Abstract
Cytoplasmic protein tyrosine phosphatase nonreceptor type 11 (PTPN11) and Drosophila homolog Corkscrew (Csw) regulate the mitogen-activated protein kinase (MAPK) pathway via a conserved autoinhibitory mechanism. Disease-causing loss-of-function (LoF) and gain-of-function (GoF) mutations both disrupt this autoinhibition to potentiate MAPK signaling. At the Drosophila neuromuscular junction glutamatergic synapse, LoF/GoF mutations elevate transmission strength and reduce activity-dependent synaptic depression. In both sexes of LoF/GoF mutations, the synaptic vesicles (SV)–colocalized synapsin phosphoprotein tether is highly elevated at rest, but quickly reduced with stimulation, suggesting a larger SV reserve pool with greatly heightened activity-dependent recruitment. Transmission electron microscopy of mutants reveals an elevated number of SVs clustered at the presynaptic active zones, suggesting that the increased vesicle availability is causative for the elevated neurotransmission. Direct neuron-targeted extracellular signal-regulated kinase (ERK) GoF phenocopies both increased local presynaptic MAPK/ERK signaling and synaptic transmission strength in mutants, confirming the presynaptic regulatory mechanism. Synapsin loss blocks this elevation in both presynaptic PTPN11 and ERK mutants. However, csw null mutants cannot be rescued by wild-type Csw in neurons: neurotransmission is only rescued by expressing Csw in both neurons and glia simultaneously. Nevertheless, targeted LoF/GoF mutations in either neurons or glia alone recapitulate the elevated neurotransmission. Thus, PTPN11/Csw mutations in either cell type are sufficient to upregulate presynaptic function, but a dual requirement in neurons and glia is necessary for neurotransmission. Taken together, we conclude that PTPN11/Csw acts in both neurons and glia, with LoF and GoF similarly upregulating MAPK/ERK signaling to enhance presynaptic Synapsin-mediated SV trafficking.
- Drosophila
- glia
- Noonan syndrome (NS)
- NS with multiple lentigines (NSML)
- SH2 domain-containing protein tyrosine phosphatase-2 (SHP2)
- synapse
Significance Statement
Noonan syndrome (NS) is, by far, the most common RASopathy; a group of clinically-classified genetic syndromes caused by MAPK pathway alterations: it affects 1 in every 1,000–2,000 people. Patients present with cognitive deficits caused by PTPN11 mutations, with gain-of-function the most common basis for NS and loss-of-function resulting in NS with multiple lentigines (NSML). We find NS/NSML patient–derived LoF/GoF PTPN11 mutations, as well as Drosophila homolog corkscrew LoF/GoF mutations, all increase presynaptic MAPK signaling, synapsin turnover, and synaptic vesicle availability at presynaptic release sites. Surprisingly, we find PTPN11/corkscrew to be required in both glia and neurons to control neurotransmission strength. These findings suggest disease interventions manipulating presynaptic vesicle trafficking mechanisms, as well as therapeutic strategies targeting both glia and neurons.
Introduction
Noonan syndrome (NS) and NS with multiple lentigines (NSML) are autosomal dominant disorders caused by mutations in protein tyrosine phosphatase nonreceptor type 11 (PTPN11) in ∼50% of NS and >95% of NSML patients (El Bouchikhi et al., 2016; Gelb and Tartaglia, 2022). The NS/NSML disease states share most symptoms, including cognitive impairments in 30–50% of patients (Pierpont et al., 2009; Wingbermühle et al., 2022). Surprisingly, NS is caused by gain-of-function (GoF) and NSML by loss-of-function (LoF) mutations in PTPN11, encoding the Src homology-2 (SH2) domain-containing PTP-2 (SHP2) cytoplasmic phosphatase that positively modulates MAPK/ERK signaling (Zhu et al., 2020). SHP2 function is regulated by autoinhibition, with phosphatase and N-SH2 domains interacting during inactivation, and an open conformation adopted only with signaling activation (Hof et al., 1998). Both GoF/LoF disease mutations favor the SHP2 open conformation with an exposed catalytic domain, leading to elevated MAPK/ERK signaling (Yu et al., 2013). In the mouse NS disease model, long-term potentiation and memory deficits are rescued by MAPK/ERK inhibition (Lee et al., 2014). In Drosophila NS and NSML disease models, MAPK/ERK-dependent long-term memory is likewise disrupted (Pagani et al., 2009; Das et al., 2021). Although these previous investigations clearly indicate MAPK/ERK-dependent deficits in the nervous system, the underlying neuronal mechanism that causes these impairments has yet to be fully determined.
At the Drosophila neuromuscular junction (NMJ) glutamatergic synapse, we previously tested PTPN11 homolog corkscrew (csw) LoF/GoF mutations as well as human patient–derived PTPN11 LoF/GoF transgenes targeted to neurons for synaptic phenotypes (Leahy et al., 2023). All of these mutations increase presynaptic MAPK/ERK signaling and elevate neurotransmission, with defects corrected by MAPK/ERK inhibitors. Electrophysiological recordings suggest synaptic vesicle (SV) release is heightened to increase signaling strength in basal resting conditions, with compensatory reductions during activity-dependent synaptic depression and short-term plasticity (Leahy et al., 2023). Vesicles can be functionally classified into the readily releasable pool (RRP), rapid recycling pool, and reserve pool (Rizzoli and Betz, 2005). The RRP available for immediate release represents vesicles physically docked at the presynaptic active zone (AZ) and primed for exocytosis (Hur et al., 2018). The RRP is replenished from the vesicle recycling pool, with greater replacement during conditions of elevated demand (Kuromi and Kidokoro, 1998). The reserve pool is recruited by high usage levels; for example, during high-frequency stimulation. The synapsin phosphoprotein tethers vesicles well away from the membrane, with phosphorylation causing SV disassociation and SV recruitment to AZ fusion sites (Akbergenova and Bykhovskaia, 2010). Loss of this synapsin regulation results in an inability to sustain proper release dynamics (Pieribone et al., 1995). Given synapsin is a MAPK target (Cheng et al., 2017), we hypothesized a causal mechanism in our NS/NSML disease models.
To test this hypothesis, this study employs confocal imaging of SV-associated synapsin, electron microscopy synaptic ultrastructure analyses, and two-electrode voltage-clamp (TEVC) electrophysiology recordings of PTPN11/csw mutants. We first test SV-associated synapsin in both basal and stimulated conditions. In PTPN11/csw LoF and GoF mutants, we find elevated synapsin colocalized with SVs under basal conditions and a dramatic loss of SV-associated synapsin with acute stimulation. These findings suggest strongly altered synapsin-dependent SV trafficking in the NS/NSML disease models. We next use transmission electron microscopy to visualize synaptic ultrastructure. In mutants, we find an increase in SVs clustered at presynaptic active zones. These findings suggest that increased vesicle availability underlies the elevated mutant neurotransmission strength. We next test a neuron-targeted ERKGoF line to assay the presynaptic MAPK/ERK mechanism. We find higher MAPK signaling drives increased presynaptic function, and synapsin loss blocks elevated neurotransmission with presynaptic NS/NSML mutations and ERKGoF. To rescue these phenotypes, we reintroduced neuron-targeted wild-type csw into csw null mutants. Surprisingly, we find absolutely no improvement. Only rescue in both neurons and glia restores neurotransmission, indicating an unexpected dual requirement in both cell types. However, we find that glial-targeted PTPN11/csw LoF and GoF alone elevate neurotransmission, showing that disrupted glial function is sufficient to increase synaptic strength.
Materials and Methods
Drosophila genetics
All Drosophila stocks were reared on standard cornmeal/agar/molasses food at 25°C in 12 h light/dark cycling incubators. All animals were reared to the wandering third instar stage, with all genotypes confirmed with a combination of sequencing, Western blots, and transgenically marked balancer chromosomes. Due to the corkscrew gene being on the X chromosome, all experiments utilizing csw5 mutants were conducted on males only. All other experiments were done on both sexes (males and females together). The two genetic background controls were (1) w1118 for all of the mutants back-crossed in the w1118 background and (2) the Transgenic RNAi Project (TRiP) RNAi third chromosome background control for the TRiP RNAi lines (Perkins et al., 2015). The csw5 null (Perrimon et al., 1985), syn97 (Ojelade et al., 2019 Godenschwege et al., 2004), ERK1 hypomorph (Beck et al., 2015), wild-type transgenic UAS-cswWT line (Johnson Hamlet and Perkins, 2001), gain-of-function UAS-ERKSEM line (Ariss et al., 2020), and UAS-csw RNAi lines (Perkins et al., 2015) were all obtained courtesy of the Drosophila Bloomington Stock Center (BDSC, Indiana University). All of the NS/NSML patient–derived UAS-PTPN11 mutant lines (Das et al., 2021) were a kind gift from Dr. Tirtha Das (Department of Cell, Developmental, and Regenerative Biology, Icahn School of Medicine at Mount Sinai). Transgenic driver studies were performed with the pan-neuronal elav-Gal4 (Ogienko et al., 2020), glutamatergic neuron-specific vglut-Gal4 (Mahr and Aberle, 2006), and glial-specific repo-Gal4 (Ogienko et al., 2020), which were all obtained from the BDSC. Recombinant lines were confirmed via PCR, chromosome markers, and cell markers (e.g., membrane mCD8::GFP).
Immunocytochemistry imaging
Wandering third instars were dissected in physiological saline containing the following (in mM): 128 NaCl, 2 KCl, 4 MgCl2, 0.2 CaCl2, 70 sucrose, and 5 HEPES (pH 7.2). Preparations were fixed in 4% paraformaldehyde (Electron Microscopy Sciences, 15714) diluted in phosphate-buffered saline (PBS; Corning, 46–013-CM) for 10 min at room temperature (RT). Preparations were then washed and permeabilized in PBS containing 0.2% Triton X-100 and 1% bovine serum albumin (3×, 10 min), followed by blocking for 30 min at RT in the same solution. Preparations were incubated with primary antibodies overnight at 4°C. Primary antibodies used included: rabbit anti–phospho-ERK1/ERK2 (Thr185, Tyr187) polyclonal antibody (Thermo Fisher Scientific, 44-680G, 1:100), rabbit anti-vesicular glutamate transporter (vglut, 1:1,000, Daniels et al., 2004), mouse anti-synapsin (Developmental Studies Hybridoma Bank, 3C11, 1:100), goat Cy3–conjugated anti-horseradish peroxidase (HRP; Jackson ImmunoResearch, 123-165-021, 1:200), and goat 647–conjugated anti-HRP (Jackson ImmunoResearch, 123-605-021, 1:200). Following primary incubations, preparations were again washed (3×, 10 min) and then incubated with secondary antibodies for 2 h at RT. Secondary antibodies used included donkey 488 anti-rabbit (Invitrogen, A21206, 1:250) and donkey 555 anti-mouse (Invitrogen, A31570, 1:250). Preparations were again washed (3×, 10 min) and then mounted in Fluoromount G (Electron Microscopy Sciences) on 25 × 75 × 1 mm slides (Thermo Fisher Scientific, 12–544-2) with 22 × 22 mm coverslips (Thermo Fisher Scientific, 12–542-B), sealed with clear nail polish (Sally Hansen). All NMJ imaging was performed using a Zeiss LSM 510 META laser-scanning confocal microscope, with images projected in Zen (Zeiss) and analyzed using ImageJ (NIH open source). All imaging was done with identical settings and kept consistent between all genotypes and stimulation conditions. Fluorescence intensity was calculated from the maximum projection of the full HRP signal-delineated z-stack using ImageJ threshold and wand-tracing tools. Thresholds were set consistently to avoid image saturation and maximize the fluorescence signal range (0–255). To compare across different trials, we normalized NMJ intensity values to the mean intensity value of basal controls for each trial (absolute units).
Colocalization assays
Synaptic vesicle (SV) colocalization was done with vesicular glutamate transporter (vglut) and synapsin antibody colabeling at the wandering third instar NMJ (as above). Image settings were identical in all conditions. Glutamatergic SVs were marked with anti-vglut labeling (Daniels et al., 2006), which has been used previously in anti-synapsin colocalization analysis (Vanlandingham et al., 2014). With ImageJ, an NMJ synaptic bouton region of interest (ROI) was determined by selection with anti-HRP presynaptic membrane labeling (Jan and Jan, 1982). Before analyzing, acquisition images were split by channel, and fluorescent background from outside the NMJ region was subtracted to ensure no false overlap. Using the HRP channel, the synaptic bouton ROI was converted to a mask, and colocalization analyses of anti-vglut and anti-synapsin within this defined mask ROI were then performed using the Coloc2 plugin, with the Pearson’s correlation coefficient (PCC) quantified (Vanlandingham et al., 2014).
Electron microscopy
Wandering third instars were dissected in physiological saline (as above) and fixed overnight at 4°C in 2.5% glutaraldehyde (Electron Microscopy Sciences, 16020) in 0.1 M sodium cacodylate (SC) buffer (EMS, 11,650), followed by a secondary fixation in 1% osmium tetroxide (Electron Microscopy Sciences, 19172) in SC buffer for 1 h at RT. Preparations were washed in 0.1 M SC buffer (3×, 10 min) and then ddH2O (3×, 15 min). Labeling was done en bloc using 2% uranyl acetate (Electron Microscopy Sciences, 22400) for 2 h in the dark, and then preparations were rinsed in ddH2O (3×, 15 min). Preparations were next dehydrated through an ethanol series (30, 50, 70, 90, 95, 100, 100%), followed by propylene oxide (Electron Microscopy Sciences, 20401) infiltration and then resin embedding (Embed-812; Electron Microscopy Sciences, 14121). Body wall muscles 6/7 from abdominal segments 3/4 were dissected free and then embedded into a semihardened resin block. The muscles from 10 animals were put into each block. Blocks were polymerized at 60°C for 48 h. Blocks were thick sectioned for ∼150 µm (to NMJ depth) using a DiATOME diamond knife on a Leica Ultracut UCT ultramicrotome. Thin sections were then cut at 65 nm and collected on uncoated 200 mesh copper grids (Electron Microscopy Sciences, T200H-Cu). Five sections were collected per grid, with two consecutive grids collected at a time. Blocks were then thick sectioned an additional 10 µm before collecting on grids again to prevent reimaging the same bouton. Only muscle 6 type 1b boutons surrounded by subsynaptic reticulum (SSR) containing a presynaptic AZ t-bar were analyzed to quantify bouton area, synaptic vesicle (SV) number, and distribution in ImageJ. All imaging was done with a Philips/FEI T-12 TEM operating at 100 kV, with images collected using a 4 megapixel AMT CCD camera.
Synaptic electrophysiology
All NMJ TEVC recordings were done as previously reported (Leahy et al., 2023). Briefly, wandering third instar dissections were done at 18°C in physiological saline (as above). Animals were dissected longitudinally along the dorsal midline, and all internal organs and ventral nerve cord (VNC) were removed. The peripheral motor nerves were cut at the base of the VNC, and the body walls were glued down (Vetbond, 3M). Dissected preparations were imaged with a Zeiss 40× water-immersion objective on a Zeiss Axioskop microscope. Muscle 6 in abdominal segments 3/4 was impaled with two intracellular electrodes (1 mm outer diameter borosilicate capillaries; World Precision Instruments, 1B100F-4) of ∼15 MΩ resistance (3 M KCl). The muscle was clamped at −60 mV with an Axoclamp-2B amplifier (Axon Instruments), and the motor nerve was stimulated with a fire-polished glass suction electrode using 0.5 ms suprathreshold voltage stimuli at 0.2 Hz (Grass S88 stimulator). Nerve stimulation-evoked evoked excitatory junction current (EJC) recordings were filtered at 2 kHz. To quantify EJC amplitudes, we averaged 10 consecutive traces and recorded the average peak value. Spontaneous miniature EJC (mEJC) recordings were made in continuous 2 min sessions and low-pass filtered at 200 Hz. The quantal content for each evoked recording was calculated by dividing the EJC amplitude by the mean mEJC amplitude. Clampex 9.0 was used for all data acquisition, and Clampfit 10.6 was used for all data analyses (Axon Instruments).
Western blots
Wandering third instar neuromusculature from 10 dissected larvae was homogenized in 100 μl lysis buffer [20 mM HEPES, 10 mM EDTA, 100 mM KCl, 0.1% (v/v) Triton X-100, 5% (v/v) glycerol] with protease inhibitor (Roche, 04693132001) combined with protease and phosphatase inhibitor cocktail (Abcam, ab201119). Samples were sonicated and run on 4–15% Mini-PROTEAN TGX Stain-Free Precast Gels (Bio-Rad, 4568083) alongside Precision Plus Protein all blue prestained protein standards (Bio-Rad, 1610373). The total protein was transferred to PVDF membranes using a Trans-Blot Turbo system (Bio-Rad), and the membrane was blocked by TBS intercept blocking buffer (LI-COR, 927–60000) for 1 h at RT. Blocked membranes were incubated with primary antibodies for 1.5 h at RT. Antibodies used included the following: rabbit anti–phospho-ERK1/ERK2 (Thr185, Tyr187) polyclonal antibody (Thermo Fisher Scientific, 44-680G, 1:1,000) and goat anti-GAPDH (Abcam, ab157157, 1:2,000). The membrane was washed with Tris-buffered saline with 0.1% Tween 20 (TBST) and then incubated with secondary antibodies for 40 min at RT. Secondary antibodies used included the following: Alexa Fluor 680 donkey anti-goat (Invitrogen, A21084, 1:10,000) and Alexa Fluor 800 goat anti-rabbit (Invitrogen, A32735, 1:10,000). Membranes were washed in Tris-buffered saline with 0.1% Tween (TBST, 3×, 10 min) and then imaged using the LI-COR Odyssey CLx system.
Statistical analyses
All statistics were performed using GraphPad Prism software (version 9.5). Data sets were subject to normality tests, with D’Agostino–Pearson’s tests utilized (n > 10) or Shapiro–Wilk tests (n < 10). With normal data either (1) two-tailed student’s t tests for two-way comparison with 95% confidence (two data sets) or (2) one-way ANOVA followed by a Tukey’s multiple-comparisons test (three or more data sets, comparing all samples). If data were not normal, Mann–Whitney tests (two data sets) or Kruskal–Wallis followed by a Dunn’s multiple-comparisons test (three or more data sets) were performed. Data sets with multiple sources of variation were analyzed with a two-way ANOVA, followed by Tukey’s multiple-comparisons test. Figures show all individual data points as well as the mean ± SEM, with significance displayed in figures as p ≤ 0.05 (*), p ≤ 0.01 (**), p ≤ 0.001 (***), p ≤ 0.0001 (****), and p > 0.05 (n.s., not significant). Exact p values for all comparisons are provided in the text.
Results
Csw regulates presynaptic synapsin levels and activity-dependent dynamic maintenance
At the Drosophila glutamatergic NMJ, corkscrew null mutants (csw5) have elevated neurotransmission under basal resting conditions and decreased synaptic depression with heightened stimulation (Leahy et al., 2023). We hypothesize these MAPK/ERK-dependent alterations arise from changes in synaptic vesicle (SV) availability in these different synapse activity states. Synapsin is an SV-associated phosphoprotein that regulates availability in an activity-dependent mechanism (Winther et al., 2015; Mirza and Zahid, 2018; Vasin et al., 2019). Reserve pool SVs are tethered in the bouton interior by synapsin and released upon phosphorylation during strong stimulation (Akbergenova and Bykhovskaia, 2007; Winther et al., 2015; Zhang and Augustine, 2021). In this traffic mechanism, MAPK/ERK acts to phosphorylate synapsin, leading to vesicle disassociation and mobilization for exocytosis (Akbergenova and Bykhovskaia, 2010; Zhang and Augustine, 2021). We therefore hypothesize that the Csw) phosphatase modulates MAPK/ERK signaling to regulate this synapsin function in presynaptic boutons. To test this hypothesis, we triple-labeled NMJ boutons with the presynaptic membrane marker anti-HRP, the SV marker anti-vesicle glutamate transporter (vglut), and anti-synapsin (syn; Daniels et al., 2004; James et al., 2019). Synapsin fluorescent intensity and SV colocalization were measured under resting conditions (basal) and following 10 min depolarization stimulation with 90 mM [K+] (stimulated; Kopke et al., 2017; Leahy et al., 2023). We compared the matched genetic background control (w1118) to the csw5 null mutant. Representative images and quantifications are shown in Figure 1.
Csw loss elevates synapsin and causes stimulation-dependent synapsin loss. A, Representative wandering third instar NMJ synaptic boutons labeled for synapsin (syn, magenta), the vesicular glutamate transporter (vglut, green), and HRP (blue) in w1118 genetic background control and csw5 null mutant. NMJs without stimulation (basal, left) and after acute, 10 min 90 mM [K+] depolarizing stimulation (stimulated, right). Scale bar: 2.5 µm. Higher-magnification bouton images are shown below. Scale bar: 1 µm. B, Quantification of synapsin fluorescence intensity in all four conditions (basal and stimulated). Statistical comparisons done using two-way ANOVA, followed by Tukey’s multiple-comparisons test. C, Quantification of synapsin and vglut colocalization in basal (left) and stimulated (right) conditions using PCC analyses. Statistical comparisons done using two-sided t tests. Scatter plots show all the data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.05 (*), p < 0.01 (**), p < 0.0001 (****), and p > 0.05 not significant (n.s.).
In w1118 controls at rest, synapsin is highly enriched in synaptic boutons and strongly associated with SV markers (Fig. 1A, top left). When controls are acutely stimulated (10 min), synapsin is maintained at nearly indistinguishable levels (Fig. 1A, top right). In contrast, csw nulls (csw5) exhibit sharply increased synapsin levels under basal conditions (Fig. 1A, bottom left). When stimulated, csw mutants show a striking decrease in synapsin levels (Fig. 1A, bottom right). In high-magnification images of single boutons, synapsin colocalization with SVs is decreased with stimulation in csw nulls (Fig. 1A, bottom). Quantification of synapsin fluorescence intensity shows a highly significant interaction between genotype and stimulation with a two-way ANOVA (F(1,66) = 15.5, p = 0.0002; Fig. 1B). Under resting (basal) conditions, synapsin in csw5 null mutants (1.67 ± 0.18, n = 19) is highly elevated normalized to controls (1.0 ± 0.08, n = 19), a significant upregulation based on Tukey’s multiple-comparisons test (p = 0.002; Fig. 1B). In controls, normalized synapsin levels after stimulation do not change significantly (0.83 ± 0.14, n = 16, p = 0.7989; Fig. 1B). However, in csw5 nulls, synapsin after stimulation is very significantly decreased (0.47 ± 0.08, n = 16, p = 8.07 × 10−8; Fig. 1B). Imaging limitations prevent identification of where synapsin goes in the stimulated null condition, but it is presumed dispersed in the cytosol (Winther et al., 2015).
To further assay synapsin dynamics, we quantified synapsin–SV colocalization using PCC (Adler and Parmryd, 2010). Quantification shows higher colocalization in csw5 null mutants (0.80 ± 0.01, n = 19) compared with controls (0.74 ± 0.02, n = 19), a significant change based on an unpaired t test (t(36) = 2.277, p = 0.029; Fig. 1C, left). With stimulation, synapsin–SV colocalization in controls (0.71 ± 0.05, n = 16) and mutants (0.65 ± 0.05 n = 16) is no longer significantly different based on an unpaired t test (t(30) = 0.7635, p = 0.4511; Fig. 1C, right). Quantification of vglut fluorescence intensity shows a significant interaction between genotype and stimulation in a two-way ANOVA (F(1,65) = 19.8, p = 3.37 × 10−5; Fig. 1B). Under resting (basal) conditions, vglut fluorescence intensity in csw5 null mutants (1.33 ± 0.01, n = 19) is highly elevated relative to control (0.78 ± 0.07, n = 18), with significant upregulation in Tukey’s multiple-comparisons test (p = 4.45 × 10−5). In controls, normalized vglut fluorescence intensity level after stimulation does not change significantly (0.78 ± 0.08, n = 16, p > 0.999). In contrast, csw5 null vglut fluorescence intensity after stimulation is significantly decreased (0.59 ± 0.07, n = 16, p = 1.2 × 10−7). Taken together, these results indicate that Csw regulates synapsin and synapsin–SV colocalization. We therefore next investigated impacts on presynaptic vesicle pools.
Csw regulates the distribution of SV around presynaptic active zones
The above results suggest altered synaptic vesicle pools in csw5 null mutants. Previous electrophysiological recordings indicate csw mutants have elevated neurotransmission strength from heightened vesicle fusion probability, suggesting a larger population of available vesicles (Leahy et al., 2023). We hypothesize csw mutants would have an increased SV number in close proximity to presynaptic AZ fusion sites (Kaeser and Regehr, 2017; Hur et al., 2018). To test this hypothesis, we examined synaptic ultrastructure using transmission electron microscopy (Dear et al., 2016; Kopke et al., 2017; Bhimreddy et al., 2021). As previous electrophysiology recordings were done on ventral longitudinal muscle 6 in abdominal segments 3/4, we restricted our ultrastructural examination to the same NMJ terminals. Muscles were isolated from dissected wandering third instars, embedded in resin, and sectioned in 65 nm increments (Dear et al., 2016; Kopke et al., 2017). NMJ type 1b boutons were classified based on the surrounding muscle folded SSR, and bouton sections containing a single electron-dense t-bar AZ were selected for all analyses (Hur et al., 2018; Hong et al., 2020; Justs et al., 2022). This identification method is well-established, as SSR around type 1b boutons clearly differentiates them from the smaller type Is boutons (Karunanithi et al., 2002). Comparing genetic background controls (w1118) and csw null mutants (csw5), we quantified bouton area and synaptic vesicle size, number, and distribution. To measure vesicle docking, we counted all vesicles in direct proximity (<20 nm; ½ SV diameter) to the presynaptic density containing an AZ t-bar (Bao et al., 2005; Long et al., 2008; Mohrmann et al., 2008, Jetti et al., 2023). To measure internal SV distribution, we counted all vesicles in 0–200 and 200–400 nm domains from the AZ t-bar (Hur et al., 2018). Representative images and accompanying quantifications are shown in Figure 2.
Csw loss increases presynaptic AZ clustered and docked vesicle pools. A, Representative transmission electron microscope (TEM) images of wandering third instar NMJ presynaptic active zones from w1118 genetic background control (left) and csw5 null mutant (right). Abbreviations: synaptic vesicle (SV), active zone (AZ), SSR. Scale bar: 200 nm. B, Higher-magnification AZ images showing SVs near the electron-dense t-bar. The arrows indicate the vesicles <20 nm from the AZ membrane. Scale bar: 100 nm. C, Quantification of the percentage of the bouton area occupied by SVs with a two-sided t test. D, Quantification of SV density (number of SVs/µm2) within 0–200 nm and 200–400 nm from t-bar with two-sided t tests. E, Quantification of docked SVs (number <20 nm to the presynaptic membrane density containing an AZ t-bar) with two-sided t test. Scatter plots show all data points with mean ± SEM. Data points: bouton number (C, D), AZ number (E). Significance: p < 0.05 (*), p > 0.05 not significant (n.s.).
The synaptic ultrastructure in controls and csw5 null mutants is largely indistinguishable. The genetic background controls (w1118) and csw5 nulls have a similar bouton appearance, with the characteristic AZ t-bar and surrounding vesicles (Fig. 2A). Compared to controls, mutants have no significant change in synaptic bouton area or perimeter, vesicle density, or size (Table 1).
Synaptic bouton ultrastructure parameters
However, there is an increase in the SV number clustered around the AZ in the mutants (Fig. 2A), with a greater number of docked SVs (Fig. 2B, arrows). To quantify these parameters, we first measured the bouton area occupied by SV. Compared to the control coverage (68.67 ± 6.89%, n = 7), vesicles in the csw5 null mutants occupy less area (41.46 ± 9.07%, n = 9), which is significant based on an unpaired t test (t(14) = 2.272, p = 0.0394, Fig. 2C). Thus, vesicles are more spatially cohesive in the mutant boutons. We next measured SV density in concentric rings around the AZ. In the 0–200 nm region, the control (16.01 ± 1.754 SV/µm2, n = 11) and csw5 null (15.27 ± 1.3 SV/µm2, n = 12) densities are not significant with an unpaired t test (t(21) = 0.3427, p = 0.7353, Fig. 2D, left). In sharp contrast, the 200–400 nm region shows a strikingly higher SV density in the controls (13.96 ± 1.47 SV/µm2, n = 11) compared with csw5 nulls (8.75 ± 1.62 SV/µm2, n = 12), with a significant elevation based on an unpaired t test (t(21) = 2.375, p = 0.0271, Fig. 2D, right). This indicates that the reserve pool is specifically disrupted in csw5 null mutants (Hur et al., 2018). In addition, AZ docking measured as SVs with ½ a vesicle diameter from the presynaptic density (<20 nm, arrowheads) is lower in controls (2.083 ± 0.288 SVs, n = 12) compared with csw5 nulls (3.33 ± 0.373 SVs, n = 9), which have a significantly more docked SVs based on an unpaired t test (t(19) = 2.702, p = 0.0141, Fig. 2E). Taken together, these results reveal a disrupted reserve pool and more docked SVs at AZ release sites in csw null mutants. We next turned to investigate this presynaptic mechanism in NS/NSML patient–derived point mutations.
PTPN11 regulates synapsin synaptic vesicle association under basal and stimulated conditions
Drosophila csw5 null mutants have defects in presynaptic synapsin dynamics and vesicle pool regulation. To further test a presynaptic regulatory mechanism in the NS/NSML disease states, we next analyzed neuron-targeted, patient–derived point mutations. These mutations include gain-of-function (GoF) PTPN11N308D and loss-of-function (LoF) PTPN11Q510P associated with NS and NSML, respectively (Das et al., 2021). We have previously discovered that both of these GoF/LoF mutations targeted to neurons elevate neurotransmission strength based on electrophysiology recordings (Leahy et al., 2023). Moreover, like csw5 nulls, we have found that these patient–derived point mutations increase presynaptic MAPK/ERK signaling and decrease synaptic depression during heightened activity (Leahy et al., 2023). We therefore hypothesized that neuronal PTPN11 mutations with increased neurotransmission would phenocopy the csw5 null increased basal synapsin levels and activity-dependent synapsin loss. We once again utilized triple-labeled antibody imaging at the NMJ to investigate synapsin levels at rest (basal) and in the 10 min high [K+] depolarizing stimulation condition (stimulated). The presynaptic bouton membrane was again marked with anti-HRP, and the total SV population was again marked with anti-vglut. Anti-synapsin fluorescence intensity and SV colocalization measurements were done in single synaptic boutons. Tests were done with the neuronal driver alone control (elav-Gal4/w1118) compared with neuron-targeted PTPN11 GoF (elav-Gal4 > PTPN11N308D) and LoF (elav-Gal4 > PTPN11Q510P) mutations. Representative basal/stimulated images and quantifications for all three genotypes are shown in Figure 3.
NS/NSML transgenes increase synapsin levels with stimulation-dependent loss. A, Representative NMJ synaptic boutons labeled for synapsin (syn, magenta), vesicular glutamate transporter (vglut, green), and HRP (blue) in neuronal transgenic driver control (elav-Gal4/w1118, left) and driving PTPN11N308D (elav-Gal4 > PTPN11N308D, middle) and PTPN11Q510P (elav-Gal4 > PTPN11Q510P, right). NMJs without stimulation (basal, top) and 10 min 90 mM [K+] (stimulated, bottom). Scale bar: 2.5 µm. Higher-magnification bouton images are shown below. Scale bar: 2.5 µm. B, Quantification of synapsin fluorescence intensity in all six conditions (basal and stimulated). Statistical comparison done using a two-way ANOVA, followed by Tukey’s multiple-comparisons test. C, Quantification of synapsin and vglut colocalization in basal (left) and stimulated (right) conditions using PCC analyses. Statistical comparisons done using one-way ANOVAs followed by Tukey’s multiple-comparisons tests. Scatter plots show all data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****), and not significant (n.s.).
Like the w1118 genetic background, neuronal driver controls (elav-Gal4/w1118) maintain the SV-associated synapsin at the same levels in basal and stimulated conditions (Fig. 3A, left). In contrast, GoF (PTPN11N308D) and LoF (PTPN11Q510P) both increase SV-associated synapsin at rest and exhibit a sharp decrease in synapsin with stimulation (Fig. 3A, middle and right). In high-magnification single boutons, synapsin colocalization with SVs is decreased with stimulation in both mutants (Fig. 3A, bottom). Quantification shows a significant interaction between genotype and stimulation in a two-way ANOVA (F(2,119) = 7.538, p = 0.0008; Fig. 3B). At rest (basal), neuronal PTPN11N308D (1.47 ± 0.11, n = 22, p = 0.0002) and PTPN11Q510P (1.35 ± 0.10, n = 17, p = 0.0193) have significantly elevated synapsin normalized to driver controls (1.0 ± 0.04, n = 19; Fig. 3B). With stimulation, synapsin levels in controls do not change significantly (0.95 ± 0.06, n = 22, p = 0.9954; Fig. 3B). However, PTPN11N308D (0.87 ± 0.04, n = 22, p = 1.796 × 10−7) and PTPN11Q510P (1.04 ± 0.06, n = 23, p = 0.0438) are significantly decreased from basal levels (Fig. 3B). This effect is confirmed with PCC synapsin–SV colocalization analyzed by one-way ANOVA (F(2,54) = 5.779, p = 0.0053; Fig. 3C, left). Quantification shows significantly increased colocalization in PTPN11N308D (0.60 ± 0.03, n = 22, p = 0.0068) and PTPN11Q510P (0.59 ± 0.03, n = 16, p = 0.0301) versus driver control (0.49 ± 0.02, n = 19; Fig. 3C). Compared to synapsin–SV colocalization following stimulation (0.45 ± 0.03 n = 21), neither PTPN11N308D (0.4 ± 0.03, n = 23) or PTPN11Q510P (0.47 ± 0.04, n = 24) are not significantly different based on a one-way ANOVA (F(2,65) = 1.296, p = 0.2805; Fig. 3C, right). Taken together, these results indicate neuronal GoF/LoF PTPN11 mutations increase synapsin, with aberrant activity-dependent synapsin loss. To test MAPK signaling dependence, we next analyzed a neuronal extracellular signal–regulated kinase (ERK) GoF condition.
Neuronal ERK gain-of-function phenocopies PTPN11/csw mutant synapse regulation
In our previous studies, we showed PTPN11/csw mutants elevate both local presynaptic MAPK/ERK signaling and presynaptic neurotransmission, with the defects prevented by feeding pharmaceutical MAPK/ERK inhibitors (Leahy et al., 2023). These drug analyses demonstrate a requirement for MAPK/ERK signaling in both GoF and LoF mutants, but do not rule out other pathway contributions. To start testing the specificity of the MAPK/ERK-dependent presynaptic mechanism, we first assay whether neuron-targeted ERKGoF (elav-Gal4 > ERKSEM; Kim et al., 2006) recapitulates the PTPN11/csw mutant synaptic phenotypes. To confirm elevated activation we test the MAPK signaling endpoint of phospho-ERK (pERK; Guo et al., 2020) with Western blots and look for specific local pERK activation with NMJ double-label imaging with anti-pERK (green) and anti-HRP (magenta). Using HRP to delineate presynaptic membranes, we measure pERK fluorescence intensity normalized to the transgenic driver control (elav-Gal4/w1118). We next use TEVC electrophysiological recording to test elav-Gal4 neuron–targeted ERKGoF compared with the transgenic control by assaying EJC responses driven by motor nerve stimulation. We also test spontaneous release with mEJC recordings, assessing both event frequency and amplitude. Representative images, recordings, and quantified results are shown in Figure 4.
Neuronal ERKGoF recapitulates PTPN11 pERK and neurotransmission defects. A, Representative Western blot for pERK (42 kDA, top) and GAPDH loading control (35 kDA, bottom) in driver control (elav-Gal4/w1118) and ERKGoF (elav-Gal4 > ERKSEM). B, Quantification of pERK levels normalized to GAPDH with a two-sided t test. C, Representative NMJ images co-labeled for anti–phospho-ERK (pERK, green) and presynaptic membrane marker anti-HRP (magenta) in driver control and ERKGoF. Scale bar: 2.5 µm. D, Quantified pERK presynaptic fluorescence levels normalized to control with a two-sided t test. E, Representative TEVC recordings showing EJC traces with 10 superimposed evoked responses (1.0 mM Ca2+) from neuronal driver control (left) and ERKGoF (right). F, Quantification of EJC amplitudes with a two-sided t test. G, Representative mEJC traces (1.0 mM Ca +2) from neuronal driver control (left) and ERKGoF (right). H, Quantification of the mEJC frequency using a two-sided t test. I, Quantification of mEJC amplitude using a two-sided t test. J, Quantification of quantal content using a two-sided t test. Scatter plots show all data points with mean ± SEM. Data points: animal number (B) and NMJ number (D-J). Significance: p < 0.05 (*), p < 0.01 (**), p < 0.0001 (****), and not significant (n.s.).
We first confirmed increased signaling activation in Western blots measuring pERK levels with neuronally targeted elav-Gal4 > UAS-ERKSEM (shown as ERKGoF) compared with driver controls (elav-Gal4/w1118). At the predicted pERK molecular weight (42 kDa), there is a clearly increased band with ERKGoF (Fig. 4A). Normalized to a GAPDH loading control, the ERKGoF band (1.37 ± 0.14, n = 11) is increased over control (1.0 ± 0.04, n = 11), a significant elevation based on an unpaired t test (t(20) = 2.478, p = 0.0222, Fig. 4B). We next confirmed pERK is locally increased in NMJ boutons similar to NS/NSML mutants. In elav-Gal4 controls, pERK is only weakly detectable in boutons (Fig. 4C, top). In contrast, pERK levels are elevated with neuronal ERKGoF (Fig. 4C, bottom). Quantification of pERK with neuronal ERKGoF (1.33 ± 0.05, n = 31) normalized to control (1.0 ± 0.05, n = 29) shows a significant elevation in an unpaired t test (t(58) = 4.626, p = 2.1375 × 10−5; Fig. 4D). We next tested effects on neurotransmission. Compared to the driver controls, neuron-targeted ERKGoF causes elevated neurotransmission (Fig. 4E). EJC amplitudes in controls (123.40 ± 12.79 nA, n = 10) are increased with neural ERKGoF (188.25 ± 16.86 nA, n = 11), a significant strengthening with a two-sided t test (t(19) = 3.019, p = 0.0071; Fig. 4F). In spontaneous mEJC recordings, SV fusion events are greatly elevated by neuronal ERKGoF (Fig. 4G). Compared to the control frequency (0.79 ± 0.1 Hz, n = 11), there are many more events with ERKGoF (2.55 ± 0.1 Hz, n = 10), a significant increase based on a two-sided t test (t(19) = 12.45, p = 1.39 × 10−10; Fig. 4H). The mEJC amplitude is not altered (t(19) = 0.3096, p = 0.7602, Fig. 4I). Compared to the control quantal content (124.01 ± 12.86, n = 10), quanta are increased by neuronal ERKGoF (184.25 ± 16.5, n = 11), a significant elevation with a two-sided t test (t(19) = 2.838, p = 0.0105, Fig. 4J). Taken with above results, we conclude heightened presynaptic pERK signaling elevates vesicle fusion probability, phenocopying PTPN11/csw mutants in a synapsin-dependent mechanism.
Removal of synapsin blocks the elevated neurotransmission in PTPN11/csw and ERK mutants
To further test a mechanistic connection between PTPN11, ERK, and synapsin elevating neurotransmission, we next expressed neuronally driven PTPN11Q150P or ERKGoF transgenes in a synapsin null mutant. Our goal was to investigate the effects of preventing aberrantly elevated synapsin on neurotransmission strength in these two mutant conditions. In our previous studies, we found that neuronally targeted PTPN11 mutations strongly elevate basal neurotransmission, which can be rescued by introducing pharmaceutical MAPK/ERK inhibitors (Leahy et al., 2023). These previous findings, alongside the highly increased synapsin levels in csw/PTPN11 mutants (Figs. 1, 3) and elevated presynaptic ERK activity and strengthened neurotransmission in ERKGoF (Fig. 4), led us to hypothesize that PTPN11, ERK, and synapsin act together to regulate basal neurotransmission amplitudes. Notably, however, synapsin has been previously reported to have no effect on basal NMJ function (Sun et al., 2006), generating an apparent quandary. To test this question, we again use TEVC electrophysiology to assay the neuronal PTPN11 point mutant (elav-Gal4 > PTPN11Q510P), a synapsin null mutant (syn97), and, in this null mutant background, neuronal PTPN11Q510P (syn97; elav-Gal4 > PTPN11Q510P) and ERKGoF (syn97; elav > ERKSEM) compared with the driver control (elav-Gal4/w1118). To further test these genetic interactions, we also assay the csw5 null mutant, ERK1 hypomorph (ERK1/+), and the double mutant (csw5; ERK1/+) compared with the genetic background control (w1118). We measured spontaneous mEJC events and evoked EJC responses from motor nerve stimulation (Kopke et al., 2020; Leahy et al., 2023). Representative traces and quantifications are shown in Figures 5 and 6.
synapsin null blocks elevated neurotransmission in PTPN11 and ERK mutants. A, Representative TEVC recordings of EJC traces showing 10 superimposed evoked synaptic responses (1.0 mM Ca2+). From left to right: neuronal driver control (elav-Gal4/w1118), neuronal PTPN11 (elav-Gal4 > PTPN11Q510P), synapsin null (syn97), neuronal PTPN11Q510P with synapsin null (syn97; elav-Gal4 > PTPN11Q510P), and neuronal ERKGoF with synapsin null (syn97; elav-Gal4> ERKSEM). B, EJC amplitude quantification using one-way ANOVA followed by Tukey’s multiple-comparisons test. C, Representative TEVC recordings of EJC traces showing 10 superimposed evoked synaptic responses (1.0 mM Ca2+). From left to right: control (w1118), csw5 null (csw5), ERK heterozygote (ERK1/w1118) and csw5 with ERK heterozygote (csw5; ERK1/w1118). D, Quantification of EJC amplitudes using one-way ANOVA followed by Tukey’s multiple-comparisons test. Scatter plots show all the data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.05 (*), p < 0.01 (**), and p > 0.05 not significant (n.s.).
synapsin null blocks elevated vesicle fusion in PTPN11 and ERK mutants. A, Representative mEJC traces (1.0 mM Ca + 2) from top to bottom: the neuronal driver control (elav-Gal4/w1118), neuronal PTPN11 point mutant (elav-Gal4 > PTPN11Q510P), synapsin null (syn97), neuronal PTPN11Q510P in the synapsin null (syn97; elav-Gal4 > PTPN11Q510P), neuronal ERKGoF in the synapsin null (syn97; elav-Gal4> ERKSEM), genetic background control (w1118), csw5 null (csw5), ERK heterozygous mutant (ERK1/w1118) and ERK heterozygote in the csw5 null (csw5; ERK1/w1118). B, Quantification of mEJC frequencies in all nine genotypes using one-way ANOVAs followed by Tukey’s multiple-comparisons test. C, Quantification of mEJC amplitudes in all 9 genotypes using Kruskal–Wallis followed by Dunn’s multiple-comparisons test (left) or a one-way ANOVA (right). D, Quantification of the evoked neurotransmission quantal content in all nine genotypes using one-way ANOVAs followed by Tukey’s multiple-comparisons tests. Scatter plots show all the data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****), and not significant (n.s.).
Driver control (elav-Gal4/+) and synapsin nulls (syn97) show similar neurotransmission, and neuronally driven PTPN11Q510P and ERKGoF no longer elevate neurotransmission amplitudes in the synapsin null background compared with the neuron-targeted PTPN11 mutant (Fig. 5A). Quantification reveals controls (164.8 ± 14.81 nA, n = 13) are not significantly different from syn97 nulls (141.1 ± 15.41 nA n = 10; p = 0.7542) nor, in this background, neuronal PTPN11Q510P (148.1 ± 13.94 nA n = 10; p = 0.9167) or neuronal ERKGoF (142.7 ± 11.67 nA n = 10; p = 0.7997) when compared via a one-way ANOVA and Tukey’s multiple-comparisons test (F(4,47) = 5.878, p = 0.0006; Fig. 5B). In contrast, neuronal PTPN11Q510P (230.7 ± 16.43 nA n = 9) is highly elevated compared with driver control (p = 0.0188), syn97 nulls (p = 0.0013), and neuronal PTPN11Q510P (p = 0.0035) and neuronal ERKGoF (p = 0.0017) in this background. This genetic suppression suggests heightened MAPK/ERK signaling to elevate synapsin function and increase neurotransmission strength. To further test this idea, we next assayed whether reduced ERK could prevent the csw5 null elevated neurotransmission (Leahy et al., 2023). As reported, csw5 nulls have elevated neurotransmission compared with background controls (w1118; Fig. 5C, left). ERK1 heterozygotes (ERK1/+) appear similar to controls, but csw5 with heterozygous ERK1 no longer have elevated neurotransmission (Fig. 5C, right). Quantification indicates significant differences between genotypes based on a one-way ANOVA (F(3,36) = 6.435, p = 0.0013; Fig. 5D). Compared to the control EJC amplitude (142.9 ± 12.9, n = 10), csw5 nulls (234.6 ± 16.1, n = 8) have significantly elevated EJC amplitudes (p = 0.0013; Fig. 5D). In sharp contrast, there is no significant elevation based on a Tukey’s multiple-comparisons test in either ERK1/+ (156.8 ± 8.84, n = 12; p = 0.9029) or csw5; ERK1/+ (163.3 ± 21.3, n = 10; p = 0.7705; Fig. 5D). There is also no significant difference in neurotransmission between ERK1 and csw5; ERK1/+ (p = 0.9881), but csw5 nulls are significantly elevated compared with csw5; ERK1/+ (p = 0.0155) and ERK1/+ heterozygotes (p = 0.005; Fig. 5D). Thus, reducing ERK genetically in csw nulls effectively restores neurotransmission strength back to normal levels.
We next used mEJC recordings to test spontaneous neurotransmission events (Fig. 6A). Quantification reveals the driver control mEJC frequency (0.76 ± 0.07 Hz, n = 12) is not significantly different from syn97 null mutants (1.12 ± 0.11 Hz, n = 14, p = 0.1076) or, in this background, neuronal PTPN11Q510P (1.11 ± 0.09 Hz, n = 14, p = 0.1273) and ERKGoF (0.98 ± 0.13 Hz, n = 10, p = 0.6468) when compared with one-way ANOVA and Tukey’s multiple-comparisons test (F(4,63) = 8.384, p = 1.733 × 10−5; Fig. 6B, top). In contrast, neuronal PTPN11Q510P (1.51 ± 0.09 Hz n = 18) is elevated compared with the driver control (p = 7.19 × 10−6), syn97 null mutant (p = 0.0297), and, in this background, neuronal PTPN11Q510P (p = 0.0236) and neuronal ERKGoF (p = 0.0039; Fig. 6B, top). Similarly, when analyzed via a one-way ANOVA and Tukey’s multiple-comparisons test (F(3,38) = 11.371, p = 1.83 × 10−5) csw5 nulls (1.83 ± 0.17 Hz, n = 9) have a significantly elevated mEJC frequency compared with the background control (0.81 ± 0.1 Hz, n = 13; p = 1.29 × 10−5), ERK1/+ heterozygote (1.09 ± 0.14 Hz, n = 10; p = 0.0025), and csw5; ERK1/+ double mutant (0.95 ± 0.12 Hz, n = 10; p = 0.0003; Fig. 6B, bottom) In contrast, there is no significant elevation compared with the control frequency of either ERK1/+ (p = 0.4139) or csw5; ERK1/+ (p = 0.8628; Fig. 6B, bottom). In all nine conditions, there is no change in mean mEJC amplitudes compared with their controls (Fig. 6C). Thus, specific changes in spontaneous release frequencies are consistent with the evoked neurotransmission strengths.
When analyzing the stimulation-evoked quantal content, quantification reveals that the elav/+ driver control (187 ± 16.8, n = 13) is not significantly different from the syn97 null mutant (194.6 ± 21.24, n = 10, p = 0.9981; Fig. 6D, left). In this syn97 background, there is also no significant difference in quantal content with neuron-targeted PTPN11Q510P (199.8 ± 18.79, n = 10, p = 0.9859) or ERKGoF (194 ± 15.86 n = 10, p = 0.9986; Fig. 6D, left). In contrast, neuronal PTPN11Q510P alone does elevate quantal content (265.9 ± 18.93 n = 9, p = 0.0279) compared with using a one-way ANOVA followed by Tukey’s multiple-comparisons test (F(4,47) = 2.858, p = 0.0336; Fig. 6D, left). Similarly, the csw5 null mutant quantal content (263.6 ± 18.09, n = 8) is significantly elevated when analyzed via a one-way ANOVA and Tukey’s multiple-comparisons test (F(3,36) = 7.526, p = 0.0005) compared with w1118 background controls (145.1 ± 13.09, n = 10; p = 0.0002), ERK1/+ heterozygotes (187.5 ± 10.57, n = 12; p = 0.017, and csw5; ERK1/+ double mutants (186.1 ± 24.26, n = 10; p = 0.0198; Fig. 6D, right). In sharp contrast, there is no significant change in quantal content in either ERK1/+ heterozygotes (p = 0.2593) or csw5; ERK1/+ double mutants (p = 0.325). Thus, genetically reducing ERK levels in the csw null mutant effectively restores presynaptic release function back toward normal levels. To confirm this neuronal presynaptic mechanism, we next aimed to rescue csw null elevated neurotransmission by reintroducing wild-type csw into motor neurons.
Csw is necessary in both neurons and glia to rescue csw null neurotransmission elevation
To simply confirm the PTPN11/csw role is neuronal as expected, we next tested genetic rescue of the csw null elevated neurotransmission by expressing wild-type csw (cswWT) only in the motor neurons (Johnson Hamlet and Perkins, 2001). We previously showed that ubiquitous cswWT expression fully rescues the csw5 null neurotransmission (Leahy et al., 2023). To drive neuronal cswWT, we utilized vglut-Gal4, a glutamatergic neuronal driver with strong expression in the motor neurons (Mahr and Aberle, 2006). To test this genetic rescue condition, we again employed TEVC electrophysiology recordings to measure stimulation-evoked EJC amplitude. In the neuronal driver control (vglut-Gal4/w1118), motor nerve stimulation drives normal, consistent neurotransmission amplitudes (Fig. 7A, left). To our enormous surprise, csw5 null mutants with neuronally targeted cswWT (csw5; vglut-Gal4 > cswWT) still display strongly elevated mutant neurotransmission with the same clear increase in EJC amplitudes (Fig. 7A, left). Quantified csw5 null with neuronal cswWT neurotransmission amplitudes (206.80 ± 17.61 nA, n = 10) remains obviously elevated compared with neuronal driver controls (149.0 ± 7.79 nA, n = 15), a very significant increase based on a two-sided t test (t(23) = 3.369, p = 0.0027, Fig. 7B, left). This elevated neurotransmission is comparable with the csw null phenotype, indicating that neuronal expression of cswWT has no effect in rescuing the neurotransmission strength phenotype.
Both neuronal and glial Csw are required to rescue csw null neurotransmission. Null csw5 elevated neurotransmission is rescued only with wild-type Csw in both the motor neurons and glia, indicating a requirement in both cell types. A, Representative EJC recordings showing 10 superimposed traces in csw5 null and driver controls, and with motor neuron, glial, or combined driven wild-type Csw (cswWT) to test neurotransmission rescue. From left to right: motor neuron transgenic driver control (vglut-Gal4/w1118) and neuronal rescue (csw5; vglut-Gal4 > cswWT), glial transgenic driver control (repo-Gal4/w1118) and glial rescue (csw5; repo-Gal4 > cswWT), and the dual neuron + glial driver control (vglut-Gal4; repo-Gal4/w1118) and the dual combined cell type rescue (csw5; vglut-Gal4;repo-Gal4 > cswWT). B, Quantification of the mean EJC amplitudes in all six conditions with two-sided t tests. Scatter plots show all the data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.01 (**) and p > 0.05 not significant (n.s.).
The NMJ is a tripartite synapse consisting of the presynaptic neuron, postsynaptic muscle, and perisynaptic glia (Brink et al., 2012; Calderon et al., 2022). In our previous work, we found the postsynaptic muscle is unaffected in csw mutants (Leahy et al., 2023). Additionally, glial mutations are known to lead to behavioral defects in Drosophila (Petley-Ragan et al., 2016). Therefore, we tried to reintroduce cswWT in glia to rescue neurotransmission. For this test, we utilized the glial driver repo-Gal4, which is expressed in all glial cells (Ogienko et al., 2020; Vita et al., 2021). As above for motor neurons, csw5 null mutants with glial-targeted cswWT (csw5; repo-Gal4 > cswWT) still display highly elevated neurotransmission with a clear increase in EJC amplitude compared with the driver control (Fig. 7A, middle). The EJC amplitudes of csw5 null with glial cswWT (207.70 ± 8.47 nA, n = 11) remain elevated compared with glial driver control (175.20 ± 7.48 nA, n = 14), a significant elevation based on an unpaired t test (t(23) = 2.879, p = 0.0085; Fig. 7B, middle). As a consequence of these unexpected results, we next tried reintroducing cswWT in both motor neurons and glia in the csw null mutant. Finally, csw5 nulls with neuronal and glial cswWT together (csw5; vglut-Gal4; repo-Gal4 > cswWT) show fully rescued transmission compared with the dual driver control (vglut-Gal4; repo-Gal4/w1118; Fig. 7A, right). When quantified, csw5 nulls with neuronal and glial cswWT (148.50 ± 11.52 nA, n = 10) are indistinguishable from the dual driver transgenic control (145.30 ± 11.04 nA, n = 10), with no significant difference based on a two-sided t test (t(18) = 0.2039, p = 0.8407; Fig 7B, right). Taken together, these results indicate a requirement for Csw in both neurons and glia to control neurotransmission strength.
We next used mEJC recordings to test spontaneous neurotransmission events (Fig. 8A). In csw5 null mutants, neither neuronal nor glial wild-type csw expression (cswWT) can rescue the elevated mEJC frequency (Fig. 8A, top). Compared to control frequency (1.02 ± 0.14 Hz, n = 11), mEJCs remain elevated with neuronal cswWT (1.97 ± 0.16 Hz, n = 11), a significant two-sided t test increase (t(20) = 4.612, p = 0.0002; Fig. 8B, left). Likewise, relative to the glial driver control (0.74 ± 0.1 Hz, n = 11), frequency is still heightened with glial cswWT (1.47 ± 0.43 Hz, n = 8), a significant increase based on a two-sided t test (t(17) = 4.194, p = 0.0006; Fig. 8B, middle). However, neuronal and glial cswWT together show full rescue of the csw5 null elevated mEJC frequency (Fig. 8A, bottom). When quantified, csw5 nulls with neuronal and glial cswWT (0.61 ± 0.09 Hz, n = 11) were indistinguishable from dual driver controls (0.77 ± 0.12 Hz, n = 10), with no significant difference based on a two-sided t test (t(19) = 1.105, p = 0.2829, Fig. 8B, right). In all six conditions, there is no change in mean mEJC amplitudes (Fig. 8C). When analyzing quantal content with two-sided t tests, csw5 null mutants with neuronal cswWT (199.28 ± 16.97, n = 10) and glial cswWT (239.46 ± 9.76, n = 11) are significantly elevated from their respective neuronal (144.48 ± 7.56, n = 15; t(23) = 3.306, p = 0.0031; Fig. 8D, left) and glial driver controls (177.06 ± 7.56, n = 14; t(23) = 5.141, p = 3.29 × 10−5; Fig. 8D, middle). This shows that neither neuronal nor glial csw expression alone can rescue the neurotransmission defect. In sharp contrast, the quantal content of csw5 null mutants with dual neuronal and glial cswWT (172.03 ± 13.34, n = 10) is rescued compared with the dual driver controls (158.03 ± 12.01, n = 10), with no significant change remaining based on a two-sided t test (t(18) = 0.7798, p = 0.4456, Fig. 8D, right). The results indicate both neuronal and glial function is needed. We next tested whether glial-targeted PTPN11/csw mutations alone cause neurotransmission defects.
Both neuronal and glial Csw are required to rescue spontaneous transmission. Null csw5 elevated spontaneous neurotransmission frequency is only rescued with wild-type Csw in both motor neurons and glia, indicating a dual requirement. A, Representative mEJC traces (1.0 mM Ca+2) in csw5 null and driver controls and with motor neuron, glial, or combined wild-type Csw (cswWT). From top to bottom: motor neuron driver control (vglut-Gal4/w1118), neuronal rescue (csw5; vglut-Gal4 > cswWT), glial driver control (repo-Gal4/w1118), glial rescue (csw5; repo-Gal4 > cswWT), neuron and glial driver control (vglut-Gal4; repo-Gal4/w1118), and combined cell type rescue (csw5; vglut-Gal4; repo-Gal4 > cswWT). B, Quantification of the mEJC frequency in all six conditions using two-sided t tests. C, Quantification of mEJC amplitude in all six conditions using two-sided t tests. D, Quantification of the quantal content in all six conditions using two-sided t tests. Scatter plots show all the data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.01 (**), p < 0.001 (***), p < 0.0001 (****), and not significant (n.s.).
Glial-targeted PTPN11/csw LoF and GoF independently increase neurotransmission strength
To test the unexpected glial function for PTPN11/csw, we first drove csw RNAi in glia to assay synaptic transmission strength (Perkins et al., 2015). To further test the PTPN11 glial role, we also targeted the GoF (PTPN11N308D) NS mutation and LoF (PTPN11Q510P) NSML mutation exclusively to glia (Das et al., 2021). We had previously established each of these three conditions causes elevated neurotransmission strength when targeted exclusively to presynaptic neurons (Leahy et al., 2023). Compared to glial driver controls (repo-Gal4/w1118), both glial-targeted PTPN11 GoF (repo-Gal4 > PTPN11N308D) and PTPN11 LoF (repo-Gal4 > PTPN11Q510P) conditions strongly elevate neurotransmission strength (Fig. 9A, left). Quantification indicates significant differences between genotypes based on a one-way ANOVA (F(2,34) = 7.502, p = 0.002; Fig. 9B, left). Compared to the glial driver control EJC amplitude (170.20 ± 9.39 nA, n = 15), both PTPN11N308D (221.0 ± 10.32 nA, n = 11, p = 0.006) and PTPN11Q510P (219.20 ± 13.51 nA, n = 11, p = 0.008) show a very significant elevation based on a Tukey’s multiple-comparisons test (Fig. 9B, left). There is no significant difference in EJC amplitudes between GoF and LoF conditions (p = 0.994, Fig. 9B). Similarly, glial-targeted csw knockdown (repo-Gal4 > csw RNAi) causes clearly elevated evoked neurotransmission compared with the glial driver control (repo-Gal4/TRiP BDSC 36303 control; Fig. 9A, right). Quantification shows the glial-targeted csw RNAi amplitude (226.9 ± 11.3 nA, n = 11) is elevated compared with the glial driver control (177.60 ± 9.59 nA, n = 12), with a significant increase based on a two-sided t test (t(21) = 3.343, p = 0.0031; Fig. 9B, right). These results show glial-targeted PTPN11/csw LoF and GoF mutations drive elevated evoked neurotransmission strength, as we previously established also occurs with neuronal manipulations (Leahy et al., 2023).
Glial-targeted PTPN11/csw loss- and gain-of-function elevate neurotransmission. Glial human PTPN11 mutations from NS/NSML disease states and csw knockdown, all increase synaptic function, indicating that glia-specific changes are sufficient to elevate neurotransmission. A, Representative EJC recordings showing 10 superimposed traces for each of the conditions. From left to right: the glial transgenic driver control (repo-Gal4/w1118), the glial-driven PTPN11 patient–derived point mutations PTPN11N308D (repo-Gal4 > PTPN11N308D) and PTPN11Q510P (repo-Gal4 > PTPN11Q510P), glial driver RNAi background control (repo-Gal4/TRiP Ctl), and glial-driven csw RNAi (repo-Gal4 > csw RNAi). B, Quantification of the EJC amplitudes in all five conditions, using one-way ANOVA followed by Tukey’s multiple-comparisons test (PTPN11) and two-sided t tests (RNAi). C, Representative mEJC traces (1.0 mM Ca + 2) from top to bottom: glial driver control (repo-Gal4/w1118), and glial-driven PTPN11N308D, PTPN11Q510P, and csw RNAi. D, Quantification of mEJC frequency in all five conditions; one-way ANOVA followed by Tukey’s multiple-comparisons test (PTPN11) and two-sided t tests (RNAi). E, Quantification of mEJC amplitudes, using one-way ANOVA (PTPN11) and two-sided t tests (RNAi). F, Quantification of the quantal content in all five conditions, using one-way ANOVA followed by Tukey’s multiple-comparisons test (PTPN11) and two-sided t tests (RNAi). Scatter plots show all data points with mean ± SEM. Data points: NMJ number. Significance: p < 0.05 (*), p < 0.01 (**), p < 0.001 (***), and not significant (n.s.).
We finally wanted to test whether the glial-mediated increase in neurotransmission is due to higher presynaptic vesicle release probability. Spontaneous mEJC recordings reveal clearly elevated event frequencies in all of the above glial-targeted PTPN11/csw mutations (Fig. 9C). Compared to glial repo-Gal4/+ driver control frequency (1.13 ± 0.16 Hz, n = 17), there is an increase with glial PTPN11N308D (2.14 ± 0.34 Hz, n = 15, p = 0.0104) and glial PTPN11Q510P (2.02 ± 0.19 Hz, n = 17, p = 0.0207), significant elevations based on a one-way ANOVA followed by Tukey’s multiple-comparisons test (F(2,46) = 5.771, p = 0.0058; Fig. 9D, left). This increase in mEJC frequency also occurs with glial csw RNAi (1.57 ± 0.29 Hz, n = 10) compared with glial driver controls (0.92 ± 0.14 Hz, n = 12), significantly elevated with a two-sided t test (t(20) = 2.117, p = 0.047; Fig. 9D, right). In contrast, there is no significant change in mEJC amplitudes compared with controls (0.86 ± 0.06 nA, n = 17) in glial PTPN11N308D (0.82 ± 0.04 nA, n = 15) or PTPN11Q510P (0.88 ± 0.03 nA, n = 17) analyzed via a one-way ANOVA (F(2,46) = 0.4970, p = 0.6116; Fig. 9E, left) or the glial driver control (0.80 ± 0.06 nA, n = 12) versus csw RNAi (0.88 ± 0.07 nA, n = 10) compared with a two-sided t test (t(20) = 0.8183, p = 0.4228; Fig. 9E, right). Moreover, compared with the quantal content in the glial driver control (198.3 ± 10.94, n = 15), the number of stimulus-evoked quanta is elevated with both glial-targeted PTPN11N308D (269.4 ± 12.58, n = 11, p = 0.0009) and PTPN11Q510P (248.3 ± 15.3, n = 11, p = 0.0217) analyzed with a one-way ANOVA (F(2,34) = 8.746, p = 0.0009; Fig. 9F, left). This increase in quantal content also occurs with csw RNAi (258.91 ± 12.9, n = 11) compared with control (221.35 ± 11.95, n = 12) based on a two-sided t test (t(21) = 2.139, p = 0.0443; Fig. 9F, right). Taken together, we conclude that PTPN11/csw regulates presynaptic neurotransmission via dual roles in both neurons and glia, with targeted LoF/GoF mutations in either cell type sufficient to cause strongly elevated neurotransmission, but a necessary requirement for function in both neurons and glia.
Discussion
Csw regulates MAPK/ERK signaling in critical neuronal functions, such as the proactive interference modulation of learning and memory (Zhao et al., 2023). Mutation of the human homolog PTPN11 causes NS through gain-of-function (GoF) and NS with Multiple Lentigines (NSML) through loss-of-function (LoF), with striking cognitive impairments (El Bouchikhi et al., 2016). Both directions elevate MAPK/ERK signaling to cause similar disease symptoms (Das et al., 2021; Wingbermühle et al., 2022). Similarly, we discovered Drosophila NS/NSML models both elevate presynaptic MAPK/ERK signaling to heighten neurotransmission and reduce synaptic depression (Leahy et al., 2023). These results suggested a presynaptic mechanism of vesicle recruitment. To test this hypothesis, we began by analyzing synapsin in the strongest neurotransmission condition: csw nulls (Fig. 1). Synapsin phosphoprotein tethers regulate neurotransmission strength by restraining vesicles in the reserve pool (Akbergenova and Bykhovskaia, 2007, 2010), which replenishes the RRP upon sustained exocytosis (Zhang and Augustine, 2021). In mammals, synapsin is encoded by three genes, whereas Drosophila has one single homolog (Mirza and Zahid, 2018). In mouse glutamatergic neurons, synapsin triple knock-out increases synaptic depression rate (Gitler et al., 2004), and uniquely synapsin IIA rescues this phenotype (Gitler et al., 2008). Likewise, synapsin IA/B overexpression increases the short-term depression rate, indicating that synapsin dysregulation in either direction can increase depression (Vasileva et al., 2013). In Drosophila csw null mutants, synapsin levels and SV colocalization are elevated at rest and strongly depleted with acute stimulation (Fig. 1). This is presumed to represent cytosolic dispersion, although we cannot verify this hypothesis with immunocytochemical imaging.
With stimulation, synapsin is phosphorylated by several kinases, including MAPK/ERK, to cause SV disassociation and enable RRP recruitment (Rizzoli and Betz, 2005; Akbergenova and Bykhovskaia, 2010). With stimulation, csw nulls display dramatic loss of anti-synapsin signal and colocalization with SV markers, whereas matched genetic controls do not change (Fig. 1). This suggests that the absence of Csw causes aberrant synapsin disassociation from SV. The increased activity-dependent recruitment of synapsin-bound vesicles in csw nulls (Fig. 1) accounts for sustained resistance to synaptic depression (Leahy et al., 2023). Consistent with csw results, Drosophila synapsin nulls have reduced facilitation (Blanco-Redondo et al., 2019). A transgenic mouse model with increased ERK-dependent synapsin I phosphorylation displays not only an increased frequency of synaptic vesicle fusion events but also increased paired-pulse facilitation (Kushner et al., 2005). However, mouse synapsin triple knockouts also show reduced short-term synaptic plasticity (Cheng et al., 2018). Thus, there is good overlap between Drosophila and mouse results regarding the ERK-dependent regulation of synapsin controlling presynaptic vesicle fusion probability, but a difference in the consequences affecting synaptic plasticity, which is likely due to the multiple synapsin isoform interactions in mammals (Kushner et al., 2005; Cheng et al., 2018; Leahy et al., 2023). The very striking changes in synapsin dynamics in csw null mutants (Fig. 1) suggest altered presynaptic vesicle pools.
We tested synaptic vesicle pools in csw null mutants by imaging synaptic ultrastructure with transmission electron microscopy (Kopke et al., 2017; Bhimreddy et al., 2021). The csw nulls display normal synapse architecture, including presynaptic bouton and postsynaptic SSR (Fig. 2, Table 1). Overall, synaptic bouton and vesicle morphology are all unchanged, with SV distribution specifically disrupted. SVs cluster closer together, possibly accounting for the elevation in the SV marker vglut under the same conditions (Fig. 1). This is in agreement with the constitutively active ERK H-rasG12V mouse model, which also displays no defects in gross synaptic ultrastructure (Kushner et al., 2005). Likewise, synapsin overexpression does not alter gross SV number or density, but does change the SV distribution (Vasileva et al., 2013). In analyzing SV distribution relative to presynaptic active zones, we discovered a decrease at 200–400 nm removed from the t-bar (Fig. 2). This alteration in vesicle density associated with the reserve pool is consistent with direct synapsin involvement. In Drosophila synapsin nulls, SV are spread toward the bouton interior, with reduced vesicle density in the center of the bouton (Akbergenova and Bykhovskaia, 2010). Docked vesicles within ½ of a vesicle diameter (<20 nm) of the AZ membrane are elevated in csw null mutants (Fig. 2). This is consistent with an increase in the RRP (Hur et al., 2018), and we measured increases in the functional RRP previously via electrophysiology in csw nulls (Leahy et al., 2023). Likewise, the constitutively active ERK H-rasG12V mouse model displays an increase in docked SV (Kushner et al., 2005). These results reveal altered MAPK/ERK-dependent presynaptic vesicle pool distributions that provide an explanation for the csw null mutants’ neurotransmission strength elevation.
To confirm this presynaptic mechanism in NS and NSML disease models, we next tested synapsin in NS (PTPN11N308D, GoF) and NSML (PTPN11Q510P, LoF) conditions (Das et al., 2021; Wingbermühle et al., 2022). PTPN11N308D occurs in ∼25% of NS patients (Pierpont et al., 2009), and PTPN11Q510P confers NSML symptoms (Das et al., 2021). Like csw nulls, both patient–derived mutations increase synapsin and SV association (Fig. 3). With stimulation, synapsin-bound vesicles are aberrantly recruited, providing the basis for sustained synaptic depression resistance (Leahy et al., 2023). Inhibition of MAPK/ERK signaling alleviates both NS and NSML disease model phenotypes (Lee et al., 2014; Das et al., 2021; Leahy et al., 2023). However, this pathway is highly regulated, and intersecting pathways can affect function (Guo et al., 2020). Therefore, to confirm the specificity for neurotransmission, we tested targeted ERK gain-of-function (ERKGoF). This constitutively active ERK mutant at the endpoint of the MAPK pathway allows for the most specific analysis of MAPK signaling on neurotransmission (Ariss et al., 2020). Constitutively active ERK recapitulates the elevated presynaptic pERK levels and strengthened neurotransmission in the PTPN11/csw mutants (Fig. 4), to place activated ERK on site for presynaptic regulation of vesicle dynamics (Giachello et al., 2010; Leahy et al., 2023). To further confirm this mechanism, we targeted neuronal PTPN11 and ERKGoF mutations in a synapsin null background to find a block of the elevated neurotransmission (Figs. 5, 6). This is internally consistent, but unexpected due to synapsin loss not affecting basal neurotransmission strength (Sun et al., 2006). However, unlike synapsin knock-out, NS/NSML mutations likely affect a wide range of differential pathways throughout development, so other mechanisms may also be going awry. To finally confirm this presynaptic mechanism, we turned to genetic rescue by expressing wild-type Csw within glutamatergic neurons in an otherwise global csw null mutant.
Cell-targeted genetic rescue is the gold-standard for demonstrating cellular requirements (Lutz et al., 1996). Based on our previous work and the above results, we were confident that driving wild-type Csw in motor neurons (cswWT) would demonstrate a neuron-specific requirement. To our astonishment, neuron expression fails to provide any rescue whatsoever of the csw null elevated neurotransmission strength (Figs. 7, 8). Our previous work showed no postsynaptic muscle role for Csw (Leahy et al., 2023), so we next examined the third cell type at the tripartite NMJ synapse; glia (Brink et al., 2012). However, glial cswWT expression also fails to provide any rescue of the csw null elevated neurotransmission strength (Figs. 7, 8). We therefore next turned to drive cswWT in motor neurons and glia simultaneously, to find this fully rescues the csw null elevated neurotransmission strength (Figs. 7, 8). Thus, Csw is needed in both neurons and glia together to regulate neurotransmission amplitude. However, we previously found that neuron-targeted PTPN11/csw LoF and GoF transgenes are both sufficient to replicate csw null mutant elevated neurotransmission (Leahy et al., 2023). Likewise, we find glial-targeted transgenes also sufficient to elevate synaptic strength (Fig. 9). These results show that PTPN11/csw mutations in either glia or neurons alone cause the defect but that PTPN11/csw function in both cell types is necessary to properly regulate neurotransmission. PTPN11/csw has known glial roles for the regulation of neurogenesis and gliogenesis (Gauthier et al., 2007; Zhu et al., 2018) and in response to injury (Servidei et al., 1997; Kim et al., 2003; Logan et al., 2012), but this is the first work to reveal a glial role for PTPN11/csw in neurotransmission.
Glial roles in regulating neurotransmission have been previously discovered, such as the Repo transcription factor regulation of glutamate neurotransmitter cycling (Mazaud et al., 2019). However, based on both spontaneous and quantal content analyses, glial PTPN11/Csw acts in the regulation of glutamatergic signaling via a common pathway with the neuronal presynaptic MAPK/ERK-dependent regulation of synapsin. There are three characterized glial subclasses at the NMJ; perineurial (PG), subperineurial (SPG), and wrapping glia (WG; Brink et al., 2012). Each glial subclass has functions at the NMJ that could contribute to the elevation of neurotransmission strength. Based on accumulating current evidence, it is likely that PTPN11/csw functions within the glial subclasses with the most active roles in modulating neurotransmission (SPG or WG), and future experiments will focus on elucidating the exact glial subclass(es) involved. In this study, we establish that PTPN11/csw acts both within motor neurons and glia, with LoF/GoF similarly upregulating MAPK/ERK signaling to control synapsin-mediated synaptic vesicle trafficking in an activity-dependent mechanism. Future interventions to improve cognitive outcomes in NS and NSML patients can utilize these findings by targeting cojoined glial and neuronal mechanisms regulating presynaptic vesicle trafficking mechanisms.
Footnotes
We thank Dr. Tirtha Das (Icahn School of Medicine at Mount Sinai, NY, USA) for providing the transgenic UAS-PTPN11 lines. We thank the Bloomington Drosophila Stock Center (Indiana University, Bloomington, IN, USA) for the genetic lines, and the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA, USA) for the antibodies. We thank the Broadie Lab members for their constant input throughout the course of this study. This work was supported by National Institutes of Health Grants MH084989 and NS131557 to K.B.
The authors declare no competing financial interests.
- Correspondence should be addressed to Kendal Broadie at kendal.broadie{at}vanderbilt.edu.
