Gut Microbiota-Targeted Photobiomodulation Ameliorates Alzheimer’s Pathology via the Gut–Brain Axis: Comparable Efficacy to Transcranial Irradiation

All animal work was approved by the animal ethical and welfare committee at the Institute of Radiation Medicine, Chinese Academy of Medical Sciences & Peking Union Medical College (IRM-DWLL-2022239) and performed with strict adherence to the guidelines set out in the National Institutes of Health Guide for the Care and Use of Laboratory Animals. A total of 24 six-month-old female APPswe/PS1dE9 transgenic mice and 8 ten-month-old female wild-type (WT) C57BL/6 mice were purchased from Beijing HFK Bioscience Co., Ltd. (Beijing, China). Genotyping verification for the APPswe/PS1dE9 line was provided by the supplier. At approximately 6 months of age, APPswe/PS1dE9 mice typically exhibit prominent amyloid-β plaque deposition in the cortex and hippocampus, accompanied by mild microglial activation, early synaptic alterations, and subtle cognitive impairment. Mice were housed under a standard 12-h light/dark cycle, with four animals per cage, and were given free access to food and water. APPswe/PS1dE9 AD mice were randomly assigned to groups by the experimenter, with 8 mice in each group. Random numbers were generated using the standard = RAND () function in Microsoft Excel. Sample sizes were chosen in accordance with power analysis and with similar previously published experiments. Each mouse was used as one experimental unit.

The PBM devices used in this study were custom-developed in our laboratory for delivering tc-PBM and gm-PBM in APPswe/PS1dE9 mice. All LED modules were assembled and calibrated prior to experimentation.

A continuous-wave 810 nm LED was used as the light source. The output power was adjusted to achieve a stable irradiance of 25 mW/cm² at the surface of the target region. Each mouse received irradiation for 20 min per day over a period of four weeks. The optical output was verified before each session to ensure consistent device performance.

For tc-PBM, the LED array was mounted onto a 3D-printed head-fixed helmet structure, designed using SolidWorks software (Version 2022, Dassault Systèmes, France) to ensure precise alignment over the prefrontal cortex (PFC) and hippocampal (HIPP) regions. The LED array was driven by a parallel–series hybrid circuit configuration, powered by an LT3465AES6 DC-DC constant-current driver. The output current was regulated using an external feedback resistor (R_FB). A passive heat sink was installed on the back of the LED board to prevent thermal buildup and to maintain the surface temperature below 37 °C during the irradiation period.

For gm-PBM, a 3D-printed arch-shaped mounting platform was used to hold a 4 × 3 LED array positioned above the abdominal region, enabling repeatable, stable, and non-contact illumination of the gut area. This configuration ensured consistent body positioning and irradiation alignment throughout the intervention.

The mice were restrained and fixed in an awake state (with limbs taped), and the irradiation sites were treated with depilation cream. The exposed areas were then shaved and fully exposed before the LED device was placed on the abdomen (supine position) or the head (prone position) for irradiation. During the light exposure, the position of the light source was monitored to ensure that no displacement occurred.

The LEDs emitted light at a wavelength of 810 nm in continuous output mode, with a power density of 25 mW/cm² and an irradiation time of 20 min.

AD mice in the sham group underwent the same procedures as the treatment groups, except that no light was applied, including the same duration of restraint. The control group consisted of healthy 10-month-old wild-type (WT) mice. To minimize subjective bias, cage labels indicating group names were replaced with serial number codes during the evaluation phase after the intervention.

The Morris Water Maze (MWM) was used to assess cognitive ability in mice. After a 4-week course of light intervention, mice were rested for 1 week before MWM testing was started. First, mice were trained for 3 days to find a platform in the pool. This was followed by a 5-day learning ability assessment. The platform was placed in the first quadrant of the pool, and mice entered the water from each of the other three quadrants. Behavioral tracking and swim-path analysis were performed using the Tracking Master System, Version 4.0-MWM (Shanghai Fanbi Intelligent Technology Co., Ltd., Shanghai, China). A trial was considered successful when the mouse reached the hidden platform and remained on it for at least 3 s. The ≥3-s dwell criterion was adopted to distinguish intentional platform recognition from accidental contact, and this approach has been widely implemented in Morris water maze protocols to ensure that escape latency accurately reflects spatial learning performance. Escape latency was defined as the time from entry into the water to the moment the mouse continuously remained on the platform for ≥3 s. If a mouse failed to locate the platform within 120 s, a latency of 120 s was recorded, and the mouse was gently guided to the platform and allowed to remain there for 30 s. On the final day, the platform was removed and mice entered the water from any quadrant. The percentage of time spent in the platform area and the number of times the mice crossed the platform quadrant were calculated to assess the memory ability of the mice.

All antibodies and reagents used in this study were commercially sourced. Primary anti bodies included Iba1 (Abcam, Cambridge, UK; ab178846), β-Amyloid 1–42 (Abcam, Cambridge, UK; ab201061), phospho-Tau (Ser396) (Abcam, Cambridge, UK; ab109390), F4/80 (Abcam, Cambridge, UK; ab300421), synaptophysin (ABclonal, Wuhan, China; A6344), COX-2 (Cell Signaling Technology, Danvers, MA, USA; 12282T), and CD163, ZO-1, and occludin (Proteintech, Wuhan, China; 16646-1-AP, 21773-1-AP, and 27260-1-AP).

Secondary antibodies included Cy^TM3-conjugated donkey anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA, USA; 711-165-152) and Alexa Fluor^® 488 donkey anti-rabbit IgG (Jackson ImmunoResearch, 711-545-152). HRP-polymer detection reagents (anti-rabbit and anti-mouse) were obtained from Fuzhou New Step Biotechnology (Fuzhou, China; KIT-5005 and KIT-5002). Tyramide signal amplification (TSA) fluorophores, including Cy5-tyramide, Cy3-tyramide, and fluorescein-tyramide, were obtained from AAT Bioquest (Sunnyvale, CA, USA; 11066, 11065, and 11062).

Mouse tissue was prepared after MWM testing. Brain and colon tissues were rapidly removed and fixed for 24 h in 4% paraformaldehyde. The tissues were then dehydrated, embedded, and prepared as paraffin sections with a thickness of 5 μm. The tyramide signal amplification technique was used for multi-label immunofluorescence staining. Aβ1–42 and Iba1 were stained together to observe the pathological changes in AD plaques and brain inflammation. Co-staining of p-tau and synaptophysin was used to visualize nerve fibers and nerve axons. Aβ1–42, COX2, and CD163 were co-stained to observe the relationship between inflammation and plaques. For colon tissue, F4/80, ZO-1, and occludin were co-stained to observe intestinal inflammation and the intestinal barrier. For all image quantification, ImageJ 1.54f (http://imagej.org/↗, accessed on 10 February 2025) was used. The positive area percentage, plaque counts, and cell counts were analyzed for all experiments. To quantify the fraction of microglia overlapping with Aβ, we used clolc2 in ImageJ to analyze colocalization and acquire the MCC scores.

Stool samples from the colons of mice were collected (n = 5). 16S rRNA gene amplicon sequencing analysis was then used to analyze the classification and abundance of mouse gut microbiota. Bacterial V3/V4 hypervariable regions of the bacterial 16S rRNA gene were amplified using a polymerase chain reaction (PCR) system with the following primers: 341F (5′-CTACGGGNGGCWGCAG) and 805R (3′-GACTACHVGGGTATCTAATCC). The PCR products were detected using electrophoresis and purified using magnetic beads. After purification, the PCR products were used as a two-round PCR template. The DNA was then detected and purified again before being quantified using a Qubit (Invitrogen Qubit 4 Fluorometer, Thermo Fisher Scientific, Waltham, MA, USA). Finally, the samples were mixed according to the concentration of PCR products and sequenced on the MiSeq platform (Illumina, San Diego, CA, USA).

Qiime 2 was used to analyze the data, and DADA2 was called to denoise the raw data. The redundancy of the denoised sequence was removed to obtain the feature information of the amplicon sequence variants (ASV). Species annotation was performed for each ASV to obtain the corresponding species information and species-based abundance distribution. The relative abundance, alpha diversity, beta diversity, and differences between groups were analyzed using the ASV_table. Principal coordinates analysis, principal component analysis, non-metric multidimensional scaling, and other dimensionality reduction diagrams and sample clustering trees were used to display the bacteria. The differential bacteria were screened based on linear discriminant analysis effect size analysis and a Kruskal–Wallis test. Furthermore, PICRUSt2 was used to predict the function of the flora; the databases included MetaCyc, Enzyme Commission (EC) Kyoto Encyclopedia of Genes and Genomes (KEGG), and Cluster of Orthologous groups (COG) of proteins.

All data with error bars are represented as the mean ± SD. For comparisons between two groups, an unpaired two-tailed Student's t-test was applied. For comparisons among more than two groups, one-way ANOVA was performed. p < 0.05 was considered significant. The GraphPad Prism (Version 9.5, GraphPad Software, San Diego, CA, USA) was used for the statistical analysis of data and figure production. For image quantification, ImageJ version 1.54f (http://imagej.org/↗, accessed on 10 February 2025) was used.

To evaluate the effects of gm-PBM and tc-PBM on cognitive functions in APPswe/PS1dE9 mice, we administered daily light irradiation (wavelength: 810 nm; power density: 25 mW/cm²; duration: 20 min/day) to 6-month-old APP/PS1 mice for 4 consecutive weeks, followed by Morris water maze (MWM) testing during the 5th week (Figure 1A).

During the 5-day spatial acquisition training phase (Figure 1B–D), escape latencies among all groups did not significantly differ on day 2 (WT: 66.88 ± 21.96 s; AD: 79.96 ± 11.85 s). However, by day 5, the escape latency in AD mice significantly increased to 71.21 ± 17.99 s compared to WT controls (33.78 ± 13.34 s). gm-PBM treatment moderately reduced the latency to 48.80 ± 18.91 s, although this difference did not reach statistical significance when compared to the untreated AD group (p = 0.17). In contrast, tc-PBM intervention significantly shortened the latency further to 39.12 ± 14.85 s.

In the subsequent spatial probe test (Figure 1E,F), AD mice exhibited a significantly reduced number of entries into the target quadrant (2.0 ± 1.22 times) compared to WT mice (7.8 ± 1.3 times, p < 0.01). The gm-PBM treatment group showed a clear trend toward improvement in entry frequency (3.8 ± 1.48 times); although the difference did not reach statistical significance versus the AD group (p = 0.18), the numerical increase may hold biological relevance. Regarding time spent in the target quadrant, gm-PBM significantly prolonged the duration (24.92 ± 8.15 s) compared to the AD group (12.23 ± 6.72 s, p < 0.05), suggesting a beneficial effect on spatial memory. In comparison, tc-PBM treatment significantly increased both the number of entries (5.8 ± 1.48 times) and the duration spent (26.94 ± 4.67 s) in the target quadrant, demonstrating a more robust cognitive improvement in this behavioral paradigm.

To evaluate the therapeutic efficacy of gm-PBM and tc-PBM in alleviating AD pathology, we first examined whether these two light-based interventions could reduce β-amyloid (Aβ) plaque burden in the hippocampus (HIPP) and prefrontal cortex (PFC) of APP/PS1 transgenic AD mice.

Our results revealed that photobiomodulation exerted a robust effect in attenuating Aβ deposition. As shown in Figure 2A,F, the AD group exhibited dense red fluorescent Aβ plaques in both the hippocampus and PFC, while both gm-PBM and tc-PBM treatments markedly reduced plaque density. Quantitative analysis confirmed that both PBM modalities significantly decreased plaque burden, albeit to varying extents (Figure 2B–E). Specifically, in the hippocampus, gm-PBM and tc-PBM reduced total plaque number by 46% and 67%, and total plaque area by 42% and 61%, respectively, compared to the AD group. Notably, both treatments also significantly diminished the number and area of large plaques (>100 μm²).

Similar trends were observed in the PFC (Figure 2G–J). Overall, both gm-PBM and tc-PBM effectively reduced Aβ plaque accumulation in brain regions critical for cognition. However, gm-PBM exhibited superior clearance efficiency in several quantitative parameters, suggesting that remote gm-PBM may offer enhanced therapeutic benefits compared to direct transcranial irradiation.

Neuronal apoptosis and synaptic loss are two additional key pathological features of AD. To determine whether different PBM strategies promote neuronal survival and synaptic restoration, we further examined their effects in relevant brain regions.

In the hippocampus (Figure 3A–C), WT mice showed dense and continuous synaptophysin (Syn, green; a presynaptic marker) labeling, whereas p-Tau (red) was nearly undetectable. In contrast, APPswe/PS1dE9 mice exhibited a marked reduction in Syn signal, along with a mild increase in p-Tau labeling. It is important to note that this model does not typically develop classic neurofibrillary tangles; however, previous studies have reported age-dependent, Aβ-associated early elevations in p-Tau at this stage, which are interpreted as indicators of neuronal stress and synaptic dysfunction rather than mature tau aggregation. Therefore, the measurement of p-Tau in this study serves primarily to evaluate downstream Aβ-related neurotoxic responses rather than canonical tauopathy.

Compared with untreated AD mice, both gm-PBM and tc-PBM groups showed increased Syn expression accompanied by reduced p-Tau labeling, suggesting partial restoration of synaptic structure and cellular homeostasis. A similar tendency was observed in the prefrontal cortex (Figure 3D–F).

Collectively, both PBM strategies effectively restored synaptic marker expression and suppressed Tau hyperphosphorylation. Notably, gm-PBM demonstrated superior efficacy in repairing synaptic integrity specifically within the hippocampus, a brain region critical for memory formation, compared to its effects in the prefrontal cortex.

Neuroinflammation plays a dual role in the pathology of AD, contributing to amyloid clearance during early stages but exacerbating disease progression in later phases. To further elucidate the effects of PBM on AD-associated neuroinflammation, we examined the area fraction of ionized calcium-binding adapter molecule 1 (Iba1)-positive microglia in the PFC and hippocampus, as well as their spatial association with Aβ plaques.

As shown in Figure 4A–C, AD mice exhibited substantial microglial activation, with Iba1-positive areas reaching 19.38 ± 2.06% in the hippocampus and 11.96 ± 1.77% in the PFC. Both gm-PBM and tc-PBM significantly reduced Iba1 coverage in these regions. No significant differences were observed between the two PBM treatment groups, indicating that both PBM strategies effectively attenuated pathological microglial aggregation.

Mander's colocalization coefficient (MCC) analysis further revealed that in AD mice, the Iba1/Aβ colocalization coefficients were 0.54 ± 0.04 in the hippocampus and 0.46 ± 0.07 in the PFC (Figure 4D,E). tc-PBM treatment increased these values to 0.66 ± 0.04 and 0.66 ± 0.05, respectively, while gm-PBM yielded coefficients of 0.61 ± 0.07 (HIPP) and 0.63 ± 0.04 (PFC). These results suggest that PBM enhances the spatial overlap between microglia and residual Aβ plaques, potentially reflecting elevated microglial phagocytic activity toward Aβ following intervention.

The polarization of microglia toward either the M1 (pro-inflammatory) or M2 (anti-inflammatory) phenotype is influenced by pathological insults, inflammatory conditions, and therapeutic interventions. In the untreated AD group, the proportion of M2-type microglia—labeled by CD163—was significantly reduced in the hippocampus compared to the WT group, while the expression of cyclooxygenase-2 (COX-2), a marker of M1-type microglia, was markedly elevated. These findings indicate a phenotypic shift toward pro-inflammatory activation, characterized by a decrease in anti-inflammatory microglia and upregulation of inflammatory mediators (Figure 5A–C).

Following gm-PBM and tc-PBM treatments, the number of CD163-positive cells in the hippocampus increased significantly, accompanied by a notable reduction in COX-2 expression. Similar trends were observed in the PFC (Figure 5D–F).

These results suggest that PBM intervention effectively restores the M2 anti-inflammatory microglial phenotype (CD163) and suppresses the expression of the pro-inflammatory enzyme COX-2. Both PBM modalities demonstrated comparable efficacy, indicating that gm-PBM is as effective as tc-PBM in modulating neuroimmune responses in the AD brain.

In addition to its beneficial effects on the central nervous system, whether PBM can modulate peripheral gut homeostasis via the gut–brain axis represents a crucial aspect of its therapeutic potential. To address this, we further evaluated the impact of different PBM strategies on intestinal inflammation and gut barrier integrity. The results demonstrated that gm-PBM exhibited a notable advantage in alleviating intestinal inflammation in AD model mice.

Using F4/80 as a macrophage marker, we observed a marked increase in F4/80-positive area in the colons of AD mice, indicating extensive infiltration of inflammatory cells (Figure 6A). tc-PBM treatment significantly reduced this area to 3.77 ± 1.1%, whereas gm-PBM achieved a further reduction to 1.03 ± 0.36%, suggesting that remote modulation via the gut–brain axis is more effective in suppressing localized intestinal inflammation (Figure 6B).

Moreover, tight junction proteins ZO-1 and Occludin were significantly disrupted in the AD group: the ZO-1-positive area decreased from 27.57 ± 5.04% in WT mice to 2.51 ± 0.89%, and Occludin from 39.72 ± 5.64% to 8.46 ± 2.8%. tc-PBM treatment partially restored ZO-1 (13.58 ± 2.63%) and Occludin (26.66 ± 3.42%) levels. In contrast, gm-PBM led to more robust recovery, increasing ZO-1 and Occludin expression to 20.83 ± 2.31% and 44.09 ± 1.54%, respectively (Figure 6C–E).

Collectively, these findings indicate that gm-PBM not only significantly mitigates AD-associated intestinal inflammation but also more effectively restores epithelial barrier integrity compared to tc-PBM, further supporting its therapeutic potential in modulating the gut–brain axis.

Venn diagram analysis (Figure 7A) revealed 114 ASVs shared between the AD and WT groups, while gm-PBM and tc-PBM interventions selectively enriched 45 and 66 unique ASVs, respectively. Linear discriminant analysis (LDA) (Figure 7B–D) further indicated that, compared to WT mice, the AD group was enriched with potentially pathogenic genera such as Parabacteroides, Pseudomonas, and Romboutsia, while short-chain fatty acid SCFA-producing taxa including Butyricicoccus, Lachnospiraceae_NK4A136_group, and Eubacterium_siraeum_group were markedly reduced. gm-PBM intervention (Figure 7C) significantly enriched beneficial SCFA-producing genera such as Butyricicoccus, Lactococcus, and multiple members of the Eubacterium genus (LDA > 3.5). In contrast, tc-PBM (Figure 7D) primarily enriched genera like Aerococcus, Corynebacterium, and Jeotgalicoccus.

Bar plots further illustrated (Figure 7E) a significant reduction in SCFA-producing bacteria including Eubacterium_coprostanoligenes_group and Butyricicoccus in the AD group, accompanied by a pronounced increase in pathogenic Pseudomonas. gm-PBM treatment (Figure 7F) markedly restored the abundance of these SCFA-producing genera and significantly reduced Pseudomonas levels. tc-PBM (Figure 7G) also exerted partial restorative effects on these taxa, although to a lesser extent compared to gm-PBM.

Finally, we assessed microbial functional potential via heatmap-based predictive analysis. The COG heatmap (Figure 7H) showed that the AD group exhibited upregulation of functions related to "multidrug efflux pumps" and "cell wall-associated antibiotic resistance", whereas both gm-PBM and tc-PBM interventions activated beneficial functional modules such as "ABC transporters" and "signal transduction components", indicative of improved metabolic regulation. The KEGG pathway heatmap (Figure 7I) revealed that AD mice had suppressed activities in pathways such as "fatty acid biosynthesis" and "apoptosis". Notably, gm-PBM treatment significantly restored key metabolic pathways including "secondary bile acid biosynthesis" and "glucose metabolism", as indicated by red upregulation signals. These findings suggest that gm-PBM more effectively remodels gut microbial metabolic function, offering a mechanistic basis for its superior capacity to mitigate gut–brain axis-related inflammation.

Encouraged by the finding that gm-PBM significantly restored microbial diversity and enriched SCFA-producing taxa in AD mice, we further conducted fecal metabolomics analysis across WT, AD, and gm-PBM groups. This was aimed at elucidating how microbial remodeling translates into potential metabolic mechanisms for ameliorating AD pathology.

Compared to WT mice, the AD group exhibited widespread disturbances in the fecal metabolite profile (Figure 8A,D), with 14 metabolites significantly upregulated and 43 downregulated (p < 0.05, |log₂FC| > 1). Following gm-PBM intervention (Figure 8B,E), 63 metabolites were significantly upregulated and 99 downregulated. Notably, in the WT vs. gm-PBM comparison (Figure 8C,F), 31 metabolites were restored to WT levels, while 157 showed inverse regulation relative to the AD group, indicating that gm-PBM effectively reversed AD-associated metabolic disruptions.

KEGG level 1 pathway classification (Figure 8G–I) revealed that AD mice showed marked downregulation in critical pathways including lipid metabolism, amino acid metabolism, and energy metabolism. gm-PBM intervention significantly reactivated pathways such as fatty acid biosynthesis, secondary metabolite biosynthesis, and signal transduction. Enrichment analysis (Figure 8J–L) further demonstrated that the WT vs. AD comparison showed significant enrichment in pathways related to MAPK signaling, tryptophan metabolism, and ABC transporters. After gm-PBM treatment, the AD vs. gm-PBM comparison revealed enrichment in fatty acid metabolism, PPAR signaling, and AMPK signaling pathways. Importantly, the pathway profile of WT vs. gm-PBM highly overlapped with that of WT vs. AD, suggesting a restoration of metabolic network organization toward a healthy state following gm-PBM treatment.

Collectively, gm-PBM markedly reversed fecal metabolic dysregulation in AD mice by bidirectionally correcting over 160 key metabolites involved in lipid, amino acid, and energy metabolism, and reactivating core metabolic pathways such as fatty acid biosynthesis and the PPAR/AMPK signaling axes. This metabolic reprogramming provides robust molecular support for the neuroprotective effects observed with improved gut microbiota diversity and highlights a potential mechanism by which gm-PBM alleviates AD pathology through a coordinated "microbiota–metabolite–signaling" axis.