Abstract
Destabilization of neural activity caused by failures of homeostatic regulation has been hypothesized to drive the progression of Alzheimer's Disease (AD). However, the underpinning mechanisms that connect synaptic homeostasis and the disease etiology are yet to be fully understood. Here, we demonstrated that neuronal overexpression of amyloid β (Aβ) causes abnormal histone acetylation in peripheral glia and completely blocks presynaptic homeostatic potentiation (PHP) at the neuromuscular junction in Drosophila. The synaptic deficits caused by Aβ overexpression in motoneurons are associated with motor function impairment at the adult stage. Moreover, we found that a sphingosine analog drug, Fingolimod, ameliorates synaptic homeostatic plasticity impairment, abnormal glial histone acetylation, and motor behavior defects in the Aβ models. We further demonstrated that perineurial glial sphingosine kinase 2 (Sk2) is not only required for PHP, but also plays a beneficial role in modulating PHP in the Aβ models. Glial overexpression of Sk2 rescues PHP, glial histone acetylation, and motor function deficits that are associated with Aβ in Drosophila. Finally, we showed that glial overexpression of Sk2 restores PHP and glial histone acetylation in a genetic loss-of-function mutant of the Spt-Ada-Gcn5 Acetyltransferase complex, strongly suggesting that Sk2 modulates PHP through epigenetic regulation. Both male and female animals were used in the experiments and analyses in this study. Collectively, we provided genetic evidence demonstrating that abnormal glial epigenetic alterations in Aβ models in Drosophila are associated with the impairment of PHP and that the sphingosine signaling pathway displays protective activities in stabilizing synaptic physiology.
SIGNIFICANCE STATEMENT Fingolimod, an oral drug to treat multiple sclerosis, is phosphorylated by sphingosine kinases to generate its active form. It is known that Fingolimod enhances the cognitive function in mouse models of Alzheimer's disease (AD), but the role of sphingosine kinases in AD is not clear. We bridge this knowledge gap by demonstrating the relationship between impaired homeostatic plasticity and AD. We show that sphingosine kinase 2 (Sk2) in glial cells is necessary for homeostatic plasticity and that glial Sk2-mediated epigenetic signaling has a protective role in synapse stabilization. Our findings demonstrate the potential of the glial sphingosine signaling as a key player in glia–neuron interactions during homeostatic plasticity, suggesting it could be a promising target for sustaining synaptic function in AD.
- Alzheimer's disease
- epigenetic regulation
- glia
- histone acetylation
- presynaptic homeostatic plasticity
- sphingosine kinase
Introduction
Alzheimer's disease (AD) is among the most debilitating neurodegenerative conditions, yet its complex etiology remains to be elucidated (Knopman et al., 2021). Homeostatic regulation serves as a protective mechanism providing stability to neural activity and maintaining overall brain functionality (Davis, 2006; Turrigiano, 2008). Impaired homeostatic regulation may lead to synaptic and neural network instability, a common feature linked to various neurologic disorders (Wondolowski and Dickman, 2013; Bourgeron, 2015; Styr and Slutsky, 2018). While a compromised homeostatic regulation could potentially induce unbalanced excitation and inhibition, leading to neurodegeneration in AD, the underlying molecular and cellular mechanisms associating homeostatic plasticity with disease pathology remain largely undefined (Busche and Konnerth, 2016).
The homeostatic modulation of presynaptic neurotransmitter release, termed presynaptic homeostatic potentiation (PHP), is a fundamental form of neural regulation (Davis and Müller, 2015). PHP can be induced rapidly by pharmacological inhibition of postsynaptic glutamate receptors (Frank et al., 2006). During this process, a compensatory increase of presynaptic neurotransmitter release precisely counteracts the perturbation of glutamate receptors, thereby stabilizing postsynaptic excitation at baseline levels (Fig. 1A). PHP can also be sustained chronically when postsynaptic glutamate receptors are genetically deleted (Petersen et al., 1997). As a highly conserved form of homeostatic plasticity, PHP is observed across different species, including Drosophila, mice, and humans, and functions in both the CNS and peripheral nervous system (Cull-Candy et al., 1980; Wang et al., 2014, 2018; Delvendahl et al., 2019; Chipman et al., 2022).
Glia play an instrumental role in neurodevelopment, synaptic transmission, and plasticity, and are closely associated with the pathology of neurodegenerative disorders (Chung et al., 2015; Keren-Shaul et al., 2017; Stogsdill and Eroglu, 2017; Habib et al., 2020). Both astrocytes and microglia have been observed to exert significant signaling control over neuronal function in AD (Hong et al., 2016; McAlpine et al., 2021; Shah et al., 2022). However, our understanding of glial function in synaptic homeostatic plasticity under both physiological and pathologic conditions is still in its infancy (Stellwagen and Malenka, 2006; Wang et al., 2020). In addition, while the importance of epigenetic regulation in neurons for Hebbian and homeostatic plasticity has been established, how the glial epigenome contributes to maintaining brain function in the presence of perturbations remains largely unknown (Vecsey et al., 2007; Guan et al., 2009; Benevento et al., 2016; Wang et al., 2020).
Sphingosine-1-phosphate (S1P) is a signaling lipid produced through the phosphorylation of sphingosine by sphingosine kinases (Alemany et al., 2007). These kinases are essential for regulating neural development, synaptic physiology, and synaptic plasticity (Mizugishi et al., 2005; Kanno et al., 2010; Chan et al., 2012; Hait et al., 2014). Previous studies have shown that disruption of sphingosine kinases leads to impaired hippocampal long-term potentiation and compromised learning memory, suggesting a crucial role of sphingosine signaling in regulating brain function (Kanno et al., 2010; Hait et al., 2014). Moreover, there is an increasing body of evidence suggesting the involvement of the sphingosine signaling pathway in neurologic disorders, including schizophrenia, anxiety, and neurodegeneration (Panayotis et al., 2018; Grassi et al., 2019; Lei et al., 2019; Esaki et al., 2020). Fingolimod, a synthetic sphingosine analog and an oral drug used to treat multiple sclerosis, has been suggested to alleviate memory loss in AD mouse models (Angelopoulou and Piperi, 2019; Kartalou et al., 2020; Fagan et al., 2022). However, the underlying mechanisms through which Fingolimod offers protective effects in AD remain unclear.
In this study, we examined the roles of homeostatic plasticity, sphingosine kinases, and glial epigenetic signaling in the stabilization of synaptic physiology using Drosophila models of AD. We found a complete disruption of PHP and a dramatic reduction in glial histone acetylation at the lysine 9 site (H3K9ac) in Drosophila amyloid β (Aβ) models. We identified that sphingosine kinase 2 (Sk2) functions within glia for both the rapid induction and long-term maintenance of PHP. Remarkably, we found that either glial overexpression of Sk2 or treatment with Fingolimod restores H3K9ac levels, PHP, and motor function in Drosophila Aβ models. Additionally, we revealed that Sk2 overexpression rescues PHP and glial H3K9ac deficits in a Spt-Ada-Gcn5 acetyltransferase (SAGA) complex mutant. Overall, our findings provide compelling genetic evidence that glial sphingosine-mediated epigenetic regulation is crucial for the stabilization of synaptic physiology in Drosophila.
Materials and Methods
Drosophila melanogaster strains and husbandry.
Drosophila stocks were raised at room temperature on standard molasses food. Drosophila alleles used for experiments were raised at 25°C. For motoneuron-specific Gal4 expression we used OK371-Gal4 on the second chromosome. For muscle-specific expression, we used BG57-Gal4. For glial specific expression, we used perineurial glial-Gal4 (NP6293-Gal4; Stork et al., 2012) drivers as indicated in figure legends. For eye-specific expression, we used GMR-Gal4 (BL27930). Unless otherwise noted, the w1118 strain was used as the wild-type (wt) control. All genetic mutant, transgenic, and RNAi alleles are from the Bloomington Drosophila Stock Center or the Vienna Drosophila Resource Center. Stock numbers are listed as follows: UAS-human-Aβ42 (BL33769), UAS-human-Aβ42Ar (BL33773), UAS-human-APPAr, Sw (BL33794), UAS-human-MAPT (BL64389), Sk2 (BL14133), UAS-Sk2 RNAi #1 (v101018), UAS-Sk2 RNAi #2 (BL35570), UAS-Redstinger.nls (BL8546). For pan-neuronal expression using the Q system, we used nSyb-QF2 (BL51960) and QUAS-Aβ42 (BL83348). Sk1 mutant, UAS-Sk1, and UAS-Sk2 transgenic alleles (Yonamine et al., 2011) are provided by Jairaj Acharya [National Institutes of Health (NIH)/Neuroscience of Interoception and Chemosensation Study Section)]. Ada2b mutant allele (Qi et al., 2004) is provided by Jerry Workman (Stowers Institute, Kansas City, MO). NP6293-Gal4 is provided by Marc Freeman (Vollum Institute, Portland, OR). For Fingolimod hydrochloride (FTY720; catalog #SML0700, Sigma-Aldrich) treatment, eggs were laid on molasses food supplemented with or without FTY720 and were maintained at 25°C; third-instar larvae or adult flies with the right genotypes were used for electrophysiological, imaging, and behavior experiments.
Electrophysiology.
Sharp-electrode recordings were made from muscle 6 at abdominal segments 2 and 3 in third-instar larvae using an Axoclamp 900A Amplifier (Molecular Devices). HL3 saline was used as follows (in mm): 70 NaCl, 5 KCl, 10 MgCl2, 10 NaHCO3, 115 sucrose, 5 trehalose, 5 HEPES, and 0.3 CaCl2. EPSP and miniature EPSP (mEPSP) traces were analyzed in Stimfit (version 3.7.12, Python) with previously published routines and MiniAnalysis (6.0.3, Synaptosoft). For the rapid induction of synaptic homeostasis, larvae were incubated in 20 μm Philanthotoxin-433 in an unstretched, partially dissected preparation (PhTX; catalog #AOB0876, Aobious) for 10 min (Frank et al., 2006). For each neuromuscular junction (NMJ), the average amplitudes of evoked EPSPs are based on the mean peak amplitudes in response to 20–30 individual stimuli. Spontaneous mEPSPs were recorded continuously for 60–90 s. Quantal content (QC) was estimated for each NMJ as the ratio of EPSP amplitude/mEPSP amplitude. The mean value across all NMJs for a given genotype is reported. Two-electrode voltage-clamp recordings was performed as previously described (Wang et al., 2016). Briefly, after dissection in Ca2+-free HL3, larval preparations were then switched to HL3 solution with 1.5 mm Ca2+ before electrophysiological recording. Muscles were clamped to −65 mV in a two-electrode voltage configuration for EPSC recordings using the Axoclamp 900A Amplifier. Apparent quantal content was estimated for each NMJ as the ratio of EPSC amplitude/mEPSP amplitude. For each NMJ, the average amplitude of evoked EPSCs is based on the mean peak amplitudes in response to 20 individual stimuli. All statistical analyses were performed in GraphPad Prism (version 9.4.1).
Immunohistochemistry.
Standard immunocytochemistry was performed as previously described (Wang et al., 2014). Briefly for total immunostaining: dissected third-instar larvae were fixed with 4% PFA in PBS for 15 min at room temperature and incubated with primary antibody diluted in PBST (PBS with 0.5% Triton X-100) overnight at 4°C after six brief washes with PBST. For H3K9ac staining, samples were incubated in blocking solution (PBST with 5% normal goat serum) for 2 h at room temperature before primary antibody incubation. Larvae were then incubated with secondary antibody diluted in PBST for 2 h at room temperature and mounted in VECTASHIELD Antifade Mounting Medium with or without DAPI (Vector Laboratories) after six brief washes with PBST. The following antibodies were used: mouse anti-Bruchpilot (Brp; 1:100; catalog #Nc82, Developmental Studies Hybridoma Bank); rabbit anti-Discs large (Dlg; 1:1000; Wang et al., 2014); rabbit anti-H3 acetyl K9 (1:500; ChIP grade; catalog #ab10812, Abcam); goat anti-mouse Alexa Fluor 488 (1:300; catalog #A11034, Thermo Fisher Scientific); goat anti-mouse Alexa Fluor 555 (1:300; catalog #A21422, Thermo Fisher Scientific); goat anti-rabbit Alexa Fluor cy3 (1:300; catalog #A10520, Thermo Fisher Scientific); and Alexa Fluor 647 conjugated goat anti-HRP (1:500; catalog #123-605-021, The Jackson Laboratory).
Image acquisition and analysis.
Confocal imaging for synapse morphology and H3K9ac was performed on a laser-scanning confocal microscope (model LSM 880, Carl Zeiss) with a 63× objective [Plan-Apochromat 63×/1.4 Oil DIC (differential interference contrast) M27] at 1200 × 1200 pixels. Z-stacks of the NMJ on muscle 6/7 were acquired and maximum projections were used for analyses. Confocal images for H3K9ac in Sk2 overexpression in Aβ models were acquired on a laser-scanning confocal microscope (model FluoView 3000, Olympus) with a LUMPLFLN 40×/0.8 water objective. Samples were scanned with 405/561/640 diode lasers in line sequential mode, and signals were collected with HSD Detectors. Images for eye degeneration were acquired on a fluorescence stereoscope (model SZX16, Olympus) with an SDF PLAPO 0.8×/0.12 objective (Olympus). Maximum projections of confocal images were used for analyses. Quantifications of Brp and bouton number were performed as previously described with the Watershed Segmentation plugin in Fiji software (NIH; Wang et al., 2014). For H3K9ac average fluorescence intensity analyses, a universal intensity threshold mask was applied to H3K9ac channel to select nuclei, and non-glial cell nuclei were manually excluded from the selection. Particle analysis was then performed on the channel and the average intensity of each nucleus was recorded in Fiji software (NIH). The mean of the average fluorescence intensity of each nucleus quantified was reported.
Climbing assay.
For climbing behavior experiments, 8–10 adult flies (23–25 d posteclosion) were inserted into a glass cylinder that was capped with Parafilm with small holes. Flies habituated in the cylinder for 5 min before experiments. The flies were then tapped to the bottom and were recorded using a commodity camera for 1 min. Both male and female flies were used.
Eye degeneration analysis.
The level of degeneration was quantified based on the size and number of degenerating retinal areas, as follows: 0, no degeneration; 1, few subtle rough areas; 2, ≤2 small rough/white patches (the size of each path is <10% of the total area of a single eye); 3, one large rough/white patch (area for each patch is >10% of the total area) or three small rough/white patches; and 4, ≥2 large rough/white patches or ≥4 small rough/white patches. Both eyes of each animal and both male and female animals are included in the analysis.
Protein structure alignment.
The protein structures used for alignment were as follows: human sphingosine kinase 1 (SPHK1) protein structure (3VZB, X-ray, 2.00Å) and AlphaFold (Jumper et al., 2021; Mirdita et al., 2022) predicted structures for human SPHK1 (AF-Q9NYA1-F1), human SPHK2 (AF-Q9NRA0-F1), and Drosophila Sk2 (AF-Q9VZW0-F1). The pair-wise alignment, comparison, and presentation of protein structures were performed using pair-wise structure alignment function (jFATCAT rigid) in RCSB PDB.
Single-cell RNA-sequencing and single-nucleus RNA-sequencing data analysis.
Drosophila single-cell RNA-sequencing (scRNA-seq) dataset was published in GSE107451 (Gene Expression Omnibus, NCBI; Davie et al., 2018). Cell clusters (57,000 cells, scenic_tsne1 and scenic_tsne2) are plotted as in the published study. For Sk1 and Sk2 expression in perineurial glia, we plotted Perineurial_glia cell with scenic_tsne1 between −26.5 and −21.5, and scenic_tsne2 between 10 and 14. A published single-nucleus RNA-sequencing (snRNA-seq) dataset from human cortex was used for SPHK1 and SPHK2 glial expression analysis (Hodge et al., 2019). Comparisons of human SPHK1 and SPHK2 glial expression in pathologic versus nonpathologic conditions were performed using a published snRNA-seq dataset (Mathys et al., 2019). All analysis of scRNA-seq and snRNA-seq was performed using custom scripts in Python version 3.7.12.
Experimental design and statistical analysis.
Data were analyzed in Stimfit (version 3.7.12; Python), MiniAnalysis (version 6.0.3; Synaptosoft), or Fiji (NIH). Both male and female animals are used in the experiments and analyses. An α-level of 0.05 was used to determine statistical significance. The data are presented as the mean ± SEM, with the precise sample sizes indicated in the figure legends. Statistical analysis was performed using Prism (version 9.5.1; GraphPad). We conducted an assumption check by assessing the normality of residuals from ANOVA tests across all datasets using the Shapiro–Wilk test. For datasets that exhibited non-normal ANOVA residuals, we used the nonparametric Mann–Whitney U test for comparisons between two conditions and the Kruskal–Wallis test with post hoc Dunn's test for comparisons involving more than two conditions. Sample sizes and Kruskal–Wallis scores with the degree of freedom are reported in the figure legends. Furthermore, p-values and test scores obtained from the Mann–Whitney U test and Dunn's test are detailed in the Results section.
Results
PHP is disrupted in Drosophila neuronal Aβ overexpression models
Transgenic Drosophila models of AD have been well characterized and widely used to study the molecular etiology of AD (Iijima et al., 2004; Crowther et al., 2005; Prüßing et al., 2013; Wang et al., 2021). To address whether homeostatic plasticity is affected in Drosophila models of AD, we examined basal synaptic transmission and PHP at the NMJ in third-instar larvae that overexpressed human-Aβ42 (UAS-Aβ42), human-Aβ42 with Arctic mutation (UAS-Aβ42Ar), human-APP with Arctic and Swedish mutations (UAS-APPAr, Sw), and human-MAPT (UAS-MAPT) in motoneurons (OK371-Gal4; Fig. 1A–F).
PHP is disrupted in Drosophila Aβ models. A, Schematic of PHP at the wild-type NMJ in Drosophila (left panel). Pharmacological inhibition of postsynaptic glutamate receptors induces a compensatory increase of presynaptic transmitter release, which offsets the reduction in mEPSP amplitude and recovers the excitation (EPSP) to the baseline value. Glial epigenetic signaling is required for PHP. It remains to be elucidated whether PHP is affected in Drosophila models of AD (right panel). B, Representative EPSP and mEPSP traces in wt (black) and overexpression of Aβ42 (OK371-Gal4>UAS-human-Aβ42, red) in motoneurons, in the absence (–PhTX) and presence (+PhTX) of philanthotoxin. C, mEPSP amplitudes (open bars) and presynaptic release (quantal content, filled bars). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Genotypes and sample sizes are as follows: wt (–PhTX, n = 12; +PhTX, n = 9); OK371-Gal4>UAS-human-Aβ42 (Aβ42; –PhTX, n = 10; +PhTX, n = 10); OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar; –PhTX, n = 13; +PhTX, n = 13); OK371-Gal4>UAS-APPAr, Sw (APPAr, Sw; –PhTX, n = 10; +PhTX, n = 10); and OK371-Gal4>UAS-MAPT (MAPT; –PhTX, n = 11; +PhTX, n = 13). Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test were used. Non-normalized raw data were used for statistical tests. D–F, Non-normalized values for genotypes as in C: average mEPSP amplitude (D), EPSP amplitude (E), and presynaptic release (quantal content; F) in the absence (open) and presence (filled bars) of PhTX. Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. The levels of significance when comparing baseline values in the mutants to those in the wild type are labeled at the top in the graph. G, Representative EPSC traces wt (black) and overexpression of Aβ42 (OK371-Gal4>UAS-human-Aβ42, red) in motoneurons, in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin. H, mEPSP amplitudes (open bars) and presynaptic release (apparent quantal content, filled bars). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Genotypes and sample sizes are as follows: wt (–PhTX, n = 12; +PhTX, n = 9) and OK371-Gal4>UAS-human-Aβ42 (Aβ42; –PhTX, n = 11; +PhTX, n = 17). Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. Non-normalized raw data were used for statistical tests. I–K, Non-normalized values for genotypes as in H: average mEPSP amplitude (I), EPSC amplitude (J), and presynaptic release (apparent quantal content; K) in the absence (open) and presence (filled bars) of PhTX. Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. L, Representative confocal images of the NMJ immunolabeled with anti-Brp (green, presynaptic), anti-Dlg (red, postsynaptic), and the neuronal membrane (anti-HRP; blue) in wt, OK371-Gal4>UAS-human-Aβ42 (Aβ42), OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar), and OK371-Gal4>UAS-APPAr, Sw (APPAr, Sw). M, L, The total number of presynaptic Brp puncta, total Dlg area, Brp density (Brp number/Dlg area), total number of synaptic boutons at muscle 6/7 in abdominal segment 2 for indicated genotypes (M) as in L. Sample sizes are as follows: wt, n = 4; Aβ42, n = 8; Aβ42Ar, n = 4; APPAr, Sw, n = 8. Mean ± SEM; *p < 0.05, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. N, Schematic to show climbing behavioral analysis. The percentage of adult flies (day 25) that climb above 5, 10, 15, and 20 cm lines in 1 min are quantified. O, Quantification of the number of animals that climb above 5, 10, 15, and 20 cm lines in 1 min. Data for each genotype are normalized to the total number of animals used in each trial. Genotypes and sample sizes are as follows: wt (n = 12 trials), OK371-Gal4>UAS-human-Aβ42 (Aβ42, n = 12 trials), OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar, n = 12 trials). Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test.
First, we assessed the rapid induction of PHP elicited by pharmacological inhibition of postsynaptic glutamate receptors using current-clamp recordings in 0.3 mm extracellular calcium. The bath application of 20 μm PhTX led to an approximately 50% reduction of the average mEPSP amplitude at the NMJ (Fig. 1B–D). In the wild-type control, an increase in presynaptic neurotransmitter release (QC; Fig. 1C,F) was observed to counteract the changes in mEPSP amplitude, thereby restoring postsynaptic excitation back to the baseline value (Fig. 1E). In contrast, in transgenic models overexpressing Aβ42, Aβ42Ar, and APPAr, Sw in motoneurons, there was no change in quantal content, in the presence of PhTX [Fig. 1C,F; p = 0.09 (z = 1.67), p = 0.27 (z = 1.09), and p = 0.15 (z = 1.42), respectively; Kruskal–Wallis test with post hoc Dunn's test]. Consequently, the average EPSP amplitudes at the synapse in these three AD models were significantly decreased in the presence of PhTX compared with the baseline values [Fig. 1E; p = 0.02 (z = 2.35), p = 3.9e-7 (z = 5.07), p = 4.8e-4 (z = 3.49) respectively; Kruskal–Wallis with post hoc Dunn's test]. Thus, all three transgenic models that express Aβ42, Aβ42Ar, and APPAr, Sw in motoneurons all exhibited impairment of PHP.
Although we observed a reduction of EPSP amplitude in the Aβ42 overexpression model at baseline [Fig. 1E; p = 0.002 (z = 3.09); Kruskal–Wallis with post hoc Dunn's test], none of the AD models displayed substantial deficits of basal synaptic transmission (Fig. 1D–F), suggesting that the PHP impairment is not directly caused by basal synaptic transmission defects. Interestingly, although a decrease in EPSP amplitude was observed in the presence of PhTX [p = 0.01 (z = 2.48); Kruskal–Wallis with post hoc Dunn's test], when human-MAPT was overexpressed in motoneurons, there was still a significant increase in quantal content [p = 8.3e-4 (z = 3.34), Kruskal–Wallis with post hoc Dunn's test]. This suggests that PHP is expressed, albeit at a reduced magnitude, in the human-MAPT model (Fig. 1C,F). The distinct effects of Aβ42 and MAPT on PHP suggest that PHP is sensitive to Aβ-mediated signaling.
We proceeded to examine PHP in AD models under physiological calcium concentrations (1.5 mm) using the two-electrode voltage-clamp technique. We induced PHP with 20 μm PhTX in a transgenic model overexpressing human-Aβ42Ar in motoneurons (OK371-Gal4>UAS-human-Aβ42Ar) and conducted the recording in 1.5 mm extracellular Ca2+. Consistent with the observations made in low calcium conditions (0.3 mm Ca2+) using the current-clamp method, the rapid induction of presynaptic homeostatic potentiation was completely disrupted in the Aβ42Ar overexpression model in physiological calcium (Fig. 1G–K). Moreover, baseline EPSC amplitude was not altered in the Aβ42Ar overexpression model when compared with the wild-type control in physiological calcium [Fig. 1J; p = 0.66 (z = 0.43); Kruskal–Wallis with post hoc Dunn's test], confirming the findings made in 0.3 mm Ca2+ conditions (Fig. 1B–F).
To investigate whether the impairment of PHP directly results from synapse loss in Drosophila Aβ models. We examined the synapse morphology of the NMJ in transgenic flies overexpressing human-Aβ42, Aβ42Ar, and APPAr, Sw in motoneurons. We assessed the total presynaptic active zone number (estimated by quantification of Brp puncta number; Marie et al., 2004), the total postsynaptic area (region of Dlg immunostaining), Brp density (Brp puncta number/Dlg area), and bouton number at muscle 6/7, segment 2 (Fig. 1L,M). Despite a slight increase in Brp puncta number [p = 0.03 (z = 2.23); Kruskal–Wallis with post hoc Dunn's test] and Dlg area [p = 0.05 (z = 2.00); Kruskal–Wallis with post hoc Dunn's test] in the Aβ42Ar model, the synaptic morphology remains largely unaltered across all the AD models (Fig. 1L,M). These results suggest that PHP deficits do not directly result from synapse loss in the Aβ models. To validate the behavioral deficits that are associated with AD, we examined the motor function in adult flies that overexpress Aβ in motoneurons (Fig. 1N,O). We found that the climbing behavior was significantly impaired in Aβ transgenic flies compared with the wild-type control at day 25 posteclosion (Fig. 1O). In conclusion, PHP is disrupted when human-Aβ42, Aβ42Ar, and APPAr, Sw are overexpressed in motoneurons, but these deficits do not appear to be directly caused by synapse loss.
Fingolimod rescues PHP in Drosophila Aβ models
Previous studies suggest a role for the sphingosine pathway in the pathology of AD (He et al., 2010; Ceccom et al., 2014; Couttas et al., 2014), yet the function of endogenous sphingosine in neurodegeneration and cognitive impairment remains to be fully understood (Takasugi et al., 2011; but see Lei et al., 2019). Fingolimod, an orally active medication used to treat relapsing-remitting multiple sclerosis, is a structural analog of sphingosine (Brinkmann et al., 2010). While evidence suggests that Fingolimod may help in mitigating memory loss by restoring synaptic plasticity in mouse models of AD, the exact mechanisms through which it operates are yet to be unraveled (Aytan et al., 2016; Angelopoulou and Piperi, 2019; Kartalou et al., 2020; Fagan et al., 2022). Fingolimod is phosphorylated by endogenous sphingosine kinases within the cell to produce its biologically active form, (s)-Fingolimod-phosphate [(s)-Fingolimod-P], which triggers two primary signaling pathways. At the cellular surface, it serves as an agonist for sphingosine receptors, facilitating the endocytosis of these receptors (Chun and Hartung, 2010; Brunkhorst et al., 2014). Intracellularly, (s)-Fingolimod-P binds to the active sites of histone deacetylase 1 (HDAC1) and HDAC2 within the nucleus, thereby preventing them from removing acetyl groups from histone lysine residues (Hait et al., 2009). In Drosophila, direct homologs to sphingosine receptors have not been identified, suggesting that in flies, sphingosine possibly functions through the inhibition of HDAC activity.
We sought to determine whether Fingolimod stabilizes synaptic function via homeostatic plasticity in Drosophila AD models. First, we treated Aβ42, Aβ42Ar, and APPAr, Sw overexpression flies with FTY720 (300 μm in the food) and assessed PHP at the NMJ in third-instar larvae. We found that FTY720 treatment rescued PhTX-induced synaptic homeostasis in Aβ42, Aβ42Ar, and APPAr, Sw overexpression conditions, while PHP remained normal in the wild-type control with FTY720 treatment (Fig. 2A,B). We performed a series of control experiments to confirm the effects of FTY720. We examined baseline synaptic transmission in Aβ42Ar and found that 300 μm FTY720 did not change baseline mEPSP, EPSP amplitudes, and quantal content, compared with the wild-type control, indicating that its effect is specific to PHP [Fig. 2D–F; p = 0.52 (z = 0.64), p = 0.15 (z = 1.42), and p = 0.89 (z = 0.13), respectively; Kruskal–Wallis with post hoc Dunn's test]. We examined dose responses of FTY720 in the Aβ42Ar overexpression condition (Fig. 2C–F). Notably, treating Aβ42Ar with 300 and 500 μm FTY720 resulted in a dramatic increase in quantal content in the presence of PhTX [Fig. 2C,F; p = 5.8e-4 (z = 3.44) and p = 1.7e-4 (z = 3.75), respectively; Kruskal–Wallis with post hoc Dunn's test]. However, a high concentration of FTY720 (500 μm) reduced animal body size (data not shown; see Discussion) and baseline EPSP amplitude [Fig. 2E; p = 1.9e-4 (z = 3.73); Kruskal–Wallis with post hoc Dunn's test]. As such, we used 300 μm FTY720 in the remainder of the study.
Fingolimod rescues PHP deficits in Drosophila Aβ models. A, Representative EPSP and mEPSP traces in wt (black) and overexpression of Aβ42Ar (OK371-Gal4>UAS-human-Aβ42Ar, red) in the motoneuron, in the absence (–PhTX) and presence (+PhTX) of philanthotoxin. Data from Fingolimod (FTY720) treated and untreated conditions for each genotype are shown. B, mEPSP amplitudes (open bars) and presynaptic release (quantal content, filled bars) with and without FTY720 (300 µm) treatment. Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Genotypes and sample sizes are as follows: wt (–PhTX, n = 12; +PhTX, n = 9), wild type with FTY720 (wt FTY720; –PhTX, n = 10; +PhTX, n = 10), OK371-Gal4>UAS-human-Aβ42 (Aβ42; –PhTX, n = 10; +PhTX, n = 10), OK371-Gal4>UAS-human-Aβ42 with FTY720 (Aβ42 FTY720; –PhTX, n = 11; +PhTX, n = 11), OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar; –PhTX, n = 13; +PhTX, n = 13), OK371-Gal4>UAS-human-Aβ42Ar with FTY720 (Aβ42Ar FTY720; –PhTX, n = 10; +PhTX, n = 9), OK371-Gal4>UAS-APPAr, Sw (APPAr, Sw; –PhTX, n = 10; +PhTX, n = 10), and OK371-Gal4>UAS-APPAr, Sw with FTY720 (APPAr, Sw FTY720; –PhTX, n = 11; +PhTX, n = 10). Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. C, mEPSP amplitudes (open bars) and presynaptic release (quantal content, filled bars) with different concentrations of FTY720 treatment. Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Genotypes and sample sizes are as follows: wt (–PhTX, n = 12; +PhTX, n = 9), wild type with 300 µm FTY720 (wt+300 µM FTY720; –PhTX, n = 10; +PhTX, n = 10), OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar; –PhTX, n = 13; +PhTX, n = 13), OK371-Gal4>UAS-human-Aβ42Ar with 100 µm FTY720 (Aβ42Ar+100µM FTY720; –PhTX, n = 12; +PhTX, n = 10), OK371-Gal4>UAS-human-Aβ42Ar with 300 µm FTY720 (Aβ42Ar+300µM FTY720; –PhTX, n = 10; +PhTX, n = 9), OK371-Gal4>UAS-human-Aβ42Ar with 500 µm FTY720 (Aβ42Ar+500µM FTY720; –PhTX, n = 9; +PhTX, n = 9). Mean ± SEM; **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. D–F, Non-normalized data for synaptic transmission in the Aβ models treated with different concentrations of FTY720. The average mEPSP amplitude (D), EPSP amplitude (E), and quantal content (F) are presented as in C. Mean ± SEM; **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. G, Representative confocal images of the NMJ immunolabeled with anti-Brp (green, presynaptic), anti-Dlg (red, postsynaptic), and the neuronal membrane (anti-HRP; blue) in wt treated with 300 µm FTY720 (wt+FTY720), OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar), and OK371-Gal4>UAS-human-Aβ42Ar treated with 300 µm FTY720 (Aβ42Ar+FTY720). H, Quantification of the total number of presynaptic Brp puncta, total Dlg area, Brp density (Brp number/Dlg area), total number of synaptic boutons at muscle 6/7 in abdominal segment 2 for indicated genotypes as in C. Sample sizes are as follows: wt, n = 4; wt+FTY720, n = 4; Aβ42, n = 8; Aβ42+FTY720, n = 6; Aβ42Ar, n = 4; Aβ42Ar+FTY720, n = 4; APPAr, Sw, n = 8; APPAr, Sw+FTY720, n = 8; mean ± SEM; *p < 0.05, **p < 0.01; N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test.
Next, we examined synapse morphology in the wild-type and Aβ models with and without FTY720 treatment (Fig. 2G,H). We found that FTY720 caused a slight reduction in active zone number (estimated by Brp puncta number) and postsynaptic area (Dlg area) in the Aβ42, Aβ42Ar, and APPAr, Sw models (Fig. 2H). However, Brp density remained unaltered across all AD models following drug treatment compared with untreated groups (Fig. 2H). We only observed a mild reduction in bouton number in APPAr, Sw after drug treatment [p = 0.02 (z = 2.24), Kruskal–Wallis with post hoc Dunn's test], while FTY720 did not change bouton numbers in other Aβ models or in the wild-type control (Fig. 2H). Importantly, despite minor changes in synapse morphology observed in Aβ42, Aβ42Ar, and APPAr, Sw following FTY720 treatment compared with the untreated group of the same genotype, all drug-treated AD models showed unchanged synapse morphology compared with the treated wild-type animals (Fig. 2H). Together, FTY720 has minor effects on gross synapse morphology.
Fingolimod rescues glial H3K9ac and motor behavior deficits in Drosophila Aβ models
Previously we demonstrated that SAGA-dependent histone acetylation at H3K9 and H3K14 sites in peripheral glia are critical for PHP (Wang et al., 2020). Therefore, we explored the mechanisms underlying the Aβ-mediated PHP defects by asking whether peripheral glia, distributed throughout the peripheral nerves, exhibit abnormal histone acetylation. We overexpressed human-Aβ42, Aβ42Ar, and APPAr, Sw in motoneurons and assessed the abundance of H3K9ac in peripheral glial nuclei localized on peripheral nerves by immunostaining (Fig. 3A,B). We found that the average fluorescence intensity of H3K9ac in peripheral glia was reduced in Aβ42, Aβ42Ar, and APPAr, Sw models compared with the wild-type control [p = 0.007 (z = 2.71), p = 1.2e-9 (z = 6.08), p = 6.3e-5 (z = 4.00) respectively; Kruskal–Wallis with post hoc Dunn's test]. FTY720 increased H3K9ac average intensities in peripheral glial nuclei in all AD models tested (Fig. 3A,B). Collectively, the impairment of synaptic homeostasis caused by neuronal overexpression of Aβ42, Aβ42Ar, and APPAr, Sw is associated with dysregulated glial epigenetic signaling. Intriguingly, FTY720 treatment rescued deficits in both PHP and glial H3K9ac, suggesting a significant role of FTY720-depedent glial epigenetic signaling in modulating homeostatic plasticity.
Fingolimod treatment rescues glial H3K9ac defects, motor function, and eye degeneration in Drosophila Aβ models. A, Representative confocal images of acetylated H3K9 (H3K9ac) in glial nuclei on the peripheral nerves. Glial nuclei (DAPI, green), acetylated H3K9 (H3K9ac, red), and neuronal membrane (HRP, blue) are shown for wt, OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar), and OK371-Gal4>UAS-APPAr, Sw (APPAr, Sw) with (+FTY720) or without (–FTY720) treatment. B, Quantification of the average H3K9ac fluorescence intensity within glial nuclei. Genotypes and sample sizes are as follows: wt (n = 269 nuclei; n = 4 animals); wild type with FTY720 (wt+FTY720, n = 199 nuclei; n = 3 animals); OK371-Gal4>UAS-human-Aβ42 (Aβ42; n = 122 nuclei; n = 4 animals); OK371-Gal4>UAS-human-Aβ42 with FTY720 (Aβ42 FTY720; n = 198 nuclei; n = 4 animals); OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar; n = 160 nuclei; n = 4 animals); OK371-Gal4>UAS-human-Aβ42Ar with FTY720 (Aβ42Ar FTY720; n = 246 nuclei; n = 4 animals); OK371-Gal4>UAS-APPAr, Sw (APPAr, Sw; n = 234 nuclei; n = 4 animals); and OK371-Gal4>UAS-APPAr, Sw with FTY720 (APPAr, Sw FTY720; n = 210 nuclei; n = 3 animals). Mean ± SEM; **p < 0.01, ***p < 0.001; N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test (Kruskal–Wallis score = 374.0, df = 7). C, Schematic to show climbing behavioral analysis with and without FTY720 treatment. The percentage of adult flies (day 25) that climb above 5, 10, 15, and 20 cm lines in 1 min are quantified. D, Quantification of the percentage of animals that climb above 5, 10, 15, and 20 cm lines in 1 min in the climbing assay. Genotypes and sample sizes are as follows: wt (n = 12 trials), wild type with FTY720 (wt+FTY720; n = 12 trials), OK371-Gal4>UAS-human-Aβ42Ar (Aβ42Ar; n = 12 trials), and OK371-Gal4>UAS-human-Aβ42Ar with FTY720 (Aβ42Ar+FTY720; n = 12 trials). Mean ± SEM; *p < 0.05, **p < 0.01, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. E, Cumulative percentage of flies with various degrees of eye degeneration (0–4: none to severe degeneration; see Materials and Methods for details). Cumulative percentage of flies with more than 0 (≥0) to more than 4 (≥4) degree of degeneration is shown. Genotypes and sample sizes are as follows: GMR-Gal4>UAS-human-Aβ42Ar (–FTY720, n = 15; +FTY720, n = 16) and GMR-Gal4>UAS-human-APPAr, Sw (–FTY720, n = 15; +FTY720, n = 14).
We further examined whether FTY720 rescues behavioral and neurodegeneration deficits that occur at the adult stage in the Aβ models. We fed Aβ flies that overexpressed human-Aβ42Ar in motoneurons with FTY720 throughout larval and adult stages and assessed their motor function at day 25 posteclosion (Fig. 3C,D). We found that the climbing behavior was significantly improved in the Aβ42Ar flies following FTY720 treatment (Fig. 3D). We also examined eye degeneration in adult flies that overexpress Aβ42Ar (GMR-Gal4>UAS-human Aβ42Ar) and APPAr, Sw (GMR-Gal4>UAS-human APPAr, Sw) specifically in the eye (Fig. 3E). Eye degeneration caused by Aβ42Ar and APPAr, Sw overexpression was partially rescued when the animals were treated with FTY720 (Fig. 3E). These results demonstrate that FTY720 improves motor behavior and partially ameliorate eye degeneration phenotype in Drosophila Aβ models.
Sk2 is necessary for both the rapid induction and long-term maintenance of PHP
Based on our findings that FTY720 rescues PHP and glial H3K9ac deficits in Drosophila models of AD (Figs. 2B, 3B), we hypothesize that the endogenous sphingosine pathway is crucial for regulating PHP. In Drosophila, there are two homologs of sphingosine kinases, Sk1 and Sk2 (Herr et al., 2004; Yonamine et al., 2011). These kinases phosphorylate endogenous sphingosine and FTY720 to their active forms, S1P and (s)-Fingolimod-P, respectively (Fig. 4A). These sphingosine kinases are important players in the sphingolipid metabolism, and their downstream lipid messengers have been found to have critical roles in many cellular processes including synaptic transmission and synaptic plasticity (Kanno et al., 2010; Hait et al., 2014). Therefore, understanding how sphingosine kinases regulate PHP can provide important insights into the mechanisms of synaptic plasticity in the context of neurodegenerative diseases.
Sk2 is necessary for the rapid induction and long-term expression of PHP. A, Schematic to show sphingosine kinase-mediated phosphorylation of sphingosine and FTY720 in wild type and Sk2 mutant. B, Superimposed AlphaFold predicted structures of human SPHK2 (blue) and Drosophila Sk2 (orange). rmsd and TM-score are shown. C, Representative EPSP and mEPSP traces in wt (black), Sk1 (green), and Sk2 (blue), in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin. D–G, Normalized data (D) for mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars) and non-normalized data of mEPSP amplitudes (E), EPSP amplitudes (F), and quantal content (G) for wt (–PhTX, n = 48; +PhTX, n = 21), Sk1 (–PhTX, n = 22; +PhTX, n = 13), and Sk2 (–PhTX, n = 27; +PhTX, n = 16). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX in D. Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. H, Representative EPSC traces in wt (black) and Sk2 mutant (Sk2, blue), in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin. I–L, Normalized data (I) for mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars) and non-normalized data of mEPSP amplitudes (J), EPSC amplitudes (K), and apparent quantal content (L) for wt (–PhTX, n = 12; +PhTX, n = 9) and Sk2 mutant (Sk2; –PhTX, n = 10; +PhTX, n = 11). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX in I. Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. M, Representative EPSP and mEPSP traces in the wt (gray), GluRIIA mutant (GluRIIA, black), Sk2 (light purple), and GluRIIA;Sk2 double mutant (GluRIIA;Sk2). N–Q, Normalized data (N) for mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars) and non-normalized data of mEPSP amplitudes (O), EPSP amplitudes (P), and quantal content (Q) for wt (n = 28), GluRIIA mutant (GluRIIA, n = 18), Sk2 (n = 10), and GluRIIA;Sk2 double mutant (GluRIIA;Sk2, n = 10). Mean ± SEM; *p < 0.05, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test.
Given that we observed a decrease in H3K9ac in Aβ models and that Fingolimod could restore this deficit (Fig. 3A,B), we initially concentrated on studying the function of Sk2, which has been shown to phosphorylate Fingolimod and inhibit HDAC activity (Hait et al., 2009, 2014). First, we examined the structural similarity between human SPHK2 and Drosophila Sk2. The structure of human SPHK2 or Drosophila Sk2 protein has not been experimentally determined. Therefore, we used the AlphaFold algorithm-predicted structures in the comparison (Jumper et al., 2021; Mirdita et al., 2022). To validate the prediction, we compared the predicted structure with the X-ray structure of human SPHK1. We used two well established measures of protein topological similarity and homology, the template modeling score (TM-score) and the root mean square deviation (rmsd). The TM-score ranges from 0 to 1, with a score of 1 indicating a perfect alignment between two structures, and a score >0.5 suggesting high structural similarity (Xu and Zhang, 2010). Conversely, a rmsd value of 0 suggests that two structures are identical, while larger rmsd values indicate less similarity between the structures (Kufareva and Abagyan, 2012). Our comparison yielded a TM-score of 0.93 and an rmsd of 0.49 Å, indicating a high degree of similarity between the experimentally determined and predicted structures of human SPHK1, thus confirming the reliability of the AlphaFold prediction.
After successfully validating the AlphaFold predictions using the structure of human SPHK1, we proceeded to align the predicted 3D structures of Drosophila Sk2 with those of human SPHK1 and SPHK2. Upon comparing the Drosophila Sk2 structure to human SPHK1, we found a TM-score of 0.5 and an rmsd of 3.11 Å. When we compared Drosophila Sk2 to human SPHK2, the TM-score improved to 0.64 and the rmsd was slightly reduced to 3.04 Å. Given that the similarity between protein pairs significantly increases when the TM-score is >0.5, the TM-score of 0.64 between Drosophila Sk2 and human SPHK2 indicates that these proteins are highly likely to belong to the same topology family (Xu and Zhang, 2010). While Drosophila Sk2 share residue similarities with both human SPHK1 and SPHK2, its protein structure aligns more closely with that of SPHK2.
Next, we assessed whether Sk1 and Sk2 are required for PHP. We found that PhTX-induced synaptic homeostasis was completely blocked at the NMJ in the loss-of-function mutant of Sk2 (Fig. 4C–G). However, both the baseline EPSP amplitude and quantal content remained normal in the Sk2 mutant compared with the wild type [p = 0.08 (z = 1.77) and p = 0.94 (z = 0.07), respectively; Kruskal–Wallis with post hoc Dunn's test], suggesting that Sk2 is specifically required for PHP but not for basal synaptic transmission (Fig. 4F,G). Although we observed a reduction in EPSP amplitude in the Sk1 mutant in the presence of PhTX [p = 0.03 (z = 2.21); Kruskal–Wallis with post hoc Dunn's test], there was a significant increase of quantal content with PhTX application [p = 0.02 (z = 2.35); Kruskal–Wallis with post hoc Dunn's test], suggesting that Sk1 is not required for the rapid induction of PHP (Fig. 4C–G). These results align with our finding that FTY720, a substrate of sphingosine kinases, modulates PHP (Fig. 2A,B).
To validate the role of Sk2 in regulating PHP under physiological calcium conditions, we used a two-electrode voltage-clamp method to assess PHP at 1.5 mm Ca2+ (Fig. 4H–L). We observed a significant decrease in EPSC amplitude in Sk2 homozygous mutants in the presence of PhTX compared with the baseline (Fig. 4H,K). Notably, the expected compensatory increase in quantal content was completely abolished in the Sk2 mutants [p = 0.10 (z = 1.65); Kruskal–Wallis with post hoc Dunn's test], suggesting that Sk2 is necessary for PHP in physiological calcium conditions (Fig. 4H,L). Interestingly, we found that the baseline EPSC amplitude at 1.5 mm Ca2+ did not differ between the Sk2 mutant and wild-type control, mirroring our earlier findings at 0.3 mm Ca2+ (Fig. 4F,K). This indicates that while Sk2 is essential for PHP, it is not required for maintaining basal transmission at the NMJ in Drosophila.
We then assessed whether Sk2 is necessary for the long-term maintenance of PHP, using a genetic deletion mutant of the postsynaptic glutamate receptor subunit GluRIIA (Fig. 4M–Q; Petersen et al., 1997). We observed a significant reduction of mEPSP amplitude in both the GluRIIA and GluRIIA;Sk2 double mutants (Fig. 4M–O). There was a significant increase in quantal content in the GluRIIA mutant compared with the wild-type control [Fig. 4M,N,Q; p = 9.8e-6 (z = 4.42); Kruskal–Wallis with post hoc Dunn's test]. However, in the GluRIIA;Sk2 double mutant, the compensatory increase in presynaptic release was completely diminished [Fig. 4M–Q; p = 0.10 (z = 1.64); Kruskal–Wallis with post hoc Dunn's test]. Thus, we concluded that Sk2 is necessary for both the rapid induction and maintenance of PHP.
Sk2 expressed in perineurial glia is necessary for PHP
To examine the specific cell type in which Drosophila Sk2 functions during PHP, we assessed Sk2 activity in motoneurons, muscles, and in perineurial glia using tissue-specific RNAi knock-down approaches (Fig. 5A–E). We expressed UAS-Sk2 RNAi in the motoneuron (OK371-Gal4), muscles (BG57-Gal4), and perineurial glia (NP6293-Gal4). Although the EPSP amplitude was reduced when Sk2 was knocked down by RNAi in the motoneuron in the presence of PhTX [Fig. 5D; p = 1.2e-4 (z = 3.84); Kruskal–Wallis with post hoc Dunn's test], there was still a dramatic increase in quantal contents in motoneuron-Gal4>Sk2 RNAi [p = 0.008 (z = 2.64); Kruskal–Wallis with post hoc Dunn's test] and muscle-Gal4>Sk2 RNAi conditions [p = 6.2e-4 (z = 3.42); Kruskal–Wallis with post hoc Dunn's test; Fig. 5A,B,E]. However, when UAS-Sk2-RNAi was expressed in perineurial glia, PHP was completely diminished (Fig. 5A,B,D,E). We confirmed the glial function of Sk2 in PHP using a second UAS-Sk2-RNAi allele (Fig. 5B–E). As a control, heterozygous perineurial glial Gal4 driver alone without the UAS-RNAi transgene did not affect PHP (Fig. 5B,E). Next, we performed tissue-specific rescue experiments. We expressed UAS-Sk2 specifically in glia in the Sk2 mutant background and found that it fully restored PHP in the Sk2 mutant (Fig. 5F–I).
Sk2 functions in perineurial glia for PHP. A, Representative EPSP and mEPSP traces in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin. Genotypes are as follows: wt (black), OK371-Gal4>UAS-Sk2 RNAi (Motoneuron Sk2 RNAi, orange), BG57-Gal4>UAS-Sk2 RNAi (Muscle Sk2 RNAi, red), and NP6293-Gal4>UAS-Sk2 RNAi (Glial Sk2 RNAi, blue). B–E, Normalized data (B) for mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars), and non-normalized data of mEPSP amplitudes (C), EPSP amplitudes (D), and quantal content (E) for wt (–PhTX, n = 32, +PhTX, n = 17), NP6293-Gal4/+ (Glial Gal4/+; –PhTX, n = 10; +PhTX, n = 9), OK371-Gal4>UAS-Sk2 RNAi #1 (Motoneuron Sk2 RNAi #1; –PhTX, n = 13; +PhTX, n = 11), BG57-Gal4>UAS-Sk2 RNAi (Muscle Sk2 RNAi #1; –PhTX, n = 10; +PhTX, n = 10), NP6293-Gal4>UAS-Sk2 RNAi (Glial Sk2 RNAi #1; –PhTX, n = 19; +PhTX, n = 13), and NP6293-Gal4>UAS-Sk2 RNAi #2 (Glial Sk2 RNAi #2; –PhTX, n = 10; +PhTX, n = 8). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX in B. Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. F–I, Normalized data (F) for mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars), and non-normalized data of mEPSP amplitudes (G), EPSP amplitudes (H), and quantal content (I) for wt (–PhTX, n = 48; +PhTX, n = 24), Sk2 (–PhTX, n = 27; +PhTX, n = 17), and NP6293-Gal4>UAS-Sk2;Sk2 (Sk2;Glial rescue; –PhTX, n = 11; +PhTX, n = 10). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX in F. Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. J, Perineurial glia cell clusters are shown in t-SNE (t-distributed stochastic neighbor embedding) plots using a published scRNA-seq dataset from Drosophila adult brains (57,000 cells; as published in Davie et al., 2018; GSE107451). The perineurial glia cluster is highlighted in red (left panel). Individual perineurial glia expressing Sk1 (green) or Sk2 (blue) are shown on the right. K, L, The total number of perineurial glia expressing Sk1 (green) and Sk2 (blue; K), and the average expression level of Sk1 (green) and Sk2 (blue; L) in perineurial glia. CPM, Counts per million.
Finally, we repurposed a published Drosophila scRNA-seq dataset (Davie et al., 2018) to validate Sk2 expression in perineurial glia (Fig. 5J–L). Our analysis revealed that Sk2 is indeed expressed in perineurial glia (Fig. 5J,K), with the average expression level for Sk2 being higher than that for Sk1 in perineurial glia (Fig. 5L). These results indicate that perineurial glia serve as an endogenous source of Sk2. Together, our findings establish a critical role of glial Sk2 in PHP.
Glial Sk2 rescues PHP, glial H3K9ac, and motor behavior in Drosophila Aβ models
We examined whether we can restore PHP in Aβ transgenic flies by increasing the expression of Sk2 in glia. First, we sought to determine whether FTY720 functions via the Sk2-mediated sphingosine pathway. We treated the Sk2 mutant with FTY720 throughout the larval stage and examined PhTX-induced homeostatic plasticity. We found that FTY720 treatment did not rescue PHP defects in the Sk2 mutant, suggesting that Sk2 is required for the effect of FTY720 in restoring PHP (Fig. 6A,B,D). Notably, glial-specific overexpression of Sk1 or Sk2 in the wild-type background does not alter PHP, suggesting that the reduction but not the increased expression level of Sk2 disrupts PHP (Fig. 6C,E).
Glial-specific expression of Sk2 rescues PHP, H3K9ac, and motor deficits in Drosophila Aβ models. A, Representative EPSP and mEPSP traces in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin in wt, wt with FTY720 (wt+FTY720), Sk2 mutant (Sk2), and Sk2 mutant with FTY720 (Sk2+FTY720). B, mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars) in genotypes as in A. Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Sample sizes are as follows: wt (–PhTX, n = 28; +PhTX, n = 15), wild type with FTY720 (wt FTY720; –PhTX, n = 10; +PhTX, n = 10), Sk2 mutant (Sk2; –PhTX, n = 17; +PhTX, n = 13), and Sk2 mutant with FTY720 (Sk2 FTY720; –PhTX, n = 10; +PhTX, n = 10). Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. C, mEPSP amplitudes (open bars) and presynaptic release (quantal content, filled bars) for glial-specific overexpression of Sk1 and Sk2 in the wild-type background. Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Genotypes and sample sizes: wt (–PhTX, n = 28; +PhTX, n = 15), NP6293-Gal4>UAS-Sk1 (Glial UAS-Sk1; –PhTX, n = 12; +PhTX, n = 10), NP6293-Gal4>UAS-Sk2 (Glial UAS-Sk2; –PhTX, n = 11; +PhTX, n = 11). Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. D, Non-normalized data of mEPSP amplitudes (left panel), EPSP amplitudes (middle panel), and quantal content (right panel) for genotypes shown in A and B. Mean ± SEM; *p < 0.05, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. E, Non-normalized data of mEPSP amplitudes (left panel), EPSP amplitudes (middle panel), and quantal content (right panel) for genotypes shown in C. Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. F, Schematic to show neuronal overexpression of Aβ42 and glial overexpression of Sk2 at the same synapse. G, Representative EPSP and mEPSP traces in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin in heterozygous nSyb-QF2/+ (Neural QF/+), nSyb-QF2>QUAS-human-Aβ42 (Neural Aβ42), and nSyb-QF2>QUAS-human-Aβ42;NP6293-Gal4>UAS-Sk2 (Neural Aβ42+Glial Sk2 rescue). H, mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars) for wt (–PhTX, n = 32; +PhTX, n = 15), heterozygous nSyb-QF2/+ (Neural QF/+; –PhTX, n = 11; +PhTX, n = 9), nSyb-QF2>QUAS-human-Aβ42 (Neural hAβ42; –PhTX, n = 12; +PhTX, n = 14), nSyb-QF2>QUAS-human-Aβ42;UAS-Sk2/+ (Neural hAβ42;UAS-Sk2/+; –PhTX, n = 10; +PhTX, n = 8), and nSyb-QF2>QUAS-human-Aβ42;NP6293-Gal4>UAS-Sk2 (Neural hAβ42;Glial Sk2; –PhTX, n = 11; +PhTX, n = 9). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX. Mean ± SEM; **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. I–K, Non-normalized data of mEPSP amplitudes (I), EPSP amplitudes (J), and quantal content (K) for genotypes presented in H. Mean ± SEM; *p < 0.05, **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. L, Representative confocal images of acetylated H3K9 (H3K9ac) in peripheral glial nuclei on the peripheral nerves. Peripheral glial nuclei (DAPI, green), acetylated H3K9 (H3K9ac, red), and neuronal membrane (HRP, blue) are shown for the wt, nSyb-QF2>QUAS-human-Aβ42 (Neural QF>QUAS-hAβ42), nSyb-QF2>QUAS-human-Aβ42;NP6293-Gal4/+ (Neural QF>QUAS-hAβ42;Glial-Gal4/+), and nSyb-QF2>QUAS-human-Aβ42;NP6293-Gal4>UAS-Sk2 (Neural QF>QUAS-hAβ42;Glial-Gal4>UAS-Sk2). M, Average H3K9ac fluorescence intensity within peripheral glial nuclei. Genotypes are shown as in L. Sample sizes are as follows: wt: n = 429 nuclei; n = 5 animals; Neural QF>QUAS-hAβ42: n = 429 nuclei; n = 5 animals; Neural QF>QUAS-hAβ42;Glial-Gal4/+: n = 192 nuclei; n = 6 animals; Neural QF>QUAS-hAβ42;Glial-Gal4>UAS-Sk2: n = 276 nuclei; n = 6 animals. Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test (Kruskal–Wallis score = 98.25; df = 3). N, The number of adult animals (day 23–25) that climb above 5, 10, 15, and 20 cm lines in 1 min in the climbing assay. Data for each genotype are normalized to the total number of animals used in each trial. Genotypes and sample sizes: wt, n = 21 trials; Neural QF>QUAS-hAβ42, n = 59 trials, Neural QF>QUAS-hAβ42;NP6293-Gal4/+, n = 56 trials, Neural QF>QUAS-hAβ42;NP6293-Gal4>UAS-Sk2, n = 47 trials. Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. O, The total numbers of astrocyte, microglia, OPC, and oligodendrocyte that express SPHK1 (green) and SPHK2 (blue) using a published human scRNA-seq dataset (Hodge et al., 2019). P, Quantification of the average expression level of human SPHK1 (green) and SPHK2 (blue) in astrocyte, microglia, OPC, and oligodendrocyte using a published human scRNA-seq dataset (Hodge et al., 2019). CPM, Counts per million. Q, The mean expression level of SPHK1 and SPHK2 in nonpathological (no patho) and pathological (patho; AD) from human brain samples as published in Mathys et al. (2019).
Then, we directly assessed whether we can manipulate Sk2 expression to rescue synaptic homeostasis in the Aβ model (Fig. 6F). When we overexpressed QUAS-Aβ42 in all neurons (nSyb-QF2>QUAS-Aβ42; Potter et al., 2010), we found that PHP was completely diminished, while the heterozygous pan-neuronal QF2 driver alone without QUAS-Aβ42 had no effect on PhTX-induced PHP (Fig. 6G–K). Next, we overexpressed UAS-Sk2 in glia specifically (NP6293-Gal4) in the Aβ42 background (nSyb-QF2>QUAS-Aβ42;NP9293-Gal4>UAS-Sk2) and found that PHP was rescued, demonstrating that overexpression of Sk2 in glia is sufficient to rescue PHP in the AD model (Fig. 6G–K). As a control, we showed that heterozygous UAS-Sk2 transgene alone without a Gal4 driver did not rescue PHP in the Aβ model (Fig. 6G–K).
We noted a reduction in baseline mEPSP and EPSP amplitude, when UAS-Sk2 was expressed in glia in the Aβ42 background (nSyb-QF2>QUAS-Aβ42;NP9293-Gal4>UAS-Sk2) compared with the wild-type control [Fig. 6I,J; p = 0.004 (z = 2.87) and p = 5.6e-6 (z = 4.54), respectively; Kruskal–Wallis with post hoc Dunn's test]. However, these baseline deficits may be because of the nonspecific effects of the UAS-Sk2 insertion. This is suggested by the observation that the UAS-Sk2/+ transgene, in the absence of the Glial-Gal4 driver, exhibited similar deficits in mEPSP and EPSP amplitude as the Neural Aβ42;Glial Sk2 rescue (Fig. 6I,J). Furthermore, the reduction in the basal EPSP amplitude in the Neural Aβ42;UAS-Sk2/+ and Neural Aβ42;Glial Sk2 rescue conditions is likely because of a decrease in mEPSP amplitude, while the quantal contents in both genotypes remain unaffected at baseline compared with wild-type [Fig. 6K; p = 0.78 (z = 0.27) and p = 0.22 (z = 1.23), respectively; Kruskal–Wallis with post hoc Dunn's test]. In conclusion, glial-specific expression of Sk2 rescued PHP in the Aβ42 background.
Next, we examined histone acetylation in glia in the Aβ model (nSyb-QF2>QUAS-Aβ42; Fig. 6L,M). We probed the average intensity of H3K9ac in peripheral glial nuclei by immunostaining (Fig. 6L,M). We found that the average H3K9ac intensity was significantly reduced in peripheral glia in the Aβ42 model compared with the wild-type control [p = 1.5e-19 (z = 9.04); Kruskal–Wallis with post hoc Dunn's test]. Glial-specific overexpression of Sk2 significantly increased the average H3K9ac intensity in the Aβ42 background [nSyb-QF2>QUAS-Aβ42;NP9293-Gal4>UAS-Sk2; Fig. 6L,M; p = 4.2e-12 (z = 6.93); Kruskal–Wallis with post hoc Dunn's test]. Finally, we overexpressed Sk2 in glia and assessed climbing behavior in the Aβ model. Interestingly, we found that glial overexpression of Sk2 was sufficient to improve climbing behavior in Aβ42 at the adult stage (day 23–25 posteclosion; Fig. 6N). Together with the electrophysiology and H3K9ac immunostaining data, these results suggest that Sk2 rescues PHP and improves motor function in Drosophila Aβ models via regulating glial epigenetic signaling.
Increasing evidence suggests that abnormal epigenetic and transcriptomic alterations are closely associated with AD pathology (Lord and Cruchaga, 2014; Nativio et al., 2018; Klein et al., 2019; Mathys et al., 2019). We asked whether human sphingosine kinases are expressed in glia and whether they are dysregulated in human AD patients. We examined the expression of SPHK1 and SPHK2, in different glial populations in human brains using a published scRNA-seq dataset (Hodge et al., 2019). We found that higher numbers of human glial cells, including astrocytes, microglia, oligodendrocyte precursor cells (OPCs), and oligodendrocytes, express SPHK2 than SPHK1 (Fig. 6O). The expression level of SPHK2 is higher than SPHK1 in all glial cell types (Fig. 6P). To examine whether SPHK expression level is affected in AD brains, we performed an analysis using a published snRNA-seq dataset (Mathys et al., 2019), which was generated from postmortem human brain samples (prefrontal cortex) from participants in the Religious Order Study or the Rush Memory and Aging Project, two longitudinal cohort studies of aging and dementia (Fig. 6Q). Interestingly, the expression levels of SPHK1 and SPHK2 in all glial types in pathologic samples of AD are not significantly different compared with nonpathological samples (Fig. 6Q). Although it is likely that the dysregulation of glial SPHK2 itself is not a primary risk factor that contributes directly to pathologic changes in AD, our data suggest that the sphingosine pathway stabilizes synaptic function via homeostatic regulation (Fig. 6A,G–N).
Fingolimod and glial Sk2 overexpression rescues PHP in a SAGA mutant
We further explored the activity of FTY720 as an HDAC inhibitor in regulating PHP. Glial histone acetyltransferase protein complex, SAGA complex, is necessary for synaptic homeostasis (Wang et al., 2020). We sought to determine whether FTY720 rescues PHP and histone acetylation deficits in a genetic mutant of ada2b, an essential component of the SAGA complex (Fig. 7A). As reported previously, we observed a complete disruption of PhTX-induced PHP in the ada2b mutant (Fig. 7B–F; Wang et al., 2020). Interestingly, we found that FTY720 treatment fully restored PHP in the ada2b mutant (Fig. 7B,C,F). This observation is consistent with the previous report that FTY720 increases histone acetylation by inhibiting HDAC activities (Hait et al., 2014). When we overexpressed UAS-Sk2 in perineurial glia in the ada2b mutant background, PHP was also fully rescued, indicating that glial Sk2 modulates synaptic homeostasis through histone acetylation (Fig. 7B–F).
Fingolimod and glial-specific expression of Sk2 rescues PHP and H3K9ac in a SAGA complex mutant. A, Schematic to show histone acetylation is reduced in the ada2b (a SAGA component) mutant (bottom left) compared with wild type (top left). Sphingosine kinase-mediated phosphorylation of sphingosine and FTY720 can modulate histone acetylation (right panel). B, Representative EPSP and mEPSP traces in the absence (–PhTX) and the presence (+PhTX) of philanthotoxin in wt, ada2b mutant (ada2b), ada2b with FTY720 (ada2b+FTY720), and ada2b;NP6293-Gal4>UAS-Sk2 (ada2b;Glial UAS-Sk2). C–F, Normalized data (C) for mEPSP amplitudes (open bars) and presynaptic release (QC; filled bars), and non-normalized data of mEPSP amplitudes (D), EPSP amplitudes (E), and quantal content (F) for wt (–PhTX, n = 28; +PhTX, n = 15), ada2b (–PhTX, n = 10; +PhTX, n = 10), ada2b with FTY720 (ada2b+FTY720; –PhTX, n = 10; +PhTX, n = 9), and ada2b;NP6293-Gal4>UAS-Sk2 (ada2b;Glial UAS-Sk2; –PhTX, n = 9; +PhTX, n = 10). Data for each genotype are presented as the percentage change in PhTX compared with the same genotype recorded in the absence of PhTX in C. Mean ± SEM; **p < 0.01, ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. G, Representative confocal images of acetylated H3K9 (H3K9ac) in perineurial glial nuclei on the peripheral nerves. Perineurial glial nuclei (NP6293-Gal4>UAS-Redstinger.nls, green), acetylated H3K9 (H3K9ac, red), and neuronal membrane (HRP, blue) are shown for NP6293-Gal4>UAS-Redstinger.nls (wt), ada2b;NP6293-Gal4>UAS-Redstinger.nls (ada2b), and NP6293-Gal4>UAS-Redstinger.nls;ada2b with FTY720 (ada2b+FTY720). H, Average H3K9ac fluorescence intensity within perineurial glial nuclei (left panel) and all peripheral glial nuclei (right panel) for genotypes as in G. Sample sizes for perineurial glia: wt: n = 310 nuclei; n = 4 animals; ada2b: n = 203 nuclei; n = 3 animals; ada2b+FTY720: n = 98 nuclei; n = 3 animals. Kruskal–Wallis score = 281.6, df = 2. Sample sizes for all peripheral glia are as follows: wt: n = 439 nuclei; n = 4 animals; ada2b: n = 361 nuclei; n = 3 animals; ada2b+FTY720: n = 228 nuclei; n = 3 animals. Kruskal–Wallis score = 362.5, df = 2. Mean ± SEM; ***p < 0.001, N.S. not significant; Kruskal–Wallis test with post hoc Dunn's test. I, Representative confocal images of acetylated H3K9 (H3K9ac) in glial nuclei on the peripheral nerves. Glial nuclei (DAPI, green), acetylated H3K9 (H3K9ac, red), and neuronal membrane (HRP, blue) are shown for ada2b (ada2b) and ada2b;NP6293-Gal4>UAS-Sk2 (ada2b;Glial Sk2). J, Average H3K9ac fluorescence intensity within peripheral glial nuclei for genotypes as in I. Sample sizes are as follows: ada2b, n = 204 nuclei; n = 3 animals; ada2b;Glial-Gal4>UAS-Sk2; n = 427 nuclei; n = 3 animals. U score = 15815. Mean ± SEM; ***p < 0.001; Mann–Whitney U test.
As H3K9ac is a direct target of SAGA complex-dependent acetylation (Qi et al., 2004; Carré et al., 2005; Weake et al., 2008), next, we assessed the effect of FTY720 and overexpression of Sk2 on glial H3K9ac in the ada2b mutant. We examined the average intensity of H3K9ac in peripheral glia nuclei by immunostaining. As expected, we found a significant decrease in H3K9ac intensity in the perineurial glia in ada2b mutant compared with the wild-type control [Fig. 7G,H; p = 1.5e-55 (z = 15.70); Kruskal–Wallis with post hoc Dunn's test]. In ada2b mutants treated with FTY720, we observed a mild increase in H3K9ac fluorescence intensity in perineurial glia when compared with untreated ada2b mutants, though this increase was not statistically significant [Fig. 7G,H; p = 0.10 (z = 1.64); Kruskal–Wallis with post hoc Dunn's test]. However, when we expanded our analysis to include the effect of FTY720 on H3K9ac levels across all peripheral glia, not just perineurial glia, we observed a significant increase in H3K9ac intensity in the ada2b mutant treated by FTY720 compared with the untreated ada2b group [Fig. 7G,H; p = 1.7e-5 (z = 4.92); Kruskal–Wallis with post hoc Dunn's test]. This suggests that FTY720 treatment enhances glial histone acetylation at the H3K9 site in the ada2b mutant. Furthermore, we found that the intensity of H3K9ac was significantly increased when we overexpressed UAS-Sk2 specifically in glia in the ada2b background [ada2b;NP6293-Gal4>UAS-Sk2; Fig. 7I,J; p = 2.3e-38 (U = 15,815); Mann–Whitney U test]. These results demonstrated that FTY720 and Sk2 directly regulate histone acetylation in glia (Fig. 7G–J). Together, we showed that FTY720 and overexpression of Sk2 are sufficient to rescue PHP and peripheral glial histone acetylation in the ada2b mutant.
Discussion
We demonstrated that neuronal overexpression of Aβ in Drosophila caused abnormal histone acetylation (H3K9ac) in glia and a complete disruption of PHP (Figs. 1, 3). We found that Fingolimod not only rescued glial histone acetylation and PHP during the larval stage, but also improved neurodegeneration and motor function during the adult stage in Drosophila Aβ models, highlighting a beneficial role of FTY720 in neurodegenerative disorders (Figs. 2, 3). Consistent with the effects of Fingolimod, we found that the endogenous Sk2, a kinase required for FTY720 activation, is required for both the rapid induction and maintenance of PHP (Figs. 4, 5). When we overexpressed Sk2 in glia, we found that glial H3K9ac, PHP, and motor function were all rescued in the Aβ overexpression models, mirroring the protective effects of FTY720 (Fig. 6).
Moreover, we explored the direct connection between Sk2 and histone acetylation. We found that FTY720 treatment or glial overexpression of Sk2 rescued PHP and H3K9ac deficits in a loss-of-function mutant of ada2b, which is an essential component in the SAGA complex (Fig. 7). Collectively, we demonstrated that glial sphingosine-mediated epigenetic regulation is critical in modulating homeostatic plasticity in Drosophila models of AD (Fig. 8). By leveraging genetic and electrophysiological methods, we not only provided evidence that PHP is a fundamental form of homeostatic regulation in maintaining synaptic function, but also revealed a protective role of sphingosine signaling pathway in synapse stabilization in AD.
Glial sphingosine signaling modulates PHP in Drosophila Aβ models. Neuronal overexpression of Aβ results in decreased histone acetylation in peripheral glia, leading to a disruption of glial epigenome-dependent downstream signaling (right side of the figure). This aberrant histone acetylation and subsequent transcriptomic alterations in glia impair PHP, resulting in synapse destabilization. Conversely, glial Sk2 can activate either endogenous sphingosine or exogenous FTY720, which act as HDAC inhibitors, thereby restoring histone acetylation and homeostatic plasticity in Drosophila models of AD (left side of the figure). We propose that glial epigenome-mediated processes, including the synthesis and secretion of signaling factors, are required for glial–neuron interactions, thus enabling PHP and synapse stabilization.
Impairment of homeostatic plasticity in Drosophila AD models
We found that while neuronal overexpression of Aβ completely disrupts PHP, homeostatic plasticity is still expressed when we overexpressed MAPT in neurons (Fig. 1C–F). This contrast in effects of Aβ and tau on PHP suggests that these two hallmarks might affect distinct signaling pathways involved in homeostatic plasticity. Furthermore, we demonstrated that glial histone acetylation (H3K9ac) is sensitive to neuronal Aβ, suggesting that extracellular Aβ could impair neuron–glial communications that are critical for homeostatic plasticity (Fig. 3A,B). The exact glial sensor for Aβ and the signal transducers that underpin the Aβ-mediated disruption of PHP remain compelling areas of inquiry for future investigations.
Existing evidence suggests that PHP can be directly induced by synapse loss and that synaptic excitation stability is maintained through transcriptional regulation of ion channels localized at presynaptic terminals (Orr et al., 2020). Interestingly, we observed PHP deficits without significant synaptic morphologic changes in conditions of neuronal Aβ42 and APP overexpression. Given that Aβ deposition occurs before the onset of neuronal loss or the manifestation of tau neurofibrillary tangles in AD (Long and Holtzman, 2019), our findings suggest a potential role of glial signaling in regulating homeostatic plasticity at the early stages of neurodegenerative disorders.
Fingolimod and Sk2-mediated sphingosine pathway in PHP
While previous studies suggest dysregulation of SPHK and S1P within AD brains, their exact functions in AD are still not fully understood (Takasugi et al., 2011; Ceccom et al., 2014; Couttas et al., 2014; Lei et al., 2019). In this study, we highlight the beneficial roles of Fingolimod and Sk2 in stabilizing synaptic physiology within Drosophila AD models. Not only did we find that glial Sk2 is crucial for PHP, but we also demonstrated that overexpression of Sk2 in glia is sufficient to restore PHP and glial histone acetylation under Aβ conditions (Figs. 4–6). Notably, expressing Sk2 RNAi to knock it down in motoneurons led to a decrease in EPSP amplitude in the presence of PhTX (Fig. 5D), despite a significant increase in quantal content (Fig. 5E). These tissue-specific knock-down experiments suggest that Sk2 primarily functions in glia for PHP, though we cannot exclude a partial defect in acute PHP when Sk2 is disrupted in motoneurons.
When we analyzed published snRNA-seq datasets from AD patients, we found that the expression level of SPHK2 in glia is not significantly altered in human AD brains (Fig. 6Q). Nevertheless, it is plausible that the SPHK2-mediated sphingosine signaling pathway may function as a key protective modulator of synaptic physiology in AD. It remains an intriguing question whether SPHK1 and SPHK2, expressed in mammalian glial cell types in the CNS, modulate synaptic homeostatic plasticity. Our findings shed light on the role of the glial sphingosine pathway in synapse stabilization. Molecules such as sphingosine that regulate histone acetylation could serve as potential therapeutic targets for treating neurodegenerative disorders.
We observed a reduced body size in Aβ models treated with 500 μm FTY720, a prodrug of Fingolimod, leading us to question whether these beneficial effects result from sphingosine pathway modulation or general metabolic changes. Previous studies suggest that caloric restriction may improve cognitive function and neurodegeneration pathology in AD (Morris et al., 2015; Van Cauwenberghe et al., 2016). However, our study presents genetic evidence that contradicts the hypothesis that the effects of FTY720 are solely caused by caloric restriction. First, we found that overexpressing Sk2 specifically in glial cells in the Aβ42 background rescued deficits in PHP (Fig. 6F–K) and H3K9ac in glial cells (Fig. 6L,M) and improved climbing behavior (Fig. 6N). Notably, these beneficial effects from glial Sk2 overexpression mirror the effects of FTY720 treatment. However, we did not observe significant body size changes when Sk2 was overexpressed in glial cells in the Aβ42 background. Second, treating the homozygous Sk2 mutant with 300 μm FTY720 did not rescue PHP. This observation argues that FTY720 functions through the Sk2-mediated signaling pathway (Fig. 6A,B,D). Additionally, previous studies have shown that acute fasting, but not amino acid restriction, disrupts PHP at the NMJ in Drosophila (Kauwe et al., 2016). While we cannot completely rule out the possibility that FTY720-mediated modulation of sphingosine signaling interacts with other pathways regulated by caloric restriction, it seems unlikely that the protective effects of FTY720 are solely because of caloric restriction.
Histone acetylation and glial signaling in AD
Recent advances in epigenetic studies underscore the complex roles of histone acetylation in AD (Nativio et al., 2018; Klein et al., 2019). Histone acetylation changes have been noted to occur in a brain region-specific, cell type-specific, and disease stage-dependent manner in neurodegenerative disorders (Marzi et al., 2018; Schueller et al., 2020; Santana et al., 2022). Mirroring human findings, we detected abnormal histone acetylation in Drosophila Aβ models (Figs. 3A,B, 6M). Remarkably, FTY720 treatment that rescued deficits in glial histone acetylation also restored homeostatic plasticity and motor function in these Drosophila models of AD (Figs. 2, 3). While the role of glial cells in regulating synaptic plasticity, neuroinflammation, and neurodegeneration is increasingly recognized, a comprehensive understanding of the glial epigenetic signaling network in AD remains elusive. To probe the role of glial sphingosine signaling pathway in AD, we used an integrative approach encompassing genetics, electrophysiology, imaging, and computational methods. Transcriptomic studies in AD patients offer insights into the downstream target genes directly regulated by histone acetylation. Whether rectifying individual target genes will be sufficient to alleviate synaptic plasticity deficits and neurodegenerative phenotypes associated with AD is an interesting question awaiting further investigation. Regardless, dissecting the epigenetic signaling pathways through genetic, electrophysiology, imaging, and computational approaches remains crucial for unraveling the mechanisms underlying neurodegeneration.
Footnotes
Work in the laboratory of Tingting Wang was supported by National Institutes of Health (NIH) Grants R01-NS-117372 and R21-NS-121284 (to Tingting Wang); the Simons Foundation Autism Research Initiative (SFARI) Bridge to Independence Award 551354 (to Tingting Wang); and the Brain and Behavior Research Foundation Young Investigator Award 27792 (to Tingting Wang). We thank Wang laboratory members Dr. Ken Kellar, Dr. Bill Rebeck, Dr. Dan Pak, and Sneha Thandra for discussions. We also thank Dr. Jairai Acharya (NIH/Neuroscience of Interoception and Chemosensation Study Section), Dr. Jerry Workman (Stowers Institute), and Dr. Marc Freeman (Vollum Institute) for sharing research materials.
The authors declare no competing financial interests.
- Correspondence should be addressed to Tingting Wang at tw652{at}georgetown.edu
