The Ethics Committee for Animal Research at China Medical University in Shenyang has sanctioned experiments involving animals. We procured neonatal Sprague-Dawley rodents from the Liaoning Changsheng Biotechnology company, situated likewise in Shenyang. These rodents were accommodated in a habitat that maintained a consistent 12-h cycle of light and darkness, with uninterrupted access to food and water.

To induce cerebral WMD in neonatal rats, we followed the procedure described below which was previously employed by our research team (Vannucci et al., 1999; Cheng et al., 2015). At the age of 3 days, both male and female Sprague-Dawley rodents were sedated using isoflurane before being positioned on their backs on the surgery table. Under an anatomical microscope, we exposed the left common carotid artery. The rats were randomly assigned to one of three groups: the Sham group, HI group, and Caffeine group. In the HI group, we permanently ligated both ends of the left common carotid artery using sterile needle sutures and then severed the segment between the ligations. The incision was closed with sutures, and the entire surgery lasted for ~8–10 min. Following recovery from anesthesia, the rats were reunited with their dams for a period of 1 h before being transferred to a low oxygen box maintained at a constant temperature of 37°C. In the low oxygen box, a gas mixture composed of 8% oxygen and 92% nitrogen was continuously administered at a flow rate of 2 L/min for a duration of 2.5 h, resulting in an oxygen concentration of 8%. The Sham group underwent left common carotid artery dissection without ligation or exposure to hypoxia.

From post-natal (P) days 2 to 6, the Caffeine group were administered 20 mg/kg/day caffeine citrate via abdominal cavity injections once daily for five consecutive days. In the Sham and HI group, an equivalent volume of physiological saline was injected instead. The caffeine citrate used in the study was sourced from Casey Pharmaceuticals (Casey Pharmaceuticals, Parma, Italy). Furthermore, in the Caffeine+AK-7 group, rats were subjected to treatment with both caffeine and the sirtuin 2 (SIRT2) inhibitor 3-(1-azepanylsulfonyl)-N-(3-bromphenyl) benzamide (AK-7) (s591401, Selleck Chemicals, USA). AK-7 was administered intraperitoneally at a dosage of 20 mg/kg/day for a period of 5 days, commencing from day 2 to day 6 after birth.

Two weeks after the model was established (on the 17th postnatal day), we harvested and immediately preserved the brain's white matter by freezing it on a precooled platform using liquid nitrogen. For preservation, we stored the brain samples at −80°C until further analysis. Each sample was gently thawed, subjected to ultrasonic disruption for finer separation, and then centrifuged at 12,000 revolutions per minute for 10 min at a temperature of 4 degrees Celsius. After centrifugation, we carefully extracted the supernatant from the cellular waste and placed it into a new tube. To deplete the most common 14 proteins, we used a depletion kit from Thermo Fisher Scientific's Pierce™ brand. The protein levels were then measured with a bicinchoninic acid assay kit from Beyotime. For protein digestion, we initially reduced the proteins with 5 mM dithiothreitol at 56 degrees Celsius for half an hour. The alkylation step followed by adding 11 mM iodoacetamide and incubating the solution for 15 min at ambient temperature shielded from light. We diluted the urea concentration to under 2 M using 100 mM tetraethylammonium bromide before adding trypsin at a protein-to-trypsin mass ratio of 1:50, allowing digestion to occur overnight. A secondary digestion was performed for an additional 4 h at a ratio of 1:100. In order to avoid interference in mass spectrometry analysis, we try to avoid the introduction of surfactants during protein extraction and pretreatment of proteome samples. At the same time, before mass spectrometry, we remove non peptide components from the samples through C18 solid-phase extraction to minimize matrix effects.

Tryptic peptides were dissolved in solvent A, which consisted of 0.1% formic acid and 2% acetonitrile in water. The peptide solution was then loaded directly onto a homemade reversed-phase analytical column with dimensions of 25 cm length and 75 μm i.d. The separation of peptides was achieved using solvent B, a gradient ranging from 5% to 25% consisting of 0.1% formic acid in 90% acetonitrile, over a period of 60 min. Subsequently, the gradient was adjusted to range from 25% to 35% over 22 min, followed by a sharp increase to 80% over 4 min. The system was maintained at 80% for the final 4 min. These separations were performed at a constant flow rate of 450 nL/min on an EASY-nLC 1200 UPLC system manufactured by Thermo Fisher Scientific.

The separated peptides were analyzed using a Q ExactiveTM HF-X mass spectrometer manufactured by Thermo Fisher Scientific. The analysis was performed using a nano-electrospray ion source, with an applied electrospray voltage of 2.0 kV. Full MS scans were conducted at a resolution of 60,000, covering a scan range of 350–1,600 m/z. For MS/MS analysis, up to 20 of the most abundant precursor ions were selected, with a dynamic exclusion time of 30 s. HCD fragmentation was utilized, with a normalized collision energy of 28%, and the resulting fragments were detected using an Orbitrap at a resolution of 30,000. The first fixed mass was set to 100 m/z. The automatic gain control target was set to 1E5 with an intensity threshold of 3.3E4, and the maximum injection time was set to 60 ms.

MaxQuant software (v.1.6.15.0) was utilized to examine the gathered tandem mass spectrometry data. By employing the SwissProt human protein database, which consists of 20,422 protein records, we successfully matched peptides. A reversed decoy database was integrated to evaluate the rate of incorrect identifications. The enzyme trypsin/P was assumed to have specificity, with provisions for potential missed cleavages capped at two. The initial search allowed for a mass deviation of 20 parts per million (ppm) for precursor ions, which was later narrowed to 5 ppm. A quality deviation of 20 ppm basically covers the detection of the vast majority of peptide segments while maintaining the accuracy of the results. Due to the independence of quality deviation from peptide abundance, quality accuracy has no effect on the results of low abundance peptide segments. A mass variance of 0.02 Dalton was permitted for fragment ions.

Criteria set to pinpoint acetylated proteins with altered expression included a p-value under 0.05 and a fold change above 1.2 or below 0.83. To characterize conserved sequences around acetylation sites, a 21-residue span (10 residues on either side of the acetylation site) was examined using the motif discovery application Soft motif-x (Chou and Schwartz, 2011). This application is adept at detecting recurring motifs within protein sequences. The secondary structure of the proteins was predicted using NetSurfP, with protein sequences from the database providing the background for predictions and t-tests being conducted for statistical evaluations (Klausen et al., 2019). Subcellular localization predictions were made with WoLF PSORT, using default settings and the organism-specific background (Horton et al., 2007). We analyzed the functional pathways and gene ontology of the altered proteins using OmicsBean, applying hypergeometric tests for enrichment with a significance cut-off of p < 0.05. The functions disrupted by these protein changes were explored using the Ingenuity Pathway Analysis tool, with significance determined by Fisher's exact test (p < 0.05) and p-scores calculated accordingly (Krämer et al., 2014). The STRING database (v9.1) was employed for analyzing protein interaction networks and the functional domains of acetylated lysines. Fuzzy c-means clustering was used to elucidate gene clusters while GO, KEGG, and protein domain analyses provided insight into the biological processes for each protein cluster, considering terms with p < 0.05 as significant.

A fortnight subsequent to the creation of the animal model, a sample of six rodents from each group was euthanized in a manner that minimized suffering. The brains were then extracted, and the cerebral slices from the left side were precisely cut into cubic millimeter sections, ranging from 1.5 mm anterior to 0.5 mm posterior to the bregma landmark. These samples were preserved in a fixative composed of 4% paraformaldehyde and 2.5% glutaraldehyde. Further processing of the samples included treatment with 1% osmium tetroxide, a stepwise dehydration in an acetone series, and finally, embedding in an epoxy resin for subsequent slicing. Sections of extreme thinness were obtained using an ultramicrotome, then stained with uranyl acetate and lead citrate, and examined under a JEM-1400 transmission electron microscope manufactured by JEOL, located in Tokyo, Japan.

Fourteen days post-model creation, the rats were humanely terminated. The cortex was removed, and brain samples were extracted from the ligation-adjacent periventricular zones, immediately chilled, and preserved at −80°C. Samples were lysed in RIPA buffer, supplemented with phosphatase inhibitors from Sigma-Aldrich, France. Nuclear proteins were sourced utilizing a Solarbio kit, and protein levels quantified via a BCA assay from Sigma-Aldrich. Proteins (50 μg each) underwent SDS-PAGE (4–20%) and were then applied to PVDF sheets. After blocking with 5% skim milk for 2 h, the membranes were incubated at 4°C overnight with a selection of primary antibodies from Abcam and Proteintech, followed by corresponding secondary antibodies from Proteintech, and visualized using ECL substrates from Thermo Fisher Scientific. Band intensities were quantified using ImageJ and normalized to β-tubulin, β-actin, or Lamin B markers.

The isolated tissue was resuspended in RIPA buffer in the presence of protease inhibitors. For immunoprecipitation assays, a Pierce Co-IP Kit was employed. Lysates were precleared with control Ig and protein A/G beads at 4°C for 1 h. Post-centrifugation, supernatants were collected; a portion was reserved for Western blot analysis, while the remainder was incubated with Tau antibodies overnight at 4°C. Protein A/G beads were introduced for a subsequent 4 h incubation. The precipitates were centrifuged, washed with PBS, eluted, and analyzed via Western blot for acetyl-lysine.

The assessment of Mapt gene expression in the tied-off section of the paraventricular brain area was performed using RT-PCR. To prepare RNA samples, the left segment of the midbrain within this region was pulverized, homogenized, and then spun down at a temperature of 4°C (14,000 rpm for 15 minutes). The process of extracting the total RNA was executed with the aid of a TRIzol reagent from Takara Bio, located in Dalian, China. Subsequently, the conversion of RNA to cDNA was accomplished using the HiScript QRT SuperMix specifically designed for qPCR, which includes a gDNA removal feature (product R123-01 from Vazyme Biotech Co., Ltd., Nanjing, China). The RT-PCR analysis was conducted with the ChamQ universal SYBR qPCR master mix (product code Q711 from Vazyme Biotech Co., Ltd.), and the Mapt gene's relative expression was quantified by employing the 2^−ΔΔCT technique, utilizing the Gapdh gene as a normalization control. Primer sequences for this analysis were custom-developed by the Shanghai Biotechnology Service Co., Shanghai, China.Mapt forward 5′-ACACATCTCCACGGCACCTCAG-3′ and reverse 5′-GCGGACACTTCATCGGCTAACG-3′ and Gapdh forward 5′-GATGGTGAAGGTCGGTGTGA-3′ and reverse 5′-TGAACTTGCCGTGGGTAGAG-3′.

To investigate the effect of caffeine on MMP in brain white matter tissue cells after damage, cells were collected by brain tissue grinding, centrifugation, and resuspended in staining buffer containing JC-1 fluorescent probe (C2006, Beyotime, China). After 20 min of dark staining at 37 °C, the proportion of red fluorescence to green fluorescence in brain tissue cells was detected and analyzed using FCM to reflect the changes of MMP in brain tissue cells.

Data were expressed as mean ± SEM. Group differences were determined by one-way ANOVA, using GraphPad Prism, with a significance threshold set at a p < 0.05.

To investigate the impact of caffeine on acetyl group dynamics in the white matter of HI neonatal rats, we utilized an extensive methodology including Sample preparation, tissue protein extraction, Kac antibody affinity enrichment, LC-MS/MS analysis, mass spectrometry, differential analysis, and subsequent identification and validation processes (Figure 1A). The consistency of the experimental data was confirmed through a repeatability analysis of the three replicates (Figure 1B). A total of 1999 lysine acetylation sites were identified across 1,123 proteins, with 1,342 sites quantified in 689 proteins. The quality errors of the identified peptides were assessed, revealing an error distribution centered around zero, with the majority of errors below 5 ppm, demonstrating high accuracy and compliance to the requirements of MS data (Figure 1C). The length distribution of the peptide segments ranged from 7 to 20 amino acids, consistent with enzymatic hydrolysis and MS fragmentation patterns (Figure 1D), indicating appropriate sample preparation. These proteins displayed varying numbers of acetylation sites (Figure 1E). Among the acetylated proteins, 48% had two modification sites, 40% had three sites, and 20% had four sites. Notably, only 2% of the acetylated proteins had one or five acetylation sites.

To analyze the sequence surrounding the acetylation site in rats, we compared the frequencies of 10 amino acids upstream and downstream of the acetylation site (Figure 1F). Downstream of the acetylated lysine, there was an enrichment of aspartic acid (D), histidine (H), methionine (M), asparagine (N), and serine (S), while upstream of the modification site, phenylalanine (F), tyrosine (Y), and valine (V) were enriched. Both the upstream and downstream regions of the modification site showed enrichment of alanine (A), glycine (G), and threonine (T). Notably, acidic amino acids appeared in these motifs, such as G at +5, +3, and +1, and valine (V) at +5 and +3. The acetylated lysine modification site exhibited F, G, T, and Y at position 1; H, N, S, and T at position 1; or A, D, G, M, and N at position 2. These characteristic sequences are likely to represent motifs of acetylated modified peptides. We identified three significantly enriched acetylation site motifs from the quantified lysine acetylation sites, namely YKacN, FKacN, and G^***GKacS, where Kac represents acetylated lysine and ^* represents random amino acid residues. These motifs may indicate specific protein-protein interaction domains, where amino acids (such as Y, F, N) near acetylated lysine (Kac) may have a significant impact on the specificity or affinity of the interaction. Y. F is usually a hydrophobic amino acid that may contribute to hydrophobic interactions within or between proteins. The regulation of protein function by these specific motifs and their acetylation modifications is multifaceted. They regulate various biological processes within cells by affecting protein activity, stability, localization, interactions, and participating signaling pathways (Huntley and Golding, 2002; Iakoucheva et al., 2004). The specific role and importance of these mechanisms depend on the type of protein, cell type, and biological context. These findings highlight the potential functional importance of the amino acids surrounding lysine during acetylation (Figure 1G). To understand the relationship between lysine acetylation and the local secondary structures of acetylated proteins, we conducted secondary structure predictions.

A significant proportion of the acetylation sites were observed within the coil configurations of proteins (69.82%), with less in the α-helix (24.0%) and even fewer in the β-strand (6.17%) regions (Figure 1H). Surface analysis indicated that around 39.24% of these sites were on the exterior of the proteins (Figure 1I). Such a distribution hints at a proclivity for acetylation within the coiled regions of rat proteins.

We conducted a thorough functional annotation of the identified proteins to gain a comprehensive understanding of their functional characteristics (Figure 1J). This annotation involved several aspects including GO analysis, Protein domain analysis, KEGG pathway analysis, COG/KOG functional classification, subcellular localization analysis. Our findings demonstrated that protein acetylation has a significant involvement in various physiological processes.

In exploring the influence of caffeine on acetylation within the white matter of neonatal rats affected by hypoxia-ischemia (HI), we conducted a comparative analysis across three groups, using a significance threshold of p < 0.05 and a minimum 1.2-fold change. This led to the identification of 113 acetylation sites across 96 proteins, with specific sites showing increased or decreased acetylation in the HI group. Compared with the Sham and Caffeine groups, 3 acetylation sites in 2 proteins were upregulated in the HI group, and 11 acetylation sites in 10 proteins were downregulated in the HI group (Figures 2A, B). We have described the specific roles of these upregulated and downregulated proteins in neurological diseases (Table 1).

To investigate the biological functions and associated networks of the 12 proteins with differential expressions in lysine acetylation, we conducted subcellular localization analysis. Initially, we analyzed the subcellular localization patterns of these differentially expressed acetylated proteins. The differential expression of acetylated subcellular localization was mainly observed in the cytoplasm (4, 33.33%), cytoplasmic nucleus (3, 25%), and mitochondria (2, 16.7%). This suggests that the neuroprotective effect of caffeine is primarily achieved through the alteration of acetylation in proteins located in the cytoplasm, cytoplasmic nucleus, and mitochondria (Figure 2C).

Our examination of protein functions affected by acetylation changes highlighted their involvement in a variety of biological activities, from cell size regulation to the development of the central nervous system. These activities are particularly essential during the generation of action potentials (Figure 2D). Furthermore, the enriched cellular components included the primary endosome membrane, neuronal axons, intervertebral discs, and cell junction regions (Figure 2E). Molecular functions with significant enrichment in the analysis included GTP-dependent protein binding, ATPase binding, protein binding, and other molecular functions (Figure 2F).

By delving into the KEGG pathways through functional enrichment analysis, we pinpointed considerable pathway involvement, including neurotrophin and PI3K-Akt signaling (Figure 2G). This analysis corroborated the acetylation patterns we had initially uncovered (Figure 2H).

In summary, these findings suggest that protein acetylation may play a crucial role in various cellular processes occurring in the cytoplasm, nucleus, and mitochondria. Many of the important classical pathways identified are closely associated with energy metabolism and production, suggesting their significance, and warranting further research.

Further analysis of the differentially acetylated proteins, which consisted of 10 upregulated proteins and two downregulated proteins, was performed using the IPA database. The analysis revealed a significant enrichment of these proteins in diseases of the nervous system. To gain a better understanding of these differentially expressed proteins, three types of analysis were conducted, namely regulatory network (Figure 3A), disease function (Figure 3B), and classical pathway (Figure 3C) analyses. In the regulatory network analysis, it was found that Mapt was associated with AGRN and LY6H, while RAC1 and CRYM formed a regulatory network with ADCY, ARHGDIA, LEP, PTGS2, SH2D5, and SLC12A5 (Figure 3A). The classical IPA pathway and disease functional pathway were enriched within the 12 differentially expressed proteins. Among the downregulated genes, significant changes related to nervous system diseases such as microglial transformation and neurite neurodegeneration were observed. Additionally, diseases related to mitochondrial transport and degeneration were also noted (Figure 3B).

Moreover, statistically significant pathways identified in the analysis included the synaptogenesis signaling pathway and mitochondrial dysfunction and semaphorin neuronal recursive signaling pathway. Both of which are related to Mapt (Figure 3C). These enriched pathways observed in the IPA analysis are consistent with the results from the KEGG pathway analysis. Many of these pathways are crucial and closely associated with mitochondria. It is worth noting that the downregulated proteins primarily participate in these pathways, suggesting their potential significance and highlighting the need for further investigation.

To investigate the impact of caffeine on the acetylation of lysine in the white matter of HI neonatal rats, we performed an expression pattern clustering analysis on 1999 lysine acetylation sites derived from three groups, namely the Sham, HI, and Caffeine groups. We focused on proteins that exhibited significant changes in the abundance of acetylation modifications across these groups. First, we transformed the relative expression levels of protein acetylation sites using a logarithmic conversion with a base of 2 (log2). Next, we selected proteins with a standard deviation > 0.3 to ensure substantial variation. After this screening process, we identified 432 modified proteins for further analysis. We utilized the Mfuzzy method to cluster the expression patterns of modified proteins. The clustering parameters included six clusters (k) and a fuzzy degree (m) of 2. Notably, Cluster1, Cluster2, and Cluster3 exhibited distinctive patterns that reflected the impact of caffeine treatment, indicating its potential to improve acetylation modification of lysine in the white matter of HI neonatal rats.

We performed GO function, KEGG pathway, and protein domain enrichment analyses for the proteins in each cluster (Figure 3D). The results revealed that Cluster1 was predominantly enriched in the mitochondrial matrix and secondary endosomes, with protein domains including isopropyl citrate/isopropyl malate dehydrogenase and hexokinase. Its function was closely related to nicotinamide adenine dinucleotide (NAD) binding, coenzyme binding, cytokine production, and positive regulation of cytokine secretion. Many functions and pathways were linked to propionate metabolism and lysine degradation, which are crucial for mitochondrial activity. Cluster2 was mainly localized in cell junctions and intercellular connections, with SH2 and SH3 protein domains. Its functions and pathways were related to gap linking, cell entry, white cell migration, axon guidance, insulin receptor binding, transferase activity, and glycosyl transfer. Cluster3 played a role in the nuclear and inner mitochondrial membrane protein complexes, with specialist repeats and Ca²⁺-sensitive EF hand protein domains. Its functions and pathways were related to thermogenic effects, oxidative phosphorylation, isomerase activity, and silk protein-based movement. These pathways were consistent with the KEGG pathway and IPA analyses of differentially expressed proteins, suggesting that caffeine improves mitochondrial dysfunction through protein acetylation, a concept which requires further exploration.

Further examination of acetylation's role post-caffeine treatment in neonatal rats with HI-induced brain damage involved mapping the protein interactions within three clusters. In this comprehensive network, we included 159 acetylation sites and 133 proteins, assessing the interactions with a stringent threshold. The resulting network visualization displayed the regulatory patterns of these proteins in response to treatment (Figure 3E).

To characterize the complexes between acetylated proteins, 133 proteins from Cluster1, Cluster2, and Cluster3 were used as protein interaction networks. The acetylated modified proteins related to the citric acid cycle, tricarboxylic acid cycle, mitochondrial phagocytosis, oxidative phosphorylation, inner mitochondrial membrane proteins, mitochondrial organelles, and other keywords were classified by clustering.

Upon additional analysis, we found that compared to the HI group, Mapt was downregulated in both the Sham group and the Caffeine group. This gene is significantly differentially expressed and is closely related to mitochondrial function. Further verification of this finding is warranted in future investigations.

In previous research, we noted alterations in histone acetylation following caffeine treatment in neonatal rats with HI-induced damage. We then extended our study to assess caffeine's impact on oxidative stress and inflammation, as well as its effect on mitochondria in this context. Myelination status was gauged using MBP staining, which revealed disarray in nerve fiber organization and reduced myelination in the HI group (P < 0.001) (Figure 4A), with some recuperation observed in the Caffeine group (P < 0.05) (Figure 4B).

Western blotting analysis were conducted to measure the expression levels of NOX-1, NOX-2, IL-1β, and TNF-α proteins. Our results showed that compared to the Sham group, the levels of NOX-1, NOX-2, IL-1β, and TNF-α proteins were increased in the HI group. However, administration of caffeine reduced the levels of these proteins following HI-induced WMD (Figures 4C–G, p < 0.01 or p < 0.001).

Next, we detected the levels of apoptotic proteins Caspase-3 and Bcl-2. The results showed that the HI group exhibited a significant increase in Caspase-3 protein levels compared to the Sham group. However, caffeine reduced Caspase-3 protein levels after WMD. On the contrary, compared with the Sham group, the HI group had a decrease in Bcl-2 protein levels, while the Caffeine group had an improvement in Bcl-2 protein levels (Figures 4H–J, p < 0.01 and p < 0.001, respectively). Then we examined the structure and function of the mitochondria to evaluate the role of caffeine in rescuing damaged mitochondria. Cytochrome C (Cyt-C) levels, used as an indicator of mitochondrial damage, were analyzed in tissues, and the results showed that the HI group exhibited a significant increase in Cyt-C levels compared to the Sham group. However, caffeine reduced Cyt-C levels after HI-induced WMD (Figures 4H, K, p < 0.01). The mitochondria were observed by transmission electron microscopy. In the Sham group, the mitochondria had a distinct cristae appearance. Following HI, the mitochondria showed abnormalities including swelling, ridge collapse, and membrane rupture in the HI group. Importantly, caffeine treatment increased the mitochondrial tubular network and maintained normal mitochondrial morphology (Figure 4L). Similar results were observed in terms of improving mitochondrial dysfunction, including ATP and mitochondrial DNA (mtDNA) levels. The results showed that the HI group reduced in ATP and mDNA levels compared to the Sham group. However, caffeine increased ATP and mDNA levels (Figures 4M, N, p < 0.05, p < 0.01, and p < 0.001, respectively). The decrease in cell membrane potential can be easily detected through the transition of JC-1 from red fluorescence to green fluorescence. In normal cells, this dye accumulates and aggregates in mitochondria, forming bright red fluorescent clusters. When the membrane potential of the cell mitochondria decreases, the red color changes to green. Therefore, we observed a decrease in membrane potential in the HI group and an improvement in membrane potential in the Caffeine group compared to the Sham group (Figures 4O, P, p < 0.001).

To ensure the accuracy of our proteomic analysis of lysine acetylation, we conducted a Western blot analysis to confirm the overall acetylation modifications of the three groups. The results showed differences in acetylated proteins among the Sham, HI, and Caffeine groups. Specifically, the HI group had an increased rate of lysine acetylation near a molecular weight of 50 KD compared to the Sham and Caffeine groups (Figures 5A, B, p < 0.01 or p < 0.001). This suggests that caffeine may have a positive effect on protein deacetylation during HI-induced WMD.

According to our previous research, proteomic analysis identified 47 proteins that were differentially expressed among the Sham, HI, and Caffeine groups. These proteins are potentially linked to SIRT2-mediated myelin development and synaptic formation (Yang et al., 2022a) (Figure 5C). SIRT2 is an enzyme known as a nicotinamide adenine dinucleotide-dependent deacetylase. SIRT2, a key player in the myelination process and implicated in various neurological conditions, was scrutinized (Fourcade et al., 2017; Chen et al., 2021). Western blot analysis confirmed a decrease in SIRT2 expression in the HI group (p < 0.001), with a marked increase observed in the Caffeine group, suggesting a potential protective or restorative effect of caffeine (p < 0.001) (Figures 5D, E).

Cluster analysis revealed that the acetylated modified proteins in Cluster3 primarily participate in the protein complex of the inner mitochondrial membrane. Within Cluster3, Mapt exhibited significant positive changes in deacetylation. Through the PPI analysis of acetylated and proteomic proteins (Yang et al., 2022a), we found that there is a protein interaction between SIRT2 and Mapt (Figure 5F). Therefore, we applied WB and PCR to validate the levels of full and acetylated Mapt proteins. The results showed that Mapt mRNA levels did not differ among the three groups (Figure 5G). The acetylation level of Mapt protein in the HI group was significantly upregulated compared to that in the Sham group, and this increase was observed to be inhibited in the Caffeine group (Figures 5H, J, p < 0.01 or p < 0.001). However, there was no difference in total Mapt protein levels among the three groups (Figures 5H, I). This result is consistent with findings from studies analyzing acetylated proteins using multiphoton spectroscopy, which further verifies the accuracy and feasibility of omics research results.

To further investigate the relationship between Mapt deacetylation, SIRT2, and the prevention and treatment of caffeine in WMD, we conducted experiments on neonatal rats. The rats were treated with caffeine and the SIRT2 inhibitor AK-7 to evaluate changes in Mapt acetylation in WMD, as well as their impact on mitochondrial function. Western blot analysis was performed to measure the levels of total Mapt, acetylated Mapt, Cyt-C, mitochondrial transcription factor A (TFAM), Caspase-3 and Bcl-2 proteins.

The results revealed that when AK-7 was applied, the levels of Cyt-C and Caspase-3 in the caffeine+AK-7 group were significantly increased compared to those in the caffeine-only group (Figures 6A, C, D, p < 0.05 or p < 0.001). Compared with the Sham group, TFAM and Bcl-2 protein levels were downregulated after HI and upregulated after caffeine treatment, but this improvement was inhibited after the application of AK-7 (Figures 6A, B, E, p < 0.01 or p < 0.05). Moreover, the degree of acetylation of the Mapt protein was also upregulated in the caffeine+AK-7 group (Figures 6F–H, p < 0.05). Meanwhile, compared with the Sham group, Mapt nuclear translocation was reduced after HI. Caffeine treatment can promote nuclear translocation, but this effect is inhibited after the application of AK-7 (Figures 6I, J, p < 0.05 or p < 0.01). Similar findings were observed in terms of rescuing mitochondrial dysfunction. When AK-7 was applied, the caffeine+AK-7 group exhibited a decrease in ATP and mtDNA levels compared to the Caffeine group (Figures 6K, L, p < 0.05). MMP detection revealed a decrease in membrane potential in the Caffeine+AK-7 group compared to the Caffeine group (Figures 6M, N, p < 0.01). The effect of caffeine on improving membrane potential is inhibited by AK-7.