Early molecular layer interneuron hyperactivity triggers Purkinje neuron degeneration in SCA1

Abstract

Toxic proteinaceous deposits and alterations in excitability and activity levels characterize vulnerable neuronal populations in neurodegenerative diseases. Using in vivo two-photon imaging in behaving spinocerebellar ataxia type 1 (Sca1) mice, wherein Purkinje neurons (PNs) degenerate, we identify an inhibitory circuit element (molecular layer interneurons [MLINs]) that becomes prematurely hyperexcitable, compromising sensorimotor signals in the cerebellum at early stages. Mutant MLINs express abnormally elevated parvalbumin, harbor high excitatory-to-inhibitory synaptic density, and display more numerous synaptic connections on PNs, indicating an excitation/inhibition imbalance. Chemogenetic inhibition of hyperexcitable MLINs normalizes parvalbumin expression and restores calcium signaling in Sca1 PNs. Chronic inhibition of mutant MLINs delayed PN degeneration, reduced pathology, and ameliorated motor deficits in Sca1 mice. Conserved proteomic signature of Sca1 MLINs, shared with human SCA1 interneurons, involved the higher expression of FRRS1L, implicated in AMPA receptor trafficking. We thus propose that circuit-level deficits upstream of PNs are one of the main disease triggers in SCA1.

Keywords: GABAergic neurons; Purkinje neurons; cerebellar circuit; chemogenetics; circuit modulation; excitation/inhibition; iPSCs; in vivo imaging; molecular layer interneurons, spinocerebellar ataxia type 1; mouse models; neurodegenerative disease.

Figure 1

Simultaneous monitoring of diverse cerebellar neuronal subtypes in behaving mice (A) Experimental design (fb, feedback; dark, darkness; iso, isoflurane anesthesia). (B) Scheme, depicting sites of adeno-associated virus (AAV)-injection and imaging and the cerebellar cortex layers. (C) Average projection of a quiet wakefulness (QW, left) episode and locomotion (loco, right) of the same field of view (FOV) in WT and Sca1 mice (dashed lines-cerebellar layers). (D) Calcium traces of molecular layer interneurons (MLINs, magenta), Gol (Gol, yellow), and Purkinje neurons (PNs, gray) during fb and darkness in WT (upper) and Sca1 mice (lower). Gray areas-locomotion epochs. (E–J) Baseline fluorescence and spontaneous neuronal activity (area under the curve (AUC)/min) during QW. (E) MLIN baseline fluorescence, (F) MLIN neuronal activity (WT 1,386, Sca1 714), (G) Gol baseline fluorescence and (H) neuronal activity (WT 114, Sca1 48), and (I) PN baseline fluorescence and (J) spontaneous neuronal activity (WT 112, Sca1 132). (K) Heatmap depicting average neuronal activity in response to running onset for all MLINs in WT (left, blue frame) and Sca1 mice (right, red frame). (L and M) (L) Same as in (K) for Gol and (M) for PN. (N–S) (N) Average population activity as function of instantaneous running velocity in MLINs (superimposed by sigmoidal fit), (O) fraction running-responsive MLINs per FOV (WT: 16 FOV [8 mice], Sca1: 9 FOV [6 mice]), (P) as in (N) for Gol, (Q) as in (O) for Gol (WT: 16 FOV [8 mice], Sca1: 6 FOV [4 mice]), (R) as in (N) for PN, (S) as in (O) for PN (WT: 14 FOV [8 mice], Sca1: 9 FOV [6 mice]). Data in (N), (P), and (R) mean ± SEM. Data presented as box plots throughout the manuscript display the median and interquartile range and whiskers extend to the minimum and maximum, unless differently stated. Scale bars, (B) 1 mm, (C) 100 μm, ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. (ML, molecular layer; PNL, Purkinje neuron layer; GCL, granule cell layer). Statistical test description and F/t/R values are in STAR Methods. See also Figures S1–S4.

Figure 2

Compromised coding space of behavioral parameters in cerebellar cortex of Sca1 mice (A) Exemplary input matrix consisting of 0-centered and Z scored ΔF/F traces used for dimensionality reduction derived from all cells in one FOV of a feedback session of a WT mouse aligned with behavioral parameters (lower traces, magenta, MLIN; gray, PN; yellow, Gol). (B) PC1, PC2, and PC3 time-series derived from input matrix shown in (A). (C) Variance explained of the neuronal activity as a function of number of principal components in WT (blue, thin lines, individual FOV; thick line, mean) and Sca1 mice (red). PC1 explains ∼50% of the variance in both WT and Sca1 (WT 51.9% ± 8.6%, 16 FOV, 8 mice; Sca1 54.63% ± 11.6%, 14 FOV, 7 mice). (D) Variance of PC1 (derived from all cells in a FOV or individual cell types), explained by the respective behavioral parameters in WT and Sca1. (E) Representative manifold of QW and active states (as function of locomotion speed, color-coded) for WT and Sca1 mice. (F) Euclidean distance in PC space within and across brain states (WT: 12 FOV, 8 mice, Sca1: 10 FOV, 5 mice). Data in (D) and (E) are median + interquartile range. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. See also Figures S5–S7.

Figure 3

Altered inhibitory connectivity in the adult Sca1 cerebellar cortex (A) Representative confocal images (R.I.) of MLINs stained for parvalbumin (PV) from P30, WT, and Sca1 cerebellum and quantitative analysis (Q.A.) of the percentage of neurons expressing low, medium, and high levels of PV at P30 (presymptomatic), P90 (symptomatic), and P200 (end-stage). Chi-squared test. (B) No difference in MLIN numbers between WT and Sca1 at different ages. (C) R.I. of MLIN stained for PV, VGAT (inhibitory), or VGluT1 (excitatory) synaptic markers. (D) R.I. of PNs stained for Calbindin and VGAT illustrating increased density of inhibitory synapses onto Sca1 PNs soma. (E) Representative 3D-SBF-SEM reconstructions of WT and Sca1 PNs (blue), MLINs (green), inhibitory axosomatic synapses (red), excitatory synapses (magenta). Right: 2D-SBF-SEM image: inhibitory axosomatic synapse (magenta) and PN soma (cyan), showing elevated inhibitory axosomatic inputs onto mutant PNs. No. of animals: 3-6/genotype/age. Graphs (B–D) depict mean values/animal, (E) depicts mean value per PN. Scale bars, (A) 20, (C) 2, (D) 10, (E) 2 and 0.5 (μm). ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. See also Figure S8.

Figure 4

Acute MLIN activity reduction influences cerebellar network responses and alleviates motor deficits (A) Experimental design showing acute MLIN inhibition in WT::PV and SCA1::PV mice via AAV2/8-DREADD(Gi). Right: R.I. of mCherry expression within WT::PV MLINs (yellow arrowheads), note the absence of mCherry staining in PNs (blue arrowheads). (B) R.I. of comparable PV expression in WT::PV and Sca1::PV-expressing DREADD(Gi) MLINs after saline/CNO injection, n = 3 mice/group. (C) Rotarod measurement reveals improved motor performance of Sca1::PV-DREADD(Gi)+CNO mice vs. saline group. WT::PV+saline n = 9; WT::PV+CNO n = 8; Sca1::PV+saline n = 8; Sca1::PV+CNO n = 10 mice. (D) In vivo imaging experiment design. Same neurons were recorded (fb, iso) at baseline (BL, w/o CNO, P110) and upon CNO application. (E) Fraction of neurons with increased, decreased, or no change in activity in Sca1::PV MLIN+CNO. (F) Absolute activity changes in MLINs upon CNO. (G) Maximum intensity projection of a Sca1::PV mouse before and after CNO. (H–K) (H) MLIN activity under anesthesia before and after CNO in WT::PV (left, 1,088 MLINs) and in Sca1::PV (right, 548 MLINs), (I) same as in (H) for Gol in WT::PV (78 Gol) and Sca1::PV (38 Gol), (J) and in PN in WT::PV (102 PN) and Sca1::PV (83 PN) and (K) calcium signals in PN dendrites in WT::PV (83 PN dendrites) and Sca1::PV (88 PN dendrites), data in (H)–(K) from WT::PV: 12 FOV, and Sca1::PV: 8 FOV, 5 mice/genotype). (L–O) (L) QW spontaneous activity of MLINs before and after CNO in WT::PV (1,316 MLINs) and Sca1::PV (813 MLINs), (M) of Gol in WT::PV (82 Gol) and Sca1::PV (56 Gol), (N) of PN in WT::PV (49 PN) and Sca1::PV (76 PN) and (O) calcium signals in PN dendrites in WT::PV (12 PN dendrites) and Sca1::PV (19 PN dendrites), data from WT::PV: 17 FOV, and Sca1::PV: 17 FOV, 7 mice/genotype). (P) Example manifold of all neurons in a FOV from a Sca1::PV mouse, depicting less discrete segregation of QW and active states (running speed color-coded), which is improved after CNO application (right). (Q) Euclidean distance in PC-space between the center of mass of QW and locomotion (before CNO: WT::PV: 10 FOV, Sca1::PV: 12 FOV, after CNO: WT::PV: 9 FOV, Sca1::PV 11 FOV, (5–6 mice/genotype). Graph (C) represents mean ± SEM, data in (H–O) and (Q) represent median and interquartile range. Scale bars, (A) 100, (B) 20, (G) 50 (μm). ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p<0.001. See also Figures S9–S11.

Figure 5

Mimicking mutant MLIN hyperresponsiveness in WT::PV mice (A) Experimental design for acute cerebellar MLIN activation in WT::PV mice. (B) R.I. of mCherry expression in WT::PV, lobule VIII. (C) Mimicking mutant MLIN hyperresponsiveness in WT::PV MLINs by CNO stimulation of excitatory DREADD (Gq), increases PV levels. (D) WT::PV animals display impaired motor performance on rotarod vs. saline group after CNO injection WT::PV+saline n = 7, WT::PV+CNO n = 10 mice). (E) Experimental design for chronic MLIN activation in WT::PV animals. (F) R.I. and Q.A. of reduced Homer-3 expression in WT::PV-mCherry and WT::PV-DREADD(Gq) mice after 30 or 60 days of CNO treatment. (G) Q.A. of reduced P-CAMKII expression in WT::PV-DREADD(Gq) and WT::PV-mCherry PNs, no change in total CAMKII. (H) Experimental design for chronic WT::PV MLIN activation, analyzed at P200 (bold arrow: longitudinal rotarod measurement throughout). (I) Chronic CNO administration impairs motor performance in WT::PV-DREADD(Gq) mice, n = 8 animals/group. (J) R.I. and Q.A showing reduced Homer-3 expression in WT::PV-DREADD(Gq) vs. WT::PV-mCherry. (K) Q.A. of P-CAMKII and CAMKII expression after chronic activation of MLINs in WT::PV mice. n = 3–5 mice/analysis. Graphs (D and I) represent mean ± SEM, (F, G, J, and K) depict mean values/animal. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. Scale bars, (B) 200, (C) 10, (F) 10, (J) 10 (μm).

Figure 6

Chronic suppression of MLIN activity delays SCA1 pathology (A) Experimental design for chronic MLIN inhibition. (B) R.I. of PV expression within WT::PV and Sca1::PV MLINs. Chronic inhibition of Sca1::PV MLINs via DREADD(Gi), normalizes PV expression to WT levels. (C) R.I. and Q.A. of restored P-CAMKII expression in PNs. (D) 3D-isosurface reconstructions showing elevated Homer-3 and Calbindin expression within the molecular layer in Sca1::PV-DREADD(Gi) mice. (E) R.I. of Calbindin labeled PNs from four conditions and Q.A. of sustained PN numbers in Sca1::PV-DREADD(Gi) vs. Sca1::PV-mCherry group. (F) CNO chronic administration delays appearance of motor symptoms in Sca1::PV-DREADD(Gi) mice, WT::PV-mCherry n = 13; WT::PV-DREADD(Gi) n = 11; Sca1::PV-mCherry n = 13; Sca1::PV-DREADD(Gi) n = 10 mice. (G) Clasping phenotype across disease progression in Sca1::PV-mCherry and Sca1::PV-DREADD(Gi), showing reduced clasping severity in DREADD(Gi) cohort. Graphs (C–E) depict mean values/animal, (F) represents mean ± SEM. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. Scale bars, (B) 20, (C and E) 20, (D) 4 (μm). See also Figure S12.

Figure 7

Frrs1l likely determines MLIN hyperresponsiveness (A) Experimental design for isolating mCherry labeled MLINs from adult WT::PV and Sca1::PV cerebellum (n = 9 mice/genotype). (B) MLIN-selective markers identified via mass spectrometry (MS) from one experiment (3 cerebella/genotype). (C) Downregulated (iLFQ Log2FC 1.4) proteins in Sca1::PV MLINs. (D) PANTHER overrepresentation test: downregulated protein classes in Sca1::PV MLINs. (E) Upregulated (iLFQ Log2FC 1.4) proteins in Sca1::PV MLINs. (F) PANTHER overrepresentation test: upregulated protein classes in Sca1::PV MLINs. (G) MS-based iLFQ values of Frrs1l. R.I. of increased Frrs1l and PV expression in MLINs from WT::PV and Sca1::PV, and Pearson's correlation. (H) Q.A. of anterior (lobules II–V) and posterior (lobules VI–X) cerebellum for Frrs1l expression in MLINs. (I) R.I. of GluR2 and Q.A. displaying increased GluR2 expression in Sca1::PV MLINs vs. WT::PV. n = 3–5 mice for analysis. Graphs (B and G) represent mean ± SEM, (I) depicts mean values/animal, ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. Scale bars, (G) 20, (I) 5 (μm). See also Figure S13 and Table S1.

Figure 8

iGNs activity in Control and SCA1 correlates with PV and FRRS1L expression (A) R.I. of FRRS1L, GluR2 and PV expression in 3 human SCA1 patients' iPSC-derived iGNs vs. 3 CNTRL lines and Q.A. of elevated. GluR2 expression. (B) Q.A. of PV expression binned into expression classes: high, medium, and low (chi-squared test). Pearson’s correlation between PV expression and FRRS1L in SCA1 vs. CNTRL. Q.A. of FRRS1L expression level. (C) Experimental timeline for iGNs calcium imaging and Q.A. of differentially responding populations after 50 mM KCl stimulation (low-responding cut off: 1–1.5 at 70 s; high-responding < 1.5 at 70 s) and corresponding heatmaps. Q.A. of calcium transients showing increase in ΔF at 70 s in SCA1 high-responding neurons vs. CNTRL. (D) Significant increase in ΔF amplitude at 70 s in SCA1 high-responding neurons. Time to max ΔF was significantly higher in high-responding SCA1 populations. n = 27–33 neurons. (E) Timeline for calcium imaging of iGNs transduced at DIV4 with LV:FRRS1L-myc (transduction efficiency, mean: 83.06) or LV:shFRRS1L-GFP (transduction efficiency mean: 89.60) or LV-GFP. Subsequently at DIV7 iGNs were infected with AAV2/1-GCaMP6 or AAV2/1-RCaMP. R.I. of FRRS1L overexpression or knockdown in iGNs. (F) Q.A. of calcium transients in CNTRL lines overexpressing FRRS1L, showing increase in ΔF after KCl stimulation, n = 12–15 neurons/condition. (G) Q.A. of calcium transients in SCA1 lines after FRRS1L knockdown, displaying decline in ΔF after KCl stimulation. n = 12–14 neurons/condition. Box and whisker plot in (A and B) and graphs in (C and F) show mean ± SEM, (E) percentage of mean/coverslip. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001. Scale bars, (A, B, and D) 15 μm. See also Figures S14 and S15.