholinergic interneurons (ChINs). Projections from contralateral M1 evoked the strongest responses in LTSIs but none in ChINs, whereas thalamus provided the strongest excitation to ChINs but none to LTSIs. In addition, inputs varied in their glutamate receptor composition and their short-term plasticity. Our data revealed a highly selective organization of excitatory striatal afferents, which is determined by both pre- and postsynaptic neuronal identity.

Introduction

The basal ganglia constitute a highly interconnected group of subcortical nuclei that allow organisms to adjust their behavior according to changes in their internal state or environment (Graybiel and Grafton, 2015, Graybiel et al., 1994, Grillner et al., 2005). One essential prerequisite for the selection and execution of appropriate movements is the convergence of inputs conveying sensory information, motor commands, reward value, and more. This integration is performed by the striatum, the primary input structure of the basal ganglia, which receives almost all cortical and subcortical afferents (Doig et al., 2010, Flaherty and Graybiel, 1993, Wall et al., 2013, Yeterian and Van Hoesen, 1978). Corticostriatal inputs (Guo et al., 2015) are divided into two pathways: pyramidal tract (PT)-type neurons extend their axons ipsilaterally to downstream targets such as thalamus and brainstem, with collaterals in striatum, while intra-telencephalic (IT)-type neurons project to ipsi- and contralateral cortex and to striatum to varying degrees (Alloway et al., 2006, Kress et al., 2013, Lei et al., 2004, Wilson, 1987). Frontal areas have denser projections to contralateral striatum than posterior primary sensory areas (Hooks et al., 2018, Reig and Silberberg, 2016). Besides cortex, thalamus constitutes a major source of excitatory input, which is believed to underlie behavioral switching by providing sensory, attentional, and salience information to the striatal network (Aoki et al., 2015, Aosaki et al., 1994). The principal structures giving rise to thalamostriatal fibers in rodents are the centrolateral/centromedian (CL/CM) nucleus and the parafascicular (PF) nucleus (Ellender et al., 2013, Lacey et al., 2007, Mandelbaum et al., 2019, Sadikot et al., 1992, Smith and Wichmann, 2015). Both thalamic nuclei target striatal medium spiny neurons (MSNs), but only the PF additionally targets striatal interneurons (Lapper and Bolam, 1992, Rudkin and Sadikot, 1999, Sidibé and Smith, 1999).

Most studies on cortico- and thalamostriatal projections have focused on the synaptic inputs to the two subtypes of MSNs expressing D1 or D2 dopamine receptors (D1- or D2-MSNs), because they account for most striatal neurons (Ding et al., 2008, Fino and Venance, 2010, Rothwell et al., 2015). However, alongside MSNs, the striatum contains numerous interneuron types, such as parvalbumin (PV) expressing fast-spiking interneurons (FSIs), somatostatin (SOM) expressing low threshold-spiking interneurons (LTSIs), and cholinergic interneurons (ChINs) (Kawaguchi, 1993). Although FSIs and LTSIs are sparse, they exert decisive inhibitory control over striatal output by releasing γ-aminobutyric acid (GABA) and mediating feedforward inhibition onto MSNs (Garas et al., 2016, Gittis et al., 2010, Koós and Tepper, 1999, Planert et al., 2010, Szydlowski et al., 2013). While FSIs are thought to promote sequence learning (Owen et al., 2018), LTSIs are involved in goal-directed learning (Holly et al., 2019). ChINs are tonically active, but their firing is periodically interrupted by cue and reward-related pauses (English et al., 2011, Kimura et al., 1984, Ravel et al., 1999, Schulz and Reynolds, 2013).

Apart from ChINs, all striatal neurons are GABAergic; therefore, excitatory inputs to the striatum constitute the primary driving force for the operations performed by the basal ganglia (Tepper et al., 2007). A growing body of anatomical studies has uncovered numerous striatal input structures (Guo et al., 2015, Klug et al., 2018, Wall et al., 2013), and recent studies have shown that synaptic properties differ according to presynaptic region (Assous and Tepper, 2019a, Ding et al., 2008, Lee et al., 2019, Parker et al., 2016). However, the impact of these afferent pathways on the different striatal cell types is not well described. In vivo recordings from striatal neurons have shown that their responses to cortical stimulation vary in a cell-type-specific manner (Doig et al., 2014, Reynolds et al., 2004, Sharott et al., 2012). In hippocampus and cortex, target cell type specificity of synaptic dynamics allows the extraction of distinct features from one common afferent input (Cruikshank et al., 2007, Éltes et al., 2017, Kloc and Maffei, 2014, Lefort and Petersen, 2017, Pelkey and McBain, 2007, Silberberg and Markram, 2007, Tsodyks et al., 1998). Overall, these findings suggest that the synaptic properties of inputs to striatal neurons may be finely tuned according to both the afferent pathway and the target neuron type.

In this study, we test whether synaptic transmission from cortico- and thalamostriatal projections occurs in an input- and target cell-type-specific manner. In total, we map the connections and characterize the synaptic properties of five input pathways (PF, ipsi- and contralateral primary somatosensory cortex [S1I and S1C], and ipsi- and contralateral primary motor cortex [M1I and M1C]) onto five striatal cell types (D1- and D2-MSNs, FSIs, LTSIs, and ChINs). Each input was selectively activated using optogenetics while obtaining simultaneous whole-cell recordings from MSNs and neighboring interneurons, thus revealing how the input is processed by the different components of the striatal microcircuit. Our findings show that all five afferent pathways target both types of MSNs, while the classes of co-activated interneurons are unique for each input. Moreover, we find input- and cell-type-specific differences in synaptic strength, receptor composition, and short-term plasticity, resulting in highly distinctive spatiotemporal excitation patterns. Thus, our study provides novel insights into the roles of several striatal cell types in integrating excitatory inputs within the striatal microcircuit.

Results

Corticostriatal Projections from S1 Preferentially Excite Striatal MSNs and FSIs

To explore the relative strengths of S1 inputs to striatal neuron populations, we injected AAV2-CamKIIa-YFP-ChR2 unilaterally in S1 to express channelrhodopsin (ChR2) selectively in excitatory neurons (Figures 1A–1C). Whole-cell patch-clamp recordings in acute slices confirmed that S1 layer V pyramidal cells expressed the construct (Figure 1E). Axonal projections of S1 pyramidal cells densely innervated the ipsilateral dorsolateral striatum (S1I) and induced reliable responses in striatal neurons (Figures 1, S3A, and S3B). We recorded only from pairs and triplets that were in close vicinity (intersomatic distances < 150 μm; Figures 1H–1J and S1A–S1F; Table S1). Optogenetic wide-field stimulation of S1I terminals elicited excitatory postsynaptic potentials (EPSPs) that were blocked by bath application of NBQX and D-APV (Figures S1H and S2C). Based on their onset latency, responses were monosynaptic (Figure S2F). In a subset of experiments, this was additionally confirmed by bath applying TTX and 4-AP, which failed to abolish responses (Figure S2A) (Petreanu et al., 2009).

Figure 1

S1 Input Preferentially Excites Striatal MSNs and FSIs

(A) Schematic of virus injection in S1 and the recording site in dorsolateral striatum.

(B) Confocal image of the injection site in a D2-tdTomato mouse. Red, D2-MSNs; green, virally transduced cells. Scale bar: 1 mm.

(C) High magnification of cortex in (B) showing a neurobiotin-filled pyramidal cell at the injection site in S1. Scale bar: 100 μm.

(D) High magnification of the striatum in (B) showing representative expression of ChR2-YFP in S1 axon terminals and D2-MSNs in striatum. Scale bar: 100 μm.

(E) Left: schematic of the control experiment with virus injection and recording in S1. Center: whole-cell recordings of the pyramidal cell (PYR) shown in (C) and its response to step current injections. Right: light response in the presence of synaptic blockers. Scale bars: 20 mV, 200 pA, 200 ms.

(F) Schematic of simultaneous whole-cell recordings of three striatal neurons in a parasagittal slice within the area of S1 axon terminals (green).

(G) Triplet whole-cell recordings in striatum. Differential interference contrast (DIC, top), epifluorescent image of YFP-expressing S1 axon terminals (center), and overlay (bottom) of a parasagittal slice with recording pipettes. Scale bar: 500 μm.

(H) Schematic of recordings: tdTomato-positive and tdTomato-negative neurons were recorded simultaneously, while S1 fibers were stimulated through the objective.

(I) DIC and fluorescent images of simultaneous patch-clamp recordings from two tdTomato-negative cells (putative D1-MSNs) and one tdTomato-positive D2-MSN. Scale bar: 10 μm.

(J) Characteristic responses of different striatal neuron types to increasing step current injections. Scale bars: 20 mV, 200 pA, 200 ms.

(K) Relative strength of EPSPs in striatal neurons evoked by stimulation of S1 afferents in the presence of gabazine. Left to right: representative traces of triplet recordings. Simultaneously recorded EPSPs are overlaid. D1-MSNs and D2-MSNs are identified in transgenic mouse lines, while the dopamine receptor subtype of "MSNs" is unknown. Scale bars: 2 mV, 20 ms.

(L) Summary graph of EPSP amplitudes obtained from pairs and triplets of striatal neurons in response to S1 stimulation. Solid lines indicate paired recordings in which both neurons responded, dashed lines indicate pairs in which only one neuron responded. The proportions of responding cells are shown in pie charts (inserts). Center values represent mean ± SEM (n = 17 D1-MSN-D2-MSN pairs, N = 8; n = 20 FSI-MSN pairs, N = 14; n = 16 LTSI-MSN pairs, N = 12; n = 14 ChIN-MSN pairs, N = 8; two-tailed paired t test/Wilcoxon signed-rank test).

See also Figures S1 and S2 and Table S1.

Optogenetic stimulation of S1I elicited reliable EPSPs in 100% of recorded D1-MSNs, D2-MSNs, and FSIs (Figure 1L). No difference was found in the EPSP amplitudes recorded in pairs of D1- and D2-MSNs (4.54 ± 0.96 mV in D1-MSNs versus 4.55 ± 0.88 mV in D2-MSNs, p = 0.7583, n = 17 pairs, N = 8 mice), whereas FSIs responded with larger EPSPs than neighboring MSNs (7.80 ± 1.01 mV in FSIs versus 2.89 ± 0.62 mV in MSNs, p < 0.0001, n = 20 pairs, N = 14; Figures 1K and 1L). The latency to the peak of the EPSPs was also shorter for FSIs (Figures S2E and S2G). In contrast, only 50% of LTSIs and 21% of ChINs responded to S1I stimulation (n = 8/16 LTSIs; n = 3/14 ChINs; Figures 1K and 1L). Furthermore, EPSP amplitudes of LTSIs and ChINs were significantly smaller than those of surrounding MSNs (0.92 ± 0.43 mV in LTSIs versus 4.36 ± 0.59 mV in MSNs, p = 0.0006, n = 16 pairs, N = 12; 0.14 ± 0.10 mV in ChINs versus 2.76 ± 0.75 mV in MSNs, p = 0.0001, n = 14 pairs, N = 8; Figures 1K and 1L). There were no differences in the intrinsic properties of responding and non-responding LTSIs and ChINs (data not shown). Because all interneurons were recorded simultaneously with neighboring, responding MSNs, small responses, or a lack thereof, reflect true differences in synaptic input and cannot be due to variability in virus injections, tissue preparation, or other experimental conditions. Altogether, these results suggest that S1I provides a highly selective input to the striatal microcircuit consisting of strong excitation of MSNs and FSIs, while LTSIs and ChINs are more rarely innervated by S1I afferents and have weaker synaptic EPSP amplitudes.

Anatomical studies on the IT tract of S1 (S1C) have shown that the projection to the contralateral striatum is particularly sparse (Alloway et al., 2006, Reig and Silberberg, 2016), and using standard procedures for virus expression, we were unable to evoke S1C responses in D1- and D2-MSNs previously (Ketzef et al., 2017). In line with this, we observed few S1 fibers terminating in contralateral striatum, and striatal neurons did not respond to optogenetic stimulation within the first three to five weeks after virus injections (n = 5 MSNs, 3 FSIs, 2 LTSIs, and 1 ChIN; N = 2; Figures S3A–S3E). Despite the absence of responses to S1C input, we recorded robust EPSPs ipsilateral to the injection site in the same coronal slices (Figure S3E). To elucidate the role of S1 fibers targeting the contralateral striatum, we increased the time window for expression of ChR2 in a subset of experiments beyond five weeks. In these mice, activation of S1C input occasionally evoked responses in striatal neurons (Figures S3F–S3H). Importantly, after five weeks, only a fraction of MSNs responded within the area of axonal projections (81% of MSNs, n = 26/32 MSNs, N = 5), whereas 100% of MSNs responded in the innervated areas ipsilateral to the injection site as early as two weeks after the virus injections. As observed for S1I, paired recordings of MSNs and adjacent interneurons indicate that S1C input primarily targets MSNs and FSIs, while only 50% of LTSIs are innervated by this input (3.78 ± 2.10 mV in FSIs versus 1.27 ± 0.82 mV in MSNs, p = 0.25, n = 3 pairs, N = 3; n = 3/6 LTSIs; 0.26 ± 0.04 mV in LTSIs versus 0.85 ± 0.19 mV in MSNs, p = 0.0313, n = 6 pairs, N = 3; Figures S3F–S3H). No ChINs responded to S1C input (n = 0/5 ChINs; n = 5 pairs, N = 3; Figure S3F). These results indicate that S1 afferents primarily target the ipsilateral striatum, while there are considerably fewer contralateral projections that form fewer and weaker functional synapses with MSNs and interneurons.

M1I Targets Striatal Interneurons More Frequently Than S1I

Projection neurons in M1 also extend their axons in the PT and IT tract, but fiber bundles crossing to the contralateral hemisphere are denser than the ones from S1 (Hooks et al., 2018). In accordance, we found yellow fluorescent protein (YFP) expressing axon terminals in equivalent parts of the striatum in both hemispheres following unilateral virus injections in M1 (Figures 2A–2D). We first performed whole-cell recordings from striatal neurons ipsilateral to the injection site (M1I) and found that optogenetic activation of M1I afferents evoked monosynaptic glutamatergic EPSPs (Figures S2A–S2F). In contrast to S1I, D1-MSNs responded with larger EPSPs than D2-MSNs (4.45 ± 0.87 mV in D1-MSNs versus 3.21 ± 0.78 mV in D2-MSNs, p = 0.0386, n = 16 pairs, N = 5). Yet the responses of all three types of interneurons followed the same pattern as observed for S1I: FSIs responded with significantly larger and faster EPSPs than neighboring MSNs to M1I activation (3.91 ± 0.57 mV in FSIs versus 1.11 ± 0.17 mV in MSNs, p < 0.0001, n = 15 pairs, N = 4; Figures 2F, 2H, S1H, S2E, and S2G), and both LTSIs and ChINs received weaker inputs than adjacent MSNs (0.98 ± 0.24 mV in LTSIs versus 5.81 ± 1.05 mV in MSNs, p = 0.0005, n = 18 pairs, N = 8; 0.21 ± 0.05 mV in ChINs versus 3.32 ± 0.53 mV in MSNs, p < 0.0001, n = 35 pairs, N = 14; Figures 2F and 2H). However, 89% of LTSIs (n = 16/18) and 60% of ChINs (n = 21/35) responded to stimulation of M1I input (Figure 2H), which constitutes a larger proportion of responding interneurons than for S1 input.

Figure 2

Target Cells and Synaptic Strengths of Ipsi- and Contralateral M1 Inputs Differ

(A) Schematic of virus injections in M1 labeling pyramidal cells projecting to both ipsi- and contralateral striatum.

(B) Confocal image of the injection site in a D2-tdTomato mouse. Red, D2-MSNs; green, virally transduced cells. Scale bar: 1 mm.

(C) Confocal image of YFP-expressing axon terminals (green) innervating ipsi- and contralateral striatum in a D2-tdTomato mouse. Scale bar: 1 mm.

(D) High magnification of (C) showing M1 axon terminals projecting to contralateral striatum (IT tract) and two recorded neurons filled with neurobiotin. Scale bar: 250 μm.

(E) High magnification of (D) showing two MSNs that responded to contralateral M1 input. Scale bar: 50 μm.

(F) Top: schematic of unilateral virus injections in M1 and striatal multi-neuron recordings ipsilateral to the injection site. Bottom: relative strength of EPSPs in striatal neurons evoked by stimulation of M1I afferents in the presence of gabazine. Left to right: representative traces of triplet recordings. Simultaneously recorded EPSPs are overlaid. Scale bars: 2 mV, 20 ms.

(G) Same as in (F) for recordings obtained contralateral to the injection in M1.

(H) Summary graph of EPSP amplitudes obtained from pairs and triplets of striatal neurons in response to M1I stimulation. Solid lines indicate paired recordings in which both neurons responded; dashed lines indicate pairs in which only one of the two neurons responded. The proportions of responding cells are shown in pie charts (inserts). Center values represent mean ± SEM (n = 16 D1-MSN-D2-MSN pairs, N = 5; n = 15 FSI-MSN pairs, N = 4; n = 18 LTSI-MSN pairs, N = 8; n = 35 ChIN-MSN pairs, N = 14; two-tailed t test).

(I) Same as in (H) for recordings obtained contralateral to the injection in M1 (n = 10 D1-MSN-D2-MSN pairs, N = 3; n = 12 FSI-MSN pairs, N = 5; n = 20 LTSI-MSN pairs, N = 9; n = 19 ChIN-MSN pairs, N = 7; two-tailed paired t test/Wilcoxon signed-rank test).

See also Figures S1 and S2 and Table S1.

M1C Input Preferentially Excites LTSIs, but Not ChINs

We then recorded in the striatal hemisphere contralateral to the injection in M1 (M1C), which allowed us to study the synaptic properties of the IT tract specifically (Figures 2C–2E). These inputs were also monosynaptic and glutamatergic (Figures S2A, S2C, and S2F). MSNs and FSIs responded robustly to M1C input, and the synaptic strength was equal for both types of MSNs, while FSIs responded with larger EPSPs than MSNs (0.71 ± 0.17 mV in D1-MSNs versus 1.15 ± 0.26 mV in D2-MSNs, p = 0.2754, n = 10 pairs, N = 3; 3.84 ± 0.31 mV in FSIs versus 1.58 ± 0.37 mV in MSNs, p = 0.0024, n = 12 pairs, N = 5; Figures 2G and 2I). The percentages of LTSIs responding to M1 input were equally high for both hemispheres (90%, n = 18/20 LTSIs). Although LTSIs responded with smaller EPSPs than simultaneously recorded MSNs, M1C input evoked the largest responses in LTSIs compared with the other inputs (1.33 ± 0.32 mV in LTSIs versus 2.79 ± 0.93 mV in MSNs, p = 0.0479, n = 20 pairs, N = 9; Figure 2I). The reliable occurrence of functional synapses of M1C terminals onto LTSIs and their relatively larger responses contrast with the properties of M1I and S1 input. Overall, this suggests that M1C input is an important source of excitation for the LTSI population. Recordings from ChINs and neighboring MSNs revealed that M1C input selectively avoids ChINs (n = 0/19 ChINs; 2.21 ± 0.33 mV in MSNs, p < 0.0001, n = 19 pairs, N = 7; Figures 2G and 2I).

PF Projections Preferentially Excite ChINs, but Not LTSIs

Next, we investigated the thalamostriatal pathway by injecting PF with the same virus used for cortical inputs. The viral infection was always centered on PF with negligible or no co-infection of adjacent thalamic nuclei (Figures 3A–3C). In striatum, YFP-labeled axon terminals were found ipsilateral to the injection site without converging onto a single confined area (Figures 3A, 3C, and 3D). Whole-cell recordings of MSNs revealed that optogenetic stimulation of PF afferents induced glutamatergic, monosynaptic EPSPs (Figures S2A, S2C, and S2F). Striatal responses to thalamic activation differed in several aspects from those evoked by cortex: Thalamostriatal inputs evoked smaller EPSPs in D1-MSNs than in D2-MSNs (2.30 ± 0.37 mV in D1-MSNs versus 3.51 ± 0.43 mV in D2-MSNs, p = 0.0032, n = 23 pairs, N = 5; Figures 3E and 3F). EPSPs in FSIs were larger than in adjacent MSNs, but the difference was not as large as for cortical inputs (3.83 ± 0.81 mV in FSIs versus 2.20 ± 0.3 mV in MSNs, p = 0.0329, n = 19 pairs, N = 9; Figures 3E and 3F). Interestingly, our recordings revealed that thalamus constitutes the most reliable input to ChINs (100%, n = 20/20 ChINs), whereas LTSIs failed to respond (n = 0/21 LTSIs, 3.29 ± 0.60 mV in MSNs, n = 21 pairs, N = 7; Figures 3E and 3F). Nonetheless, the EPSP amplitudes in ChINs were significantly smaller than in neighboring MSNs (0.65 ± 0.08 mV in ChINs versus 2.74 ± 0.65 mV in MSNs, p = 0.0003, n = 20 pairs, N = 8; Figures 3E and 3F). The robust innervation of ChINs, the absence of synaptic inputs to LTSIs, and comparatively weak inputs to FSIs demonstrate that thalamostriatal inputs differ from corticostriatal inputs in terms of their preferences for postsynaptic striatal targets.

Figure 3

Thalamic Input Excites All Striatal Cell Types Except of LTSIs

(A) Schematic of virus injection in PF and its projections innervating striatum.

(B) PF projection neurons express ChR2. Top: response of a neuron in PF to step current injections. Bottom: light responses in the presence of synaptic blockers. Scale bars: 20 mV, 200 pA, 200 ms.

(C) Confocal image of the injection site in a D2-tdTomato mouse. Red, D2-MSNs; green, virally transduced cells. Scale bar: 1 mm.

(D) Confocal image of YFP-expressing axon terminals (green) innervating ipsilateral striatum. Scale bar: 1 mm.

(E) Relative strength of EPSPs recorded in striatal neurons during stimulation of PF afferents in the presence of gabazine. Left to right: representative traces of triplet recordings. Simultaneously recorded EPSPs are overlaid. Scale bars: 2 mV, 20 ms.

(F) Summary graph of EPSP amplitudes obtained from pairs and triplets of striatal neurons in response to M1I stimulation. Solid lines indicate paired recordings in which both neurons responded, dashed lines indicate pairs in which only one of the two neurons responded. The proportions of responding cells are shown in pie charts (inserts). Center values represent mean ± SEM (n = 23 D1-MSN-D2-MSN pairs, N = 5; n = 19 FSI-MSN pairs, N = 9; n = 21 LTSI-MSN pairs, N = 7; n = 20 ChIN-MSN pairs, N = 8; two-tailed paired t test/Wilcoxon signed-rank test).

See also Figures S1 and S2 and Table S1.

Altogether, these results suggest that the connectivity and synaptic strength of striatal inputs depends on their exact origin (Figure 4). Ipsilateral inputs excite the striatum with similar target cell-type-specific patterns of synaptic strength but differ in the proportion of innervated LTSIs and ChINs. Moreover, our results show that ipsi- and contralateral projections from the same cortical area (M1, in this case) exhibit different connectivity patterns at their respective striatal targets (Figure 4A). While ChINs are only innervated by ipsilateral M1 input, LTSIs receive stronger input from the contralateral cortex, suggesting differences in postsynaptic target selectivity between PT and IT pathways (Figure 4B).

Figure 4

Synaptic Strength and Innervation of Four Inputs to Five Striatal Cell Types

(A) Comparison of the relative synaptic strength and innervation probability provided by each input to different striatal cell types. Top: schematic of the four inputs, including their injection sites in S1, M1, and PF and their corresponding recording sites in striatum. Center: relative strength of responses in D2-MSNs compared with D1-MSNs and FSIs, LTSIs, and ChINs compared with MSNs. Each circle represents the ratio of the EPSP amplitudes of two simultaneously recorded responding neurons. Center values represent mean ± SEM (one-way ANOVA with Tukey’s multiple comparison test). The proportions of responding cells are shown in pie charts (inserts). Bottom: schematic illustrating which cell types are most robustly excited by each input.

(B) Comparison of the relative synaptic strength and innervation probability of different inputs for each striatal cell type. Top: relative strength as shown in (A) rearranged to compare how different inputs excite and innervate each cell type. The statistics reflect how strong responding neurons are excited by each input (one-way ANOVA with Tukey’s multiple comparison test). The proportions of responding cells are shown in pie charts (inserts). Bottom: schematic summarizing which input is most robustly exciting each cell type.

Short-Term Plasticity Varies in an Input- and Cell Type-Specific Manner

The strength of a synaptic input is not uniform but varies depending on the frequency of activation, allowing a dynamic regulation of synaptic transmission (Abbott et al., 1997, Markram and Tsodyks, 1996, Zucker, 1999, Zucker and Regehr, 2002). Differences in short-term plasticity are often target cell type specific, including individual axons forming synapses onto several postsynaptic cell types that express different forms of plasticity (Markram et al., 1998, Pelkey and McBain, 2007, Silberberg, 2008). We examined the short-term plasticity of striatal inputs by activating them at 20 Hz, which is sufficiently fast to reveal dynamic properties while remaining within the activation/inactivation range of ChR2(H134R). Control experiments using Chronos, a faster excitatory opsin, confirmed the results (Klapoetke et al., 2014, Mattis et al., 2011) (Figure S4). We found that the short-term plasticity expressed by D1- and D2-MSNs primarily depended on which cortical or thalamic input was activated (Figure 5). The strongest depression was observed for S1 input, while thalamic input was characterized by a facilitatory component (Figures 5D, 5J, and S5A–S5F). The facilitation and the slow kinetics of EPSPs recorded in response to PF stimulation suggests an involvement of N-methyl-D-aspartate (NMDA) receptors, which was confirmed by bath application of D-APV (Figures S2H and S5P). M1I input was the only input that evoked target cell-type-specific differences between the two types of MSNs, eliciting stronger depression in D1-MSNs (Figures 5F–5H).

Figure 5

Short-Term Plasticity Varies in an Input- and Cell Type-Specific Manner

(A) Representative whole-cell recording of EPSPs in an LTSI evoked by optogenetic stimulation (20 Hz) of striatal input in the presence of gabazine. Scale bars: 2 mV, 100 ms.

(B) High magnification of (A): EPSPs were fitted with a double-exponential function (red), and their amplitudes were extracted after subtracting the decay of preceding EPSPs. Scale bars: 2 mV, 20 ms.

(C) Extracted EPSP amplitudes were normalized to the first pulse, and the paired-pulse ratio (PPR) and the steady-state ratio (SSR) were extracted.

(D) Quantification of responses of paired D1- and D2-MSNs (left, n = 28 pairs, N = 13) and interneuron-MSN pairs (right, n = 16 FSI-MSN pairs, N = 12; n = 7 LTSI-MSN pairs, N = 4) to 20 Hz stimulation of S1 input. MSNs recorded in parallel with interneurons are not shown. Example traces are shown in inserts for all responding cell types (top right, scale bars: 2 mV, 100 ms). Data are presented as mean ± SEM (†p < 0.05, °p < 0.01, ∗p < 0.001; one-way ANOVA for repeated measures corrected with Dunnett’s multiple comparison test).

(E) Comparison of PPR and SSR of data presented in (D) across cell types. Circles represent individual cells; center values represent mean ± SEM (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001; one-way ANOVA corrected with Tukey’s multiple comparison test). D1-MSNs were recorded with D2-MSNs, and all interneurons were recorded with MSNs (#p < 0.05, ##p < 0.01, ###p < 0.001; two-tailed paired t test/Wilcoxon signed-rank test).

(F) Same as in (D) for stimulation of M1I input (left: n = 16 D1-MSN-D2-MSN pairs, N = 4; right: n = 13 FSI-MSN pairs, N = 3; n = 17 LTSI-MSN pairs, N = 7; n = 13 ChIN-MSN pairs, N = 8).

(G) Same as in (E) for stimulation of M1I input.

(H) Same as in (D) for stimulation of M1C input (left: n = 9 D1-MSN-D2-MSN pairs, N = 3; right: n = 11 FSI-MSN pairs, N = 5; n = 12 LTSI-MSN pairs, N = 8)

(I) Same as in (E) for stimulation of M1C input.

(J) Same as in (D) for stimulation of PF input (left: n = 21 D1-MSN-D2-MSN pairs, N = 5; right: n = 17 FSI-MSN pairs, N = 8; n = 18 ChIN-MSN pairs, N = 8)

(K) Same as in (E) for stimulation of PF input.

See also Figure S5 and Table S1.

Excitatory inputs to FSIs have been shown to exhibit significant use-dependent depression in numerous brain areas (Galarreta and Hestrin, 1998, Gibson et al., 1999, Silberberg and Markram, 2007, Thomson, 1997). Similarly, ipsilateral cortical and thalamic afferents targeting FSIs showed short-term depression. However, M1C input (IT tract) provided remarkably stable excitation to FSIs, as seen in the steady-state ratio (SSR) (Figures 5H and 5I). Activation of M1I inputs was found to induce short-term facilitation in LTSIs, whereas M1C inputs were subject to short-term depression (Figures 5F–5I and S5J–S5L). ChINs exhibited short-term facilitation for M1I and PF inputs (Figures 5F and 5J). We restricted our analysis to these two inputs, because most ChINs did not respond to optogenetic stimulation of S1 inputs and the few responding neurons did not reliably follow trains of photostimulation.

Altogether, our findings show that each striatal input evokes a highly unique excitation pattern in the striatal microcircuit. In addition to synaptic strength, the presynaptic region and the postsynaptic cell type define the short-term plasticity that is expressed by the synapse between them. These diverse synaptic properties provide a neural substrate for cell-type-specific filtering, because they allow a common input to govern the temporal recruitment of different striatal populations.

Glutamate Receptor Composition Differs across Striatal Inputs

At the postsynaptic site, one key determinant of the short-term integration of synaptic inputs is the expression of NMDA and α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptors (Myme et al., 2003, Thomson, 1997). We therefore compared the NMDA to AMPA ratio of different striatal inputs and found that it was significantly larger at thalamostriatal synapses than at corticostriatal synapses for both D1- and D2-MSNs (p < 0.0001 for D1- and D2-MSNs for S1, M1I, and M1C versus PF input; n ≥ 10 D1-MSNs per input, N = 2–5; n ≥ 13 D2-MSNs per input, N = 2–6; Figures 6A–6C, S6A, and S6B). The relatively larger NMDA component is in line with the short-term facilitation observed for PF input. The contribution of NMDA currents was detected even at −70 mV in MSNs, suggesting that thalamostriatal synapses are sufficiently powerful to depolarize distal dendrites locally and release NMDA receptors from their magnesium block. Bath application of D-APV shortened the decay kinetics of the excitatory postsynaptic currents (EPSCs), confirming that NMDA currents had been activated (Figures S6C and S6D). All corticostriatal inputs were predominantly mediated by AMPA receptors in MSNs, indicating that the observed differences in corticostriatal short-term plasticity (Figure 5) are not caused by differences in receptor composition (Figures 6B, 6C, S6A, and S6B).

Figure 6

Cell Type-Specific Expression of Glutamatergic Receptors

(A) Extraction of the NMDA to AMPA ratio with cesium-based intracellular solution in the presence of gabazine: average response (dark gray) and individual trials (pale gray) in a D2-MSN showing the measurement of the AMPA current (IAMPA) and the NMDA current (INMDA). Scale bars: 200 pA, 50 ms.

(B) Representative average AMPA and NMDA currents obtained in striatal neurons evoked by activating S1 (top), M1I (center), and PF (bottom) inputs. Scale bars: 200 pA, 50 ms. The fractions of ChINs that responded, failed to respond, or expressed silent synapses are shown in pie charts (right).

(C) Summary graph of the NMDA to AMPA ratio of responding neurons (n = 10–25 D1- and D2-MSNs per input, N = 2–6; n = 8–16 FSIs per input, N = 3–8; n = 4–14 ChINs per input, N = 2–6; one-way ANOVA with Tukey’s multiple comparison test for S1, M1I M1C, and PF inputs).

(D) Scatterplot of absolute AMPA and NMDA currents of responding neurons. Responses to different inputs are pooled for each cell type.

(E) I-V curves of AMPA-mediated currents in MSNs and FSIs in response to activation of M1I inputs (n = 11 MSNs, n = 9 FSIs, N = 6; two-tailed unpaired t test). Recordings were obtained with a spermine-based intracellular solution in the presence of gabazine and D-APV. Insets show representative AMPA currents from MSNs and FSIs normalized to the response at −70 mV. Scale bar: 10 ms.

(F) Same as in (E) for activation of PF inputs (n = 7 MSNs, n = 6 FSIs, N = 5; two-tailed t test).

See also Figure S6 and Table S1.

FSIs exhibited large AMPA currents, while the NMDA component was negligible or absent, independent of the input (Figures 6B–6D). This finding points to a presynaptic mechanism underlying differences in short-term plasticity in FSIs. Activation of S1 inputs evoked the largest NMDA currents in FSIs, but the overall NMDA to AMPA ratio was significantly lower than the ratios obtained from MSNs and ChINs (Figures 6B–6D and S6A). The predominant expression of AMPA receptors provides a mechanism for fast synaptic transmission commonly used by FSIs in other brain regions (Angulo et al., 1999, Collingridge et al., 1988, Jonas et al., 1994, Koh et al., 1995, Nyíri et al., 2003, Sah et al., 1990). The temporal precision of postsynaptic AMPA currents can be sharpened by expressing calcium-permeable AMPA receptors (CP-AMPA), which are characterized by faster kinetics than calcium-impermeable ones (Burnashev et al., 1992, Hollmann et al., 1991, Jonas and Burnashev, 1995). CP-AMPA receptors either lack the GluA2 subunit or contain an unedited version and can be identified by a characteristic inward-rectifying current-voltage (I-V) relationship (Hollmann et al., 1991). Previous studies have suggested that striatal FSIs express CP-AMPA receptors, but it is unknown whether their expression is regulated in an input-specific way, as in hippocampus (Gittis et al., 2011, Topolnik et al., 2005, Tóth and McBain, 2000). Using a spermine-based intracellular solution, we found strong inward rectification of AMPA currents in FSIs, but not MSNs, confirming the expression of CP-AMPA receptors in these interneurons (Figures 6E and 6F). Furthermore, our data show that the expression of CP-AMPA receptors in FSIs and the expression of calcium-impermeable ones in MSNs are maintained across different presynaptic inputs, including cortical and thalamic afferents, suggesting cell-type-specific regulation of AMPA receptor subtypes in the striatum (Figures 6E and 6F).

ChINs displayed large NMDA currents in response to activation of PF, while their AMPA component was comparatively small (Figures 6A–6D). The same ratio of NMDA to AMPA receptors was observed in a subset of ChINs when activating cortical S1I or M1I inputs (Figures 6A–6D). Stimulation of S1I terminals evoked current amplitudes in the same range as those evoked by thalamostriatal afferents in 50% of the recorded ChIN-MSN pairs (n = 4/8 pairs, N = 3; Figures 6B and 6D), while there was only a small NMDA current and almost no detectable AMPA current in the remaining 50% of the recorded ChINs (n = 4/8 pairs, N = 3; Figures S6E and S6F). These responses resemble those observed at silent synapses (Faber et al., 1991, Isaac et al., 1995). Recordings of ChINs during M1I stimulation revealed large AMPA currents in all neighboring neurons, but only 44% of ChINs responded (n = 7/16 ChINs), 44% of ChINs failed to respond (n = 7/16 ChINs), and the NMDA-mediated responses in the remaining 12% of ChINs again resembled silent synapses (n = 2/16 ChINs, N = 7; Figures 6B–6D, S6G, and S6H). The failure of 44% of ChINs to respond to M1I input confirms that M1I afferents form sparse synapses onto ChINs. These findings suggest that excitatory inputs are predominantly mediated by NMDA currents in ChINs and that these NMDA receptors are most reliably activated by PF.

Our results show that NMDA to AMPA ratios at corticostriatal synapses targeting MSNs, FSIs, and ChINs are primarily determined by the postsynaptic cell type, not the presynaptic cortical region. Altogether, this indicates a homogeneous regulation of NMDA and AMPA receptors across different corticostriatal synapses. We further show that FSIs and ChINs express a stable ratio of these receptors not only for different cortical inputs but also for thalamic afferents. In contrast, D1- and D2-MSNs expressed different ratios of NMDA and AMPA receptors between cortical and thalamic synapses.

Inputs from PF and M1 Differentially Modulate the Firing of Tonically Active Striatal Interneurons

Compared with MSNs and FSIs, our data show that ChINs and LTSIs receive significantly weaker inputs for all tested pathways, raising the question of whether such weak inputs are sufficient to modify their activity. Both ChINs and LTSIs are characterized by a depolarized membrane potential and often spontaneous spiking (Beatty et al., 2012, Bennett and Wilson, 1999) (Figure S1C), suggesting that relatively weak synaptic inputs can shape their ongoing activity. To test this, we obtained cell-attached recordings from ChINs and LTSIs to minimize interference with their intrinsic properties and firing. A train (20 Hz) of thalamic photostimuli evoked a burst of spikes in 6/6 ChINs, which was followed by a pause in spiking (n = 8 ChIN-MSN pairs, N = 2; Figure 7B). ChINs increased their activity only in response to repetitive stimulation, showing that PF can efficiently modulate their firing despite low levels of AMPA receptor expression (Figures 6C and 7D). In contrast to PF input, stimulation of M1I afferents failed to affect ongoing spontaneous ChIN activity, although it did evoke large EPSPs in simultaneously recorded neighboring neurons (n = 8/8 ChINs: n = 11 pairs, N = 5; Figures 7A and 7C).

Figure 7

Opposing Effects of Thalamic and M1I Input on Activity of ChINs and LTSIs

(A) Representative traces of a simultaneous recording from an MSN (whole-cell mode, top) and an adjacent ChIN (cell-attached mode, below) during 20 Hz stimulation of M1I inputs in a ChAT-tdTomato mouse. Scale bars: 2 mV, 200 ms. A raster plot and a histogram of the spikes recorded in the ChIN are shown at the bottom. Recordings were done in modified artificial cerebrospinal fluid (ACSF) containing 5 mM potassium to increase spontaneous activity.

(B) Same as in (A) for stimulation of PF inputs. Scale bars: 0.5 mV, 200 ms.

(C) Average population response following M1I stimulation. The histogram shows the spiking of individual ChINs in gray and the average response in purple (20 ms bins). The insert shows spiking before (t1), during (t2), and after (t3) stimulation and during the recovery pulse (t4). Gray circles represent individual ChINs: colored circles represent the mean ± SEM. All ChINs were recorded in parallel with one or two responding neurons (n = 8 ChINs; n = 11 pairs, N = 5 ChAT-tdTomato mice; repeated-measures one-way ANOVA with Tukey’s multiple comparison test).

(D) Same as in (C) for PF-evoked changes in ChIN firing (n = 6 ChINs; n = 8 pairs, N = 2 ChAT-tdTomato mice).

(E) Same as in (A) for a simultaneous recording of an MSN, together with an LTSI in response to optogenetic stimulation of M1I. Scale bars: 1 mV, 200 ms. Recordings were done in modified ACSF containing 5 mM potassium and gabazine.

(F) Same as in (E) for stimulation of PF inputs. Scale bars: 1 mV, 200 ms.

(G) Average population response as in (C) for changes in LTSI firing following M1I stimulation. All LTSIs were recorded in parallel with one or two responding neurons (n = 5 LTSIs; n = 9 pairs, N = 2 SOM-tdTomato mice).

(H) Same as in (G) for PF-evoked changes in LTSI firing (n = 7 LTSIs; n = 9 pairs, N = 3 SOM-tdTomato mice).

See also Figure S7 and Table S1.

Next, we assessed the effect of weak synaptic inputs on the intrinsic activity of LTSIs. Because the spontaneous firing of LTSIs is tightly controlled by ambient GABA in acute slices, these experiments were performed in gabazine (Elghaba et al., 2016). In contrast to ChINs, activation of PF did not affect the intrinsic activity of LTSIs (n = 0/7 LTSIs; n = 9 LTSI-MSN pairs, N = 3), whereas stimulation of M1I and M1C inputs significantly increased their firing (M1I: n = 5/5 LTSIs; n = 9 pairs, N = 2; Figures 7E–7H and S7). Interestingly, LTSIs already showed a reliable increase in spiking at the first pulse, suggesting that LTSIs and ChINs differ not only with regard to the source of afferent excitation but also in the dynamic integration of synaptic inputs (Figures 7D and 7G). These results show that both ChINs and LTSIs act as effective integrators of afferent activity, with their depolarized membrane potential and spontaneous activity allowing weak excitatory synaptic inputs to reliably shape their activity.

Discussion

In this study, we present a comprehensive map of the connectivity and synaptic properties of four excitatory striatal afferent pathways onto five major striatal neuron populations. We dissect each of these pathways individually, characterizing synaptic properties such as strength, short-term plasticity, and postsynaptic glutamate receptors, as well as the anatomical target area within the striatum. Using simultaneous multi-neuron recordings, we are able to demonstrate that our observations constitute robust cell-type-specific differences. Our results reveal that each input structure extends diverging projections that target specific subsets of striatal neuron populations while avoiding others. MSNs and FSIs are robustly innervated by cortical and thalamic afferents, with FSIs receiving the strongest and fastest synaptic inputs. In contrast, the two types of tonically active interneurons, ChINs and LTSIs, receive overall weaker inputs than neighboring MSNs. LTSIs are most reliably activated by M1 inputs but do not receive inputs from PF. In contrast, ChINs are strongly driven by PF but receive relatively weak input from ipsilateral cortex and none from S1C and M1C (IT tract). We further show that the synaptic properties of each pathway are determined by both the presynaptic region and the postsynaptic cell type, thus showing a high degree of specificity in the interactions between the striatal microcircuitry and its excitatory afferents.

Striatal S1 and M1 Inputs Differ Anatomically and Functionally

Cortical inputs to the striatum are often treated as a uniform entity, and many studies use electrical or optogenetic activation of global cortical afferents (Arias-García et al., 2018, Ding et al., 2008, Doig et al., 2010, Ibáñez-Sandoval et al., 2011, Owen et al., 2018, Sciamanna et al., 2015). Apart from one recent study focusing on MSNs and FSIs (Lee et al., 2019), the synaptic properties of excitatory inputs from distinct functional cortical units to striatal neurons remain unknown. To address this question, we used local virus injections that enabled us to study the synaptic properties of S1 and M1 separately. In contrast to electrical stimulation, the viral approach allowed us to target specifically excitatory neurons while avoiding several inhibitory populations that project to the striatum (Melzer et al., 2017, Rock et al., 2016). Our data show that ipsilaterally projecting M1 and S1 inputs differ both anatomically and functionally: while S1 afferents densely target the dorsolateral striatum and rarely contact LTSIs and ChINs, M1I afferents innervate not only the dorsolateral areas but also the dorsomedial striatum and frequently form functional synapses onto LTSIs and ChINs. These anatomical findings support behavioral studies identifying the dorsolateral striatum as the neural substrate for sensorimotor integration, whereas the dorsomedial striatum acts as an associative territory central to instrumental conditioning and behavioral flexibility (Devan and White, 1999, Sabol et al., 1985). The short-term plasticity of S1I and M1I inputs to striatal neurons also varied in an input-specific manner. EPSPs evoked by stimulation of S1I depressed strongly in a use-dependent manner in MSNs, whereas M1I inputs exhibited slower use-dependent depression. The overall depressing short-term plasticity of corticostriatal inputs to MSNs is surprising given the rich literature reporting facilitatory corticostriatal synapses (Ding et al., 2008, Sciamanna et al., 2015). In these previous studies, however, it is less clear which excitatory inputs were activated because of the use of extracellular stimulation electrodes or optogenetic stimulation, without the spatial specificity obtained by local expression of ChR2.

Both S1I and M1I inputs were predominantly mediated by AMPA receptors in MSNs, suggesting that presynaptic mechanisms underlie the differences in dynamics (Smeal et al., 2007, Smith et al., 2009). Several synaptic features observed in MSNs were also observed in interneurons, suggesting that short-term plasticity and anatomical organization of cortical inputs both depend on the exact cortical origin. These findings highlight the importance of dissecting striatal inputs into separate functional pathways whose properties need to be studied individually.

Despite differences in innervation and short-term plasticity, the synaptic strength and receptor composition of S1I and M1I inputs were similar for each striatal cell type. Both cortical input structures excited FSIs more than MSNs, whereas LTSIs and ChINs received weaker and sparser inputs than MSNs. Our results are in accordance with in vivo results from Sharott et al. (2012), in which the responses to electrical stimulation in motor cortex were most pronounced in PV-expressing interneurons, followed by MSNs and lastly by LTSIs and ChINs. Our data on the corticostriatal pathway also resemble thalamocortical inputs that have been shown to strongly excite FSIs but only rarely and weakly excite LTSIs (Cruikshank et al., 2007, Gibson et al., 1999). More studies are required to assess whether the preferred excitation of FSIs compared with LTSIs and ChINs (by several orders of magnitude) is an emerging pattern that can be generalized for all corticostriatal inputs. In contrast to our study, Lee and colleagues reported that M1I input excites MSNs and FSIs equally (Lee et al., 2019). This discrepancy could be explained by different experimental conditions. Here, we used simultaneous recordings from different neuron types held at similar membrane potentials in the presence of GABAA blockers, thus enabling us to study excitatory inputs without the involvement of feedforward inhibitory responses (Assous and Tepper, 2019a, Assous et al., 2017) and without differences in driving force.

Cortical Inputs to Ipsi- and Contralateral Striatum Differ

Unlike S1, M1 appears to project equally to the ipsi- and contralateral striatum in terms of the innervated areas and the overall density of axonal arborizations (Brown et al., 1996, Lei et al., 2004, Reig and Silberberg, 2016). Yet we found clear differences between M1I and M1C inputs, including a lack of synaptic inputs from M1C to ChINs. Based on our approach, we cannot exclude that ChINs receive inputs at their distal dendrites that were not detected because of dendritic filtering (Rall, 1969). However, the large responses in simultaneously recorded MSNs show that M1C input is better positioned to excite MSNs than ChINs. The absence of detectable M1C input and the failure of M1I input to influence the intrinsic firing of ChINs strengthen the hypothesis that ChINs process attention- and salience-related inputs rather than direct motor information (Ding et al., 2010, Threlfell and Cragg, 2011). In contrast, LTSIs were most reliably innervated by M1 afferents from both hemispheres, with M1C providing the strongest input to LTSIs of all inputs tested. Although EPSPs elicited by M1 activation were small, they robustly increased spike rate and bursts in LTSIs, suggesting that the role of LTSIs is closely linked to motor function. This is in line with a previous study that showed that co-activation of M1 and secondary motor cortex can strongly activate LTSIs (Ibáñez-Sandoval et al., 2011). LTSIs release GABA and co-release several neuromodulators during high-frequency activity, including SOM, nitric oxide, and neuropeptide Y (Kawaguchi et al., 1995). These neuromodulators are key controllers of striatal excitability, synaptic plasticity, and dopamine release, which puts LTSIs in an excellent position for assisting motor learning (Adewale et al., 2007, Calabresi et al., 1999, Galarraga et al., 2007). Moreover, the simultaneous recordings of striatal MSNs and interneurons revealed that autonomously active cells such as LTSIs and ChINs receive consistently smaller excitatory inputs than do MSNs and FSIs, suggesting that the synaptic strength of striatal inputs is finely tuned to the receiving postsynaptic cell type.

Striatal FSIs responded with the largest and fastest EPSPs to all cortical inputs because of a combination of intrinsic and synaptic properties, including a short membrane time constant, the absence of NMDA receptors at all cortico- and thalamostriatal synapses, and the expression of CP-AMPA receptors. These findings are consistent with other studies on FSIs in several brain areas reporting intrinsic and synaptic properties that support fast signaling and calcium-dependent plasticity—important features for mediating precisely timed inhibition onto projection neurons (Geiger et al., 1995, McBain and Fisahn, 2001, Owen et al., 2018, Tóth and McBain, 1998). Typically, excitatory inputs to FSIs not only are characterized by fast kinetics but also express strong use-dependent depression, thereby limiting the temporal integration of FSIs and turning them into effective low-pass filters and onset detectors (Beierlein et al., 2003, Galarreta and Hestrin, 1998). However, excitatory inputs to striatal FSIs evoked comparatively weakly depressing responses across all inputs, which was particularly pronounced for excitatory inputs from M1C.

Corticostriatal inputs from S1C were found to be sparser and substantially weaker than inputs from S1I and M1C. Yet the small number of neurons responding to S1C input indicates that the same corticostriatal hierarchy applies: MSNs and FSIs show the largest responses, LTSIs respond more rarely and weakly, and ChINs fail to respond. Overall, the responses to ipsi- and contralateral inputs from the same cortical area differed in amplitude and other features such as short-term plasticity. This could be because of the co-activation of both PT and IT pathways for ipsilateral inputs as opposed to activation of the IT pathway only when stimulating contralateral inputs. Another possible reason could be the delayed expression of ChR2 in contralateral terminals.

Thalamostriatal Synapses Have Different Properties than Corticostriatal Synapses

Thalamostriatal inputs were found to excite D2-MSNs more than adjacent D1-MSNs and to evoke facilitating responses in both types of projection neurons. The responses involved a strong NMDA component that was frequently activated at rest, indicating strong dendritic excitation. The high NMDA to AMPA ratio expressed by MSNs specifically at thalamostriatal synapses, but not corticostriatal synapses, is consistent with the comparatively slow kinetics we observed for thalamic-evoked EPSPs. Furthermore, these results are in agreement with several previous studies, some of which used the same genetic approach for targeting PF (Ellender et al., 2013, Hunnicutt et al., 2016, Mandelbaum et al., 2019, Spruston et al., 1994). However, studies using VGluT2-Cre mice to target thalamic inputs, or electrical stimulation of thalamic fibers, reported lower NMDA to AMPA ratios and lower paired-pulse ratios in MSNs (Ding et al., 2010, Parker et al., 2016). One limitation of electrical stimulation is that it inherently co-activates thalamic axons arising in PF and in the CL/CM nucleus, which are known to differ both anatomically and functionally (Deschênes et al., 1996b, Deschênes et al., 1996a, Lacey et al., 2007). Moreover, Mandelbaum and colleagues provided evidence for the heterogeneity of PF by showing that this nucleus consists of at least three subpopulations that differ on genetic, anatomical, and physiological levels (Mandelbaum et al., 2019).

Previous studies using different experimental approaches (Ding et al., 2010, Guo et al., 2015, Smith et al., 2004, Threlfell et al., 2012), as well as our findings, suggest that ChINs are powerfully modulated by PF inputs. In addition, we show that PF activation of ChINs is largely mediated by NMDA receptors and that the depolarization required for NMDA receptor activation is achieved through a combination of a depolarized resting membrane potential and spontaneous activity, rather than prominent AMPA receptor activation. Compared with cortical inputs, the relative strength of PF input to FSIs was smaller, indicating that PF is less effective in activating the FSI-MSN feedforward circuitry. This is consistent with anatomical studies that have shown that PF inputs to FSIs are sparse, whereas cortical afferents target FSIs massively (Kita, 1993).

Overall, we show that each glutamatergic input to the striatum evokes input- and cell-type-specific excitation of striatal neurons, resulting in unique spatiotemporal activation of the striatal network. Our findings highlight how the amplitude, short-term plasticity, and receptor composition of striatal inputs are finely adapted to the respective postsynaptic neuron populations. Revealing these activation patterns is crucial for understanding the computations performed by the striatum and for modeling the function of the basal ganglia. The discovery of novel glutamatergic inputs arising in the pedunculopontine nucleus raises the question whether these inputs follow similar rules of functional organization (Assous et al., 2019, Klug et al., 2018). Once a detailed understanding of striatal inputs has been established, the next important aspect concerns the local interactions that these inputs evoke within the striatal microcircuit. Several studies have shown that striatal inputs can recruit local inhibition and trigger the release of neuromodulators, thereby adding another layer of complexity (Assous and Tepper, 2019b, Assous et al., 2017, English et al., 2011, Tanimura et al., 2019). In addition, more studies will be needed for assessing the synaptic properties of inputs to other types of striatal interneurons and for clarifying how each of these pathways contributes to the activity observed in vivo (Tepper et al., 2018).

STAR★Methods

Key Resources Table

Lead Contact and Materials Availability

Further information and data supporting the findings of this study are available upon reasonable request. All requests should be directed to and will be fulfilled by the Lead Contact, Gilad Silberberg (gilad.silberberg@ki.se).

Experimental Model and Subject Details

All animal procedures were performed in accordance with the national guidelines and approved by the local ethics committee of Stockholm, Stockholms Norra djurförsöksetiska nämnd, under an ethical permit to G. S. (N12/15). Both male and female mice, aged from 4 to 12-week-old were used in this study, at the first stage for viral injections followed 3 or more weeks later by ex vivo patch clamp recordings. Mice were group-housed under a 12 hr light / dark schedule and given ad libidum access to food and water. D1-Cre (EY262 line, GENSAT), D2-Cre (ER44 line, GENSAT), SOM-Cre, PV-Cre, and ChAT-Cre (stock #018973, #017320, and #006410, respectively, the Jackson laboratory) mouse lines were crossed with a homozygous tdTomato reporter mouse line (‘Ai9’, stock #007909, the Jackson laboratory) to allow identification of the respective cell types based on the expression of a fluorescent marker protein. All Cre lines were heterozygous and maintained on a wild-type C57BL/6J background (stock # 000664, the Jackson Laboratory). Additionally, heterozygous Lhx6-eGFP (MMRRC, stock number 000246-MU) mice in which both FSIs and LTSIs are labeled with eGFP, were used for some current clamp recordings. Lhx6-eGFP mice were kept on a SwissWebster background (RjOrl:SWISS, Janvier).

Method Details

Virus injections

4 to 12-week-old male and female mice were anesthetized with isoflurane and placed in a stereotaxic frame (Havard Apparatus, Holliston, MA). Injections were done with a Quintessential Stereotaxic Injector (Stoelting, Wood Dale, IL) in either S1 (coordinates AP −1.5 mm, ML 3.5 mm, DV −0.7 mm), M1 (coordinates AP +1.5 mm, ML 1.8 mm, DV −0.7 mm) or PF (AP −2.3 mm, ML 0.5 mm, DV −3.3 mm). A volume of 0.5 μL of virus (AAV2-CamKIIa-YFP-ChR2 or AAV5-Syn-chronos-GFP, UPenn) was injected at 0.1 μL / min into cortex and the pipette was held in place for 5 min after the injection. For thalamic injections 0.3 - 0.5 μL of virus were injected and because of the deep location and the small size of PF, the pipette remained for 10 min in place before being slowly retracted. The injection site was frequently imaged post recordings. Analgesics were given following surgery (Buprenorphine, 0.08 mg/kg, i.p.).

Slice preparation and electrophysiology

Three to nine weeks after virus injection, mice were deeply anaesthetized with isoflurane and decapitated. The brain was removed and immersed in ice-cold cutting solution containing 205 mM sucrose, 10 mM glucose, 25 mM NaHCO3, 2.5 mM KCl, 1.25 mM NaH2PO4, 0.5 mM CaCl2 and 7.5 mM MgCl2. Parasagittal and coronal brain slices (thickness 250 μm) were prepared with a Leica VT 1000S vibratome and incubated for 30 – 60 min at 34°C in a submerged chamber filled with artificial cerebrospinal fluid (ACSF) saturated with 95% oxygen and 5% carbon dioxide. ACSF was composed of 125 mM NaCl, 25 mM glucose, 25 mM NaHCO3, 2.5 mM KCl, 2 mM CaCl2, 1.25 mM NaH2PO4, 1 mM MgCl2. Subsequently, slices were kept for at least 30 min at room temperature before recording.

Whole-cell patch clamp recordings were obtained in oxygenated ACSF at 35°C. Neurons were visualized using infrared differential interference contrast (IR-DIC) microscopy (Zeiss FS Axioskop, Oberkochen, Germany). Fluorescent cells were identified by switching to epifluorescence using a mercury lamp (X-cite, 120Q, Lumen Dynamics). Up to three cells of different cellular identities were patched simultaneously. All striatal D1- and D2-MSNs were recorded as pairs and triplets comprising tdTomato-positive and tdTomato-negative cells in either D1- or D2-tdTomato mice. Recordings from FSIs, LTSIs and ChINs were obtained in the respective transgenic mouse lines (PV-, SOM- or ChAT-tdTomato or Lhx6-eGFP mice) and from non-fluorescent neurons that were classified according to their electrophysiological properties (Figures 1H–1J and S1A–S1F; Table S1). Borosilicate pipettes of 5 – 7 MΩ resistance were pulled with a Flaming / Brown micropipette puller P-1000 (Sutter instruments). All recordings were done in current clamp mode except of the characterization of glutamate receptors and the modulation of tonic activity in ChIN and LTSI. The current clamp intracellular solution contained 130 mM K-gluconate, 5 mM KCl, 10 mM HEPES, 4 mM Mg-ATP, 0.3 mM GTP, 10 mM Na2-phosphocreatine (pH 7.25, osmolarity 285 mOsm). In some cases, 0.3% neurobiotin was added to the intracellular solution (Vector laboratories, CA). Although striatal neurons differ with regard to their resting membrane potential, we held the cells at a membrane potential of −75 ± 2 mV to ensure that the driving force is equal for all cells when studying the synaptic strength and short-term plasticity. For voltage clamp recordings pipettes of 3-5 MΩ were used and the NMDA to AMPA ratio was measured with a caesium-based intracellular composed of 100 mM CsMeSO3, 10 mM CsCl, 10 mM HEPES, 4 mM Mg-ATP, 0.3 mM Na-GTP, 10 mM Na2-phosphocreatine, and 10 mM tetraethylammonium chloride (TEA-Cl). The AMPA receptor subunit composition can be revealed ex vivo by adding spermine to the intracellular solution, which mimics the voltage-dependent block of the channel by endogenous polyamines at positive voltages (Hollmann and Heinemann, 1994, Koh et al., 1995, Tóth et al., 2000). This solution contained 105 mM CsMeSO3, 8 mM NaCl, 10 mM HEPES, 4 mM MgATP, 0.3 mM NaGTP, 10 mM Na2-phosphocreatine, 0.3 mM EGTA, 5 mM TEA-Cl, 5 mM Qx-314, and 0.1 mM spermine. Recordings were amplified using a MultiClamp 700B amplifiers (Molecular Devices, CA, USA), filtered at 2 kHz, digitized at 10-20 kHz using ITC-18 (HEKA Elektronik, Instrutech, NY, USA), and acquired using custom-made routines running on Igor Pro (Wavemetrics, OR, USA). Liquid junction potential was only corrected for recordings with spermine-based intracellular solution since this intracellular solution had a comparatively large liquid junction potential of ~20 mV. Throughout all recordings pipette capacitance and access resistance were compensated for and data were discarded when access resistance increased beyond 30 MΩ.

Stimulation protocols and drug application

The intrinsic properties of the neurons were determined by a series of hyperpolarizing and depolarizing current steps and ramps, enabling the extraction of sub- and suprathreshold properties. Based on electrophysiological features such as resting membrane potential, input resistance, membrane time constant (tau), sag, and firing frequency, MSNs, LTSIs, FSIs, and ChINs were in some cases also identified in the absence of fluorescent marker proteins in current clamp recordings (Figure S1).

Optogenetic stimulation was generated by a 1-Watt blue LED (wavelength 465 nm) and delivered through the 64x objective lens. Duration and intensity of the light stimulation was controlled by an LED driver (Mightex Systems) connected to the ITC-18 acquisition board. Light pulses of 2 ms duration were used for activating cortical or thalamic terminals and stimuli were repeated for at least 8 sweeps with 10 s time intervals in between. Throughout all experiments, light intensity was adjusted to ensure that responses in all simultaneously recorded cells were subthreshold (< 15 mV, 0.4 mWatt / mm2, Figure S1H).

AMPA and NMDA currents were measured in the presence of gabazine at a clamping potential of −70 and +40 mV, respectively. Traces were averaged offline and baseline was subtracted. The current peak at −70 mV was extracted as the AMPA component and based on the decay of the AMPA current, the time window for the extraction of the NMDA current was set to 50 – 60 ms after stimulation. The average current obtained at +40 mV during those 10 ms was quantified as the NMDA component. The AMPA subunit composition was determined in the presence of gabazine and D-APV in the bath. For all voltage clamp recordings at least 15 sweeps, separated by time intervals of 10 s, were acquired at each holding potential.

The modulation of the spontaneous activity of ChINs and LTSIs was studied in a modified ACSF solution containing 5 mM KCl to enhance intrinsic firing rates. Spikes were recorded in cell-attached mode with ACSF in the recording pipette to avoid any interference with intrinsic spontaneous activity. Other striatal neurons were recorded in parallel in whole-cell current clamp mode with the current clamp intracellular solution described above.

SR-95531 (gabazine, 10 μM, Sigma-Aldrich), NBQX (10 μM), D-APV (50 μM), TTX (0.5 μM), 4-AP (100 μM, Tocris) were bath applied to block excitatory and inhibitory transmission and to test if optogenetically evoked responses are monosynaptic. All drugs were washed in for at least 5 min in the slice chamber before acquiring recordings.

Slice fixation

Following recordings some slices were incubated for 12h at 4°C in 4% paraformaldehyde solution containing 14% picric acid in 0.01 M PBS for fixation. Cells filled with neurobiotin were visualized by washing slices and transferring them to 0.01 M PBS containing 0.3% Triton X-100, 1% BSA, 0.1% Na-Deoxycholate, and Cy5-conjugated streptavidin antibody (1:1500, Jackson ImmunoResearch Laboratories) for at least 6 h. Confocal images of the virus injection site, projections, and filled cells were acquired.

Quantification and Statistical Analysis

All statistical analyses were performed in Prism and are reported together with the specific statistical test as well as the number of recorded cells (n) and mice (N) in the figure legends. Paired recordings were analyzed with a two-tailed paired t test if data were normally distributed or with a Wilcoxon signed-rank test in case data were not normally distributed. Statistical significance is defined as ∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001, and ∗∗∗∗ p < 0.0001, if not noted otherwise.

Data and Code Availability

This study did not generate any code. All acquired data are available upon request to G.S. (gilad.silberberg@ki.se).

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Cy5-conjugated streptavidin	Jackson ImmunoResearch	RRID: AB_2337245
Bacterial and Virus Strains
pAAV2-CaMKIIa-hChR2(H134R)-EYFP	Lee et al., 2010	Addgene AA2; 26969-AAV2
pAAV5-Syn-Chronos-GFP	Klapoetke et al., 2014	Addgene AAV5; 59170-AAV5
Chemicals, Peptides, and Recombinant Proteins
D-APV	Tocris	#0106; CAS: 79055-68-8
NBQX disodium salt	Tocris	#1044; CAS: 479347-86-9
Tetrodotoxin citrate (TTX)	Tocris	#1069; CAS: 18660-81-6
4-Aminopyridine (4-AP)	Tocris	#0940; CAS: 504-24-5
SR-95531 (GBZ)	Sigma-Aldrich	#S106; CAS: 104104-50-9
Experimental Models: Organisms/Strains
Mouse, D1-Cre, EY262	MMRRC	#030779-UCD; RRID: MMRRC_017264-UCD
Mouse, D2-Cre, ER44	MMRRC	#017263-UCD; RRID: MMRRC_017263-UCD
Mouse, SOM-Cre	The Jackson Laboratory	#018973; RRID: IMSR_JAX:018973
Mouse, PV-Cre	The Jackson Laboratory	#017320; RRID: IMSR_JAX:017320
Mouse, ChAT-Cre	The Jackson Laboratory	#006410; RRID: IMSR_JAX:006410
Mouse, Ai9 (RCL-tdT)	The Jackson Laboratory	#007909; RRID: IMSR_JAX:007909
Lhx6-EGFP	MMRRC	#000246-MU; RRID: MMRRC_000246-MU
Mouse, C57BL/6J	The Jackson Laboratory	#000664; RRID: IMSR_JAX:000664
Mouse, RjOrl:SWISS (Swiss Webster)	Janvier	RjOrl:SWISS
Software and Algorithms
Data acquisition and analysis: Igor Pro 6.37	Wavemetrics	https://www.wavemetrics.com
Statistics: Graphpad Prism	Graphpad Software	https://www.graphpad.com