Fasudil improved cognitive function in APP/PS1 mice
The MWM test was used to assess the effect of fasudil on cognitive function in APP/PS1 mice at the age of 12 months. During the 5-d training period, APP/PS1 mice treated with saline (ADNS) showed significant increases in the latency to locate the platform (Fig. 1A), latency of the first entrance into the target zone (SW) (Fig. 1B), and decreases in the time (%) spent in SW (Fig. 1C) when compared with WT + saline (WT) (p < 0.05 or p < 0.01). In contrast, APP/PS1 mice treated with fasudil (ADF) showed significant decreases in the two latencies and increases in the time (%) in the target zone, as compared with ADNS (p < 0.05 or p < 0.01; Fig. 1A-C). Representative raw traces of swimming in the probe trial test (Day 6) were shown in Fig. 1D. The traces of ADNS mice were much longer than WT, which was shortened in ADF. Similarly, the number of crossings over the target area was significantly lower in ADNS compared to WT, which was attenuated in ADF (p < 0.01; Fig. 1E). These results suggest that the impairment of learning and memory in the AD mice were reversed or attenuated by treatment with fasudil.
Fasudil treatment decreased senile plaques and neurofibrillary tangles in the hippocampus of AD mice
To identify Aβ plaques and neurofibrillary tangles in the hippocampus, we used Bielschowsky silver staining, the most commonly used method for examining senile plaques and neurofibrillary tangles[34, 35]. It was found that a large number of senile plaque deposition and neurofibrillary tangles were accumulated in the hippocampus of ADNS mice, but not in WT controls (Fig. 2A, B). This was remarkably attenuated as both senile plaques and neurofibrillary tangles were obviously decreased in the hippocampus in ADF relative to ADNS mice.
Fasudil treatment reversed the specific alterations of gut microbial diversity in AD mice
To assess the effects of fasudil on gut microbiota alterations related to the pathogenesis of AD, we used Principal Component Analysis (PCA) by variance decomposition to reflect the differences between groups using APP/PS1 mice and WT controls. PCA revealed based on species annotation abundance that ADNS showed different species abundance composition from WT and ADF, both of which had similar composition of species abundance (Fig. 3A). In addition, we performed species annotated relative abundance bar chart analysis, which showed the dominant phyla in all groups, including p-Bacteroidetes (23.7% − 44%) and p-Firmicutes (6.4% − 26.6%) (Fig. 3B and Additional file 1: the dominant phyla in all groups). In these two most dominant phyla, the relative abundance ratio of Firmicutes/Bacteroidetes was increased in ADNS (59.1%) compared to WT (31.7%), while decreased in ADF (32.8%) relative to ADNS (Fig. 3C, D), suggesting the progression of AD may be associated with a high proportion of Firmicutes/Bacteroidetes, which can be lowered to normal following fasudil treatment. Further, we analyzed the data at the family levels and observed that f-Bacteroidaceae, f-Prevotellaceae, f-Lachnospiraceae, and f-Bacteroidaceae were relatively the most abundant in all groups (Fig. 3E and Additional file 2: the level of family in all groups ).
The species levels of s_Bacteroides_dorei_CAG222 (p < 0.05), s_Bacteroidetes_bacterium_OLB8 (p < 0.05), s_Prevotella_sp_CAG1031 (p < 0.05), and s_Prevotella_sp_CAG873 (p < 0.01) in ADF were significantly higher compared to ADNS, which exhibited significantly lower abundance compared to WT (p < 0.05), suggesting that fasudil treatment reversed these species abundance (Fig. 4 upper panels and Additional file 3: the level of species in all groups ). Similarly, the abundance levels of some species in ADNS were significantly lower compared to WT, including s_Alistipes_finegoldii (p < 0.01), s_Alistipes_sp_CAG53 (p < 0.05), s_Alistipes_sp_CAG435 (p < 0.05), and s_Butyricimonas_synergistica (p < 0.01) (Fig. 4 middle panels and Additional file 3: the level of species in all groups ). The abundance levels were increased after fasudil treatment, but with no significant difference between ADF and ADNS. In contrast, ADNS showed significantly more abundance in s_Helicobacter_saguini, s_Helicobacter_typhlonius, and s_Helicobacter_sp_MIT_03-1616 compared to WT (p < 0.05) (Fig. 4 lower panels and Additional file 3: the level of species in all groups ), which was reduced in ADF, with statistical significance in s_Helicobacter_saguini (p < 0.05), suggesting that fasudil treatment blocked AD-induced increases in the abundance of these species, especially s_Helicobacter_saguini. Finally, we used Lefse (LDA Effect Size, linear discriminant analysis) combined with the statistical analysis to screen key biomarkers [36, 37]. The LDA scores (log10 > ± 3) indicating more abundance at the species level in ADNS and ADF are shown in Fig. 5. The results of metagenomics demonstrated that s_Prevotella_sp_CAG873 was identified as an ADF potential biomarker, while s_Helicobacter_typhlonius and s_Helicobacter_sp_MIT_03_1616 were identified as ADNS potential biomarkers in the fecal of APP/PS1 mice (Fig. 5).
Fasudil treatment altered gut metabolites in APP/PS1 mice
To determine the effect of fasudil on changes in gut metabolites associated with AD, we carried out metabolomic analysis using UHPLC-QTOFMS and PLS-DA cluster analysis, the latter of which was used to further determine the trends of metabolic changes in ADNS, relative to WT or ADF. All samples were analyzed with a 95% confidence interval (Hotelling's t-squared ellipse; Fig. 6A). The results from the screening of different metabolites were visualized in the form of volcano plots (Fig. 6B, C). There were 295 different metabolites in ADNS vs. WT, including 117 downregulated and 178 upregulated metabolites (Fig. 6B); 335 different metabolites in ADNS vs ADF, including 185 downregulated and 150 upregulated metabolites (Fig. 6C). Further clustering analysis of differential metabolites in each group revealed that ADNS had a different heat map from WT and ADF, both of which were presented similarly (Fig. 6D). In addition, the 295 differential metabolites in ADNS vs. WT were enriched in 28 signaling pathways (Fig. 6E and Additional file 4: the differential metabolites in ADNS vs. WT were enriched in 28 signaling pathways), focusing on the metabolisms of pyruvate, glycolysis/gluconeogenesis, fructose and mannose, citrate cycle (TCA cycle), amino sugar, and nucleotide sugar; the 335 differential metabolites in ADNS vs. ADF were concentrated in 20 signaling pathways focusing on the metabolisms of pyrimidine, purine, glycolysis/gluconeogenesis, glycerophospholipid, and fatty acid degradation (Fig. 6F and Additional file 5: the differential metabolites in ADNS vs. ADF were concentrated in 20 signaling pathways).
Association analysis of the metagenomic and metabolomic profiles
To examine the correlation between metagenomic and multiple metabolites in AD, we analyzed the metabolites from the gut microbiota and the host in APP/PS1 mice. By taking the intersection of ADNS-WT and ADNS-ADF, 83 important, differential metabolites were obtained and narrowed down to 60 different metabolites by adjustment of p < 0.04 (Additional file 6: the 60 differential metabolites by adjustment of p < 0.04 ), among which 20 important, different metabolites were screened as per available literatures (Additional file 7: the correlation analysis of the species and 20 differential metabolites (p value)).
Correlation analysis of genus levels in top 30 abundance of gut microbiota and 20 different metabolites were presented in Fig. 7A. Some gut microbiotas were correlated with a single metabolite, including g_Clostridium, which had a positive correlation with dTDP-4-oxo-2,3, 6-trideoxy-d-glucose (p < 0.05), and g_Faecalibaculum, which showed a positive correlation with DG (22:4(7Z,10Z,13Z, 16Z)/24:0/0:0) (p < 0.01). Others were associated with a variety of metabolites (Additional file 7: the correlation analysis of the species and 20 differential metabolites (p value)).
Furthermore, we examined 20 different metabolites correlation with the ADNS biomarkers (s_Helicobacter_typhlonius, s_Helicobacter_sp_MIT_03-1616) or the ADF biomarker (s_Prevotella_sp_CAG873). In addition, it was shown that s_Bacteroides_dorei_CAG222 and s_Bacteroidetes_bacterium_OLB8 were significant lower in ADNS compared to WT; this was reversed in ADF (p < 0.05; Fig. 4 upper; Additional file 3: the level of species in all groups). The correlation analysis revealed that the two species were correlated with 14 metabolites, including 9-ribofuranosyl hypoxanthine, leukotriene C5, thymine, dTDP-4-oxo-2,3,6-trideoxy-D-glucose, alpha-amino-4-carboxy-3-furanpropanoic acid, L-dopachrome, UDP-4-dehydro-6-deoxy-D-glucose, prolyl-gamma-glutamate, 2-hydroxy-3-[4-(sulfooxy)phenyl]propanoic acid, CDP-DG(18:0/18:0), leukotriene F4, TG(22:5(4Z,7Z,10Z,13Z,16Z), TG(22:1(13Z)/22:6(4Z,7Z,10Z,13Z,16Z,19Z), and hexanoyl-CoA, which were widely related to a variety of metabolisms, such as carbohydrate metabolism and fatty acid metabolism. (Fig. 7B,Additional file 7: the correlation analysis of the species and 20 differential metabolites (p value)).
Heatmap was used to show the correlation results in Fig. 7B, which showed a specific positive correlation between the ADNS biomarkers (s_Helicobacter_typhlonius, s_Helicobacter_sp_MIT_03-1616), and 5 metabolites, including leukotriene F4, CDP-DG (18:0/18:0), TG (22:5 (4Z,7Z,10Z,13Z,16Z), TG (22:1(13Z)/22:6 (4Z,7Z,10Z,13Z,16Z,19Z), hexanoyl-CoA (p < 0.01). All the 5 metabolites were significantly higher in ADNS than in WT, significantly lower in ADF than in ADNS, and returned to the WT levels in ADF (Additional file 7: the correlation analysis of the species and 20 differential metabolites (p value)). These 5 metabolites are mainly involved in lipid metabolism, indicating that fasudil might protect neurons by decreasing hexanoyl-CoA, a short-chain fatty acyl-CoA precursor. Notably, fasudil also significantly decreased leukotriene F4, which is involved in the proinflammatory pathways.
There were 8 metabolites specifically correlated with the ADF biomarker (s-Prevotella sp_CAG873), four of which were significantly higher in ADNS compared to WT or ADF, including alpha-amino-4-carboxy-3-furanpropanoic acid, prolyl-gamma-glutamate, 2-aminoadenosine, and 2-hydroxy-3-[4-(sulfooxy) phenyl] propanoic acid; in other word, ADF reduced ADNS-induced high contents of these metabolites to the WT levels (Additional file 8). The other four metabolites were significantly lower in ADNS relative to WT or ADF, including 9-Ribofuranosyl hypoxanthine, L-dopachrome, UDP-4-dehydro-6-deoxy-D-glucose, and TG (22:0/o-18:0/22:0); in other words, ADF increased these metabolites in ADNS to the WT levels (Additional file 7: the correlation analysis of the species and 20 differential metabolites (p value)). The 8 metabolites were highly related to carbohydrate, nucleotide, and fatty acid metabolism.
It was noted that two of the metabolites, i.e., leukotriene C5 and thymine, were correlated with both the ADF biomarker (s_Prevotella sp_CAG873) and the ADNS biomarkers (s_Helicobacter_typhlonius, s_Helicobacter_sp_MIT_03-1616). They were significantly higher in ADNS compared to WT or ADF; in other words, ADF decreased the two metabolites in ADNS to the WT levels (Fig. 7B, Additional file 7: the correlation analysis of the species and 20 differential metabolites (p value)).
Functional annotations of association metabolites with genes of biomarkers of ADF or ADNS
In the enrichment analysis of functional annotations of association of metabolites with genes of biomarkers of ADF or ADNS, we observed significant pathways mainly involved in amino sugar, nucleotide sugar, pyrimidine metabolisms. As shown in Table 1, UDP-4-dehydro-6-deoxy-D-glucose was correlated with the ADF biomarker (s_Prevotella sp_CAG873), while thymine was correlated with both the ADF biomarker (s_Prevotella sp_CAG873) and the ADNS biomarker (s_Helicobacter_sp_MIT_03-1616). Both the ADNS and ADF biomarkers contained genes involved in signaling pathways of the pyrimidine metabolism (ko00240), encoded various enzymes, and further influenced the production of thymine. More specifically, the thioredoxin reductase gene (trxB) in the ADNS biomarker (s_Helicobacter_sp_MIT_03-1616) encoded the enzyme thioredoxin-disulfide reductase (EC:1.8.1.9) (blue rectangles, Fig. 8; Table 1), likely affecting the refolding of post-translation protein. In the ADF biomarker (s-Prevotella sp CAG873), genes such as the RNA polymerase subunit C gene (rpoC), cytidine deaminase gene (cdd, CDA), thymidine kinase gene (tdk), holA gene-encoded one subunit of DNA polymerase III holoenzyme (holA), dUTPase gene (dut), and uridine diphosphate gene(udp) respectively encoded the following enzymes: DNA-directed RNA polymerase (EC:2.7.7.6), cytidine deaminase (EC:3.5.4.5), thymidine kinase (EC:2.7.1.21), DNA-directed DNA polymerase (EC:2.7.7.7), dUTP diphosphatase (EC:3.6.1.23), and uridine phosphorylase (EC:2.4.2.3) (pink rectangles, Fig. 8; Table 1).
Table 1
Functional annotations of associated metabolites with genes of biomarkers of ADF or ADNS mice
Name | KO_id | Gene_name | EC | metabolites |
s_Prevotella sp. CAG:873 ko00520(Amino sugar and nucleotide sugar metabolism) | K07106 | murQ | EC:4.2.1.126 | UDP-4-dehydro-6-deoxy-D-glucose |
K01835 | pgm | EC:5.4.2.2 |
K12373 | HEXA_B | EC:3.2.1.52 |
K02472 | wecC | EC:1.1.1.336 |
K01443 | nagA, AMDHD2 | EC:3.5.1.25 |
s__Prevotella sp. CAG:873 ko00240(Pyrimidine metabolism) | K03046 | rpoC | EC:2.7.7.6 | Thymine |
K01489 | cdd, CDA | EC:3.5.4.5 |
K00857 | tdk, TK | EC:2.7.1.21 |
K02340 | holA , DPO3D1 | EC:2.7.7.7 |
K01520 | dut, DUT | EC:3.6.1.23 |
K00757 | udp, UPP | EC:2.4.2.3 |
s_Helicobacter sp. MIT 03-1616 ko00240(Pyrimidine metabolism) | K00384 | trxB | EC:1.8.1.9 | Thymine |