Altered Cerebellar Short-Term Plasticity but No Change in Postsynaptic AMPA-Type Glutamate Receptors in a Mouse Model of Juvenile Batten Disease

Abstract Juvenile Batten disease is the most common progressive neurodegenerative disorder of childhood. It is associated with mutations in the CLN3 gene, causing loss of function of CLN3 protein and degeneration of cerebellar and retinal neurons. It has been proposed that changes in granule cell AMPA-type glutamate receptors (AMPARs) contribute to the cerebellar dysfunction. In this study, we compared AMPAR properties and synaptic transmission in cerebellar granule cells from wild-type and Cln3 knock-out mice. In Cln3Δex1–6 cells, the amplitude of AMPA-evoked whole-cell currents was unchanged. Similarly, we found no change in the amplitude, kinetics, or rectification of synaptic currents evoked by individual quanta, or in their underlying single-channel conductance. We found no change in cerebellar expression of GluA2 or GluA4 protein. By contrast, we observed a reduced number of quantal events following mossy-fiber stimulation in Sr2+, altered short-term plasticity in conditions of reduced extracellular Ca2+, and reduced mossy fiber vesicle number. Thus, while our results suggest early presynaptic changes in the Cln3 Δex1–6 mouse model of juvenile Batten disease, they reveal no evidence for altered postsynaptic AMPARs.


Introduction
Batten disease is the collective term for a group of rare inherited neurodegenerative disorders, the neuronal ceroid lipofuscinoses (NCLs). These result from mutations in one of 14 ceroid-lipofuscinosis, neuronal type (CLN) genes (Cotman et al., 2013;Mole and Cotman, 2015;Nita et al., 2016), the majority of which encode soluble lysosomal enzymes or lysosome-associated transmembrane proteins (Cárcel-Trullols et al., 2015). The most common NCL is juvenile CLN3 disease or juvenile Batten disease (Williams and Mole, 2012). Children with this condition first exhibit symptoms at four to seven years of age, suffer loss of vision, seizures, progressive motor and cognitive decline, and die prematurely in late adolescence (Munroe et al., 1997;Haltia, 2003).
Juvenile Batten disease is caused by mutations in the CLN3 gene, commonly a 1-kb deletion encompassing exons 7 and 8, that result in the loss of full-length CLN3 protein (The International Batten Disease Consortium, 1995;Munroe et al., 1997;Kitzmüller et al., 2008). Like other NCLs, juvenile Batten disease is considered a lysosomal storage disorder and is characterized by the accumulation within lysosomes of autofluorescent lipopigments (lipofuscin-like ceroid; Seehafer and Pearce, 2006). Although the precise function of CLN3 remains unresolved, it has been implicated in multiple cellular phenomena, including endocytosis and endocytic trafficking, lysosmal pH regulation, autophagy, proliferation, cell-cycle control, and apoptosis (Cárcel-Trullols et al., 2015).
Excess influx of Ca 2ϩ through CP-AMPARs appears to be a feature common to several neurodegenerative disorders, including stroke, motor neuron disease, and hypoxic ischemic white matter damage (Follett et al., 2000;Kawahara and Kwak, 2005;Noh et al., 2005;Van Den Bosch et al., 2006;Corona and Tapia, 2007). Increased AMPAR-mediated excitotoxicity in Cln3 ⌬ex1-6 mice has been suggested to reflect altered AMPAR trafficking, an increase in CP-AMPAR number and enhanced AMPAR function (Kovács et al., 2006). However, recent experiments have described an increase in GluA2 protein in the cerebellum of Cln3 ⌬ex1-6 mice (Kovács et al., 2015), a change which is more usually associated with increased prevalence of CI-AMPAR subtypes.
Here, we have compared AMPAR properties and excitatory synaptic transmission in cerebellar granule cells from wild-type and Cln3 ⌬ex1-6 mice. Our results suggest that loss of CLN3 results in altered mossy-fiber presynaptic behavior but no alteration in postsynaptic AMPAR function and no increase in CP-AMPAR prevalence.

Animals
We used wild-type C57BL/6J mice and Cln3 knock-out mice (Cln3 ⌬ex1-6 ) on a C57BL/6J background. Cln3 ⌬ex1-6 mice were generated via targeted disruption of the Cln3 gene involving the deletion of exons 2-6 and most of exon 1 via replacement with a neomycin resistance gene that was transcribed in reverse orientation from a mouse PGK promoter (Mitchison et al., 1999). Both male and female mice were used. All procedures for the care and treatment of mice were in accordance with the Animals (Scientific Procedures) Act 1986.

Western blotting
Cerebellar tissue was homogenized in RIPA lysis buffer with proteinase inhibitors (Roche). Protein extracts were boiled for 5 min at 95°C before loading onto 5-10% gradient gels (50 g of protein sample per lane). Gels were electrotransferred to a 0.2-m nitrocellulose membrane (GE Healthcare). Blots were blocked in 4% milk (wt/vol) in PBS-Tween 20 solution for 1 h, then incubated at 4°C overnight with one of the following antibodies: anti-GluA2 (mouse, Millipore MAB397, 1:500), anti-GluA4 (rabbit, Millipore AB1508, 1:200), anti-cofilin (rabbit, Abcam ab42824, 1:10,000). Transferred proteins were detected with appropriate horseradish peroxide-conjugated (HRP) secondary antibodies: goat anti-mouse IgG-HRP (Santa Cruz sc-2005, 1:2000 or goat anti-rabbit IgG-HRP (Santa Cruz sc-2030, 1:2000, reacted with chemiluminescent ECL substrate (Thermo Scientific Pierce), and visualized by ChemiDoc MP System (Bio-Rad Laboratories). Band intensities of GluA2 and GluA4 were normalized to the respective cofilin bands or to the total protein determined mEPSCs in cultured granule cells Miniature EPSCs (mEPSCs) were recorded at -60 mV after blocking voltage-gated sodium channels, NMDA-, GABA A -, and glycine receptors by adding 1 M tetrodotoxin (TTX), 20 M D-AP5, 20 M SR-95531, and 1 M strychnine (Ascent Scientific). Before mEPSC recording, the cells were briefly exposed (2-3 min) to 200 M LaCl 3 to increase mEPSC frequency (Chung et al., 2008). The signal was filtered at 2 kHz and sampled at 20 kHz. Event detection was performed using amplitude threshold crossing (Igor Pro 5, Wavemetrics Inc; NeuroMatic 2.02, www.neuromatic.thinkrandom.com), with the threshold (typically ϳ5 pA) set to 3ϫ the baseline current variance. The rectification index (RI CM ) was calculated by dividing the mean mEPSC peak conductance calculated using all events detected at ϩ60 mV and a matching number of the largest events at -60 mV. For fluctuation analysis (see paragraph below) and kinetic analysis, only events that exhibited a monotonic rise and an uncontaminated decay were included. Such events were aligned on their rising phase before averaging. The decay of the averaged mEPSC was fitted with a double exponential, and the weighted time constant of decay ( w, decay ) calculated as the sum of the fast and slow time constants weighted by their fractional amplitudes. In some cases, mEPSCs were adequately fit with single exponentials.
Peak-scaled non-stationary fluctuation analysis (ps-NSFA) was used to estimate the weighted mean singlechannel conductance of synaptic receptors (Traynelis et al., 1993;Hartveit and Veruki, 2007). Each mEPSC was divided into 30 bins of equal amplitude, and, within each bin, the variance of the mEPSC about the scaled average was computed. The variance was plotted against the mean current value, and the weighted mean singlechannel current was estimated by fitting the full parabolic relationship with the equation: where 2 PS is the peak-scaled variance, I is the mean current, i is the weighted mean single-channel current, N p is the number of channels open at the peak of the EPSC, and 2 B is the background variance. The weighted mean chord conductance for each cell was calculated assuming a reversal of 0 mV.

Slice electrophysiology
Slices were viewed using a fixed stage upright microscope (Olympus BX 51WI with infrared differential interference contrast or oblique illumination) and recordings were made from visually identified neurons in the internal granule cell layer (Kaneda et al., 1995). To block NMDA and GABA A receptors, 20 M D-APV and 20 M SR-95531 (Ascent Scientific) were added. The internal solution contained 128 mM CsCl, 10 mM HEPES, 10 mM EGTA, 2 mM Mg 2 ATP, 0.5 mM CaCl 2 , 2mM NaCl, 5 mM TEA, 1 mM N-(2,6-dimethylphenylcarbamoylmethyl) triethylammonium bromide (QX-314), and 0.1 mM spermine tetrahydrochloride (pH 7.3 with CsOH). Currents were recorded using an Axopatch 200B amplifier, filtered at 2 kHz and digitized at 20 kHz (pClamp 10.2 Molecular Devices or Igor Pro 5 with NeuroMatic). All currents were recorded at room temperature, with the exception of minimally evoked EPSCs (meEPSCs; see 'Quanta and evoked EPSCs' below). Series resistance and input capacitance were read directly from the amplifier settings used to minimize the current responses to 5-mV hyperpolarizing voltage steps. Series resistance was compensated (up to 75%). Measured values at room temperature were 3.9 Ϯ 0.3 pF for wild type versus 3.6 Ϯ 0.3 pF for Cln3 ⌬ex1-6 (n ϭ 11 and 12; W ϭ 76.5, p ϭ 0.54 Wilcoxon rank sum test) and 10.0 Ϯ 0.5 versus 12.4 Ϯ 1.0 M⍀ (W ϭ 40.5, p ϭ 0.12 Wilcoxon rank sum test).

Quantal and evoked EPSCs
To record quantal EPSCs (qEPSCs), the standard extracellular solution was replaced with a Ca 2ϩ -free solution containing 5 mM SrCl 2 (Goda and Stevens, 1994;Abdul-Ghani et al., 1996). Mossy fibers were stimulated (0.5 Hz) using a concentric bipolar tungsten electrode placed in the white matter tract (Digitimer DS/2A constant voltage stimulator; 100 V/200 s). Events were detected using amplitude threshold crossing, with the threshold (typically ϳ5 pA) set according to the baseline current variance. To avoid the inclusion of multiquantal events, only qEPSCs occurring Ͼ10 ms after the mossy fiber stimulus were included. When analyzing event frequency, any qEPSC with a distinct peak was included. When analyzing qEPSC amplitude, all events with a monotonic rise were included, irrespective of overlapping decays. For kinetic analysis, only events with a monotonic rise and uncontaminated decay were included; they were aligned on their rising phase before averaging. The decay of the averaged qEPSC was fitted with a double exponential, and the weighted time constant of decay ( w, decay ) calculated.
To record eEPSCs, mossy fibers were stimulated (0.5 Hz) using a concentric bipolar tungsten electrode placed in the white matter tract (Digitimer DS/2A constant voltage stimulator). Pairs of eEPSCs were recorded at room temperature with an extracellular solution containing 2 mM Ca 2ϩ /1 mM Mg 2ϩ .
To more closely approximate physiologically relevant conditions, meEPSCs were recorded at an elevated temperature (30 -34°C). Mossy fibers were stimulated using constant voltage pulses (80 -100 s; 20 -48 V) delivered through a glass electrode filled with extracellular solution positioned ϳ100 -200 m from the recorded granule cell. The criteria for minimal stimulation included an initial ϳ30% failure rate during repeated single stimuli at 0.25 Hz and invariant EPSC latency and amplitude with increased stimulus intensity. The mean voltage of the threshold stimulus was 32.7 V for wild-type cells and 34.2 V for Cln3 ⌬ex1-6 cells. For each cell, trains of five stimuli (100 Hz, ϳ2 V above threshold) were delivered at 3-s intervals and meEPSCs recorded at -70 mV in both "high" and "low" extracellular Ca 2ϩ (2 mM Ca 2ϩ /1 mM Mg 2ϩ and 1 mM Ca 2ϩ /2 mM Mg 2ϩ ). In each case the amplitudes of evoked currents were normalized to the mean amplitude of the first response (meEPSC 1 ) in 2 mM Ca 2ϩ /1 mM Mg 2ϩ .

Transmission electron microscopy
Sagittal slices (200 m) of cerebellar vermis were prepared from six P13 C57BL/6 mice and three age-matched Cln3 ⌬ex1-6 mice, as described above. Slices were cut in slicing solution, immediately transferred into 4% paraformaldehyde and 0.5% glutaraldehyde, and left overnight at 4°C. Following primary fixation, the tissue was washed and osmicated for 1 h at 4°C in 1% OsO 4 in 0.1 M phosphate buffer, enblocked, stained in 2.0% uranyl acetate buffer for 30 min at 4°C, dehydrated in ethanols, cleared in propylene oxide, and embedded in Araldite. Sections of 70 -80 nm in thickness were made. These were collected on copper mesh grids, counterstained with lead citrate, and viewed in a JEOL 1010 electron microscope.
Mossy fiber axons were identified by their structural characteristics (Xu-Friedman and Regehr, 2003). Release sites were identified by the presence of a presynaptic cluster of vesicles close to the membrane, active zone material and a postsynaptic density. Electron micrographs were analyzed by individuals blinded to the genotype and quantified using ImageJ software (v1.46; https://imagej.nih.gov/ij/). To evaluate the density of vesicles in each terminal, a grid composed of multiple squares (each with an area 0.1 m 2 ) was overlaid on the image. We counted the number of vesicles (of ϳ30 nm in diameter) within each square. Squares containing organelles, or those containing the border of the mossy fiber terminal were excluded from analysis. Vesicles were considered to be proximal to the release site if they were Ͻ100 nm from the presynaptic membrane of an active zone. The active zone vesicle density was then calculated as the number of vesicles per 50 nm of active zone length. As accurate identification of docked vesicles is demanding, even in much thinner slices than used here (Molnár et al., 2016), we opted to count those within one vesicle radius of the active zone and term them "membrane adjacent" vesicles.

Statistical analysis
Summary data are presented in the text as mean Ϯ SEM from n cells (or mossy fiber terminals). Comparisons involving two datasets only were performed using a Wilcoxon rank sum test. For the comparison of paired-pulse ratios (PPRs) at different frequencies and analysis of short-term plasticity, we used two-and three-way repeated measures ANOVA. For EM data, nested analysis was performed using a likelihood ratio test comparing two New Research linear mixed-effect models (Bates et al., 2015). Exact p values are presented to two significant figures, except when p Ͻ 0.0001. Differences were considered significant at p Ͻ 0.05. Statistical tests were performed using R (version 3.3.2; the R Foundation for Statistical Computing; http://www.r-project.org/) and R Studio (version 1.1.383; RStudio). No statistical test was used to predetermine sample sizes; these were based on standards of the field.

Levels of GluA2 and GluA4 are unaltered in cerebella of Cln3 ⌬ex1-6 mice
The increased AMPAR-mediated excitotoxicity seen in dissociated and slice cultures of cerebellum from 8-to 10-d-old Cln3 ⌬ex1-6 mice was originally suggested to reflect altered AMPAR trafficking, and a possible increase in the number of GluA2-lacking CP-AMPARs (Kovács et al., 2006). However, the same authors later described an increase in GluA2 protein in the cerebellum of one-monthold Cln3 ⌬ex1-6 mice (Kovács et al., 2015). To investigate possible AMPAR subunit changes, we initially measured protein levels for GluA2 and GluA4 in cerebellum from wild-type and Cln3 ⌬ex1-6 mice in the second postnatal week, around the age when the first structural and functional defects are observed in Cln3 ⌬ex1-6 mice (Weimer et al., 2009).
We prepared cerebellar tissue lysate from 12 wild-type and 12 Cln3 ⌬ex1-6 mice (P14 -P16). For each group, four samples were generated by pooling tissue from three littermate mice. All eight samples were run together and the membrane probed with the relevant antibodies (mouse anti-GluA2, mouse anti-GluA4, rabbit anti-cofilin; see Materials and Methods; Fig. 1A,B). We found no difference in total protein for either GluA2 (0.29 Ϯ 0.04 for wild type vs 0.29 Ϯ 0.06 for Cln3 ⌬ex1-6 , normalized to the intensity of the cofilin band; W ϭ 10, p ϭ 0.69) or GluA4 (0.33 Ϯ 0.05 vs 0.30 Ϯ 0.07 normalized to the intensity of cofilin; W ϭ 9, p ϭ 0.89; Fig. 1C,D). Similar results were obtained when values were normalized to total protein (data not shown; see Materials and Methods).

AMPA-evoked currents are unchanged in cultured Cln3 ⌬ex1-6 granule cells
To determine whether the magnitude of AMPAR-mediated currents or the prevalence of CP-AMPARs was altered in cerebellar granule cells from Cln3 ⌬ex1-6 mice, we first made recordings from cultured neurons and examined whole-cell currents evoked by bath application of AMPA (20 M). The responses were compared during voltage ramps from -90 to ϩ60 mV, with spermine (500 M) included in the pipette (intracellular) solution ( Fig. 2A). As this polyamine blocks CP-AMPARs in a voltage-dependent manner, with pronounced block at depolarized potentials, it allows their presence to be identified from the characteristic inwardly rectifying I-V relationship (Bowie and Mayer, 1995;Kamboj et al., 1995;Koh et al., 1995).
To determine whether loss of CLN3 led to an alteration in the basic properties of synaptic AMPARs in granule cells, we assessed their kinetics, voltage-dependence and mean single-channel conductance by analyzing synaptic currents. The 10 -90% risetime and weighted decay of mEPSCs (see Materials and Methods) did not differ between cells cultured from wild-type and Cln3 ⌬ex1-6 mice (0.33 Ϯ 0.02 vs 0.34 Ϯ 0.02 ms, W ϭ 36, p ϭ 0.76 and 1.27 Ϯ 0.10 vs 1.42 Ϯ 0.12 ms, W ϭ 33, p ϭ 0.57; n ϭ 10 and 8 cells). Likewise, we found no difference in the weighted mean single-channel conductance determined using ps-NSFA (see Materials and Methods; 11.5 Ϯ 1.5 vs 11.2 Ϯ 0.9 pS, W ϭ 40, p ϭ 1.00; n ϭ 10 and 8 cells; Fig. 3F) or in mEPSC rectification (RI CM ϩ60/-60 mV; see Materials and Methods; 0.99 Ϯ 0.05 vs 1.05 Ϯ 0.08, W ϭ 36, p ϭ 0.65; n ϭ 12 and 7 cells; Fig. 3G). The fact that the mEPSCs remained non-rectifying and their underlying single-channel conductance remained low in Cln3 ⌬ex1-6 mice suggests that, in keeping with the data from whole-cell AMPA-evoked currents, CI-AMPARs are the predominant subtype present at granule cell synapses following loss of CLN3.

Quantal events at mossy fiber-granule cell synapses of Cln3 ⌬ex1-6 mice
To investigate transmission at mossy fiber to granule cell synapses formed in vivo, we next moved to acute cerebellar slices. As spontaneous mEPSCs occurred only at low frequency, we initially examined quantal events (qEPSCs) in response to mossy fiber stimulation. We made recordings in the presence of 5 mM extracellular SrCl 2 to trigger the asynchronous release of transmitter such that individual quanta could be identified (Fig. 4A,B). This approach allowed us to measure both the size and the number of quanta released per stimulus.

Unaltered paired-pulse depression of eEPSCs in Cln3 ⌬ex1-6 granule cells
Mossy fiber-granule cell synapses are known to sustain high bandwidth transmission, but the majority show an initial short-term depression during high frequency stimulation (Nieus et al., 2006;Saviane and Silver, 2006;Chabrol et al., 2015). Although the reduced number of qEPSCs in slices from Cln3 ⌬ex1-6 mice could be consistent with a decrease in release probability in 0 Ca 2ϩ /5 mM Sr 2ϩ , this was not evident when we examined eEP-SCs in 2 mM Ca 2ϩ . Responses to paired stimuli at 5, 10, 20 and 100 Hz showed no difference in PPR. For example, at 100 Hz, the PPR indicated similar magnitude of depres- . mEPSCs in granule cells from wild-type (WT) and Cln3 ⌬ex1-6 mice are indistinguishable. A, Representative recording of mEPSCs from a granule cell in a culture prepared from WT mice (-60 mV). Traces are consecutive and filtered at 1 kHz for display (mEPSCs are indicated by red dots). B, Same as A but from a granule cell in a culture prepared from Cln3 ⌬ex1-6 mice. Scale bars apply to both A, B. C, upper, Individual mEPSCs from the cell in a, aligned at their point of steepest rise. Middle, Color-coded image of all 77 events. Lower, Averaged mEPSC (black trace) with superimposed SEM (gray fill) and exponential fit to the decay (blue line). The time constant ( decay ) is indicated. D, Same as C but for mEPSCs from the Cln3 ⌬ex1-6 recording in B (scale bars apply to both C, D). E, Pooled data showing similar amplitude and frequency of mEPSCs in granule cells from WT and Cln3 ⌬ex1-6 mice. Left, Cumulative probability distributions for mEPSC amplitudes. The averaged distributions are shown in bold (WT blue; Cln3 ⌬ex1-6 red). Right, Box-and-whisker plots (as in Fig. 1) for mEPSC frequency (log 10 scale) and amplitude (n.s., non-significant; Wilcoxon rank sum test). Ca 2ϩ , a concentration thought likely to approximate more closely the situation in vivo (Borst, 2010).

Altered short-term plasticity of meEPSCs in Cln3 ⌬ex1-6 mice in reduced [Ca 2؉ ] o
We examined meEPSCs in response to brief trains of high frequency mossy fiber stimulation (five stimuli at 100 Hz) at 30 -34°C in both "normal" and reduced extracellular Ca 2ϩ (2 mM Ca 2ϩ /1 mM Mg 2ϩ and 1 mM Ca 2ϩ /2 mM Mg 2ϩ ; Fig. 5A,B). For each cell (six wild type and six Cln3 ⌬ex1-6 ), meEPSC amplitudes were normalized to that of the first event in 2 mM extracellular Ca 2ϩ . In both groups of mice, we observed a wide range of amplitudes for the first meEPSC (peak conductance of 0.56 -2.91 nS for wild-type and 0.33-1.56 nS for Cln3 ⌬ex1-6 ), within the wide range (0.11-3.33 nS) reported by Chabrol et al. (2015) for different mossy fiber input pathways. In 2 mM Ca 2ϩ , meEPSCs in granule cells from wild-type mice exhibited short-term depression (meEPSC 2 /meEPSC 1 was 0.46 Ϯ 0.07; W ϭ 36, p ϭ 0.0028). When the same cells were recorded in 1 mM extracellular Ca 2ϩ , there was no depression (meEPSC 2 /meEPSC 1 was 0.81 Ϯ 0.12; W ϭ 27, p ϭ 0.18; Fig. 5A). However, for Cln3 ⌬ex1-6 cells, paired-pulse depression was seen in both 2 and 1 mM extracellular Ca 2ϩ (0.30 Ϯ 0.05 and 0.31 Ϯ 0.06, respectively; both W ϭ 36, p ϭ 0.0028 and p ϭ 0.0022). A three-way repeated measures ANOVA was run to examine the effect of stimulus number, extracellular Ca 2ϩ concentration and genotype on meEPSC amplitude (normalized to meEPSC 1 in 2 mM Ca 2ϩ ). There was a significant three-way interaction, F (4,80) ϭ 3.67, p ϭ 0.0085. Thus, the effect of lowering extracellular Ca 2ϩ on the meEPSC amplitudes during short trains was affected by deletion of CLN3. Overall, these results suggest altered release dynamics in Cln3 ⌬ex1-6 mice, that are revealed in conditions of reduced extracellular Ca 2ϩ . Of note, the mean amplitude of meEPSC 1 in 2 mM Ca 2ϩ did not differ between genotypes (77.3 Ϯ 26.0 and 69.7 Ϯ 12.0 pA;   to both A, B). C, Box-and-whisker plots (as in Fig. 1) showing the reduced number of discrete quanta evoked in cells from Cln3 ⌬ex1-6 mice ‫‪p‬ءء(‬ Ͻ 0.01; Wilcoxon rank sum test). D, Cumulative probability distributions for qEPSC amplitudes. Data from each cell are shown together with the averaged distributions in bold (WT, blue; Cln3 ⌬ex1-6 , red). Shaded areas denote SEM. Right, Box-and-whisker plots (as in C) showing unaltered qEPSC amplitude in cells from Cln3 ⌬ex1-6 mice. E, Superimposed normalized global average qEPSC waveforms from six WT and seven Cln3 ⌬ex1-6 cells show no differences. Shaded areas denote SEM. Right, Box-and-whisker plots (as in C) for qEPSC 10 -90% risetime and w, decay (n.s., non-significant; Wilcoxon rank sum test).

Structural changes at mossy fiber-granule cell synapses in Cln3 ⌬ex1-6 mice
We next used 2D transmission electron microscopy to compare mossy fiber to granule cell synapses from Cln3 ⌬ex1-6 and wild-type mice (P13). Mossy fiber rosettes were identified from their characteristic size and appearance (many small vesicles, and a large number of mitochondria; Xu-Friedman and Regehr, 2003;Rothman et al., 2016), and the fact that the mossy fiber makes contact with a large number of granule cell dendrites.
Initial examination revealed no striking gross anatomic differences between Cln3 ⌬ex1-6 and wild-type synapses (Fig. 6A,B). The average vesicle diameter was unchanged in Cln3 ⌬ex1-6 compared to wild type (mean vesicle diameter per mossy fiber terminal 33.5 Ϯ 0.5 vs 32.2 Ϯ 0.5 nm, n ϭ 20 and 19 terminals from three mice each; W ϭ 130, p ϭ 0.094; Fig. 6C). We observed a high average density of vesicles within each mossy fiber terminal, comparable to the values of 118 -170 m Ϫ2 reported by Rothman et al. (2016). However, the average density of vesicles per mossy fiber terminal was decreased in Cln3 ⌬ex1-6 mice,   Figure 5. meEPSCs in granule cells in slices from wild-type (WT) and Cln3 ⌬ex1-6 mice exhibit different patterns of short-term plasticity in low extracellular Ca 2ϩ . A, Averaged meEPSCs from a representative WT granule cell evoked during a five-pulse 100-Hz train in the presence of 2 and 1 mM extracellular Ca 2ϩ (-70 mV; 428 and 110 sweeps, respectively). Red arrowheads indicate timing of stimuli (stimulus artifacts are blanked). PPRs (meEPSC 2 /meEPSC 1 ) are indicated as PPR 2/1 . B, Same as A but for a representative Cln3 ⌬ex1-6 granule cell (197 and 111 sweeps). C, Plots showing normalized meEPSC amplitude in WT granule cells during five-pulse trains in 2 and 1 mM Ca 2ϩ . Symbols denote mean and error bars SEM. D, Plots (as in C) but for Cln3 ⌬ex1-6 granule cells ‫‪p‬ءء(‬ Ͻ 0.01 and n.s., non-significant; paired Wilcoxon rank sum test with Holm's sequential Bonferroni correction for multiple comparisons). from 131.7 Ϯ 8.9 to 92.6 Ϯ 6.0 m Ϫ2 (n ϭ 16 and 21 terminals; W ϭ 267, p ϭ 0.0025; Fig. 6C). Additionally, when we determined the number of vesicles proximal to each active zone (within 100 nm), the average number per 50-nm length of active zone was reduced by ϳ30% (from 2.70 Ϯ 0.19 to 1.92 Ϯ 0.16, n ϭ 9 terminals in each of three mice; W ϭ 69, p ϭ 0.013; Fig. 6D). When we considered only membrane adjacent vesicles (those within one vesicle radius of the presynaptic membrane), the number was reduced by ϳ40% (from 1.24 Ϯ 0.16 to 0.71 Ϯ 0.15 per active zone; W ϭ 67, p ϭ 0.022 and from 0.37 Ϯ 0.04 to 0.22 Ϯ 0.04 per 50 nm of active zone; W ϭ 65, p ϭ 0.034; Fig. 6D). Of note, use of nested analysis (see Materials and Methods), rather than average measures per terminal, did not qualitatively alter the outcome. Thus, while vesicle diameter was unchanged, the overall vesicle density per terminal was slightly decreased, as was the average number of vesicles proximal to active zones and the number of membrane adjacent vesicles.

Discussion
We have examined granule cell AMPARs and cerebellar mossy fiber to granule cell synapses in the Cln3 ⌬ex1-6 mouse, a widely used model of juvenile Batten disease. Our main findings are as follows: First, GluA2 and GluA4 expression in cerebellar tissue from Cln3 ⌬ex1-6 mice is unaltered. Second, AMPA-evoked currents in granule cells cultured from wild-type and Cln3 ⌬ex1-6 mice are not different. Third, the properties of synaptic AMPARs, their kinetics, voltage-dependence, and single-channel conductance, are unaltered. Fourth, loss of CLN3 leads to altered short-term plasticity in conditions of reduced extracellular Ca 2ϩ . Fifth, in mossy fiber terminal from Cln3 ⌬ex1-6 mice the density of synaptic vesicles and their proximity to active zones is reduced. Thus, our experiments reveal unanticipated presynaptic changes but no evidence for altered postsynaptic AMPARs.

Changes in synaptic transmission occur early in
Although the original studies of Cln3 ⌬ex1-6 mice reported accumulation of lysosomal storage material at approximately three months of age (Mitchison et al., 1999;Seigel et al., 2002) the mice were thought to lack clinical symptoms, even at 12 months (Mitchison et al., 1999). Subsequent studies identified deficits in motor coordination as early as P14 (Kovács et al., 2006), which were preceded by thinning of the cerebellar granule cell layer and Purkinje cell loss (Weimer et al., 2009). Our results suggest that there are indeed early changes in synaptic transmission in the cerebellum of Cln3 ⌬ex1-6 mice (P10 -P15).
Our experiments do not allow us to conclude whether the observed changes are a direct consequence of CLN3 loss or represent secondary effects. In this regard, it is of note that extensive changes in gene expression and protein levels occur in Cln3 ⌬ex1-6 mice (Brooks et al., 2003;Llavero Hurtado et al., 2017), potentially disrupting multiple neuronal pathways. Nevertheless, our findings in a mouse model of juvenile CLN3 disease complement molecular, structural, and functional studies in various animal models of infantile CLN1 disease (Virmani et al., 2005;Kim et al., 2008;Kielar et al., 2009), late infantile CLN6 disease (Kielar et al., 2009), congenital CLN10 disease (Koch et al., 2011, and late infantile CLN5 disease (Amorim  Fig. 1) showing the unaltered vesicle diameter and the reduced vesicle density. D, Box-and-whisker plots (as in Fig. 1) showing the reduced number of vesicles proximal to active zones (AZ) and reduced number of membrane adjacent vesicles in MF terminals from Cln3 ⌬ex1-6 mice ‫‪p‬ءء(‬ Ͻ 0.01, ‫ء‬p Ͻ 0.05; Wilcoxon rank sum test). et al., 2015), and suggest that early synaptic alteration is a characteristic feature of NCLs.

No change in the rectification of AMPARs in Cln3 ⌬ex1-6 granule cells
Previous studies reported increased AMPA-mediated neurotoxicity in dissociated granule cells and organotypic cultured cerebellar slices from one-week-old Cln3 ⌬ex1-6 mice (Kovács et al., 2006), and improved motor skills in one-to seven-month-old mice following AMPAR blockade (Kovács and Pearce, 2008;Kovács et al., 2011). These authors proposed an increase in the number of CP-AMPARs in Cln3 ⌬ex1-6 cerebellar granule cells, and abnormally increased AMPA receptor-mediated neurotransmission in the cerebellum. More recently, the same authors reported an increase in both total and surface GluA2 in acute cerebellar slices from one-month-old Cln3 ⌬ex1-6 mice, and proposed a decrease in the number of CP-AMPARs (Kovács et al., 2015). Our biochemical analysis and patch-clamp recordings do not support either of these proposals. We found no difference in the levels of GluA2 or GluA4 protein in cerebellar lysates of wild-type and Cln3 ⌬ex1-6 mice. Importantly, we found the magnitude and I-V relationships of AMPAR-mediated currents obtained in the presence of intracellular spermine to be similar in cultured cerebellar granule cells from wildtype and Cln3 ⌬ex1-6 mice. In both groups of mice, I-V relationships were linear, a feature characteristic of GluA2-containing calcium-impermeable AMPARs. Given that mEPSCs in Cln3 ⌬ex1-6 cells exhibited no detectable alteration in amplitude, rise time, decay time, rectification properties or underlying mean single-channel conductance, it seems highly likely that the number and composition of AMPARs at synapses was also unchanged. The reason for these disparities is unclear, but it should be noted that our studies were conducted using mice on a C57BL/6J background, whereas the work of Pearce and colleagues used mice on a 129S6/SvEv background. Importantly, while there are some background-specific differences in motor phenotype of these Cln3 ⌬ex1-6 strains, both exhibit clear motor deficits (Kovács and Pearce, 2015).

Presynaptic changes at mossy fiber-granule cell synapses in Cln3 ⌬ex1-6 mice
As with mEPSCs in cultured granule cells, the amplitude and kinetics of qEPSCs evoked at mossy fiber to granule cell synapses (in the presence of Sr 2ϩ ) were unaffected by loss of CLN3. However, we found a marked decrease in the number of quanta released per stimulus in Cln3 ⌬ex1-6 mice. This could indicate a reduction in the probability of release or simply the activation of fewer mossy fibers. Intriguingly, a recent report described increased hippocampal field excitatory post-synaptic potentials in Cln3 ⌬ex7/8 mice and suggested increased axonal excitability at the earliest age studied (one month; Burkovetskaya et al., 2017), tending to argue against the second of these possibilities. In a separate set of experiments in 2 mM Ca 2ϩ , we found the PPR of eEPSCs was not affected by loss of CLN3, suggesting no change in release probability. Thus, the effect of CLN3 loss may depend on the extracellular Ca 2ϩ concentration.
In both wild-type and Cln3 ⌬ex1-6 slices, we observed depression of eEPSC amplitudes during short trains of mossy fiber stimulation in the presence of standard extracellular divalent cations (2 mM Ca 2ϩ /1 mM Mg 2ϩ ). Surprisingly, when we reduced release probability by lowering extracellular Ca 2ϩ (1 mM Ca 2ϩ /2 mM Mg 2ϩ ), we observed loss of depression in wild-type cells (Nieus et al., 2006;Saviane and Silver, 2006) but not in Cln3 ⌬ex1-6 cells. The fact that the loss of CLN3 appeared to have a functional impact on transmission only when extracellular Ca 2ϩ was reduced suggests the possibility of an alteration in Ca 2ϩ handling or sensing. Recent studies have indeed suggested that in both neurons (Warnock et al., 2013) and neuronal progenitor cells (Chandrachud et al., 2015) calcium handling is disrupted following loss of CLN3. This has been shown to result in the aberrant elevation of intracellular Ca 2ϩ following K ϩ -induced depolarization or moderate inhibition of the sarco/endoplasmic reticulum Ca 2ϩ -ATPase by thapsigargin. Whether altered Ca 2ϩ handling in mossy fiber terminals could account for the differences in short term plasticity between Cln3 ⌬ex1-6 and wild-type mice is unclear.

Ultrastructural changes at mossy fiber terminals in Cln3 ⌬ex1-6 mice
Our 2D EM analyses revealed presynaptic structural changes in Cln3 ⌬ex1-6 mice, including a decrease in the vesicle density per mossy fiber terminal, a decrease in the number of vesicles proximal to active zones, and a decrease in membrane adjacent vesicles. Interestingly, broadly similar findings have been described in a different NCL. A reduction in vesicle number has been seen in cortical neurons from palmitoyl-protein thioesterase-1 knock-out mice (Ppt1 -/-), a model of infantile CLN1 disease (Virmani et al., 2005;Kim et al., 2008). This effect was linked with persistent membrane association of palmitoylated synaptic vesicle proteins preventing endocytosis. Conversely, in cathepsin D knock-out mice (Ctsd -/-), a model of congenital CLN10 disease, there is a reported increase at hippocampal CA1 synapses in the total vesicle number and in the number of docked vesicles (Koch et al., 2011). Thus, changes in the presynaptic vesicle pool may be a common feature of multiple NCLs. How, or if, the reduction we observe in synaptic vesicles of Cln3 ⌬ex1-6 mice relates to previously described changes in intracellular vesicular trafficking of CLN3-deficient cells (Fossale et al., 2004;Metcalf et al., 2008;Tecedor et al., 2013;Wavre-Shapton et al., 2015) remains to be determined. However, it is possible that the reduced vesicle numbers constitute a compensatory mechanism to overcome the tendency toward elevated release under physiologic conditions. This idea follows from our observation that in 1 mM Ca 2ϩ normalized amplitudes of meEPSC in Cln3 ⌬ex1-6 mice were greater than those of wild-type mice. Of note, the reduced Ca 2ϩ recordings are likely to reflect more accurately the situation in vivo, where the concentration of extracellular Ca 2ϩ is thought to be closer to 1 rather than 2 mM (Borst, 2010). Very recently, a paper was published which described disruption of supraspinal synaptic transmission in the Cln3 ⌬ex1-6 mouse due to impaired presynaptic release, and proposed this as a causative mechanism in juvenile Batten disease (Grünewald et al., 2017). CLN3 loss was found to impair inhibitory PSCs or inhibitory synaptic transmission and to cause loss of GABAergic interneurons, in amygdala, hippocampus, and cerebellum. In addition, the authors reported a reduction in the amplitude of eEPSCs in both principal neurons of the lateral amygdala and granule cells of the dentate gyrus, no change in the amplitude of mEPSCs or spontaneous EPSCs, but a reduction in their frequency. Paired-pulse facilitation during stimulation of the lateral perforant path was also reduced. Overall, the findings were interpreted as reduction of excitatory and inhibitory inputs. Our results echo these observations in identifying presynaptic changes in Cln3 ⌬ex1-6 mice.
Irrespective of the precise mechanism underlying synaptic changes in Cln3 ⌬ex1-6 mice (Cárcel-Trullols et al., 2015;Grünewald et al., 2017), our observations are potentially important in understanding the locus of early changes in juvenile Batten disease. While the recent study of Grünewald et al. (2017) examined synaptic function in symptomatic (14-month-old) Cln3 ⌬ex1-6 mice, the synaptic changes we observed in two-week-old mice occurred in the presymptomatic phase of the disease, and are thus likely to reflect the initial causative changes. Previous functional studies focused primarily on an apparent selective increase in AMPAR function in cerebellar granule cells of Cln3 ⌬ex1-6 mice, and proposed a block of AMPARs as a potential therapeutic approach . Importantly, our results argue strongly against any early change in postsynaptic AMPARs.