Gut Flora-Targeted Photobiomodulation Therapy Improves Senile Dementia in an Aß-Induced Alzheimer’s Disease Animal Model

Background: Emerging evidence suggests that the gut microbiota plays an important role in the pathological progression of Alzheimer’s disease (AD). Photobiomodulation (PBM) therapy is believed to have a positive regulatory effect on the imbalance of certain body functions, including inammation, immunity, wound healing, nerve repair, and pain. Previous studies have found that the intestinal ora of patients with AD is in an unbalanced state. Therefore, we have proposed the use of gut ora-targeted PBM (gf-targeted PBM) as a method to improve AD in an Aß-induced AD mouse model. Methods: PBM was performed on the abdomen of the mice at the wavelengths of 630 nm, 730 nm, and 850 nm at 100 J/cm 2 for 8 weeks. Morris water maze test, immunouorescence and proteomic of hippocampus, and intestinal ora detection of fecal were used to evaluate the treatment effects of gf-targeted PBM on AD rats. Results: PBM at all three wavelengths (especially 630 nm and 730 nm) signicantly improved learning retention as measured by the Morris water maze. In addition, we found reduced amyloidosis and tau phosphorylation in the hippocampus by immunouorescence in AD mice. By using a quantitative proteomic analysis of the hippocampus, we found that gf-targeted PBM signicantly altered the expression levels of 509 proteins (the same differentially expressed proteins in all three wavelengths of PBM), which involved the pathways of hormone synthesis, phagocytosis, and metabolism. The 16s rRNA gene sequencing of fecal contents showed that PBM signicantly altered the diversity and abundance of intestinal ora. Specically, PBM treatment reversed the typical increase of Helicobacter and uncultured Bacteroidales and the decrease of Rikenella seen in AD mice. Conclusions: Our data indicate that gf-targeted PBM regulates the diversity of intestinal ora, which may improve damage caused by AD. Gf-targeted PBM has the potential to be a noninvasive microora regulation method for AD patients.


Introduction
Alzheimer's disease (AD) is a typical degenerative disease of the central nervous system. It is characterized by cognitive impairment, progressive memory loss, behavioral and social impairment, and dementia (1). The 2018 World Alzheimer's Report states that AD is both a human problem and a global problem. Every 3 seconds, there will be a new case of dementia in the world. About 50 million people worldwide had AD in 2018, and that number is expected to rise to 152 million by 2050 (2). The main pathological feature of AD is senile plaques, which result from amyloid β protein (Aβ) aggregation in the brain. These plaques are tangles formed by tau protein hyperphosphorylation. Plaque formation leads to the dysfunction of neurons and synapses and chronic, long-term in ammation in the brain, which ultimately results in neural degeneration (3). Although the pathogenesis of AD is still unknown, four hypotheses have been put forward in view of these major neuropathologic features, namely the amyloid-β hypothesis, the tau protein hypothesis, the neurotransmitter disorder hypothesis, and the chronic in ammation hypothesis (4). Based on these hypotheses, several drugs based on different pathological mechanisms have been developed (5). Currently, there are ve AD drugs approved by the US Food and Drug Administration, including donepezil, rivastigmine, galantamine, huperzine, and memantine (6).
However, these drugs mainly improve symptoms, and there is no drug that will delay disease progression or cure AD.
Recent studies have shown that the intestinal ora is involved in the occurrence and development of AD through a variety of pathways. Therefore, intestinal ora modulation is now being considered as a new target for the treatment of AD (7,8). The intestinal ora mainly affects the occurrence of AD through the nervous system, the endocrine system, metabolism, and immunity (9,10). In some animal and human studies, researchers have attempted to regulate the intestinal ora to prevent and treat AD. For example, Woo et al. found that administration of sugar-free Lactobacillus plant varieties C29 by gastric perfusion to a cognitively-damaged mice model signi cantly improved memory dysfunction caused by D-galactose and increased the expression of BDNF (Brain Derived Neurotrophic Factor) in the brain. This reduced the expression of aging marker P16, in ammatory markers p-p65 and p-foxo3a, cyclooxygenase-2, and inducible nitric oxide synthase, and inhibited the expression of TNF-α induced by D-galactose (11). A randomized, double-blind, controlled clinical trial of patients with AD also con rmed that there was a signi cant difference in the degree of improvement in cognitive and metabolic functions in patients who received mixed probiotics (Lactobacillus acidophilus, Lactobacillus casei, Bi dobacterium bi dum, and Lactobacillus fermentum) in their milk compared with patients in the sterilized milk intake group (12).
Photobiomodulation (PBM) therapy, also called low-level laser therapy, is a noninvasive light therapy that uses a low power non-thermal light source to irradiate target tissues and regulate body functions. PBM has been effectively applied in many clinical diseases, including chronic shoulder and neck pain, wound healing, jaw joint disorders, oral mucositis, arthritis, muscle injury, bone regeneration, diabetic foot disease, multiple sclerosis, Parkinson's disease, and AD (13,14). The wavelength of red and infrared irradiation is most commonly used because it penetrates deep into tissue, affecting metabolic modi cations, DNA activity, ATP formation, and the mitochondrial chain. The effect of PBM is due to the absorption of photons by cytochrome C oxidase in the mitochondrial respiratory chain, which consequently increases cytochrome C oxidase activity and therefore ATP formation. ATP from injured regions or regions of impaired blood perfusion can reactivate injured cells and help treat metabolic disorders (15)(16)(17). PBM is also related to pain relief, in ammation relief, and the prevention of tissue death to avoid neurological degeneration (18,19). PBM is considered a method of adjusting the imbalance of body functions within a certain range and eventually establishing a new balance. The intestinal ora of patients with AD is in an unbalanced state, with a decrease in bene cial ora and an increase in harmful ora, which leads to the consequential imbalance of in ammation and immunity and eventually induces or aggravates AD. PBM may be useful for regulating the imbalance of intestinal ora in AD.
PBM has a bidirectional regulatory effect on bacteria. This therapy can promote bacterial proliferation or inhibit bacterial growth, depending on the wavelength and dose of PBM and the type of bacteria (20).
Under blue and green light (400-500 nm), bacterial growth was inhibited or inactivated by irradiation with 10-200 J/cm 2 light. For example, 95% of Pseudomonas aeruginosa could be inactivated by a 405 nm laser at 10J/cm 2 , and 90% of Staphylococcus aureus could be inactivated by 15 J/cm 2 (21). However, under the irradiation of red and infrared light, the proliferation of bacteria was accelerated at low (1-10 J/cm 2 ) light intensities. Interestingly, 810 nm laser irradiation of 5 J/cm 2 could accelerate the proliferation of Escherichia coli, while 23% of Pseudomonas aeruginosa was inhibited under the same conditions (22). The bidirectional regulation of PBM on bacteria makes it possible to improve the imbalance of the gut ora in AD patients. In this study, we proposed the method of gut ora-targeted photobiomodulation therapy (gf-targeted PBM) and explored its improvement of AD in an Aß-induced AD mouse model.

Animals
Male C57BL/6N mice with an average body weight of 24-28 g were used in this study. Mice were obtained from Beijing Vital River Laboratory Animal Technologies Co. Ltd. Mice were fed in cages of 7-8 animals, with free drinking water and food. The mice were anesthetized by an intraperitoneal injection of 0.1 mL 5% chloral hydrate. The mice were xed on a stereotactic brain apparatus, the head skin was shaved, disinfected, cut along the midline, and the fascia was removed. The skull surface was wiped with hydrogen peroxide, exposing the bregma (anterior fontanelle) point. The bilateral hippocampal CA1 area, 2.3 mm behind the anterior fontanelle and 1.8 mm beside the sagittal seam, was used as the drug administration point. The injection depth was 2.0 mm below the skull surface. Then, 1.0 µL Aβ 1−42 of 3M was injected bilaterally into the hippocampus over a period of 0.2 µL/min using a 1 µL microsyringe. The needle was left in for 5 min to allow for complete diffusion of the drug, followed by suturing the skin.

Gf-targeted PBM treatment
The mice were randomly assigned into four groups: The PBM process was carried out as follows. After iso urane anesthesia, the mice were xed in the supine position to fully expose the abdomen and remove the abdominal hair. The special LED irradiation device for the laboratory (with wavelengths of 630 nm, 730 nm, and 850 nm) was xed on the upper abdomen of the mice, with the LED irradiation surface facing the abdomen. The PBM device was operated in the dark. The irradiation time was 1000s (16 min and 40 s), the power density was 10 mW/cm 2 , and the energy density was 100 J/cm 2 . The treatment occurred once a day, 5 times a week, for 8 weeks.

Morris water maze test
A Morris water maze (MWM) test was performed to evaluate the improvement of learning ability (navigation test) and memory ability (spatial probe test) of mice before and after PBM interventions. The MWM test was divided into a training period (3 days), a positioning cruise period (6 days), and a space exploration period (1 day). During the training period, the platform was located in the same position (one of four quadrants of the pool), and the mouse was placed into the pool facing the platform. The mouse was removed from the water when it located the platform. If the mouse did not nd the platform after 120 s of swimming, it was gently guided to the platform or placed on the platform for 15 s before being removed from the pool. For the navigation test, the movement of the mouse was tracked by a digital tracking system. The latency, the percent time in the outer annulus, the average swimming speed, the swimming trajectory, and the search strategy were recorded to evaluate the spatial learning ability of the animals. The mice were tested twice per day. For the spatial probe test, the platform was removed from the pool and the mouse was placed in the pool in a random quadrant. The movement tracking was recorded for 2 min.

Sample preparation
After the last MWM test, the mice were sacri ced in a state of deep anesthesia (10% chloral hydrate injected intraperitoneally) for the collection of the brain tissues, mesenteric lymph nodes, blood samples, and fresh fecal samples. Blood samples were obtained by eyeball extraction. Stool samples were collected by squeezing a section of the colon removed from the mice. Cardiac perfusion was performed with normal saline before brain tissue was taken. Bilateral hippocampal tissues were removed within 10 min for frozen section preparation and proteomics detection, as were mesenteric lymph nodes for pathological section preparation.

Mesenteric blood ow observation by laser speckle
On the day after the end of PBM treatment, 3 mice were taken from each group and anesthetized. Then, the abdominal skin of each animal was cut open to expose the mesentery and small intestine, which were placed 28-30 cm under the lens. Blood ow pattern images were collected by laser speckle (Moor FLPI, Wilmington, DE, USA). The average blood perfusion value was automatically generated by the moorFLPI system software and the vessel diameter was calculated by ImageJ software.

Immune response detected by ELISA and immunohistochemistry
Serum was isolated from blood samples and used to detect in ammatory factors IL-6 (RAB0308, Sigma) and INF-γ (RAB0224, Sigma) using ELISA kits. Mesenteric lymph nodes were xed with 4% paraformaldehyde (0.01 M, pH 7.4, 4°C) for 24-48 hours, and were routinely dehydrated, transparent, waxed, and embedded. Sections with a thickness of 5 µm were sliced on the para n sectioning machine and placed on the slides treated with poly acid for baking at 60°C for 4-6 hours and stored.
Immunohistochemistry for CD45 and CD11b antigens was performed. Antigens were unmasked by microwaving sections in 10 mmol/L citrate buffer, pH 6.0 (15 min), and immunostaining was undertaken using the avidinbiotinylated enzyme complex method with antibodies against CD45 (ab10558, Abcam) at a concentration of 1 µg/ml, CD11b (ab133357, Abcam) at a concentration of 1 µg/ml, and biotinconjugated secondary antibody at a concentration of 1 µg/ml.

Immuno uorescence staining
For frozen section preparation, hippocampus tissue was dehydrated in gradient sucrose solutions (with concentrations of 10%, 20%, and 30%) for 24 h, with the brain tissue sinking to the bottom of the tube. Then, the tissue was embedded by optimum cutting temperature compound and quick-frozen for 10 min in a low-temperature freezing slicing machine. Slices of 50 µm thickness were cut along the coronal plane. Immuno uorescence staining was performed to observe the expression of Aß 1− 42 , Iba1, and staining. Immuno uorescence images were obtained using an automatic scanning system (C13210-01; Hamamatsu Photonics, Hamamatsu City, Japan).

Proteomic detection and data analysis
Proteins extracted from fresh hippocampus tissues (stored at − 80°C) were pooled from 5 mice of each group for LC-MS/MS analysis combined with tandem mass tags (TMT) labeling. Brie y, the pooled protein samples from ve groups of PMB630, PBVM730, PBM850, Ctrl B, and Ctrl M were cleaved into peptides with 1 µg/µL trypsin and isobaric labeled TMT. Equal protein amounts derived from each group were labeled with different TMT labels: Ctrl B, TMT-126; Ctrl M, TMT-127N; PMB630, TMT-128N; PMB730, TMT-128C; and PMB850, TMT-129N. The labeled peptides were mixed, dried, and then fractionated into 16 fractions by HPLC. Following fractionation, each peptide fraction was desalted and puri ed by C18 reverse phase chromatography. Then, the mass spectrometry data were analyzed using Q-Exactive MS (Thermo Fisher Scienti c, Waltham, MA, USA). The parameters for LC-MS/MS are shown in Table 1. A false discovery rate less than 1.0% was selected for peptides only. Precise quanti cation of protein was expressed as the protein ratio between samples, and differentially expressed proteins (DEPs) were selected at p < 0.05 by a Student's t test. The thresholds of up-and downregulated ratios were set at 1.2 and 0.83, respectively. The proteomic data distribution patterns, volcano plots, and heatmap were created using Thermo Fisher Proteome Discoverer 1.4. The DEPs of the PBM group were determined compared with the Ctrl M group and the cluster analysis of DEPs was performed using the gene ontology (GO) and KEGG pathway database.

Fecal DNA extraction and 16 s rRNA gene sequencing
A DNA extraction kit (DNeasy PowerSoil Kit; Qubit dsDNA Assay Kit, Qiagen, Germany) was used to extract the genomic DNA from the fecal samples (n = 5). After that, the purity and concentration of DNA were detected by agarose gel electrophoresis. Appropriate samples were placed in a centrifuge tube and diluted to a concentration of 1 ng/µL in sterile water. The diluted genomic DNA was used as the PCR template. Speci c primers with barcodes and a high-delity enzyme (Takara Ex Taq, Takara, Japan) were used for PCR according to the selected sequencing area to ensure ampli cation e ciency and accuracy. The V3-V4 hypervariable regions of the bacterial 16S rRNA gene were ampli ed with primers 343F (5′-TACGGRAGGCAGCAG-3′) and 798R (5′-AGGGTATCTAATCCT-3′) by a PCR system (Bio-Rad, Hercules, CA, USA). PCR products were detected by electrophoresis and puri ed by magnetic beads. After puri cation, the PCR products were used as a two-round PCR template. Then, the DNA was detected and puri ed again, and quanti ed using a Qubit (Invitrogen Qubit 4 uorometer, Thermo Fisher Scienti c). Finally, the samples were mixed according to the concentration of PCR products and sequenced on the Illumina MiSeq platform (San Diego, CA, USA).
The original data were in FASTQ format. Trimmomatic software (Illumina) was used to remove clutter from the original double-ended sequence. When the base mass was less than 20, the previous highquality sequence was intercepted. After the sequencing data were preprocessed to generate high-quality sequences, Vsearch software (https://github.com/torognes/vsearch) was used to classify the sequences into multiple OTUs according to the similarity of the sequences. Sequence similarities greater than or equal to 97% were classi ed as an OTU unit. The representative sequences of each OTU were selected using the QIIME software package (http://qiime.sourceforge.net/), and all the representative sequences were compared with the Silva (version 132) database for annotation. The RDP classi er software (23) was used for the species alignment annotation, leaving the annotation results with a con dence interval greater than 0.7.

Statistical analyses
All data with error bars are represented as mean ± SEM. For two group comparisons, an unpaired twotailed Student's t test was applied. For more than two group comparisons, a one-way ANOVA was performed. P < 0.05 was considered statistically signi cant. The original ab2020 (OriginLab Corporation, Northampton, MA, USA) was used for data statistical analysis and gure production. For image quanti cation, ImageJ with IHC pro ler (https://imagej.net/) was used.

The scheme of gf-targeted PBM treatment for AD mice
The purpose of this study was to explore whether gf-targeted PBM can improve the symptoms and pathology of AD by regulating the intestinal ora. We produced the AD mouse model by Aß 1−42 injection and veri ed the modeling effect by a MWM test in ten mice. The AD mice then received the PBM intervention with different wavelengths for up to two months. After PBM treatment, the cognitive function of AD mice was determined by a MWM test and the changes in Aß amyloid protein, tau protein, microglia cells, and proteomics of hippocampal tissue were detected to determine the therapeutic effect of PBM on AD. In addition, we calculated the intestinal ora diversity after gf-targeted PBM to analyze both the regulatory effect of PBM on the intestinal ora and the abundance of bene cial ora following PBMtreated AD. Finally, we measured cellular immunity in the mesenteric lymph nodes and in ammatory factors in the blood to explore the role of immunity and in ammation as mediators in the treatment of AD by PBM. The design scheme of the complete experiment is shown in Fig. 1. 3.2 Gf-targeted PBM improves the cognition impairment of AD mice The MWM experiment was divided into two parts: the location navigation test and the space probe test. The former tested the learning ability of AD mice and the latter tested the memory ability of AD mice. Figure 2A shows that all three PBM groups improved the escape latency period (the time period from the mouse entering the water to nding the platform) in the 6-day navigation experiment. In the later period of the experiment, the performance of mice in the PBM group was better than that of the Ctrl M group, which was close to the normal mice in the Ctrl B group. For the percent of time in the outer annulus, PBM630, PBM730, and PBM850 all decreased the searching time of AD mice in the outer annulus of the pool compared with the untreated AD mice in the Ctrl M group. PBM630 showed the highest effect, as the PBM630-treated AD mice performed the same as the normal mice in the Ctrl B group on the second day. Interestingly, the swimming speed of mice in the three PBM groups was higher than that of the Ctrl M and Ctrl B groups, especially those in the PBM730 group. Untreated AD mice swam as fast as the normal mice in our study, which implies that the method of AD model building by Aß 1−42 injection did not impact the athletic ability of the mice. However, there was no signi cant difference between the three PBM groups in all the measurements, including latency, percent time in the outer annulus, and swimming speed.
We also analyzed the search strategies of the mice. The results are shown in Fig. 2B. Compared with Ctrl B, most of the mice in the Ctrl M group swam on the edge of the pool (dark grey bar) and searched randomly (blue bar), while trend type (red bar) of the mice in the PBM group took up a larger proportion when swimming in the later period of the experiment (after the 3rd day). This was especially evident in the mice from groups PBM630 and PBM850, as they could swim in a straight line towards the platform (green bar). The searching strategy scores of the three PBM groups (PBM630, PBM730, and PBM850) were signi cantly higher than those of the Ctrl M group (p = 0.004, 0.013, and 0.001, respectively). Moreover, the scores of the PBM850 group recovered close to the Ctrl B group, indicating that the learning ability of AD mice after PBM850 treatment returned to a normal mouse level.
The memory ability of AD mice was measured in a space probe test. Figure 2C-F shows the swimming trajectory, times of traversing the platform, and the percent time in platform quadrants of the mice in the space probe test. The swimming trajectories of the normal mice were near the platform after they were put into the pool, while the AD mice in neither the Ctrl M group nor the PBM groups performed chaotic trajectories. By analyzing these trajectories, we found that regardless of the number of platform traversing times or the percentage of time in a platform quadrant, there was no signi cant difference between the three PBM groups and the Ctrl M group. This nding suggests that PBM had no effect on the memory ability of AD mice.

Improvement of pathology in brains of AD mice
Pathological changes in AD, such as Aß amyloid plaques, phosphorylated tau proteins (p-tau(s396)), and microglial cell activation are shown in Fig. 3. Aß amyloid plaques were clustered in the upper edge of the hippocampus in Ctrl M mice. The PBM630 and PBM730 eliminated most of the clusters of Aß amyloid plaques which should be present in the hippocampus of mice. However, PBM850 with a long irradiation time only weakly affected Aß amyloid plaques (the upper right box in Fig. 3 showing the Aß 1−42 image of PBM850). In accordance with the results of Aß amyloid plaques, a mass of microglial cells (stained by Iba1) clustered in the hippocampus of Ctrl M and PBM850 mice, whereas only a few microglial cells were clustered in the PBM630 and PBM730 groups. A number of phosphorylated tau proteins were also present in the hippocampus of Ctrl M and PBM850 mice. Encouragingly, p-tau protein was virtually absent in the hippocampus of mice in the PBM630 and PBM730 groups. These results imply that irradiation with PBM630 and PBM730 contributes to the elimination of Aß amyloid plaques and inhibits the neuroin ammation and tau phosphorylation caused by Aß amyloid plaques.

Response of mesenteric blood ow and the immune system after PBM treatment
We observed the mesenteric blood ow of mice using a laser speckle technique. The results are showed in Fig. 4A-C. We found that the blood ow of AD mice increased signi cantly compared with normal mice (p = 0.017). However, the PBM intervention did not correct this abnormal increase. We also observed that PBM increased the blood vessel diameter, which seems to have little substantial effect on AD development.
Mesenteric lymph nodes are an important part of the body's immune system. To determine whether gf-PBM regulates AD by stimulating the intestinal mucosal immune system, we performed immunostaining for CD45 (marker for leukocytes) and CD11b (marker for phagocytes) in the mesenteric lymph nodes. As shown in Fig. 4F, we observed no signi cant difference between the groups when the positive cells were counted by ImageJ. In addition, we detected proin ammatory cytokines IL-6 and INF-γ (Fig. 4D), which re ect humoral immunity. Both IL-6 and INF-γ were inhibited in AD mice. PBM630 and PBM730 signi cantly increased the levels of INF-γ in AD mice and exceeded the levels in normal mice (8.37-fold, P = 0.0004, and 3.97-fold, P = 0.018, respectively).

Proteomic changes in the hippocampus after gftargeted PBM
The original mass spectrometry data from the hippocampus of ve groups of mice were ltered and searched by false discovery rate (FDR) < 1%. A total of 3,872 proteins were identi ed, matching 17,296 peptides and 289,377 spectra. A quantitative analysis was carried out on samples according to the peak strength of tagged ions, and the number of DEPs obtained (p < 0.05) is shown in Fig. 5A and 5B. A foldchange of DEPs greater than 1.2 is upregulated (red) and a fold-change of less than 0.8 is downregulated (green). Compared with Ctrl M group, the number of DEPs in the hippocampal tissues of groups PBM630 and PBM730 were 1,209 and 1,329, respectively, while the number of DEPs in PBM850 was 634. We found that the DEPs between groups PBM630 and PBM730 groups was low, while the DEPs between PBM850 and PBM630 or PBM730 was high. This indicates that the mechanisms of action associated with the PBM630 and PBM730 treatments may be similar, while the PBM850 treatment may occur via other mechanisms. The heat map in Fig. 5B shows that the expression of DEPs (compared with Ctrl M) in the PBM groups is opposite to Ctrl M. This indicates that PBM treatment corrects the abnormal changes in hippocampal proteins induced by AD.
The DEPs (compared with Ctrl M) of the three PBM groups were analyzed by GO classi cations, including cellular components, molecular functions, and biological processes, to obtain functional annotation information from each protein. The rst ve proteins with the lowest p values are shown in Fig. 5C. In the PBM630 and PBM730 groups, DEPs were similar in terms of cellular components, molecular functions, and biological processes involved. For example, cellular components were mostly parts of proteins that composed cells and cytoplasm, molecular functions were mainly related to binding, and the biological processes involved were reorganization of cellular components and regulation of biological processes. Group PBM850 had both shared and unique DEPs compared with groups PBM630 and PBM730. For example, some DEPs in group PBM850 are components of neuronal projection. For molecular functions, some DEPs are related to G protein-coupled receptors, and for biological processes, the DEPs are more involved in the negative regulation of metabolic processes.
We also used the KEGG database to analyze the DEPs of the three PBM groups in Pathway. In all pathways with a p value of less than 0.05, the ratio of DEPs in background proteins of a pathway were sorted, as shown in Fig. 6. In the pathway of AD, the proportion of DEPs in the three PBM groups was high, with values of 58.7% (PBM630), 50.8% (PBM730), and 31.7% (PBM850). Mitochondrial respiratory chain complex enzymes were most affected by PBM (the lower right corner in Fig. 6) and Cx1 was downregulated, while CxII and CxV were upregulated under the three PBM treatments. CytC, a key member of the apoptosis pathway, was also downregulated in the PBM groups. In addition, Tau, RTN3/4, and SNCA, which all take part in AD development, were downregulated by the PBM interventions. Furthermore, some DEPs were involved in oxidative phosphorylation, phagocytosis, metabolism of some biological macromolecules, and most importantly, the secretion of many hormones, including insulin, thyroxine, and glucagon. The pathways associated with the DEPs of the PBM850 group were less than that of PBM630 and PBM730.
We also conducted a comparative analysis of the DEPs in the three PBM groups and found that there were 1,007 identical DEPs in PBM630 and PBM730, and 509 identical DEPs in PBM630, PBM730, and PBM850. We performed GO and KEGG pathway analyses on these 509 common DEPs to elucidate the possible association between the mechanism of AD treatment with gf-targeted PBM. The results are shown in Fig. 7. Many of the common DEPs are related to components in the cell and on the cell membrane, speci cally, the binding of G protein complexes and receptors on the cell membrane and the binding of the cytoskeleton, RNA, and proteins. These DEPs are mainly involved in the negative regulation of cellular processes such as metabolism, especially the metabolism of nitrogen compounds, and the regulation of transport and localization. For the KEGG pathway analysis, in addition to the DEPs associated with AD, the DEPs are mainly involved in oxidative phosphorylation, calcium signaling pathways, in ammation, phagocytosis, the secretion of various hormones (such as thyroid hormones, parathyroid hormones, aldosterone, renin, cortisol, and insulin), and the formation of synaptic structures.
These ndings imply that in ammation, repair, and reconstruction are all related to the improvement of AD with PBM.

Changes of intestinal ora diversity after gf-targeted PBM treatment
We compared the changes in intestinal ora diversity of each group after PBM treatment using 16S rRNA gene amplicon sequencing. Figure 8A shows the principal components analysis (PCA) of gut microbiome composition at the operational taxonomic unit (OTU) level for the mice. The gut ora of the ve groups clustered separately, and the clusters of the three PBM groups were located between the clusters of the AD mice and the normal mice. ANOVA and Kruskal-Wallis tests were used to analyze the species with signi cant differences between groups. As shown in Fig. 8B, there were differences at all taxonomic levels, especially at the genus level. A boxplot analysis was conducted to determine the relative abundances of four of the top ten species present at different levels among the study groups. As shown as Fig. 8C, the abundances of Helicobacter, Oscillibacter, Ruminiclostridium-5, and uncultured Bacteroidales increased, while Rikenella, Desulfovibrio, Ruminococcus-2, and Butyricicoccus decreased in AD mice compared with the normal mice. PBM treatment corrected this imbalance of bacteria to some extent. In particular, all PBM treatments (PBM630, PBM730 and PBM850) reversed the increase of Helicobacter and uncultured Bacteroidales and the decrease of Rikenella in AD mice. Some of the ndings were wavelength-speci c. For example, PBM630 decreased the abundance of Oscillibacter, PBM730 increased the abundance of Desulfovibrio and Ruminococcus-2, and PBM850 corrected the imbalance of Butyricicoccus and Ruminiclostridium-5 in AD mice. All PBMs decreased the abundance of Ruminococcus-1. However, the abundance of Ruminococcus-1 was not signi cantly different between the AD and normal mice. We also performed a KEGG function prediction based on the 16S sequencing data, conducted a statistical analysis among groups according to the Kruskal-Wallis algorithm, and homogenized the results to form a heat map (Fig. 8D). We found that the functional composition of the intestinal ora in AD mice was different from that of normal mice, especially for pyrimidine metabolism, lipopolysaccharide synthesis, and bacterial toxins. PBM630 and PBM850 performed to correct the functional abnormality, bringing the functional composition closer to that of the normal ora.

Discussion
We applied PBM to the abdomen (referred to as gf-targeted PBM) rather than the brain of AD mice to verify whether PBM could interfere with the AD process by regulating gut ora. Encouragingly, we con rmed that the gf-targeted PBM treated AD by eliminating Aβ plaques and inhibiting neuroin ammation and tau phosphorylation. The diversity and abundance of gut ora changed after long-term PBM irradiation.
The wavelengths of PBM were selected to be 630 nm, 730 nm, and 850 nm in this study. According to previous reports, the penetration depth of red light to infrared light in tissue was in the range of 5-10 mm (24), while the abdominal wall of mice was as thin as 1 mm. Therefore, we were con dent that the energy of PBM could directly act on the intestinal ora in this study. It has been reported that PBM has a bidirectional regulation effect on bacteria. The proliferation of E. coli was accelerated by a 810 nm laser irradiation at 5 J/cm 2 , while Pseudomonas aeruginosa was inhibited by 23% (22). Brian Bicknell et al. found that PBM could affect intestinal ora. They irradiated the abdomen of normal mice with a laser of 808 nm for 14 days and found that the abundance of Allobaculum (a bene cial gut bacteria) signi cantly increased (25). aged 65 and over who were initially free of dementia and followed for 20 years (28). Therefore, we speculated that PBM irradiation to reduce the abundance of Helicobacter pylori in the intestinal ora might be one of the main ways to improve AD. Moreover, uncultured Bacteroidales and Rikenella are also the possible targets of PBM for AD. The wavelength of PBM has a large in uence on the regulation of intestinal ora. In this study, 630 nm, 730 nm, and 850 nm PBM had unique effects on some ora as shown in Fig. 8C. All of these effects indicate the positive regulation of PBM on the dysbacteriosis of AD mice ora. These results suggest that the wavelength of PBM is an important and optimal physical parameter in future clinical applications.
The primary aim of this study was to verify whether PBM could improve the condition of AD when the intestinal ora were regulated. Cognitive behavior is the most direct clinical manifestation of AD. We found that PBM at three wavelengths (630 nm, 730 nm, and 850 nm) can effectively improve the impaired learning ability of AD mice, but it affected memory ability to a lesser extent (Fig. 2). The hippocampus is responsible for functions such as storage conversion and orientation in long-term memory. The hippocampus is essential to spatial navigation via a cognitive map, which is not dedicated to spatial cognition and navigation, but organizes experiences in memory. Justin D. Shin et al. also showed that the hippocampal-prefrontal region not only contributes to spatial learning, but also to memory-guided decision making (29). In our study, PBM improved the learning ability of AD mice but not the memory ability of mice, indicating that PBM could not completely repair the hippocampal networks. It is possible that our study needed a longer irradiation time or optimized conditions of PBM to observe memory improvement.
Amyloid plaques, tau phosphorylation, and neuroin ammation are the main pathological changes of AD (4). We found that mice in the PBM groups showed amyloid plaque elimination, p-tau reduction, and microglia proliferation. In addition, the PBM intervention resulted in a large number of proteins up-or downregulated in the hippocampus of AD mice (Fig. 5). Most of these common DEPs are involved in hormone secretion and in ammatory responses (Fig. 7), which are the main pathways by which the intestinal ora may affect AD (7). In the molecular pathway of AD, key proteins in the mitochondrial respiratory chain complex enzyme are up-or downregulated and tau protein is signi cantly decreased.
Mesenteric immunity is an important part of the body's total immunity, and it is also the tissue most likely exposed to PBM during abdominal PBM treatment. However, our results showed that PBM did not signi cantly increase the proliferation of lymphocytes and macrophages in the mesenteric lymph nodes except for the increase in mesenteric blood ow (Fig. 4). This suggests that cellular immunity does not participate in gf-targeted PBM. However, in humoral immunity, we found that INF-γ was signi cantly increased after PBM irradiation, especially for PBM630 and PBM730. It has been reported that the metabolites of intestinal ora can increase the serum levels of in ammatory factors such as IL-1α, IL-6, IL-10, TNF-α,and INF-γ, and promote the hyperphosphorylation of tau proteins and the excessive activation of microglia through the blood-brain barrier (30). In our Aβ 1−42 -induced AD mice, IL-6 and INF were both at a lower level compared with the normal mice. PBM did not increase IL-6 levels, but it signi cantly increased INF levels. We believe that this is due to the selective regulation of intestinal ora by PBM.

Limitations
First, Aß-induced AD mice is a simple AD model which simulates the main pathological features of AD patients through the deposition of exogenous Aß 1−42 proteins in the hippocampus. Other causes of AD are not re ected in this model, so we will attempt to verify these results in two or more AD models in future studies. Second, the modeling time was short. The PBM intervention started 2 weeks after modeling, although the intestinal ora of AD mice was signi cantly altered at the end point (10 weeks after modeling) compared to the control group. The causal relationship between intestinal ora and AD is relatively complex, but in this experiment, it was arti cially assumed that AD pathological changes would rst cause intestinal ora abnormalities. Future studies will gather more evidence to verify the effect of PBM on AD and intestinal ora. Finally, although we found that Helicobacter pylori may be the dominant PBM regulatory ora, su cient single-factor validation was still lacking.

Conclusions
In conclusion, gf-targeted PBM with light at 630 nm, 730 nm, and 850 nm reversed the imbalance of intestinal ora and improved learning ability, amyloid plaque deposition, tau phosphorylation, and microglia in ammation of Aß-induced AD mice. A large number of proteins in the hippocampus responded to gf-targeted PBM, with mitochondrial respiratory chain complex enzymes as a possible key intermediate target.
In future studies, we will con rm the effect of gf-targeted PBM on the brain-gut axis in additional AD animal models, such as APP/PS1 double-transgenic mice, and fully verify the targeted ora of PBM, which will promote PBM as a potential prevention and treatment method for AD.