Haploinsufficiency of Parkinsonism Gene SYNJ1 Contributes to Dopamine neuron Vulnerability in Aged Mice

Parkinson’s disease (PD) is an age-dependent neurodegenerative disorder characterized by the loss of substantia nigra dopaminergic (DAergic) neurons in ventral midbrain (MB). Identification of interactions between aging and the known risk variants is crucial to understanding the etiology of PD. Recessive mutations in SYNJ1 have recently been linked to familial early-onset atypical Parkinsonism. We now show an age-dependent decline of SYNJ1 expression in the striatum as well as in striatal DAergic terminals of aged mice. Heterozygous deletion of SYNJ1 in mice causes selective elevation of PIP2 in the MB, and manipulation of PIP2 levels also impairs synaptic vesicle recycling preferentially in MB neurons. SYNJ1+/− mice display progressive PD-like behavioral alterations and DAergic terminal degeneration. Furthermore, we found down-regulation of human SYNJ1 transcripts in a subset of sporadic PD brains, corroborating the role of an age-dependent decrease in SYNJ1 in predisposing DAergic neuron vulnerability and PD pathogenesis.


Introduction 22
Parkinson's disease (PD) is one of the most debilitating neurodegenerative disorders affecting 23 millions of people worldwide. Most cases are diagnosed later in life, and the incidence of PD 24 grows exponentially after the age of sixty (Driver et al., 2009). Due to the sporadic nature of 25 most cases, it has been difficult to determine the underlying pathogenic mechanism. Emergent 26 evidence suggested convergent pathogenic pathways, including dysfunctional synaptic 27 membrane trafficking, during disease progression (Trinh and Farrer, 2013;Schirinzi et al., 28 2017). Human genetic studies and genome-wide association studies (GWAS) have also 29 revealed an overlapping pool of genes, such as LRRK2 and SNCA, that contributes to both 30 familial and sporadic PD (Singleton et al., 2013;Spataro et al., 2015;Hernandez et al., 2016). 31 sporadic PD GWAS, the gene encoding endophilinA, SH3GL1, was revealed as one of 17 new 23 risk variants (Chang, et al., 2017). These results have brought increasing interest in 24 understanding if synj1, a close interacting partner of endophilinA, contributes to risks in sporadic 25 PD via common pathways at the nerve terminal. 26 For sporadic PD, aging is considered the greater contributor compared with genetic risks 27 (Driver et al., 2009;Collier et al., 2011). Major pathogenic pathways implicated in PD such as 28 mitochondrial dysfunction, membrane trafficking and autophagy-lysosomal impairment, 29 increased oxidative stress and neuroinflammation also deteriorate during normal aging. More 30 importantly, the number of substantia nigra par compacta (SNpc) DAergic neurons in the ventral 31 midbrain (MB), whose degeneration is known as the hallmark of PD, is also markedly reduced in 32 healthy aged subjects (Stark and Pakkenberg, 2004;Mortera and HerculanoHouzel, 2012). How 33 examined postmortem human data from public databases by extracting the raw measurements 23 of SYNJ1 transcripts in developing human brains (data from Allen Brain Atlas, lacking SN data) 24 and in aged human brains (data from GTEx) (see Materials and Methods / human data 25 analysis). To better elucidate age-dependent changes in transcript levels, we binned the data 26 by age groups and normalized both data sets (developing and aged) to the mean transcript level 27 in the cortex of the overlapping age group [21][22][23][24][25][26][27][28][29][30]. An accelerated decline of 28 SYNJ1 transcripts was observed in the striatum from 11-20 years (bin 3) to 41-60 years (bin 6), 29 while cortical SYNJ1 remained largely steady. The decline of SYNJ1 in the striatum is not due to 30 neuronal loss or synapse loss as β-actin (Figure 1C middle) and synapsin1 ( Figure 1C right) 31 levels do not exhibit similar trends in the striatum. However, the decrease of SYN1 (synapsin 1) 32 expression became apparent at 61-70 years in the cortex, and was accompanied by a 1 comparable reduction in the SYNJ1 level, suggesting synapse loss. The age-dependent down-2 regulation of SYNJ1 in the SN is less clear due to lack of available data, although the average 3 values in most binned groups appear similar to those in the striatum. 4 To verify the above findings in an animal model and to determine if the decline of SYNJ1 5 is reflected beyond the mRNA level to the protein level, we performed immunohistochemical 6 analyses of C57/BL6 mouse brains of both sexes at various ages. We first verified the specificity 7 of the synj1 antibody for its synapse-enriched localization and its validity for quantitative 8 measures in both cell cultures and brain slices (Figure 1-Figure supplement 1). We then 9 performed immunohistochemical staining for tyrosine hydroxylase (TH), synj1, and synapsin1/2, 10 a presynaptic marker, in coronal sections that included both the striatum and the cortex ( Figure  11 1- Figure supplement 2A). The immunofluorescence of synj1 and synapsin1/2 in the striatum 12 was normalized to that in the cortex for each slice (see Materials and Methods/Data analysis). 13 While synapsin1/2 level remained at a constant 80% relative to the cortex across all age groups, 14 synj1 expression was significantly reduced in the striatum of 18-month old mice (Figure 1- Figure  15 supplement 2B), reminiscent of the aged human brain. 16

Synj1 expression in striatal DAergic terminals is reduced in aged mice 17
The striatum is an essential part of the basal ganglia, where DAergic terminals from the ventral 18 MB and glutamatergic (GLUTergic) terminals from the cortex converge at striatal medium spiny 19 neurons to regulate motor output ( Figure 2A). To determine if DAergic terminals in the striatum 20 also exhibit an age-dependent change in synj1 expression, we performed an in-depth analysis 21 of the immunofluorescence in the coronal sections from 3-month and 18-month old mice. By co-22 labeling with TH, the rate-limiting enzyme for DA synthesis, and synapsin1/2, a presynaptic 23 marker, we were able to differentiate DAergic terminals (TH and synapsin1/2 positive) and non-24 DAergic terminals (TH negative, synapsin1/2 positive). We found that in DAergic terminals, 25 synj1 expression was 10% lower than in non-DAergic terminals at 3-month old ( Figure 2B-K); 26 however, this difference grew to nearly 50% in 18-month old animals. Consistently, the 27 cumulative probability for synj1 immunofluorescence sampled from over 1000 nerve terminals 28 revealed a greater difference between TH + and THterminals in an 18-month old mouse 29 compared to a 3-month old mouse (Figure 2 E, K). 30

MB displays specific vulnerability to PIP 2 metabolism in SYNJ1 +/mice 31
The primary function of synj1 is to regulate phosphoinositide metabolism and support the 1 normal functions of membrane trafficking. To understand the impact of synj1 expression on 2 membrane phosphoinositide levels, we measured the content of PIP 2 , PIP and PI in the cortex, 3 the striatum and the MB of 1-year old SYNJ1 +/mice and littermate wildtype (WT) mice using 4 high-performance liquid chromatography (HPLC). We first noted that the PIP 2 level in the MB is 5 nearly 2-fold higher, whereas the PI and the PIP levels were significantly lower, than those in 6 the cortex (n=23) ( Figure 3A). The PIP 2 level in the MB was further elevated by 15% in 1-year 7 old SYNJ1 +/mice than that of WT (p=0.007) ( Figure 3B). While there is a trend of increased 8 PIP 2 in the striatum (p=0.07), no obvious change in the cortex of SYNJ1 +/mice was observed 9 ( Figure 3B). The amount of PIP 2 increase in the heterozygous MB is commensurate to a 10 previous study, which found PIP 2 levels to be elevated by approximately 12% in the whole brain 11 samples of SYNJ1 +/mice (Voronov, et al., 2008). 12 To determine how PIP 2 accumulation in the MB could affect neuronal function, we 13 examined the efficiencies of SV recycling in cultured neurons by expressing an optical reporter, 14 pHluorin. PHluorin is a pH-sensitive variant of GFP whose protonation and deprotonation results 15 in a dynamic 20-fold change in fluorescence, which allows for the quantitative measurement of 16 SV exocytosis and endocytosis when conjugated to the lumenal aspect of the vesicular 17 transporter (Sankaranarayanan et al., 2000;Ariel and Ryan, 2010;Pan and Ryan, 2012). 18 VMAT2-pHluorin or vGLUT1-pHluorin was expressed in cultured MB or cortical neurons, 19 respectively, and a 10 Hz, 10 s field stimulation was applied to trigger SV recycling. We 20 previously showed that cultured MB but not cortical neurons from SYNJ1 +/mice displayed 21 slowed endocytosis (Pan et al., 2017), suggesting that insufficient conversion of PIP 2 due to 22 heterozygous deletion of SYNJ1 affects SV endocytosis preferentially in MB neurons. To verify 23 the selective effect of altered PIP 2 levels in MB neurons, we treated cultured neurons with 24 LY249002 (LY), a PI3K inhibitor, which blocks the conversion of PIP 2 to PIP 3 on the plasma 25 membrane, and found that SV endocytosis in MB neurons was substantially slower after a 10- average. Our data suggests that SV trafficking in MB neurons is more susceptible to PIP 2 30 accumulation than cortical neurons. 31

SYNJ1 +/mice display PD-like motor function deficits 32
We next evaluated the motor functions associated with the reduced expression of SYNJ1. Mice 1 with complete deletion of SYNJ1 are not viable and die shortly after birth. SYNJ1 +/mice, 2 however, appear normal with regard to body size and life span. Unlike the R258Q disease 3 mutation homozygous knock-in mice (RQ KI), some of which display severe movement 4 problems and tonic-clonic seizures (Cao et al., 2017b), SYNJ1 +/mice do not have seizures or 5 apparent gait difficulties in their lifetime. When SYNJ1 +/and SYNJ1 +/+ littermates were 6 evaluated for their general locomotor functions in the open-field assay, 7-month old SYNJ1 +/-7 mice exhibited hyperactivity ( Figure 4A) compared to their littermates. Both SYNJ1 +/+ and 8 SYNJ1 +/mice displayed a significant deterioration in their activity levels at 12-months old and 9 SYNJ1 +/mice appeared no different than WT mice at this age ( Figure 4B). To further test the 10 integrity of dopamine-regulated motor function, we challenged the mice with quinpirole, a DA D2 11 receptor (D2R) agonist. Motor inhibition induced by quinpirole was significantly impaired in 12 SYNJ1 +/mice at 12-months old ( Figure 4C), as was their motor coordination when tested on the 13 accelerated Rota-rod ( Figure 4D). Such decline in motor functions following hyperactivity in 14 SYNJ1 +/mice is reminiscent of the findings in many other PD models (Chesselet and Richter, 15 2011, review), including the LRRK2 G2019S KI mice (Volta, et al., 2017). 16

Aged SYNJ1 +/mice exhibit loss of DAergic terminals 17
To further understand the pathological consequence of reduced SYNJ1 expression, we 18 performed the following analyses: 1) the integrity of DAergic neurons and their nerve terminals terminals in striatal slices was reduced by nearly 50% as mice aged from 3-months to [18][19][20][21][22][23] months. While the number of DAergic terminals was unaltered in SYNJ1 +/mice compared to 24 SYNJ1 +/+ at 3-months old, aging led to a significantly exacerbated reduction of an additional 25 50% in 18-month old SYNJ1 +/mice ( Figure 5A-B). Consistently, striatal DA content and DA 26 metabolites, measured by HPLC, were also reduced in the SYNJ1 +/mice ( Figure 5C), 27 insinuating that SYNJ1 haploinsufficiency leads to the decline of DA release, the loss of DAergic 28 terminals, or a combination of both due to SYNJ1 haploinsufficiency. 29 30

Discussion 31
Emerging evidence has demonstrated a link between recessive point mutations in 1 SYNJ1/PARK20 and familial early-onset atypical Parkinsonism. We now show that a reduction 2 of SYNJ1 expression, which results in aberrant accumulation of PIP 2 in specific brain regions or 3 neuron populations, may contribute to the risks for age-related sporadic PD. SYNJ1 4 heterozygous mice are unable to maintain the proper function of the midbrain DAergic system, 5 driving PD related pathological processes in aged individuals. Our data, therefore, implicates a 6 role of SYNJ1 loss-of-function in the pathogenesis of age-dependent sporadic PD. 7 We show that SYNJ1 expression is down regulated in the striatum during aging in 8 both humans and mice. This is important evidence supporting SYNJ1 functional insufficiency in 9 PD risk, as most pathogenic changes, such as neuroinflammation and mitochondrial 10 malfunction, are also found to present in normal aging. Whether age-related reduction of SYNJ1 11 interacts with other sporadic PD variants has yet to be determined. We previously reported a 12 potential genetic interaction between LRRK2 disease mutation G2019S and SYNJ1 13 haploinsufficiency in mice and found LRRK2 mediates phosphorylation of synj1, which leads to 14 an impairment in SV trafficking in MB neurons (Pan et al., 2017). In addition, although several 15 large-scale genome-wide studies failed to reveal SYNJ1 at the level of significance examined, 16 the gene encoding its closest interacting partner, SH3GL1/endophilinA, is now suggested to be 17 a significant PD-risk gene. Recent studies have also shown that LRRK2 phosphorylates 18 endophilinA and synj1 (Matta et al., 2012;Arranz et al., 2015;Islam et al., 2016;Pan et al., 19 2017), further suggesting the potential involvement of the LRRK2-endophilinA-synj1 complex in 20 regulating synaptic membrane trafficking, a process which may go awry in the pathogenesis of 21

PD. 22
While nearly all reported mouse models with complete knockout of recessive PD genes fail 23 to show any PD related phenotypes, our study demonstrates for the first time that 24 haploinsufficiency of a Parkinsonism gene is sufficient to cause DA neuron vulnerability in aged 25 mice. SYNJ1 +/mice display PD-like behavioral and pathological changes. SYNJ1 +/mice show 26 hyperactivity, which was found in younger mice followed by reduced motor coordination and 27 D2R sensitivity at mid-age. Although it remains to be understood whether and how increased 28 motor activity at younger ages result in a faster decline in motor functions during aging, this 29 phenotype is often found in other PD mouse models, such as the WT and A53T α-synuclein 30 transgenic mice (Unger et al., 2006;Lam et al., 2011;Chesselet et al., 2012) as well as the 31 LRRK2 G2019S KI mice (Volta, et al., 2017). Furthermore, aged SYNJ1 +/mice display 32 advanced DAergic denervation in the striatum accompanied by reduced DA and DA metabolite 33 content. Interestingly, the combined functional decline and pathological DAergic degeneration 1 are mostly found in mouse models carrying variants with early disease onset and high clinical 2 penetrance (Tsika et al., 2014;Sumi-Akamaru et al., 2015;Cao et al., 2017b). Therefore, our 3 results indicate that SYNJ1 +/mice may be used as a potential PD model for dissecting 4 pathogenic pathways at the early stages of PD. 5 The key question is how reduced levels of SYNJ1 contribute to vulnerability selectively 6 within the DAergic system. We found that DAergic terminals express less synj1 compared to 7 neighboring non-DA terminals in the striatum. We also showed that cultured ventral MB 8 neurons, including nigral DAergic neurons that project to innervate the striatum, exhibit higher 9 vulnerability to SV recycling compared to cortical neurons in response to SYNJ1 deficiency (Pan 10 et al., 2017) and significant PIP 2 elevation. In fact, the elevated levels of PIP 2 was robust in MB 11 of SYNJ1 +/mouse, but not in the cortex, suggesting a lack of compensatory mechanism to 12 maintain the homeostasis of PIP 2 levels in MB. These impairments could expedite local 13 mechanisms for synapse elimination or axon degeneration (Stevens et al., 2007). Our study did 14 not differentiate DAergic from GABAergic neurons or VTA from SNpc neurons in the ventral MB 15 with respect to their age-dependent change in synj1 levels or their sensitivity to PIP 2 elevation. 16 Considering the lack of difference in SV endocytosis between TH + and THneurons of the 17 ventral MB at baseline (Pan et al., 2017), it is likely that disease protective mechanisms for VTA 18 and GABAergic neurons arise from signaling pathways other than synj1 and lipid alterations. 19 For example, different calcium burdens or neuroinflammatory responses (Pan et al., 2012;20 Sulzer et al., 2017;Surmeier et al., 2017) may account for their outcome. It was recently 21 reported that loss of synj1 SAC1 activity leads to impaired autophagosome maturation in flies 22 due to abnormal accumulation of PI3P (Vanhauwaert et al., 2017). The PIP levels, however, are 23 not altered in the cortex, the MB or the striatum of the SYNJ1 +/mouse in our study. It remains 24 to be tested if changes in PIP 2 levels alter autophagic signaling via alternative pathways 25 (George et al., 2016) in SYNJ1 +/mice that contribute to DAergic neurodegeneration in the 26

striatum. 27
Interestingly, the SAC1 domain of synj1 that predominantly hydrolyzes phosphatidylinositol 28 monophosphates is considered a weaker enzyme compared to the 5-phosphatase domain, 29 which mainly hydrolyzes PI(4,5)P2 to produce PI4P; yet, the ablation of the SAC1 activity leads 30 to DAergic degeneration in multiple model systems (Cao et al., 2017b;Vanhauwaert et al., 31 2017). Although the functional outcome of the disease-related mutation, R839C, in the 5-32 phosphatase domain of synj1 has not been examined; other mutations and variants that were 33 known to substantially impair 5-phosphatase activity were found in patients with early-onset 1 generalized neurological degeneration and epilepsy (Dyment, et al., 2015;Hardies, et al., 2016). 2 The combined clinical and animal data prompts an interesting possibility that reduced 3 expression level or overall function of synj1, rather than the impairment of an individual 4 functional domain is relevant to the pathogenesis of sporadic PD. 5 Taken together, our findings demonstrate that aging may predispose certain human 6 populations to PD risk via a reduction of SYNJ1 levels in the striatum. Our study thus not only 7 assists in the identification of novel biomarkers for PD, but also suggests a therapeutic idea for 8 PD by restoring SYNJ1 levels. 9 10

Materials and methods 11
Human data analysis 12 Sporadic PD brain transcriptome data was downloaded from PMID: 20926834. Among the 17 13 genome-wide expression datasets, we only examined datasets with a sample size greater than 14 15 in each group to ensure the statistical power. Three of these datasets, GSE28894, 15 GSE20168, GSE8397, were found to exhibit statistical difference (P < 0.05, two-sample 16 Student's t test) in SYNJ1 levels in multiple brain regions. Clinical information of the postmortem 17 brain tissue samples can be found in the following articles: PMID 15965975 and 16344956. 18 Normal human brain age-dependent data was downloaded from Allen Brain Atlas / developing 19 human brain (7D4BTI1R5K11_0log2, JOYAOQ1MXT11_0_SYN1_log2, and 20 N2884L1TVT11_3_ACTB_log2, referred to as "Allen brain data" hereafter) which contains data 21 from postconceptional week 8 to 40 years old; as well as from the Genotype-Tissue Expression 22 (GTEx) project, which contains brain region-specific data from 20-70 years old 23 (www.gtexportal.org, referred to as GTEx data hereafter). Raw mRNA data from different brain 24 regions was expressed in the Log2 reads per kilobase per million (RPKM) scale. For Allen brain 25 data, measurements from various parts of the cerebral cortices were averaged and expressed 26 as "cortical" mRNA for each subject. Data for the striatum was collected as a single entry and 27 data for substantia nigra was absent. For each documented age, the number of subjects varied 28 from 1 to 3 in the Allen brain data. For GTEx brain data, a single entry was found for cortex, 29 putamen/basal ganglia (denoted as striatum) and substantia nigra. Data distribution was 30 clustered in the 40-70 years range where typically over 10 subjects were measured for a 31 specific age. We used cortical expression in the 21-30 years age bin for normalization to conjoin 32 the two data sets and reveal age-dependent changes in the full spectrum, but have inevitably 1 lost absolute quantitative information on the log2 scale. Committee (IACUC). 7

Cell culture and Transfection 25
MB cultures (Mani and Ryan, 2009;Pan and Ryan, 2012) and cortical cultures (Mani et al., 26 2007) were prepared as described previously. Ventral MBs (containing both VTA and SN) or 27 cortices were dissected from P0-1 mouse pups and digested using papain (Worthington, 28 LK003178) or trypsin (Sigma, T1005) supplemented with DNase (Sigma, D5025), respectively. 29 MB neurons were then prepared according to our previously published protocol plated at a cell 30 density of 199,000 cells / cm 2 and grown in the Neurobasal-A based medium supplemented with 31 GDNF (10 ng/mL, EMD Millipore, GF030). Cortical neurons were plated at 142,000 cells / cm 2 1 and grown in the MEM-based medium supplemented with insulin (24 µg / ml, Sigma, I6634) and 2 transferrin (0.1 mg / ml, Calbiochem, 616420). Typically, four P0-P1 mouse brains are required 3 for a MB culture. Calcium phosphate was used for transfection to achieve sparse expression 4 and to ensure analysis of single neurons during the imaging experiments. Transfection was 5 carried out at DIV 3-5 for MB neurons and at DIV 5-6 for cortical neurons, after which, the 6 growth medium was replaced with a fresh medium supplemented with an antimitotic agent, 7 ARA-C (Sigma-Aldrich, C6645). 8

Optical setup and live cell imaging 9
For live cell imaging, cells were mounted on a custom-made laminar-flow stimulation chamber AP5, buffered to pH 7.40 was used to reveal total pHluorin expression for normalizing 16 exocytosis. All chemicals were purchased from Sigma-Aldrich. Temperature was clamped at 17 30.0 °C at the objective throughout the experiment. Field stimulations were delivered at 10 V / 18 cm by A310 Accupulser and A385 stimulus isolator (World Precision Instruments). A 1 ms pulse 19 was used to evoke single action potentials. Images were acquired using a highly sensitive, 20 back-illuminated EM-CCD camera (iXon+ Model # DU-897E-BV, Andor Corp., CT, USA). 21 Olympus IX73 microscope was modified for laser illumination. A solid-state 488 nm OPSL smart 22 laser at 50 mW (used at 10% and output at ~ 2 mW at the back aperture) was built into a laser 23 combiner system for millisecond on/off switching and camera blanking control (Andor Corp). 24 PHluorin fluorescence excitation and collection were through an Olympus PLAPON 60XO 1.42 25 NA objective using 525/50m emission filter and 495LP dichroic filters (Chroma, 49002). Images 26 were sampled at 2 Hz with an Andor Imaging Workstation driven by Andor iQ-CORE-FST (ver 27 2.x) iQ3.0 software. 28

Confocal Microscopy 29
An LSM780 upright confocal microscope driven by the Zeiss Zen Black software was used to 30 examine the immunofluorescence in brain slices. Images were acquired at 1024X1024 pixel 31 resolution using single scans by a 10x (for striatal and cortical synj1 expression analysis, Figure  32 Figure 2). Immunofluorescence of these images was analyzed using ImageJ. 2

HPLC lipid analysis and monoamine analysis 3
Different brain regions were dissected using rodent brain matrices (ASI, RBM-2000C). Flash 4 frozen mouse brain samples were used for lipid extraction, followed by anion-exchange high-5 pressure liquid chromatography quantification as described previously (Berman et al., 2008;6 Landman et al., 2006;Zhu et al., 2015 andCao et al., 2017a). Striatal samples were collected 7 from freshly dissected brains using 2mm reusable biopsy punch (World Precision Instrument, 8 504529) and flash frozen for further analysis of DA content and two major DA metabolites, HVA 9 and DOPAC, by the Vanderbilt University Neurochemistry Core. 10

Behavioral assays 11
Male SYNJ1 +/mice and their littermate controls were tested for general locomotor activity in an 12 open field chamber in a dark room for 60 min. Motor coordination was assessed by accelerated 13 Rota-rod assay. All mice were subjected to 1-hour habituation in the test room with food and 14 water supply prior to testing. Open-field test -each mouse was placed in the center of a 16 x 15 16-inch chamber equipped with a Versamax monitor system (Accuscan) in a quiet dark room. 16 The mouse horizontal and vertical movements were monitored and recorded for 60 minutes by 17 a grid of 32 infrared beams at ground level and 16 elevated (3 inch) beams. Quinpirole test -18 mice were divided into two groups for each genotype which were then subjected to peritoneal 19 injection of a D2R agonist quipirole (0.05 mg/kg) or 1x PBS before being placed into the open-20 field chamber. Movement was recorded for the following 60 minutes in the dark room as 21 described above. Accelerated Rota-rod test -the mouse was placed on a rotating rod with 22 increasing acceleration from 4-40 RPM over 5 minutes. Each mouse was trained for 2 trials 23 before the test. The duration a mouse spent on the accelerated Rota-rod was averaged for 24 consecutive 3 trials spaced by 15 minutes. 25

Stereology microscopy 26
Mice were perfused and fixed as described above (Cryostat, Immunofluorescence and 27 antibodies section). MB brain tissues were cryo-sectioned at 40 µm in thickness using Leica 28 CM3050s and stored in antifreeze media containing 30% ethanol glycol, 25% glycerol and 5% 29 PB. For stereological counting, one in every five slices was selected and a total of 8 brain slices 30 were used from each mouse for IHC labeling. Zeiss Axioplan2 was used for tissue slice imaging 31 with a 20X objective, and Stereo Investigator was used for data analysis using the following 1 parameters: frame sizes: 150 µm x 150 µm; grid sizes: 250 µm x 250 µm; top guard zone 2 height: 2 µm and optical dissector height: 8 µm. 3

Data analysis and statistics 4
All statistical tests were performed in OriginPro 8.2, except the Kolmogorov-Smirnov test, which 5 uses a built-in function at http://www.physics.csbsju.edu/stats/KS-test.n.plot_form.html. 6 Descriptive statistical tests were carried out to determine the distribution of the data sets. All 7 data sets conforming to the normal distribution were subjected to two-sample Student's t test or 8 multiple-sample ANOVA test followed by Tukey's post-hoc tests. P values less than 0.05 was 9 considered statistically significant. For human brain data ( Figure 1C), the size of each bin was 10 arbitrarily determined to balance sufficient samples and to represent the empirical stages of life 11 interests were manually placed on all colocalized puncta on the 135 X 135 µm image. To 28 reduce the error of this relatively arbitrary measurement, the following strategies were used: 1) 29 Analysis was performed in a double-blinded fashion. 2) The immunostaining procedure, the 30 confocal imaging settings, and the gain / contrast of the images during analysis were kept the 31 same for a matching number of SYNJ1 +/+ and SYNJ1 +/samples. 3) A matching number of 32 images from SYNJ1 +/+ and SYNJ1 +/mice were assigned for analysis at a single time. 4) Counts 33 were compared between two analysts. For pHluorin imaging study ( Figure 3D-E), data was 1 collected from 2-3 batches of cultures. Each data point represents an average of 3 stable trials 2 on a single cell before and after drug treatment (connected by a black line). Typically, 15-50 3 nerve terminals with consistent responses were selected for analysis for each cell. 4 5