Sphingosine kinase 1-associated autophagy differs between neurons and astrocytes

Abstract

Autophagy is a degradative pathway for removing aggregated proteins, damaged organelles, and parasites. Evidence indicates that autophagic pathways differ between cell types. In neurons, autophagy plays a homeostatic role, compared to a survival mechanism employed by starving non-neuronal cells. We investigated if sphingosine kinase 1 (SK1)-associated autophagy differs between two symbiotic brain cell types-neurons and astrocytes. SK1 synthesizes sphingosine-1-phosphate, which regulates autophagy in non-neuronal cells and in neurons. We found that benzoxazine autophagy inducers upregulate SK1 and neuroprotective autophagy in neurons, but not in astrocytes. Starvation enhances SK1-associated autophagy in astrocytes, but not in neurons. In astrocytes, SK1 is cytoprotective and promotes the degradation of an autophagy substrate, mutant huntingtin, the protein that causes Huntington's disease. Overexpressed SK1 is unexpectedly toxic to neurons, and its toxicity localizes to the neuronal soma, demonstrating an intricate relationship between the localization of SK1's activity and neurotoxicity. Our results underscore the importance of cell type-specific autophagic differences in any efforts to target autophagy therapeutically.

Fig. 1. SK1-associated autophagy differs between neurons and astrocytes.

a Cultured primary cortical neurons were treated with a vehicle (control, cont), or with 10-NCP (10-NCP, 5 µM), or fluphenazine (FPZ, 5 µM) for 4 h. Lysates were analyzed by western blotting with antibodies against LC3. Actin was used as a loading control. Bar graphs represent the quantification of the LC3-II intensities normalized to actin. *p (cont vs 10-NCP) = 0.042, *p (cont vs FPZ) = 0.024. b Cultured primary astrocytes were treated with a vehicle (control, cont), or with 10-NCP (10-NCP, 5 µM), or fluphenazine (FPZ, 5 µM), with or without NH₄Cl for 4 h. Lysates were analyzed by western blotting with antibodies against LC3. Actin was used as a loading control. Bar graphs represent the quantification of the LC3-II intensities normalized to actin. p LC3 (cont vs 10-NCP) = 0.0749, p LC3 (cont vs FPZ) = 0.5297 (one-way ANOVA). ***p (cont vs cont + NH₄Cl) = 0.0002 (t-student), p (cont + NH₄Cl vs 10-NCP + NH₄Cl) = 0.7699, p (cont + NH₄Cl vs FPZ + NH₄Cl) = 0.2257 (one-way ANOVA). Results were pooled from three independent experiments. c, d Cultured primary cortical neurons (c) and primary astrocytes (d) were treated with a vehicle (control, cont) or with 10-NCP (5 µM), or fluphenazine (FPZ, 5 µM) for 4 h. Lysates were analyzed by western blotting with antibodies against phosphorylated SK1 (pSK1) or pan-SK1 (SK1). Actin was used as a loading control. Bar graphs represent the quantification of the pSK1 and SK1 band intensities normalized to actin. In neurons: *p pSK1 (cont vs 10-NCP) = 0.0269, *p pSK1 (cont vs FPZ) = 0.0224; p SK1 (cont vs 10-NCP) = 0.7939, p SK1 (cont vs FPZ) = 0.208 (one-way ANOVA). In astrocytes: p pSK1 (cont vs 10-NCP) = 0.05356, p pSK1 (cont vs FPZ) = 0.9909; p SK1 (cont vs 10-NCP) = 0.1158, p SK1 (cont vs FPZ) = 0.4113 (one-way ANOVA). Results were pooled from three independent experiments. e Primary astrocytes were transfected with GFP and RFP-LC3 or with SK1-GFP and RFP-LC3. Twenty-four hours after transfection, astrocytes were imaged. Scale bar is 10 µm. A part of the image is zoomed in to visualize the SK1-GFP and RFP-LC3 puncta. f Quantification of the RFP-LC3 puncta index from e. ***p (GFP vs SK1-GFP) = 0.0001 (t test). One hundred and fifty cells were analyzed from three independent experiments. g The photoswitchable protein Dendra2 targeted to LC3 as a surrogate for the flux through autophagy. Brief irradiation with short-wave-length visible light causes Dendra2-LC3 to undergo an irreversible conformational change (“photoswitch”) and emit red fluorescence that can be tracked until altered Dendra2-LC3 is cleared. A cell is outlined to show cellular morphology. Scale bar is 10 µm. h Primary astrocytes were transfected with Dendra2-LC3 and an empty plasmid or with Dendra2-LC3 and a plasmid that encodes Beclin1, or Dendra2-LC3 and SK1. After the photoswitch, astrocytes were longitudinally imaged. The change in red fluorescence intensity over time was used to calculate the half-life of Dendra2-LC3. The single-cell half-life of Dendra2-LC3 was reduced by SK1 expression. Beclin1 was used as a positive control. Fifty astrocytes were analyzed per condition from two independent experiments. ***p (control vs Becl) = 0.0001, ***p (control vs SK1) = 0.0001; p (Becl vs SK1) = 0.084 (one-way ANOVA). i, j Cultured primary astrocytes were maintained in DMEM supplemented with 10% fetal bovine serum (control, cont) or in Hanks’ balanced salt solution (starvation, starv) for 4 h. Samples were also treated with 10 mM NH₄Cl to block the last step of autophagy-mediated degradation. Lysates were analyzed by western blotting with antibodies against LC3 (i) or with antibodies against phosphorylated SK1 (pSK1) or pan-SK1 (SK1) (j). Actin was used as a loading control. Bar graphs represent the quantification of the LC3-II, pSK1, or SK1 intensities normalized to actin. *p LC3 (control vs starv) = 0.017; p pSK1 (control vs starv) = 0.0072, p SK1 (control vs starv) = 0.331 (one-way ANOVA). Results were pooled from three independent experiments. n.s. not significant

Fig. 2. Overexpressed SK1 protects astrocytes during starvation.

a An example of longitudinal imaging of astrocytes. Primary astrocytes were transfected with mApple to visualize their morphology. After transfection, the same group of astrocytes were imaged longitudinally with an automated microscope at different time points. The first image is a montage of non-overlapping images captured in one well of a 24-well plate. Scale bar is 400 µm. The adjacent panels are zoomed in to three cells, to demonstrate longitudinal single-cell tracking. Scale bar is 50 µm. Arrows indicate two astrocytes that died before the last imaging event. b Astrocytes were transfected with mApple and GFP, as control, or with mApple and SK1-GFP, or with mApple and a plasmid that encodes a dominant-negative form of SK1 (dnSK1) tagged to GFP. SK1- or dnSK1-expressing cells were maintained in DMEM supplemented with 10% fetal bovine serum (SK1 or dnSK1) or in Hanks’ balanced salt solution (starvation, starv; SK1 + starv or dnSK1 + starv) and tracked with an automated microscope for 36 h. Risk of death curves demonstrate that SK1 expression protects astrocytes during starvation. ***p (dnSK1 vs dnSK1 + starv) = 0.0001, **p (GFP vs dnSK1) = 0.0253; p (GFP vs SK1) = 0.495, p (SK1 vs SK1 + starv) = 0.2184 (log-rank test). Fifty astrocytes per group were analyzed from three independent experiments. c Primary cortical astrocytes were maintained in basal conditions (control, cont), or in Hanks’ balanced salt solution (starvation, starv) overnight. The levels of S1P were measured by liquid chromatography and mass spectrometry. The bar graph represents relative S1P levels. ***p = 0.0001 (t test). Results were pooled from three independent experiments. d A cohort of astrocytes was transfected with mApple and SK1-GFP. Two cohorts of astrocytes were transfected with mApple and a plasmid that encodes a dominant-negative form of SK1 (dnSK1) tagged to GFP. Astrocytes were maintained in Hanks’ balanced salt solution (SK1 + starv or dnSK1 + starv). A cohort of dnSK1-expressing cells was treated with 1 µM S1P (dnSK1 + starv + S1P). Astrocytes were longitudinally imaged with an automated microscope. The addition of S1P partially restored survival of dnSK1-expressing astrocytes under starvation. ***p (SK1 + starv vs dnSK1 + starv) = 0.0001, **p (dnSK1 + starv vs dnSK1 + starv + S1P) = 0.0287, p (SK1 + starv vs dnSK1 + starv + S1P) = 0.0131 (log-rank test). Fifty astrocytes per group were analyzed from two independent experiments. n.s. not significant

Fig. 3. SK1 reduces the half-life of an autophagy substrate in astrocytes.

a The photoswitchable protein Dendra2 was fused to Htt^ex1-Q₄₆ to measure autophagy flux. Brief irradiation with short-wave-length visible light causes Dendra2 to undergo an irreversible conformational change (“photoswitch”) and emit red fluorescence that can be tracked until altered Htt^ex1-Q₄₆-Dendra2 is cleared. Scale bar is 10 µm. b Primary astrocytes were transfected with Htt^ex1-Q₄₆-Dendra2 and an empty plasmid (control, cont), or with Htt^ex1-Q₄₆-Dendra2 and a plasmid that encodes Beclin1 (Becl), or with Htt^ex1-Q₄₆-Dendra2 and a plasmid that encodes SK1 (SK1). Beclin1 was used as a positive control. After a “photoswitch,” astrocytes were longitudinally imaged. The change in the red fluorescence intensity over time was used to calculate the half-life of Htt^ex1-Q₄₆-Dendra2. The single-cell half-life of Htt^ex1-Q₄₆-Dendra2 was significantly reduced by SK1 expression. Fifty astrocytes per group were analyzed from two independent experiments. ***p (control vs Becl) = 0.0001, ***p (control vs SK1) = 0.0001; p (Becl vs SK1) = 0.255 (one-way ANOVA). n.s. not significant

Fig. 4. Overexpressed SK1 is neurotoxic for primary cortical neurons.

a An example of survival analysis in neurons. Primary cortical neurons were transfected with mApple to visualize neuronal morphology. The same group of neurons was imaged 24 h after transfection and tracked over time with an automated microscope. Images collected after 24 h demonstrate the ability to return to the same field of neurons and to follow them over time. Each image in the top panels is a montage of non-overlapping images captured in one well of a 24-well plate at different time points (24, 48, and 72 h). Scale bar is 400 μm. In the bottom panel, a region from the original images is zoomed in to demonstrate longitudinal single-cell tracking. Black arrows depict two neurons that develop differently over time. The neurons on the left degenerate and disappear before 72 h after transfection, and the neuron on the right remains alive until the end of the experiment. Scale bar is 50 μm. b Primary cortical neurons were transfected with mApple and GFP or with mApple and SK1-GFP and tracked with an automated microscope for 72 h. Risk of death curves demonstrate that SK1-GFP expression is neurotoxic. **p (GFP vs SK1-GFP) = 0.0032 (log-rank test). Two hundred neurons were analyzed from three independent experiments. c Primary cortical neurons were transfected with mApple and GFP or with mApple and SK1-GFP and imaged thereafter. Note that the SK1-expressing neuron died, while the control neuron was alive until the end of the experiment. Scale bar is 10 µm. d Primary neurons were transfected with mApple and GFP or with mApple and SK1-GFP. Neurons were imaged 24 h after transfection, and the green fluorescence intensity was measured in each neuron. To determine the dose-dependent toxicity in cortical primary neurons that express GFP or SK1-GFP, the green fluorescence intensities in individual neurons were correlated with the time at which each cell died. The bar graphs represent the correlate of average of GFP and SK1-GFP fluorescence intensities with neuronal longevity. Note that neuronal survival is not correlated with the expression of GFP. The graph bar contains the linear correlation slopes of GFP- and SK1-GFP-expressing neurons. The SK1-GFP fluorescence intensity of each neuron was correlated with the cell's risk of death. The SK1-GFP intensity is correlated with a higher risk of death. m (GFP) = −0.986; m (SK1-GFP) = −24.864. ***p = 0.0001 (t test). Two hundred neurons were analyzed from three independent experiments

Fig. 5. Localization of SK1-GFP puncta predicts neuronal death.

a Cortical primary neurons were transfected with mApple and SK1-GFP, and imaged. Images demonstrate two representative neurons with the first neuron containing SK1-GFP puncta in its neurites (top panel, scale bar is 20 µm) and the second neuron containing SK1-GFP puncta in the soma (bottom panel, scale bar is 10 µm). Note that the second neuron died, whereas the first neuron remained alive. b Cortical primary neurons were transfected with mApple and SK1-GFP, and imaged. Bar graphs represents the percentage of neurons containing SK1-GFP puncta in the soma or in neurites 24 h after transfection, which eventually died or remained alive by the end of the experiment (72 h). **p SK1-GFP in the soma (dead vs alive) = 0.0032, *p SK1-GFP in neurites (dead vs alive) = 0.0018 (t test). One hundred neurons were analyzed from two independent experiments. c Primary cortical neurons were transfected with mApple and SK1-GFP. Twenty-four hours after transfection, cells were treated with a vehicle or with 0.5 µM 10-NCP, or with 5 µM 10-NCP for 4 h. Neurons were then imaged. Scale bar is 10 µm. d Single-neuron analyses from (c). SK1-GFP fluorescence intensity was analyzed in neurites (black bars) and in the neuronal soma (gray bars). **p (cont vs 0.5) = 0.0144 and *p (cont vs 5) = 0.0041 in neurites. **p (cont vs 0.5) = 0.0144 and *p (cont vs 5) = 0.0041 in the soma. p (0.5 vs 5) = 0.848 (one-way ANOVA). Fifty neurons were analyzed from two independent experiments. A.u. arbitrary units, n.s. not significant

Fig. 6. Overexpressed SK1 induces excessive autophagy in the soma.

a Three cohorts of primary neurons were transfected with RFP-LC3, GFP, and an empty plasmid. The fourth cohort of neurons was transfected with RFP-LC3, GFP, and a plasmid that encodes SK1. The first three cohorts of transfected neurons were treated with a vehicle (control), or with 0.5 µM 10-NCP or with 5 µM 10-NCP for 4 h, respectively. Neurons were imaged 24 h after transfection. Scale bar is 20 µm. b Quantification of the RFP-LC3 puncta index from a. SK1-expressed neurons and neurons treated with 5 µM 10-NCP exhibit higher RFP-LC3 puncta index compared to control (cont). ***p (cont vs 5 µM 10-NCP) = 0.0001, **p (cont vs SK1) = 0.0019, *p (cont vs 0.5 µM 10-NCP) = 0.0342 (one-way ANOVA). Two hundred neurons were analyzed from three independent experiments. c The graph represents a correlation between RFP-LC3 puncta index from b and risk of death of four cohorts of neurons (neurons transfected with RFP-LC3, GFP, and an empty plasmid, and treated with a vehicle; neurons transfected with RFP-LC3, GFP, and a plasmid that encodes SK1; neurons transfected with RFP-LC3, GFP, and an empty plasmid, and treated with 0.5 µM 10-NCP for 4 h; and neurons transfected with RFP-LC3, GFP, and an empty plasmid, and treated with 5 µM 10-NCP for 4 h. ***p (cont vs 5 µM 10-NCP) = 0.0001, ***p (cont vs SK1) = 0.0001; p (cont vs 0.5 µM 10-NCP) = 0.352 (log-rank test). Two hundred neurons were analyzed from three independent experiments. d Four cohorts of primary cortical neurons were cultured. Three cohorts of neurons were nucleofected with GFP. The first cohort was treated with a vehicle (control, cont). The second and the third cohorts were treated with 0.5 or 5 µM 10-NCP for 4 h. The fourth cohort was nucleofected with SK1-GFP (SK1). Neurons were plated and maintained for 36 h. Neurons were then collected, and the levels of S1P were measured by liquid chromatography and mass spectrometry. The bar graph represents the S1P levels normalized with S1P content in the control group. ***p (cont vs SK1) = 0.0001, *p (cont vs 0.5 µM 10-NCP) = 0.0274, ***p (cont vs 5 µM 10-NCP) = 0.0012 (one-way ANOVA). Results were pooled from three independent experiments. A.u. arbitrary units, n.s. not significant