DSS rescued cognitive impairment in APP/PS1 mice

The Y-maze test demonstrated that the alternation rate of APP/PS1 mice (model group) was lower than that of WT mice (control group, p < 0.01), and was significantly improved after DSS administration (p < 0.01, Fig. 2A-C). We found that the three different doses of DSS revealed equivalent and even better effects than the donepezil group (positive control group, p < 0.01, Fig. 2B). The Morris water maze test was performed to examine the cognitive capabilities of the mice. During the training days, the model group showed a longer escape latency than the control group, and this trend was markedly reversed by treatment with low or high doses of DSS (p < 0.01; Fig. 2D, E). In the probe trial, mice in the model group had the lowest number of crossings. In contrast, the DSS group revealed more crossings and longer stay times in the target quadrant than the APP/PS1 mice (model group) (p < 0.01, Fig. 2F, G). Furthermore, mice in the DSS group had shorter escape latency than those in the model group (p < 0.01, Fig. 2H). Taken together, these behavioural tests suggest that DSS restores the cognitive decline in APP/PS1 mice.

Fig. 2

DSS improved cognitive function in APP/PS1 mice. (A) Schematic diagram of spontaneous alternation. (B) Alternation rates of all groups of mice in the Y-maze test. (C) Representative track images of mice in the Y-maze test. (D) Representative track images of mice in the Morris water maze (MWM) test. (E) Escape latency during training days in the MWM test. (F) Numbers of target crossings in the MWM test. (G) Time stayed in the target quadrant in the MWM test. (H) Escape latency of all groups of mice in the MWM test. Data are shown as mean ± SEM, n = 10 mice per group. ##p < 0.01 vs. the control group. *p < 0.05, **p < 0.01 vs. the model group, ns, no significant difference. Con, control group; Mod, model group; LiC, inhibitor group; Don, positive control group; DH, high dose of DSS group; DM, medium dose of DSS group; DL, low dose of DSS group. One-way ANOVA, followed by LSD method multiple comparisons tests (B, E, G, and H) or non-parametric independent samples Kruskal–Wallis test (F)

Effect of DSS on neuronal survival and Aβ plaque

Multiple studies have confirmed the vital role of neuronal apoptosis in AD pathology. Therefore, we investigated whether DSS could attenuate neuronal apoptosis in the hippocampus of AD mice. Representative hippocampal microphotographs of Nissl and H&E staining revealed that neuronal cells in the CA3 and DG areas of the hippocampus in the control group were uniformly distributed with clear edges and were neatly arranged (Fig. 3A, B). In contrast, the neuronal cells in the model group were damaged and loosely arranged, with abnormal cell morphology and wrinkled cytoplasm. After treatment with DSS, Li2CO3, or donepezil, the number of damaged neuronal cells in the hippocampus of AD mice was substantially reduced, and a remarkable improvement in the number, morphology, and structure of the neuronal cells was observed (Fig. 3A, B).

Since Aβ plaque is an important pathological hallmark of AD, we next performed TS staining to examine whether DSS treatment affected Aβ deposition in AD mice. TS staining results showed that TS-positive Aβ plaques in brain regions, including the cortex and hippocampus, decreased significantly with the administration of DSS (p < 0.01) and Li2CO3 (p < 0.01) compared to those in the model group (Fig. 3C, D), suggesting that DSS could reduce the Aβ burden in the brains of APP/PS1 mice.

Fig. 3

Effect of DSS on neuronal survival and Aβ plaque. (A) Images of Nissl’s staining and haematoxylin and eosin (H&E) staining of the CA3 and DG areas in the hippocampus of different groups. (B) Cell counts in CA3 and DG regions in Nissl’s staining and semi-quantitative analysis scores in H&E staining. (C) Images of TS staining in brain regions. (D) Quantification of the Aβ plaques in TS staining. Data are presented as mean ± SEM, n = 3 mice per group. ##p < 0.01 vs. the control group. *p < 0.05, **p < 0.01 vs. the model group. Con, control group; Mod, model group; LiC, inhibitor group; Don, positive control group; DH, high dose of DSS group; DM, medium dose of DSS group; DL, low dose of DSS group. One-way ANOVA, followed by LSD method multiple comparisons tests (B, D)

D-T network analysis and screening process of the core ingredients in DSS

We first developed a D-T network of herbal ingredients in DSS, which consisted of 6,514 DTIs that interacted with 1,118 ingredients and 218 AD genes (Fig. 4A). Multiple herbal ingredients were connected with numerous AD genes in this network, with an average ingredient degree of 5.1 for each gene. The top 10 ingredients with the highest degree numbers were as follows: augustic acid (CID15560128, D = 141), quercetin (CID5280343, D = 69), capsaicin (CID1548943, D = 63), apigenin (CID5280443, D = 44), luteolin (CID5280445, D = 44), gallic acid (CID370, D = 39), oleanolic acid (CID10494, D = 30), caffeic acid (CID689043, D = 29), kaempferol (CID5280863, D = 29), and niacinamide (CID936, D = 27). Emerging evidence indicates that these ingredients may have potent therapeutic effects against AD. Network analysis indicated that capsaicin interacts with 63 AD-related genes, suggesting its potential against AD. Indeed, an in vivo study showed that a diet high in capsaicin reduced the incidence and development of AD by altering the gut microbiome and serum metabolome [20] and that capsaicin could reduce neurodegeneration, neuroinflammation, and deterioration in spatial memory in the AD model [21]. Apigenin has also been reported to reduce AD symptoms in transgenic Drosophila models of AD [22]. Moreover, a recent study revealed that luteolin effectively improved cognitive deficits in 3 × Tg-AD mice and inhibited Aβ-induced oxidative stress, mitochondrial dysfunction, and neuronal apoptosis via a PPARγ-dependent mechanism, which may serve as a therapeutic agent for AD [23]. In summary, ingredients with a high degree number in the D-T network are likely to have good therapeutic potential against AD.

Network analysis indicated that 28 of the 218 AD genes had degree numbers greater than or equal to 30 (Fig. 4A). Among them, MAPT (K = 906) had the highest number of ingredient connections, followed by PTPN1 (K = 728), and AChE (K = 550). A growing body of literature has confirmed their vital role in AD. For instance, the miR-124/PTPN1 pathway has been identified as a critical mediator of synaptic dysfunction and memory loss and could be regarded as a promising novel therapeutic target for patients with AD [24]. Moreover, in vivo research revealed that correcting abnormalities in miR-124/PTPN1 signalling rescued tau pathology [25], also highlighting its importance in AD. Furthermore, multiple studies have confirmed that AChE inhibitors can reduce Aβ levels in AD both in silico and in vivo [26].

To further identify the core ingredients in DSS, we integrated the ingredients of each herb in DSS and evaluated the ADMET properties (Fig. 4B). The Sankey diagram shows that Danggui had the highest number of herbal ingredients (n = 616), followed by Chuanxiong (n = 461), and Baizhu (n = 199). Herbal ingredients can be classified into eight categories: benzenoids, organoheterocyclic compounds, organic acids and derivatives, phenylpropanoids and polyketides, organic oxygen compounds, hydrocarbons, lipids, and lipid-like molecules. The top three categories were lipids and lipid-like molecules (452/1118 = 40.4%), phenylpropanoids and polyketides (132/1118 = 11.8%), and benzenoids (127/1118 = 11.4%), indicating that these herbal ingredients may have great potential against AD. Moreover, the ADMET results suggested that 920 herbal ingredients had good HIA and 590 of them were predicted to cross the BBB. After PPB screening, 253 herbal ingredients were identified that could be the core ingredients of DSS (Table S3). Interestingly, we found that multiple core herbal ingredients were also consistent with ingredients with high degrees in the D-T network (Fig. 4A).

Fig. 4

Drug-target network analysis and screening process of core ingredients in DSS. (A) This network comprises 6,514 DTIs, which interacts with 1,118 ingredients and 218 AD genes. The labels of the top 10 ingredients and AD genes with degrees ≥ 30 are shown. The font size of the labels and the size of the nodes are proportional to the degree. (B) Sankey diagram showing the screening process of the core ingredients in DSS. HIA, human intestinal absorption; BBB, blood-brain barrier; PPB, plasma protein binding. Herbal ingredients are classified according to the chemical taxonomy provided by ClassyFire [37]

4D-FastDIA quantitative proteomics analysis

On the basis of these in vivo pharmacodynamic results, we selected a high dose of DSS (DH) for subsequent proteomic and metabolomic experiments to investigate the MOA of DSS against AD. To determine the quantitative repeatability of the data, we calculated the relative standard deviation (RSD) values of the control, model, and DH groups based on the relative quantitative values of the repeated samples in each group. RSD values of the three groups were < 0.2, indicating that the data were stable and reliable, with good quantitative repeatability (Fig. 5A). Next, we integrated the protein strength values of each sample and found that their distributions were comparatively centralised and that the sample means were at the same level, suggesting good quality of the samples (Fig. 5B). Furthermore, we characterised the protein alterations resulting from the control, model, and DH treatment groups by performing 4D-FastDIA quantitative proteomic analysis of hippocampal tissue derived from APP/PS1 mice. In total, 111 DEPs were identified, 90 of which were significantly upregulated, and 22 were downregulated in the control and model groups (Fig. 5C). Meanwhile, 69 DEPs were also identified, including 43 upregulated and 26 downregulated DEPs, between the model and DH groups, indicating that hippocampal tissue samples in the DH group showed significant changes in protein levels compared to those in the model group. We also performed an overlap analysis of the DEPs between the control group vs. the model group and the model group vs. the DH group and found that 10 DEPs were simultaneously regulated by three groups: Wdfy1, Slc6a20b, Mt-Cyb, Omp, Mef2c, Sphk1, Rpl14, Pcbp3, Clic6, and Scgn (Fig. 5D). Furthermore, we developed the core ingredient-AD gene network and protein-protein interaction (PPI) network of DEPs to investigate their relationships. Network analysis indicated that these core components interacted with 118 AD genes and 36 DEPs connected to 125 protein-protein interactions (PPIs). Remarkably, we found that two genes (ACHE and CD44) were regulated by the core ingredients of DSS and DEPs, suggesting that the core ingredients of DSS could act on these DEPs to exert anti-AD effect (Fig. 5E).

Fig. 5

Proteomic analysis of hippocampus tissue of APP/PS1 mice after treatment of high dose of DSS. (A) Relative standard deviation and (B) intensity of the control, model, and high dose of DSS (DH) groups based on identified proteins, n = 4 mice per group. (C) Differentially expressed proteins were identified from comparisons of the control vs. model groups, the model vs. DH groups, and the control vs. DH groups. (D) Venn diagram showing the differentially expressed proteins between the control vs. model groups and the DH vs. model groups. (E) The core ingredient-gene network and the protein-protein interactions network. N, control group; M, model group; DH, high dose of DSS group

To explore the potential MOAs in which the DEPs of the three different groups may be involved, we also performed a GO enrichment analysis of DEPs, including the biological process (BP), cellular component (CC), and KEGG pathway. These DEPs participate in multiple BPs, including mitochondrial ATP synthesis-coupled electron transport, ATP synthesis-coupled electron transport, and oxidative phosphorylation (Fig. 6A). Emerging literature has shown that these BPs are highly related to EM [27]. Furthermore, the CC annotations suggested that these DEPs were also associated with EM, such as the inner mitochondrial membrane protein complex and the mitochondrial respirasome (Fig. 6A). KEGG pathway analysis indicated that these DEPs could participate in EM (Fig. 6B). Taken together, the preliminary proteomic analysis demonstrated that DSS might relieve AD by modulating the EM.

Fig. 6

Enrichment analysis of differentially expressed proteins from three different groups. (A) Gene Ontology (GO) enrichments of differentially expressed proteins between the control, model, and DH groups. (B) KEGG functional classifications of differentially expressed proteins. *p < 0.05, **p < 0.01, ***p < 0.001. n = 4 mice per group. N, control group; M, model group; DH, high dose of DSS group

Targeted energy metabolic profiling

As proteomic analysis indicated that DSS might act on EM to exert a therapeutic effect on AD, we next performed a targeted energy metabolic profile to determine the energy metabolites regulated by DSS. The PCA scoring plot suggested that the model group was clearly separated from the control group (Fig. 7A), indicating a significant change in the type or level of the metabolites. Moreover, the model group was separated from the DH group (Fig. 7C) after treatment with DSS, implying that metabolic levels in the model group could be regulated by DSS. The reliability of the OPLS-DA classification model was confirmed using 200 permutation tests, and the results for the control and model groups were R2X = 0.764, R2Y = 0.983, and Q2 = 0.728 (Fig. 7B). The results between the model and DSS groups were R2X = 0.484, R2Y = 0.889, and Q2 = 0.687 (Fig. 7D), indicating that there was no overfitting of the OPLS-DA model, which had good predictive power. Targeted energy metabolic profiling of the three groups identified and quantified 47 energy metabolites in 18 samples (Fig. 7E). Next, we integrated some important metabolites in the EM and determined the differences in the levels of these metabolites in the three groups (Fig. 7F). Moreover, eight significant metabolite biomarkers, including 3-phenyllactic-acid, L-alanine, L-cysteine, serine, uracil, argininosuccinic acid, citric acid, and sedoheptulose-7-phosphate, were associated with DH treatment after applying the screening criteria (see Materials and Methods 2.6.2, Fig. 7G, H). We further analysed the relative peak area changes of eight significant metabolite biomarkers and found that three metabolites (serine, L-alanine, and 3-phenyllactic-acid) were downregulated (p < 0.05) in the control group vs. the model group, whereas their levels were upregulated after treatment with DSS (p < 0.05, DH group vs. the model group, Fig. 7I).

Fig. 7

Serum metabolic profile of APP/PS1 mice after DSS treatment (n = 6). (A) PCA and (B) OPLS-DA score plots between the control and model groups. (C) PCA and (D) OPLS-DA score plots between the model and high dose of DSS (DH) treatment group. (E) Heatmap of the 47 differential metabolites for three groups. (F) Schematic diagram of some detected metabolites associated with EM. (G,H) Volcano plot of differential metabolites of the control vs. model groups and the model vs. DH groups. (I) Change in the relative peak area of the metabolites. Data are expressed as mean ± SEM. n = 6 mice per group. #p < 0.05 vs. the control group. *p < 0.05 vs. the model group. **p < 0.01 vs. the model group

Metabolite-AD gene network of DSS

Because changes in gene expression might directly influence metabolite production and consumption, which mutually affect disease progression, we next explored the relationship between differential serum metabolites and AD genes. AlzGPS is a genome-wide positioning system platform that catalyzes multi-omics for Alzheimer’s drug discovery [28]. First, we integrated the metabolite-associated genes from the AlzGPS database and performed an overlap analysis between the metabolite-associated genes and AD disease genes to highlight metabolite-associated AD genes. As shown in Fig. 8, the differential metabolite-AD gene network consisted of 30 differential metabolites and 90 AD genes. Among them, L-aspartate (degree = 15) had the highest number of gene connections, followed by succinic acid (degree = 14), and arginine (degree = 12). SLC16A10 interacts with seven metabolites, whereas ASS1 interacts with five metabolites. We found that 14 AD genes, including HPRT1, ABCA1, TPI1, MDH1, OXCT1, NOS1, NOS3, GAD1, GAD2, ENO1, GPI, EGFR, CBS, and GOT1, had close metabolite connections that were simultaneously regulated by DSS, suggesting that DSS might act on these AD genes to regulate the related differential metabolites.

Fig. 8

Metabolite-AD gene network of DSS. The green diamond and the purple square represent the differential metabolites related to energy metabolism and genes regulated by DSS, while the yellow dot node denotes the genes highly associated with these differential metabolites integrated through the AlzGPS database

DSS promoted the brain glucose uptake

Brain glucose uptake has been demonstrated to play a vital role in cellular energy supply, and reduced GLUT levels in patients with AD could impair glucose availability, which may accelerate neuronal death and ultimately lead to brain dysfunction and memory loss [29]. In this section, we aimed to assess the brain glucose uptake capacity by measuring the gene and protein expression of GLUT1 and GLUT4 using qPCR and WB analysis. qPCR results showed that the mRNA levels of GLUT1 and GLUT4 in the cortex were significantly decreased in the model group compared to those in the control group, whereas their gene expression levels were remarkably upregulated after DSS (p < 0.01), Li2CO3 (p < 0.05), and donepezil (p < 0.01) administration in APP/PS1 mice (Fig. 9A, B). WB results showed that the protein expression of GLUT1 and GLUT4 was upregulated in the cortex of APP/PS1 mice after the administration of DSS, Li2CO3, and donepezil (Fig. 9D, E). In the hippocampus, a significant decrease in the protein expression levels of GLUT1 and GLUT4 was observed in the model group (p < 0.05, p < 0.01) compared to the control group, while their levels were also upregulated after treatment with DSS (p < 0.05, p < 0.01), Li2CO3 (p < 0.01), and donepezil (p < 0.05, Fig. 9F-G).

Brain-derived neurotrophic factor (BDNF) is reported to be associated with Aβ accumulation, tau phosphorylation, neuroinflammation, and neuronal apoptosis, which may play a potential role in the pathogenesis of AD [30]. Therefore, we used qPCR analysis to assess the mRNA expression level of BDNF in the cortex, and WB analysis to evaluate the protein expression levels of BDNF in the cortex and hippocampus. The qPCR results showed that the mRNA expression level of BDNF was significantly decreased in the model group (p < 0.01), whereas DSS (p < 0.05), Li2CO3 (p < 0.01), and donepezil (p < 0.05) treatments remarkably upregulated BDNF gene expression in AD mice (Fig. 9C). Meanwhile, WB results suggested that there was no statistically significant difference in the protein expression level of BDNF in the cortex (Fig. 9D, E). However, BDNF expression in the hippocampus of mice that received DSS (p < 0.05), Li2CO3 (p < 0.05), and donepezil (p < 0.05) exhibited a remarkable increase compared with that in the model group (Fig. 9F, G).

Fig. 9

DSS promoted brain glucose uptake in APP/PS1 mice. (A-C) qPCR analysis showing the GLUT1, GLUT4, and BDNF levels in the cortex. (D-G) Protein expression levels of GLUT1, GLUT4, and BDNF in the cortex and hippocampus by western blot analysis. Data are shown as mean ± SEM, n = 3 mice per group. #p < 0.05, ##p < 0.01 vs. the control group. *p < 0.05, **p < 0.01 vs. the model group, ns, no significant difference. Con, control group; Mod, model group; LiC, inhibitor group; Don, positive control group; DH, high dose of DSS group; DM, medium dose of DSS group; DL, low dose of DSS group. One-way ANOVA, followed by LSD method multiple comparisons tests (A, B, C, E, and G)

DSS improved mitochondrial function and relieved oxidative stress

As mitochondrion plays a crucial role in cellular EM, we first confirmed whether DSS could improve mitochondrial function. MMP, an important indicator of cellular EM, was detected, and a significant increase in MMP was observed after administration of DSS (p < 0.05) and Li2CO3 (p < 0.05) compared to that in the model group (Fig. 10A). Since the functional and structural integrity of mitochondria is also reflected in ATP and NADH levels [31], we next measured their content using commercial kits. ATP and NADH levels were markedly upregulated in APP/PS1 mice treated with DSS (p < 0.01) and Li2CO3 (p < 0.01) (Fig. 10B, C). Given that the mitochondrial respiratory chain is an important component of cellular EM, we investigated the effect of DSS on the levels of complexes I-IV. Consistent with the protective effect of DSS on mitochondrial membrane integrity, DSS significantly increased the levels of complexes I (p < 0.01), II (p < 0.01), III (p < 0.01), and IV (p < 0.05, p < 0.01) (Fig. 10D-G). In summary, DSS can promote EM and improve mitochondrial function in APP/PS1 mice.

Mitochondria are the main sources and generators of intracellular ROS. Overproduction of ROS results in oxidative stress and mitochondrial dysfunction [31]. To explore the effect of DSS on oxidative stress, ROS levels in mouse brain tissue and T-SOD and MDA levels in mouse serum were measured. The model group showed excessive ROS production compared to the control group (p < 0.01), and DSS treatment (p < 0.01) eliminated ROS overload in the brains of AD mice (Fig. 10H). T-SOD levels decreased significantly in the model group and were markedly upregulated by DSS treatment (Fig. 10I). Meanwhile, elevated MDA levels in the model group were significantly reduced after DSS treatment (p < 0.01) (Fig. 10J).

Fig. 10

DSS improved mitochondrial function and relieved oxidative stress in APP/PS1 mice. (A) MMP measurement. (B,C) ATP and NADH content (n = 6). (D-G) Mitochondrial respiratory chain I-IV content. (H-J) The relative content of ROS, total SOD, and MDA. Data are presented as mean ± SEM, n = 6−9 mice per group. ##p < 0.01 vs. the control group. *p < 0.05, **p < 0.01 vs. the model group, ns, no significant difference. Con, control group; Mod, model group; LiC, inhibitor group; Don, positive control group; DH, high dose of DSS group; DM, medium dose of DSS group; DL, low dose of DSS group. One-way ANOVA, followed by LSD method multiple comparisons test (A, B, C, D, E, F, H, and I) or Tamhane’s T2 method test (G, J)

DSS regulated GSK3β/PGC1α signalling pathway

Since GSK3β was shown to be a regulator of EM in the brain [10], and also played an important role in the D-T network (Fig. 2A), we next determined whether DSS could alleviate cognitive deficit through the GSK3β/PGC1α signalling pathway. The qPCR results showed that the mRNA expression level of GSK3β in the model group increased (p < 0.05), while its level was significantly downregulated after DSS treatment (p < 0.01) (Fig. 11A). Regarding the PGC1α, the reduced mRNA expression level of PGC1α in the model group (p < 0.01) was markedly reversed by treatment with DSS (p < 0.05, p < 0.01) and Li2CO3 (p < 0.01) (Fig. 11B). The WB results suggested that, in the hippocampus and cortex, the protein expression levels of p-GSK3β and PGC1α were decreased in the model group, while the increased expression levels could be detected in APP/PS1 mice after administration of DSS (p < 0.05, p < 0.01) and Li2CO3 (p < 0.01) (Fig. 11C-F).

Fig. 11

DSS regulated GSK3β/PGC1α signalling pathway. (A,B) The relative mRNA expression level of GSK3β and PGC1α in mouse cortex by qPCR analysis. (C-F) Protein expression of GSK3β, p-GSK3β and PGC1α in mouse cortex and hippocampus by western blot analysis. Data are presented as mean ± SEM, n = 3 mice per group. #p < 0.05, ##p < 0.01 vs. the control group. *p < 0.05, **p < 0.01 vs. the model group, ns, no significant difference. Con, control group; Mod, model group; LiC, inhibitor group; DON, positive control group; DH, high dose of DSS group; DM, medium dose of DSS; DL, low dose of DSS group. One-way ANOVA followed by LSD difference multiple comparison tests (A, B, D, and F)