SAL alleviates neurite and mitochondrial injury induced by toxic CCCP or Aβ42 oligomers

We hypothesized that SAL might protect neuronal processes (also known as neurites) against damage in AD pathogenesis. To test this hypothesis, we first carried out experiments for neurite differentiation of SH-SY5Y cells using an established protocol (Fig. 1A), and then checked the SAL effect on neurites with or without the stimulation of mitochondrial toxin CCCP or Aβ42 oligomers. Previous findings suggested that 50 μM SAL had no toxic effect and showed good protective efficacy in cultured cells [32, 33]. Our data also showed that 50 μM SAL had better efficacy than 5 μM SAL on neurite extension of SH-SY5Y cells without cell toxicity (Additional file 1: Fig. S1). We thus used 50 μM SAL for the next cell culture studies. The brightfield images of neurite morphology indeed showed a protective action of SAL on neurite length under CCCP stimulation (Additional file 1: Fig. S2A–C). Compared to CCCP, Aβ42 also induced a weaker neurite injury of SH-SY5Y cells shown by fragmented processes, which was also reduced by SAL pre-treatment, as indicated by immunofluorescence (IF) staining of Tuj1 (a neuritic marker) (Fig. 1B, D). Of note, in primary neuronal culture, SAL exerted a remarkable action in neurite protection upon Aβ stimulation, with 2.7-fold increase of the total length, compared to PBS control (Fig. 1C, D). We further found that simultaneous incubation of SAL along with Aβ also inhibited Aβ-induced neurite damage by ~ 27.3% (Additional file 1: Fig. S3).

Fig. 1

SAL ameliorates Aβ-induced neurite and mitochondrial damage. A Schematic overview shows the workflow of SH-SY5Y differentiation and primary neuronal culture for SAL efficacy study by assessing neurite morphology and mitochondrial dynamics. B Immunofluorescence (IF) of neuronal marker Tuj1 (green) in differentiated SH-SY5Y cells, treated with indicated Aβ42 oligomers or SAL (see “Materials and methods”). White rectangles indicate amplified images at lower panels. Scale bar, 50 μm. C As in B, except primary neurons used. D Quantification of neurite length of B and C. The average neurite length without additional treatment was regarded as control (CTL). In each condition, at least 7 random images (each image includes ≥ 20 cells) from 3 independent experiments were used for quantification. E IF of TOM20 shows the mitochondrial segments in differentiated neurites of SH-SY5Y cells with indicated treatments. Scale bar, 10 μm. The quantification shows the length of mitochondrial (mito.) segments (n ≥ 400) in F, and the mitochondrial density indicating by the ratio of mito. length to occupied neurite (shaft) length (n ≥ 20) in G. Error bars indicate the mean ± SD from at least 20 neurites of 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001; one-way ANOVA test

Neurite injury is highly correlated to the impairment of mitochondria inside the processes. To assess whether SAL prevents mitochondrial damage stimulated by CCCP or Aβ42, we used IF of TOM20 (a mitochondrial marker) to observe their morphology in SH-SY5Y cells. As expected, mitochondria were shortened and fragmented by CCCP stimulation. Again, SAL repressed mitochondrial shortening by ~ 22.1% and their fragmentation by ~ 26.1% in soma (Additional file 1: Fig. S2D–F). We further assessed the mitochondrial protection of SAL on neurites. Like its protective action against CCCP in somata, SAL also remarkably inhibited mitochondrial truncation and fractionalization in neurites induced by Aβ42 (Fig. 1E–G).

These data suggest that SAL ameliorates CCCP or Aβ induced neurite and mitochondrial impairment.

SAL promotes neuronal mitophagy in both somata and neurites upon Aβ stimulation

Recent evidence refers that activation of mitophagy reverses Aβ pathology in AD mice [19, 34]. These findings lead us to test whether SAL elevates neuronal mitophagy. We first generated a derivative of the SH-SY5Y stable cell line expressing Mito-mKeima, to measure the level of mitophagy. The mKeima assays clearly showed a promotion of mitophagy around 2.8-fold in soma after SAL treatment (Fig. 2A, B). This result was further confirmed by fluorescence images of LC3-GFP and MitoTracker co-localization (Additional file 1: Fig. S4). In addition, we extracted mitochondrial protein from SH-SY5Y cells and examined the markers of mitophagic pathway. Immunoblotting showed that SAL indeed upregulated mitochondrial levels of PARKIN, PINK1 and LC3-II (Fig. 2C, D), suggesting an increase of mitophagy.

Fig. 2

SAL promotes mitophagy in somata and neurites. A Images of mito-mKeima expressing SH-SY5Y cells treated with SAL or PBS (CTL). 440 nm excitation (green) labels the healthy mitochondria, whereas 589 nm excitation (red) represents the mitochondria delivered to lysosomes for mitophagy. B Quantification of mitophagy index (ratio of fluorescence intensity of λEx = 589 nm against λEx = 440 nm) in A. Data from at least 7 random images in each condition from 3 independent experiments. C Immunoblotting shows the levels of indicated proteins extracted from cytosol (left panels) and mitochondria (right panels) of SH-SY5Y cells treated with or without SAL. GAPDH and VDAC as cytosolic (Cyto) and mitochondrial (Mito) fraction in C. D Quantification of band intensity in C. Data from 3 independent experiments. E IF images of cortical neurons transduced with AAV particles expressing mitoGFP, with indicated treatments. The puncta are labeled with autophagic marker LC3 (red) and mitochondrial GFP (green) and shown in upper panels. Images in white rectangles are amplified and shown in lower panels. F Quantitation shows the number of puncta co-labeled with LC3 and MitoGFP. G, H As described in E and F, except images of neurites are shown and quantified. Error bars indicates mean ± SD from at least 20 cells/neurites from 3 independent experiments (n ≥ 20). *P < 0.05, **P < 0.01, ***P < 0.001; one-way ANOVA test. Scale bar, 10 μm

By using adeno-associated virus (AAV) particles to express MitoGFP, which specifically labels mitochondria (Additional file 1: Fig. S5A), we transduced primary neurons followed by the treatment with Aβ. IF images of LC3 (red) and MitoGFP (green) co-localization showed that Aβ elevated mitophagy levels, indicated by increased yellow puncta, in somata. Notably, SAL further raised the number of yellow puncta (MitoGFP/LC3 co-localization) by 1.4-fold, compared to Aβ alone, demonstrating the augmentation of mitophagy upon SAL treatment (Fig. 2E, F). To evaluate the mitophagy levels in neurites, we further analyzed the LC3 and MitoGFP co-labeling and found that SAL treatment accelerated mitophagy by 1.5-fold, compared to stimulation with Aβ alone (Fig. 2G, H).

Our data imply that SAL promotes basal mitophagy levels and also exacerbates Aβ-induced mitophagy either in somata or in neurites. Possibly, SAL-triggered elevation of mitophagy may act as a protective mechanism against neurite injury, likely by increasing the clearance of damaged mitochondria [35].

SAL-activated mitophagy is dependent on SIRT3 expression

SIRT family proteins may regulate mitochondrial homeostasis including mitophagy [36]. To elucidate the mechanism of neurite protection and mitophagy activation of SAL, we measured the expression levels of mitochondrial SIRT3-5 in SH-SY5Y cells with and without SAL treatment. Interestingly, quantitative PCR (qPCR) assays indicated that SAL upregulated both SIRT3 and SIRT4, but not SIRT5 (Fig. 3A). Given that SIRT3 is the most thoroughly studied mitochondrial SIRT and its higher expression induced by SAL, our next study mainly focuses on SIRT3. Consistently, immunoblotting also showed a dose-dependent upregulation of SIRT3 protein levels upon SAL treatment, but with no change on the levels of APP and BACE1 (Fig. 3B). Although we may not entirely exclude other possible targets of SAL involved in this process, our data strongly suggest that SIRT3 is a critical molecule responsible for SAL-mediated neurite protection.

Fig. 3

SIRT3 is indispensable for SAL-mediated promotion of mitophagy. A qPCR shows the mRNA levels of SIRT3, 4 and 5 in SH-SY5Y cells treated with or without SAL. B Immunoblotting shows the levels of indicated proteins in SH-SY5Y cell treated by SAL with each indicated concentration. GAPDH as a loading control. C IF shows the colocalization of LC3-GFP and TOM20 in somata of SH-SY5Y cells with SIRT3 KD. Images in rectangles are amplified shown in lower panels. D As in C, except images of neurites are shown. Arrows indicate co-labeled yellow puncta. E Quantification shows the number of puncta co-labeled with LC3 and TOM20 in somata. F, G As in C, D, except SIRT3flox/flox cortical neurons transduced with AAV particles encoding mitoGFP and Cre were used and indicated as SIRT3 KO (see Additional file 1: Fig. S5). H As in E, except SIRT3 KO cortical neurons were used. Error bars indicates mean ± SD from at least 20 cells/neurites from 3 independent experiments (n ≥ 20). *P < 0.05, **P < 0.01, ***P < 0.001, n.s., not significant; one-way ANOVA test. Scale bar, 10 μm

To answer that whether SIRT3 is essential for SAL-activated mitophagy, we generated two SH-SY5Y cell lines, in which, one stably expressed shRNA against SIRT3 [SIRT3 knockdown (KD)] and another overexpressed SIRT3 (SIRT3 OE) (Additional file 1: Fig. S5B, D and E). The data showed that the colocalization of LC3 and mitochondria was not increased after SAL treatment in SIRT3 KD cells, but upregulated in SIRT3 OE cells even without SAL treatment (Additional file 1: Fig. S4). Notably, not like in naïve cells, Aβ-induced mitophagy in SIRT3 KD cells was not further augmented by SAL either in somata, or in neurites (Fig. 3C–E), suggesting the exclusive function of SIRT3 needed. To further confirm the SIRT3 function, we cultured primary neurons from Sirt3floxp mice and transduced them with AAV particles encoding MitoGFP and a Cre recombinase to obtain SIRT3 knockout (KO) cells in which mitochondria were simultaneously labeled with GFP (Additional file 1: Fig. S5A). IF images showed a similar effect of SAL in SIRT3 KO primary neurons as in SIRT3 KD cells (Fig. 3F–H).

Taken together, our data demonstrate that SIRT3 expression is indispensable for SAL-mediated activation of mitophagy.

SIRT3 is required for SAL-induced neurite and mitochondrial protection

To verify the possible function of SIRT3 on neurite protection, we used both naïve SH-SY5Y and SIRT3 KD cell line to differentiate neurites and then conducted Aβ-induced neurite injury assays. Truly, SAL exhibited a powerful neurite protection in naïve SH-SY5Y cells when treated with Aβ. In SIRT3 KD cells, Aβ triggered a more severer neurite damage, however the SAL protection was completely abolished (Fig. 4A, B). Our data thus demonstrate the essentiality of SIRT3 for SAL-mediated neurite preservation.

Fig. 4

SIRT3 KD inhibits SAL-mediated neurite and mitochondrial protection. A IF of Tuj1 (green) shows neurite morphology in Naïve and SIRT3KD SH-SY5Y cells with indicated treatment. Nuclei counterstained with DAPI. B Quantification of neurite length of A. C IF of TOM20 shows mitochondrial morphology in neurites of SIRT3KD SH-SY5Y cells with indicated treatment. Orange arrows, healthy mitochondria; white arrows, damaged mitochondria. Quantitation of images in C shows the length of mitochondrial segments (D) and mitochondrial density (E). F Fluorescence images show the intensity of aggregates (red) and monomers (green) of JC-1 staining in naïve and SIRT3KD cells with indicated treatment. G Quantification of JC-1 ratio of intensity in F. H Images show the fluorescence intensity of DCFH (ROS indicator) in cells with indicated treatment. I Quantitation of fluorescence intensity in H. Error bars indicate mean ± SD. *P < 0.05, **P < 0.01, ***P < 0.001; n.s., not significant; one-way ANOVA test. White scale bar, 50 μm; orange scale bar, 10 μm

Besides the activity of SIRT3 in mitophagy (Fig. 3), previous studies have also noted its function in the maintenance of mitochondrial homeostasis [37, 38]. We thereby investigated the potential capacity of SIRT3 in the homeostasis regulation of mitochondria by analyzing their morphology, membrane potential, and ROS production. Using IF of TOM20 (a mitochondrial marker), we detected the mitochondrial morphology in SIRT3 KD cells and found that mitochondrial segments in neurite were severely fragmented either by CCCP or Aβ treatment. Clearly, in SIRT3 KD cells, SAL did not show any protection on Aβ-induced mitochondrial damage, indicating by no restored mitochondrial length and density (Fig. 4C–E). The JC-1 assays also showed that the mitochondrial membrane potential (MMP) was impaired upon Aβ stimulation. Similarly, SAL treatment largely reduced the Aβ-induced MMP dysfunction in naïve cells, but its protection was similarly diminished in SIRT3 KD cells (Fig. 4F, G). We further measured the cellular ROS levels using a fluorescence probe DCFH. In accordance, Aβ-induced ROS upregulation was inhibited by SAL, only in naïve cells, not in SIRT3 KD cells (Fig. 4H, I).

Thus, our findings refer that SIRT3 is also requisite for SAL-mediated maintenance of mitochondrial homeostasis in neurite protection.

SIRT3 is essential for SAL-mediated cognitive restoration of AD mice

We next extended our study into a 5×FAD mouse model of AD and a workflow diagram was described to assess the SAL efficacy and SIRT3 function (Fig. 5A). Morris water maze (MWM) test showed that 5×FAD had a clear deficiency of spatial memory in contrast to wild type (WT) mice. Indeed, oral administration of SAL sharply improved the learning and memory (Fig. 5B–E), which was in line with previous data from other AD animal models, in which SAL reduced cognitive impairment as well [39, 40]. We used an AAV injection approach for SIRT3 KD and successfully transduced the cells in the whole hippocampus (Additional file 1: Fig. S5A and S6A). As expected, SIRT3 KD in WT did not change the mouse behavior (Additional file 1: Fig. S6B–D). Strikingly, SIRT3 KD in hippocampi of 5×FAD, nearly completely abolished the protective efficacy of SAL (Fig. 5B–E). Moreover, we used Y-maze test to measure spatial reference memory. In accordance with the MWM test, it revealed that SAL strongly improved cognitive function in 5×FAD mice. Again, SIRT3 KD blocked the protective effect of SAL (Fig. 5F–H). Overall, these data imply that SAL can ameliorate cognitive decline in AD mice, which is dependent on SIRT3 expression.

Fig. 5

SAL-mediated cognitive protection is dependent on SIRT3. A A schematic overview shows the workflow of our animal study. B Quantitation shows the escape latency of mice from indicated groups from day 1 to day 5 during training. Significant differences are found in 5×FAD group compared to WT (**P < 0.01, ***P < 0.001) or 5×FAD+SAL group (##P < 0.01, ###P < 0.01). C Representative trajectories of mice from each group in MWM tests at day 6. Red square, starting point; blue square, ending point; platform (red circle) located in quadrant 2. Quantitation shows the traveled distance (D) and time spent (E) in the platform located quadrant. F Representative heatmap images show the visit frequency of mice in Y-maze. Arrow, the novel arm. Quantification shows the times of entry (G) and traveled distance (H) in the novel arm. All data are represented as mean ± SEM. *P < 0.05, **P < 0.01, ***P < 0.001, n.s., not significant; n ≥ 6; one-way ANOVA

SAL mediates a SIRT3-dependent reduction of Aβ pathology and neurite damage in AD mice

To elucidate whether SAL can mitigate Aβ pathology, a main cause for cognitive defect in 5×FAD, we examined the intensity of Aβ plaque, microgliosis and astrogliosis in hippocampi of 5×FAD mice. In fact, SAL greatly reduced the area of Aβ plaque (Aβ+ IHC) by 41.9%, microgliosis (Iba1+ IHC) by 49.3%, and astrogliosis (GFAP+ IHC) by 31.4%, respectively. We further analyzed the total number of Aβ plaques and the number of Aβ plaques with different sizes (diameter < 20 μm, 20–40 μm and > 40 μm). Data showed a SAL-mediated reduction of Aβ load, especially on medium and larger size of Aβ plaques (Additional file 1: Fig. S7). Consistent with the behavioral tests, SIRT3 KD inhibited the SAL-mediated reduction of these pathological features (Fig. 6A–E and Additional file 1: Fig. S7).

Fig. 6

SAL alleviates Aβ pathology and restores neurite morphology in AD mice via SIRT3 action. A IHC staining of Aβ (red) and Iba1 (green) in hippocampi of 5×FAD mice. Nuclei counterstained with DAPI (blue). White rectangles indicate amplified images at lower panels. B As in A, except Aβ (red) and GFAP (green) are stained. Quantification shows the Aβ+ areas (C), Iba1+ areas (D) and GFAP+ areas (E). F The tracing images show the morphology of Golgi-stained CA1 pyramidal neurons in mice with indicated treatments. Orange triangle indicates the soma. G Sholl analysis of images collected in F. Significant differences are found in 50 μm to 100 μm (X-axis) in 5×FAD+SAL group compared to 5×FAD (**P < 0.01) or 5×FAD+SIRT3KD+SAL group (##P < 0.01). H Quantification of total dendrite length of cells in F. I Golgi-stained images show the morphology of apical and basal dendrites of CA1 pyramidal neurons. Quantification shows the spine density on apical dendrites (J) and basal dendrites (K). Data mean ± SD from at least 12 slices of 6 mice. *P < 0.05, **P < 0.01, ***P < 0.001; one-way ANOVA. White scale bar, 500 μm; black scale bar, 50 μm; orange scale bar, 5 μm

Next, we aimed to test whether SAL could alleviate neurite damage in brains of 5×FAD mice, like it does in cell cultures. Although knowing the difficulty of the technology, we tried to use IHC and detect the LAMP1 level as a marker for dystrophic neurites. Data indeed showed that SAL reduced the levels of LAMP1 nearby Aβ plaques, while SIRT3KD partially inhibited the SAL function (Additional file 1: Fig. S8). Next, we used Golgi silver staining to observe and trace neurite morphology of CA1 pyramidal neurons. The drawings in Fig. 6F displayed that the morphology of CA1 neurons in 5×FAD was simplified in contrast to WT, while SAL treatment evidently rescued this phenotype. Likewise, SIRT3 KD minimized the SAL potency on neurite protection (Fig. 6F). The sholl analysis, as established previously [41], allowed the quantification of branching and length of the dendrites. As expected, 5×FAD had fewer dendritic intersections than WT at a distance from soma around 50–100 μm, and SAL markedly increased dendritic intersections, across the same region (P < 0.01). Of note, SIRT3 KD did significantly blocked SAL efficacy (P < 0.01) (Fig. 6G). Conformably, the quantitation of total dendritic length showed a promotive effect of SAL, that was still dependent on SIRT3 expression (Fig. 6H). Using the same Golgi staining, we further studied the spines locating at both apical and basal dendrites. In keeping with the results of dendritic arborization (Fig. 6F–H), SAL showed a favorable function on spine density, in a similar fashion, contingent on the expression of SIRT3 (Fig. 6I–K).

Collectively, these results support the notion that SAL alleviates Aβ pathology and dendritic impairment in brains of AD mice and its efficacy is dependent on SIRT3 expression.

SAL upregulates SIRT3 transcription by directly inhibiting NRF2-KEAP1 complex formation

Considering that the transcription factor NRF2 can regulate SIRT3 expression in some systems including aging process [42, 43], we thus hypothesized that SAL may regulate NFR2 to mediate SIRT3 transcription. To test our hypothesis, we first used immunofluorescence to check the NRF2 localization in SH-SY5Y cells incubated with low and high concentration of SAL, or potent NRF2 activator sulforaphane (SFN). As expected, SFN strongly increased nuclear levels of NRF2. Interestingly, either 5 μM or 50 μM SAL induced an obvious translocation of NRF2 from cytosol to nuclei, although with a milder potency compared to SFN (Fig. 7A, B). In addition, immunoblotting showed that SAL treatment upregulated the protein levels of NRF2 as well as SIRT3 (Fig. 7D). However, the mRNA levels of NRF2 did not change (Fig. 7C). These data refer that the NRF2 upregulation was not due to its transcription. By detecting nuclear and cytosolic NRF2, we further confirmed that SAL promoted the nuclear translocation of NRF2 (Fig. 7E). Given that SIRT3 mRNA was induced by SAL (Fig. 3A), a possible mechanism of NRF2-mediated SIRT3 transcription was further tested in the next study.

Fig. 7

SAL directly inhibits NRF2-KEAP1 complex and facilitates NRF2-induced SIRT3 transcription. A IF of NRF2 (green) and nuclei (blue) in SH-SY5Y cells. Nuclei counterstained with DAPI. Scale bar, 50 μm. B Quantification shows the nucleus-to-cytoplasm ratio of NRF2 levels. C qPCR shows the NRF2 mRNA levels in SH-SY5Y cells treated with indicated dose of SAL. D Immunoblotting images (left) and quantification of band intensity (right) show the protein levels of NRF2 and SIRT3 in SH-SY5Y cells with or without SAL treatment. E As in D, except cytosolic (Cyto) and nuclear (Nuc) protein were isolated and detected. F Immunoprecipitation (IP) shows the protein levels of NRF2 and KEAP1-Flag in SH-SY5Y cells with indicated treatment. An anti-Flag antibody was used for IP. 5% total protein used as input. G Surface plasmon resonance (SPR) analyses show the SAL-NRF2 interaction behavior. Quantitation shows the sensogram of each indicated titration of SAL, as expressed in RU (response units-Y axis) along time (X-axis). H Cellular thermal shift assays (CETSA) show the binding capacity of SAL-NRF2 interaction. Immunoblotting shows NRF2 levels and below graph shows the CETSA melting curve. I Immunoblotting shows the NRF2 levels in protein lysates at 55 °C. The below graph shows the ITDR-CETSA curve. Error bars indicate mean ± SD from at least 3 independent experiments. *P < 0.05, **P < 0.01, ***P < 0.001, n.s., not significant; one-way ANOVA

The E3 ligase adaptor KEAP1 is a principle negative regulator for NRF2. Thereby, we tested whether SAL could disrupt the formation of NRF2-KEAP1 protein complex. In KEAP1 immunoprecipitation (IP), 50 μM SAL definitely reduced the interaction of NRF2 with KEAP1 (Fig. 7F). We also checked whether SAL directly bound to NRF2 by using a real-time surface plasmon resonance (SPR) technology. To obtain the pure NRF2 protein for SPR use, we expressed the recombinant human NRF2 in Ecoli and purified it, showing by the Coomassie staining (Additional file 1: Fig. S9). SPR data indicated that SAL truly bound to NRF2 with a strong affinity (Fig. 7G), demonstrating NRF2 as a direct target of SAL. The interaction of SAL and NRF2 was further confirmed by the cellular thermal shift assays and data showed that induced a much stronger stability of NRF2 protein (Fig. 7H, I).

Thus, our findings demonstrate that SAL directly binds to NRF2 and promotes its nuclear translocation by inhibiting attachment of the KEAP1 E3 ligase, ultimately resulting in an elevated transcription of SIRT3.