[1] 12w Cholinergic modulation of Kir2 channels selectively elevates dendritic excitability in striatopallidal neurons

[1] 153w Dopamine-depleting lesions of the striatum that mimic Parkinson's disease induce a profound pruning of spines and glutamatergic synapses in striatopallidal medium spiny neurons, leaving striatonigral medium spiny neurons intact. The mechanisms that underlie this cell type-specific loss of connectivity are poorly understood. The Kir2 K 1 channel is an important determinant of dendritic excitability in these cells. Here we show that opening of these channels is potently reduced by signaling through M1 muscarinic receptors in striatopallidal neurons, but not in striatonigral neurons. This asymmetry could be attributed to differences in the subunit composition of Kir2 channels. Dopamine depletion alters the subunit composition further, rendering Kir2 channels in striatopallidal neurons even more susceptible to modulation. Reduced opening of Kir2 channels enhances dendritic excitability and synaptic integration. This cell type-specific enhancement of dendritic excitability is an essential trigger for synaptic pruning after dopamine depletion, as pruning was prevented by genetic deletion of M1 muscarinic receptors.

[1] 121w As described previously 7,10 , MSNs in tissue slices had a hyperpolarized resting potential (-87.4 ± 1.7 mV, n ¼ 32), strong inward rectification, and a long first-spike latency with near rheobase current injection (Fig. 1a). To study Kir2 channel gating, we held the membrane potential at a relatively depolarized potential (-60 mV) to largely inactivate Kv1 and Kv4 channels and then stepped it down to more hyperpolarized potentials. Kir2 channels opened rapidly and did not inactivate with maintained hyperpolarization (Fig. 1b). A plot of current amplitude as a function of voltage could be readily fit with a sum of two functions: a small, linear, K + -selective component and a larger, strongly rectifying, K + -selective Kir2 component (Fig. 1c).

[2] 175w Activation of M1 muscarinic receptors reduces the inwardly rectifying K + current in MSNs 17 . When perforated patch recording is used to retain the fidelity of the M1 muscarinic signaling pathway 18 , this modulation is strong in some MSNs, leading to a reduction of roughly half of the Kir2 channel currents (Fig. 1d,e). Bath application of Ba 2+ (200 mM) blocked the remaining portion of the Kir2 channel current, leaving the linear currents attributable to striatally expressed KCNK channels 19 . In agreement with inferences drawn from previous work, the selective M1 muscarinic receptor antagonist muscarinic toxin 7 completely eliminated the effects of muscarine 17 (Supplementary Fig. 1 online). To further determine whether the M1 receptor modulation was selective for Kir2 channels, we determined the current-voltage relationship of the modulated channels using voltage ramps. Subtraction of the records before and after muscarine application (10 mM) revealed a strongly inwardly rectifying, K + -selective current (Fig. 1e), indicating that Kir2 channels were the principal, if not sole, target of the M1 receptor signaling cascade.

[3] 247w In the course of these initial experiments, it was apparent that the M1 receptor modulation of Kir2 channels varied significantly between MSNs. Muscarine profoundly reduced Kir2 channel currents in some MSNs (for example, Fig. 1d) and left the current virtually unaffected in others. Although M1 muscarinic receptors are robustly expressed by both striatonigral and striatopallidal MSNs 9 , it was possible that the variation was correlated with projection phenotype. To test this hypothesis, we studied MSNs from transgenic mice in which one or the other of these populations was labeled with enhanced green fluorescent protein (eGFP) 16 (Fig. 2a). Although the apparent density and biophysical properties of the Kir2 channels in these two types of MSN were indistinguishable (Supplementary Table 1), the response to application of muscarine was markedly different (Fig. 2b). In striatopallidal MSNs, muscarine (at nominally saturating concentrations, 10 mM; Supplementary Fig. 1) typically reduced Kir2 channel currents by over 40%, whereas the median reduction was less than 10% in striatonigral neurons (Fig. 2c). The cellular specificity of the Kir2 modulation did not extend to KCNQ channels, as the same concentration of muscarine was equally effective at modulating these channels in striatonigral and striatopallidal neurons (Fig. 2d,e). The similarity in KCNQ modulation indicates that the asymmetry in the M1 receptor modulation is not likely to arise from differences in levels of receptor expression, G-protein coupling or regulation of coupling by 'regulators of G-protein signaling' (RGS) proteins, as these should affect the modulation of both channels.

[4] 127w The opening of both KCNQ and Kir2 channels depends on the membrane lipid phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P 2 ). G q -coupled receptors, such as M2 muscarinic receptors, reduce channel opening by activating phospholipase C (PLC) isoforms and lowering membrane PtdIns(4,5)P 2 levels 7,20,21 . A hallmark of this signaling mechanism is a slowing of recovery from modulation by inhibition of the enzyme that mediates PtdIns(4,5)P 2 replenishment, phosphatidylinositol 4 kinase (PI(4)K) 20 . As expected of a PtdIns(4,5)P 2 -dependent mechanism, inhibition of PI(4)K with wortmannin (50 mM) prevented the rapid recovery of Kir2 channel currents following removal of muscarine from the bathing medium (control median ¼ 91%, n ¼ 5; wortmannin median ¼ 25%, n ¼ 5; P o 0.01, Mann-Whitney rank sum test; Fig. 3a,b).

[5] 311w Kir2 channels are multimeric membrane proteins constructed from a family of at least four subunits (Kir2.1-2.4) 22,23 . The affinity of these subunits for PtdIns(4,5)P 2 varies: Kv2.1 subunits bind PtdIns(4,5)P 2 tightly, whereas Kv2.3 subunits have a significantly lower affinity 20,24 . The functional consequence of this variation is that channels with Kir2.3 subunits are much more potently modulated by receptors coupled to PLC than are channels composed of Kir2.1 subunits 20,25 . Could variation in the expression of Kir2.3 subunits explain the differences between striatopallidal and striatonigral MSNs? In agreement with previous in situ hybridization studies 26 , we found transcripts for all four Kir2 subunits when we analyzed striatal messenger RNA using PCR with reverse trancriptase (RT-PCR) (data not shown). Profiling of striatopallidal and striatonigral MSNs using qualitative single cell RT-PCR (scRT-PCR) also consistently detected mRNA for Kir2.1, 2.2 and 2.3 subunits but rarely detected mRNA for Kir2.4 subunits (Supplementary Fig. 2 online). To determine whether there were quantitative differences in Kir2.3 expression between MSN populations, we used a serial dilution scRT-PCR strategy 27 . This approach allows the relative abundance of a particular mRNA to be compared between two or more neuronal populations without having to make assumptions about the efficiency of reverse transcription or amplification. The strategy is to find the smallest fraction of the total cellular complementary DNA that gives a threshold PCR signal. The distribution of detection thresholds in a sample can typically be fit with a Gaussian function and compared to that of another cell type. The modal Kir2.3 detection threshold in striatonigral MSNs was roughly one-eighth (2 À3 ) of the total cDNA (Fig. 3c). The modal Kir2.3 detection threshold in striatopallidal MSNs was lower, less than one-sixteenth (2 À4 ) of the total cDNA (Fig. 3d). Thus, the Kir2.3 mRNA was roughly twofold more abundant in striatopallidal than in striatonigral MSNs.

[6] 355w In many types of neuron, Kir2 channel subunits are dendritically positioned 28,29 . Although it has been assumed that Kir2 channels are also primarily localized to the dendrites of MSNs, this has been difficult to demonstrate directly using immunocytochemical approaches 30 . In monolayer, primary cultures, dendritic proteins of MSNs can be visualized without the interference of nearby neurites; in these cells, Kir2.1 and Kir2.3 immunoreactivity was found in regions that were also immunoreactive for microtubule-associated protein 2 (MAP2), supporting the idea that these subunits are prominent in dendrites (Supplementary Fig. 3 online). Although these MSN cultures have optical advantages, they lack spines and glutamatergic synapses, which are normally derived from cortical neurons. To overcome this limitation, we generated corticostriatal cultures using striatal tissue from BAC (bacterial artificial chromosome) D2 mice and cortical tissue from wild-type mice 31 ; in these cultures, striatopallidal MSNs form mature spines and synapses and proteins can be readily localized using immunocytochemical approaches. Analysis of these cultures revealed that Kir2.1 subunits in striatopallidal MSNs were largely restricted to puncta along dendritic shafts (Fig. 4a,b). By contrast, Kir2.3 subunits were found primarily in spines (Fig. 4a,b), as seen previously in other forebrain neurons 28 . Scaffolding interactions, like those that presumably target Kir2.3 channels to spines 28 , are thought to be crucial for modulation of ion channels by G-protein-coupled receptors (GPCRs), bringing receptors and signaling enzymes into close physical proximity to substrates 32 , Although the scaffolding interactions of M1 muscarinic receptors have not been characterized, these receptors are found in spines of MSNs 8 , making it possible that they form a signaling complex with Kir2.3 channels. If this is the case, disruption of the carboxy-terminal PDZ (postsynaptic density 95-discs large-zona occludens)-domain interaction of Kir2.3 channels should attenuate their modulation by M1 receptor signaling. To test this hypothesis, we incubated tissue slices for 30 min with polyarginine peptides containing the ESRI interaction motif of Kir2.3 subunits. Incubation with the ESRI peptide virtually eliminated the M1 receptor modulation of Kir2 channel currents, whereas incubation with a similar peptide lacking the PDZ recognition motif failed to alter the modulation (Fig. 4c,d).

[7] 284w The positioning of Kir2.3 channels indicates that they should hold the dendritic membrane near the K + equilibrium potential and dampen responsiveness to glutamatergic excitatory synaptic transmission (excitatory postsynaptic potentials, EPSPs) 10 . Closure of these channels by M1 muscarinic receptor signaling should enhance responsiveness, particularly the integration of EPSPs. To test this hypothesis, we evoked EPSP trains by minimal local stimulation of glutamatergic afferent fibers at 20 Hz (Fig. 4e). AMPA receptor-mediated EPSPs were isolated by bath application of antagonists of NMDA receptors, metabotropic glutamate receptors (mGluR 1-5 ), and GABA A and GABA B receptors. The evoked EPSPs summed sub-linearly, with the ratio of the fifth to the first EPSP (EPSP5/EPSP1) falling typically between 1 and 2. Bath application of muscarine (10 mM) reduced the amplitude of the first EPSP in the stimulus train and significantly increased the EPSP5/ EPSP1 ratio (Supplementary Fig. 4a online). The reduction in the amplitude of the first EPSP was attributable to M2 receptor-mediated presynaptic inhibition of glutamate release 33 , whereas the enhanced summation of EPSPs was attributable in large measure to the postsynaptic downregulation of Kir2 channel opening. To isolate the postsynaptic component of the effect, we repeated the experiments in the presence of the M2 muscarinic receptor antagonist methoctramine. As expected in this situation, muscarine increased EPSP amplitude and summation in striatopallidal MSNs, although the magnitude of the effect on summation was reduced. The median EPSP5/EPSP1 ratio increased from 1.65 to 2.25 in the presence of muscarine and methoctramine (Fig. 4e) and to greater than 3 in the presence of muscarine alone (Supplementary Fig. 4a). By contrast, muscarine did not have a significant effect on EPSP amplitude or summation in striatonigral MSNs (Fig. 4e).

[8] 148w To provide an additional test of the hypothesis that the effects of muscarine were mediated by postsynaptic M1 receptors, we applied brief, iontophoretic puffs of glutamate to a region of the dendrite 50-70 mm from the soma and re-examined the effects of muscarine. We used tetrodotoxin (TTX, 1 mM) and the same mixture of ionotropic and metabotropic receptor antagonists to isolate the recorded neuron from surrounding cells. These glutamate puffs evoked EPSP-like depolarizations mediated by AMPA receptors in both striatopallidal and striatonigral MSNs (Fig. 4f). However, as with synaptic stimulation, muscarine selectively enhanced the summation of glutamate-evoked potentials in striatopallidal MSNs, confirming the results obtained with synaptic stimulation (Fig. 4f). Finally, a near complete block of Kir2 channels with Ba 2+ (200 mM) increased the summation of the simulated EPSPs in both cell types and occluded the effects of M1 receptor stimulation in striatopallidal MSNs (Supplementary Fig. 4b).

[9] 256w The cellular specificity of the modulation of Kir2 channels by M1 receptors indicates that this modulation could be a factor in triggering an elevation in dendritic excitability, Cav1.3 Ca 2+ channel opening and spine loss following DA depletion. Certainly, augmented ACh release would be expected selectively to increase basal excitability and synaptic summation in striatopallidal MSNs. However, as outlined above, there is not a consensus on whether ACh release increases in models of Parkinson's disease. Another possibility is that DA depletion increases the functional impact of ACh, without altering its release. One way in which this might happen is by increasing the relative proportion of Kir2 channels that have readily modulated Kir2.3 subunits. As first step toward testing this hypothesis, we used real-time PCR to estimate the abundance of striatal Kir2.1 and Kir2.3 mRNA before and after DA depletion. Treatment with reserpine for 5 d (which is known to produce spine pruning in striatopallidal MSNs) decreased Kir2.1 mRNA abundance by 40% and increased Kir2.3 mRNA abundance by 26% (Fig. 5a,b). Thus, the ratio of Kir2.3/Kir2.1 mRNA increased by more than 200%. To determine whether this change in the relative abundance of Kir2.3 mRNA was translated into an increase in the ability of M1 receptors to modulate channels, we recorded from tissue slices from DA-depleted mice. The magnitude of the modulation in striatonigral MSNs was unchanged after depletion (Fig. 5c). However, depletion significantly increased the potency of the M1 receptor modulation of Kir2 channel currents in striatopallidal MSNs, elevating the median modulation to nearly 60% (Fig. 5d).

[10] 207w If the modulation of Kir2 channels by M1 receptors is a significant factor in the elevation of dendritic excitability and synaptic pruning after DA depletion, then pruning should be attenuated by blocking or deleting M1 receptors. To test this hypothesis, we treated M1 knockout mice in which striatopallidal neurons expressed eGFP with reserpine for 5 d. This regimen produces a profound (B50%) and cell type-specific loss of spines and glutamatergic synapses in wild-type striatopallidal MSNs 3 . Spine density in striatopallidal MSNs from M1 knockout mice was indistinguishable from that in wild-type mice (Fig. 6a,b). More importantly, reserpine treatment had no effect on spine density in these neurons (Fig. 6a,b). To test the role of M1 receptors in a more widely accepted model of Parkinson's disease, DA neurons were lesioned by unilaterally injecting 6-hydroxydopamine (6-OHDA) into the medial forebrain bundle 34 . Neurons from mice that had a profound asymmetry in the use of their forelimbs 1 week after 6-OHDA injection 35 were then examined using two-photon laser scanning microscopy. As expected 3 , 6-OHDA lesions produced a marked reduction in spine density in striatopallidal MSNs (Fig. 6c,d). By contrast, 6-OHDA lesions had no effect on spine density in striatopallidal MSNs from M1 knockout mice (Fig. 6c,d).

[1] 145w Our results show that M1 muscarinic receptor activation potently downregulates Kir2 channel currents in striatopallidal MSNs, but not in neighboring striatonigral MSNs. As in other cell types, the downregulation was produced by membrane depletion of PtdIns(4,5)P 2 . The differential sensitivity of Kir2 channels was attributable to the increased reliance on readily modulated, PtdIns(4,5)P 2 -sensitive Kir2.3 subunits in striatopallidal MSNs, rather than a generalized alteration in M1 receptor signaling. The reduction in Kir2 channel opening elevated dendritic excitability and enhanced the temporal summation of glutamatergic EPSPs. DA depletion increased the reliance on Kir2.3 subunits, augmenting the potency of the modulation of excitability by M1 receptors. Genetic deletion of M1 receptors prevented pruning of spines after DA depletion, indicating that cholinergic signaling is an essential trigger for the pruning of glutamatergic synapses in models of Parkinson's disease and in patients with Parkinson's disease 36 .

[1] 147w Slice preparation and recordings. We took parasagittal corticostriatal slices from 18-25-d-old BAC D1/BAC D2 eGFP mice or C57BL6 mice. All animals were handled in accord with procedures approved by the Northwestern University Animal Care and Use Committee. Experiments were performed at room temperature. Electrical access was achieved through the perforated-patch method using amphotericin B 7 . Extracellular stimulation (0.2 ms, 5-30 mA) was applied with a small theta glass electrode 150-200 mm from the recording electrode. EPSP-like depolarizations mediated by activation of AMPA receptors were evoked using short iontophoretic pulses of sodium glutamate (150 mM). AMPA receptor-mediated excitatory input was isolated by application of antagonists to NMDA (D-AP5, 50 mM) GABA A (SR 95531, 10 mM) and GABA B (CGP 55845, 10 mM) receptors. Blockers of M2 (methoctramine, 0.3 mM) and mGluR 1-5 (LY 341495, 100 mM) receptors were also added to prevent possible presynaptic receptor modulation.

[2] 181w Single cell RT-PCR. Dissociated, individual MSNs were collected for scRT-PCR analysis as described 7 . Kir2.1 mRNA (accession NM_008425) was detected with a pair of primers, 5¢-GCC CTT TAT ATG ACT TGA GTA (position 1167) and 5¢-GCT TGC CTG GTT GTG GAG (position 1562), which gave a PCR product of 413 bp. Kir2.2 mRNA (accession NM_010603) was detected with a pair of primers, 5¢-ACC ACA CAG GCT CGC AGT TCC (position 1441) and 5¢-AAA TCT CCG ACT CCC GTC T (position 1780), which gave a PCR product of 358 bp. Kir2.3 mRNA (accession NM_008427) was detected with a pair of primers, 5¢-CAC GTG CCC AGG CGG AAA C (position 124) and 5¢-CCA GAA RAG GAG GCC GAA RAA (position 307), which gave a PCR product of 204 bp. Kir2.4 mRNA (accession NM_145963) was detected with a pair of primers, 5¢-GCT GGA TTC CTG AGG CTT AC (position 1689) and 5¢-GGA GAA GCC AGA CCT ACC TAC (position 2138), which gave a PCR product of 470 bp. Semiquantitative estimates of Kir2.3 transcript abundance were made using scRT-PCR serial dilution techniques 27 .

[3] 24w Real-time RT-PCR. Quantitative analysis of Kir2.1 and Kir2.3 cDNA abundance was performed on a CHROMO4 Continuous Fluorescence Detector (Bio-Rad Laboratories), as described 12 .

[4] 167w Cell culture and immunocytochemistry. Corticostriatal cultures were generated as described 31 . On day 21, cultured cells were fixed and then incubated with goat anti-GFP antibody and rabbit anti-Kir2.1 antibody or anti-Kir2.3 antibody in blocking buffer overnight at 4 1C. After several washes in PBS, the cells were incubated with Alexa 488-conjugated donkey anti-goat antibody and Alexa 555-conjugated donkey anti-rabbit antibody for 1 h at room temperature. Images were acquired with a laser-scanning microscope. Puncta of Kir2.1 and Kir2.3 on spines and shafts were quantified with Metamorph software. Two-photon laser scanning microscopy. Whole-cell recordings were obtained from eGFP-labeled striatonigral or striatopallidal MSNs. Alexa 568 (50 mM) was added to an internal solution for visualization of the cell bodies, dendrites and spines. Maximum projection images of the soma and dendritic fields were acquired with 0.36 mm 2 pixels with 10 ms pixel dwell time. High-magnification maximum projection images of dendrite segments 50-100 mm from the soma were acquired with 0.08 mm 2 pixels with 10 ms dwell time.

[5] 92w DA depletion. DA was acutely depleted by administering reserpine (5 mg kg -1 ) intraperitoneally (i.p.) for 5 successive days 3,12 or unilaterally injecting 6-OHDA (2.5 mg ml -1 in 0.02% ascorbic acid) into the medial forebrain bundle. The reserpine-treated mice were sacrificed 2-3 h after the final injection. The 6-OHDA-lesioned mice were maintained for 6-7 d before experiments. The degree of damage to nigrostriatal DA neurons was assessed by a limb-use asymmetry test 35 . Tyrosine hydroxylase immunoreactivity in the striatum was also examined to correlate with behavioral changes 12 .

[6] 22w Statistical analysis. The level of significance was taken at a probability threshold of 0.05 as determined by non-parametric Wilcoxon or Mann-Whitney tests.

[7] 22w Chemicals and reagents. All chemicals and reagents were obtained from Sigma or Tocris, except muscarinic toxin 7 (Peptides International) and wortmannin (Calbiochem).

[8] 10w Note: Supplementary information is available on the Nature Neuroscience website.

[1] 150w Parkinson's disease is a widespread and disabling neurodegenerative disorder that results from the loss of the dopaminergic innervation of the striatum 1,2 . Dopamine (DA) depletion triggers a range of biochemical and structural changes, some of which are compensatory but others of which seem to be maladaptive. One of these apparently maladaptive changes is the extensive pruning of dendritic glutamatergic synapses and postsynaptic spines in striatopallidal medium spiny neurons (MSNs). These neurons are one of the two main neuronal populations in the striatum, and use the inhibitory neurotransmitter GABA to shape the activity of neurons in the so-called indirect pathway of the basal ganglia 1 . In animal models of Parkinson's disease and in patients with Parkinson's disease, neurons in this pathway (in the globus pallidus and subthalamic nucleus) show the clearest pathophysiological activity, indicating that maladaptative striatopallidal activity could be an important determinant in the emergence of motor symptoms.

[2] 92w The only clue as to the mechanism responsible for synaptic pruning comes from the observation that genetic deletion or pharmacological block of synaptically positioned Cav1.3 Ca 2+ channels prevents pruning 3 . Because these channels open only near spike threshold, far (20-30 mV) from the resting membrane potential of MSNs, DA depletion must increase the excitability of striatopallidal dendrites. These neurons express D2 dopaminergic receptors that diminish excitability, but they do so only when neurons are already depolarized, raising doubts about whether diminished D2 receptor activity is sufficient to trigger synaptic pruning.

[3] 160w Another potent modulator of MSN excitability is acetylcholine (ACh). Striatal ACh is released by giant, aspiny interneurons with dense terminal fields that overlap those of DA neurons 4 . ACh elevates MSN excitability by promoting the closure of KCNQ (Kv7), SK and Kir2 K + channels [5][6][7] . Most, if not all, of these effects can be attributed to M1 muscarinic receptors, which are robustly expressed by both types of MSN 8,9 . Although KCNQ and SK channels are active only near spike threshold, Kir2 K + channels are constitutively active, serving to set resting membrane potential and dendritic input resistance-a key factor in the control of synaptic integration 10 . Could modulation of Kir2 channels contribute to synaptic pruning? It is well established that DA depletion leads to an increase in intrastriatal cholinergic signaling 11 . Although the mechanisms that underlie this shift are controversial [12][13][14][15] , the functional impact of ACh undeniably rises as intrastriatal DA levels fall.

[4] 199w What is not clear is whether cholinergic signaling adaptations have anything to do with the structural changes that follow DA depletion. Our studies fill this gap, showing that cholinergic signaling is essential for spine pruning in models of Parkinson's disease . Using a combination of electrophysiological, optical and molecular approaches in transgenic mice harboring D1 and D2 receptor reporter constructs 16 , we have uncovered three unappreciated aspects of striatal cholinergic signaling. First, M1 muscarinic receptor activation potently reduced Kir2 K + channel currents in striatopallidal MSNs, but had little or no effect on channel currents in striatonigral MSNs. This difference could be attributed to elevated incorporation of readily modulated Kir2.3 subunits into the Kir2 channels of striatopallidal MSNs. As a result, M1 receptor downregulation of these channels enhanced the temporal summation of glutamatergic excitatory synaptic potentials almost exclusively in striatopallidal MSNs. Second, modulation of Kir2 channels by M1 receptors was enhanced by DA depletion as a consequence of the relative upregulation of Kir2.3 expression. Third, genetic deletion of M1 muscarinic receptors and elimination of the Kir2 channel modulation prevented the loss of spines following DA depletion, attesting to the key role of cholinergic signaling in the structural adaptation.

[5] 139w Strongly rectifying Kir2 channels are the main determinants of the resting membrane potential and basal excitability of striatal MSNs 37 . Neuronal Kir2 channels are multimeric transmembrane proteins constructed from a family of at least four subunits (Kir2.1-2.4) 22,23 . These subunits are expressed widely in the brain and in striatal MSNs 26,30,37 . Although primarily dependent on Kir2 channels, the K + channels that are active at the resting membrane potential are likely to include members of the KCNK class. Members of this class-KCNK2 (TREK-1) and KCNK10 (TREK-2)-that are expressed in the striatum give rise to a linear current-voltage relationship and are relatively resistant to Ba 2+ block 19,38 . The currents evoked in MSNs by voltage steps or ramps could readily be broken down into a large, strongly rectifying Kir2 component and a smaller, linear, KCNK-like component.

[6] 254w Signaling by M1 muscarinic receptors potently downregulated the flow of current through Kir2 channels in striatopallidal MSNs, but only weakly reduced currents in striatonigral MSNs. Although the M1 receptor-mediated modulation of Kir2 channels in MSNs has been described 6 , the magnitude of the modulation and its cellular specificity has not been appreciated. The susceptibility of Kir2 channels in striatopallidal neurons to M1 muscarinic receptor signaling did not seem to be due to differences in the transduction pathway. M1 receptors are robustly expressed by both striatopallidal and striatonigral MSNs 8,9 and their activation led to similar modulation of another PtdIns(4,5)P 2 -dependent channel type, KCNQ channels. Rather, it seemed that a difference in the molecular composition of the targeted channels was responsible. Although MSNs expressed readily detectable levels of mRNA for three Kir2 subunits (Kir2.1, Kir2.2 and Kir2.3), quantitative measures of mRNA abundance revealed that striatopallidal MSNs expressed roughly twice as much Kir2.3 mRNA as striatonigral MSNs. This was of functional importance because Kir2.3 subunits have a relatively low affinity for PtdIns(4,5)P 2 , making them more susceptible than other subunits to changes in local PtdIns(4,5)P 2 levels induced by GPCR-mediated activation of PLC 20,25 . Although these subunits can form heteromeric channels 39 , our immunocytochemical assays indicate that channels containing Kir2.1 and those containing Kir2.3 are not extensively co-localized. In accord with this observation, the potency of the modulation of these currents by M1 receptors (40-50% reduction) is close to what would be predicted from the relative abundance of Kir2.3 subunits alone.

[7] 209w In other neurons, Kir2 channels are enriched in dendritic regions 29 . Localization of Kir2 subunits has been difficult in MSNs because they have a dense, irregular striatal neuropil. Use of a corticostriatal coculture preparation 31 allowed us to overcome this obstacle and revealed not only that Kir2 channels were dendritic, but that localization was subunit specific. Kir2.1 subunits were found largely in patches along dendritic shafts, whereas Kir2.3 subunits were largely restricted to dendritic spines, upon which glutamatergic synapses are formed. Perisynaptic positioning of Kir2.3 subunits has also been seen in other forebrain neurons, where it is maintained by a PDZ domain interaction with membrane-associated guanylate kinase (MAGUK) family proteins 28 . Although this scaffolding interaction is not unique to Kir2.3 subunits (both Kir2.1 and Kir2.2 can bind MAGUK proteins 40 ), only these subunits were consistently found in MSN spine heads. M1 muscarinic receptors also are localized to spine heads in MSNs 8 through an undefined anchoring mechanism. Scaffolding was critical to the modulation of Kir2 channels by M1 receptors, as this modulation was significantly attenuated by disruption of a Kir2.3-PDZ motif mime (ESRI). Thus, bringing M1 receptors (and presumably PLC) into close physical proximity with Kir2.3 channels seems to create an effective signaling microdomain 32 .

[8] 140w The modulation of Kir2 channels by M1 receptors significantly enhanced the temporal summation of glutamatergic EPSPs. These EPSPs arise primarily from synapses formed on spine heads 4 . Although Kir2.3 channels were positioned near these spinous synapses, it isn't clear that this positioning was important for enhancing summation (in contrast to signaling). Several lines of study indicate that spines do not act as electrical compartments 41,42 . Nevertheless, diminished Kir2 channel opening should enhance not only the temporal summation of EPSPs but their spatial summation as well; in so doing, the M1 receptor modulation should promote the transition to depolarized 'up-states' and the opening of Cav1.3 Ca 2+ channels 43 . It also is possible that M1 receptor signaling further promotes up-state transitions by targeting other ion channels involved in dendritic electrogenesis, such as NMDA and Kv4 channels 44,45 .

[9] 288w In animal models of Parkinson's disease, striatopallidal MSNs-but not neighboring striatonigral MSNs-lose nearly half of their spines and glutamatergic synaptic connections within a week of initiating DA depletion through either reserpine treatment or 6-OHDA lesioning 3 . This adaptation mimics the spine loss and dendritic atrophy found in MSNs of patients with advanced Parkinson's disease 36 . The synaptic pruning depends on activation of depolarization-activated, Cav1.3 Ca 2+ channels that are anchored in spines near glutamatergic synapses. What increases Cav1. laser scanning microscope images of striatopallidal MSNs in 275-mm-thick corticostriatal slices from an untreated (left) or reserpine-treated (right) M1 knockout (KO) BAC D2 mouse. Neurons were visualized with Alexa Fluor 568 (50 mM) by filling through the patch pipettes. Top, maximum projection images of the soma and dendritic field; bottom, high-magnification projections of dendritic segments taken 50-100 mm from the soma. The M1 KO MSN has a normal morphology. (b) Spine density in DA-depleted striatopallidal MSNs was indistinguishable from that in untreated control MSNs (control median ¼ 9 spines per 10 mm, n ¼ 5; DA depletion median ¼ 9 spines per 10 mm, n ¼ 5; P 4 0.05, Mann-Whitney). The dashed median lines show expected change in wild-type animals after DA depletion 3 . (c) Striatopallidal MSNs from a wildtype D2 mouse (left) or M1 KO D2 mouse (right) after injection of 6-OHDA for 6-7 d. Top, maximum projection images; bottom, high-magnification projections of dendritic segments. (d) 6-OHDA dopamine depletion produces a decrease in spine density in wild-type striatopallidal MSNs but not in M1 KO striatopallidal MSNs (wild-type median ¼ 6 spines per 10 mm, n ¼ 6 cells; M1 KO median ¼ 9 spines per 10 mm, n ¼ 5 cells; P o 0.01, Mann-Whitney).

[10] 164w dopamine receptors in striatopallidal MSNs reduces the opening probability of Cav1.3 channels 43 , raising the possibility that the same mechanisms work in spines and that the loss of DA disinhibits channels. However, there also is compelling evidence that the dendritic excitability of striatopallidal MSNs increases after DA depletion, an effect that cannot be readily explained simply by disinhibition of Cav1.3 channels. For example, the frequency of miniature excitatory postsynaptic currents (mEPSCs) seems to increase in striatopallidal MSNs after DA depletion (in spite of the frank loss of synapses) when cells are dialyzed with K + salts that preserve Kir2 channel gating 3 . The most parsimonious interpretation of this observation is that striatopallidal MSNs become more excitable and electrotonically compact after DA depletion, something that could be readily achieved by downregulation of Kir2 channels. In agreement with this conclusion, recent in vivo work has shown that cortical stimulation evokes larger responses in striatopallidal MSNs (but not striatonigral MSNs) after DA-depleting lesions 46 .

[11] 245w What triggers this elevation in dendritic excitability? Perhaps the earliest striatal adaptation described in patients with Parkinson's disease was an elevation in striatal cholinergic 'tone' , inferred from the ability of anti-muscarinic drugs to ameliorate symptoms 11 . Although widely interpreted to reflect disinhibition of cholinergic interneurons as striatal DA levels fall, direct measurements of ACh release in models of Parkinson's disease have yielded inconsistent results [13][14][15] . More recent work has found that signaling by M2/4 muscarinic autoreceptors in interneurons is attenuated after DA depletion 12 ; although the inhibitory regulation of ACh release is diminished, it is not clear that this leads to elevated basal ACh. Our results support an alternative hypothesis-that the ability of ACh to modulate the activity of MSNs, particularly striatopallidal MSNs, increases after DA depletion. As shown here, M1 muscarinic receptor signaling downregulates Kir2 K + channels, the principal determinant of MSN dendritic excitability at rest, and does so much more potently in striatopallidal MSNs than in striatonigral MSNs. DA depletion increased the potency of this modulation, apparently by increasing the proportion of Kir2 channels that incorporated readily modulated Kir2.3 subunits. The modulation of Kir2 channels significantly enhanced the summation of glutamatergic EPSPs, providing a means by which the opening of dendritic Cav1.3 channels could be increased. Attesting to the importance of this modulation to the structural adaptation that follows DA depletion, deletion of M1 receptors prevented the loss of striatopallidal MSN spines in both reserpinetreated and 6-OHDA-lesioned animals.

[12] 150w One unresolved aspect of this model is why the direct inhibitory linkage between M1 muscarinic receptors and Cav1.3 channels does not prevent spine loss 47 . By activating a calcium/calmodulindependent protein phosphatase (calcineurin), M1 receptor signaling diminishes the opening of Cav1.3 channels in MSNs. This linkage has been implicated in attenuating the long-term depression of corticostriatal synapses and might normally prevent episodic elevation in dendritic excitability from downregulating synaptic strength 48 . This signaling cascade is disrupted by a calmodulin-binding regulatory phosphoprotein-RCS-when it is phosphorylated by protein kinase A 47 . Following DA depletion, upregulation of signaling by A2a adenosine receptors 49 could increase RCS phosphorylation, effectively cutting the inhibitory linkage between M1 receptors and Cav1.3 channels and promoting synaptic pruning. If this is the case, A2a receptor antagonists, which have been shown to be beneficial in treating Parkinson's disease 49 , should reduce spine loss induced by DA depletion.

[13] 82w In summary, our results indicate that elevated cholinergic signaling accompanying DA depletion triggers a form of homeostatic plasticity 50 in striatopallidal MSNs. By downregulating dendritic Kir2 K + channels, M1 muscarinic receptors increase dendritic excitability and the impact of synaptic glutamate release. This deviation from the neuronal set point is presumably signaled by increased Ca 2+ entry through dendritic L-type Ca 2+ channels, which triggers the elimination of spines and synapses with the aim of restoring dendritic activity to a desired range.