Far Infrared Light Irradiation Enhances Microglial Aβ Clearance and Ameliorates Cognitive Decit in Alzheimer’s Disease-like Mice

Background: Exposure to sunlight may decrease the risk of developing Alzheimer’s disease (AD). However, the wavelength of the light with this therapeutic effect and the related mechanism remain elusive. Failure to clear amyloid-β (Aβ), the main component of amyloid plaque, has been considered as a key risk to cause the development of AD. As the resident immune cell in the brain, microglia is able to carry out Aβ clearance. We hypothesize that a component of sunlight improves AD-related cognitive dysfunction and that this benecial effect may be via Aβ clearance by microglia. Method: The APP/PS1 mice by 8.5 months of age were exposed to the visible light (λ = 500 nm), near infrared light (λ = 800 nm) and far infrared light (λ = 3 - 25 µm) for 60 min per day. After 5-week treatment with different light, all mice began to be subjected to Morris water maze behavior test under SPF environment. Western blotting was carried out to detect the expression of postsynaptic density-95 protein and synaptophysin in the hippocampus. The protein amount of interleukin (IL)-1β and IL-6 in the cerebral cortex were determined by using ELISA kits. Immunostaining was performed to characterize the Aβ, microglia, cluster of differentiation 68. Under the condition of Aβ existing, primary cultured microglia were treated with light for 2 h before Aβ phagocytosis assay. Results: The APP/PS1 mice with different light (VIS, NIR and FIR) treatments showed a trend of improvement in learning compared to those without light treatment during training stage. Moreover, FIR light-treated APP/PS1 mice had better spatial memory than APP/PS1 mice in the probe test. Simultaneously, FIR light treatment restored synaptophysin protein expression, promoted the recruitment of microglia to Aβ plaques, enhanced the phagocytosis of Aβ, reduced Aβ burden and alleviated neuroinammation. Conclusions: Taken together, FIR light treatment ameliorates the learning and memory impairment in AD-like mice. Our ndings uncovered a previously unappreciated function of FIR light, suggesting that FIR light treatment may be a potential therapeutic strategy quantication of Aβ plaque-associated microglia was similar to the previously reported method with minor modication [20] . Within 20 µm range from the edge of an Aβ plaque, the number of Iba1 staining positive cells was manually counted. At least 100 plaques in the cerebral cortex region and 40 plaques in the hippocampus per group were quantied. For further analysis, plaques were divided into several groups according to their sizes (< 300, 300 to 600, 600 to 1200 and > 1200 µm 2 ). The percentage of microglial CD68 + area was expressed as the co-staining area of Iba1 + and CD68 + divided by Iba1 + area. To further analyze the proximal interaction of microglia with Aβ plaque, the percentage of the co-staining area of 6E10 + and Iba1 + in total 6E10 + area was calculated. The co-staining area of 6E10 + , CD68 + and Iba1 + divided by the total 6E10 + area was calculated to analyze the proximal interaction of CD68 + cells with Aβ plaque.


Introduction
Alzheimer's disease (AD) is the most common neurodegenerative disease in the elderly, and is characterized by the presence of amyloid plaques and neuro brillary tangles and a progressive loss in memory and mental function [1,2] . To date, there is no disease process-modifying intervention for AD [3] . It is estimated that more than 100 million people worldwide will live with AD by 2050 [4] , which will be an enormous burden on the family, caregivers and society. Therefore, the prevention and treatment of AD are urgently needed. One major theory is that the imbalance between amyloid-β (Aβ) production and clearance results in the accumulation and aggregation of Aβ, which lead to many pathological processes in AD pathology, including neuroin ammation, oxidative stress and synaptic de cit [5,6] . Therefore, it has been a focus to nd drugs that can eliminate the effects of Aβ. A large number of anti-Aβ drugs including semagacestat, bapineuzumab and solanezumab had been proceeded to clinical trials but most of them had failed [7] . Various explanations for the failures have been suggested including side effect, neuroin ammation induction and wrong drug doses [7,8] . Probably the most important reason is that the exact pathophysiology for AD remains not fully understood. Current drug therapies only treat symptoms rather than the underlying disease pathology [9] .
Although the exact pathophysiology of AD remains unclear, it is widely recognized that the combination of genetic, environmental and lifestyle factors play an important role in AD progression. Interestingly, it was reported that exposure to sunlight could be linked to decreased risk of dementia [10] . Old people with non-melanoma skin cancer (NMSC) probably caused by overexposure to the sun had a markedly reduced risk of developing AD as compared to whom without NMSC [11] . Near infrared light irradiation could mitigate Alzheimer's disease-related pathology and synaptic vulnerability [12,13] . Even ultraviolet radiation exposure was able to enhance learning and memory of mice [14] . These studies suggest that non-invasive photobiomodulation has a great potential in the treatment of AD. So far, pharmacological therapeutic strategies have not taken advantage of the brain's endogenous clearance mechanisms. As the resident immune cells in the brain, microglia are able to carry out Aβ clearance [15,16] . However, very little is known on how to manipulate these cells for enhanced Aβ clearance due to the lack of noninvasive methods to measure this process [17] . It is worthwhile mentioning that sunlight has transcranial potential, especially for infrared light [18] . Based on the ndings of its bene cial effects on dementia, sunlight exposure probably has a protective effect on brain cells, including microglia. Sunlight has three major components: ultraviolet light, visible light and infrared radiation [19] . However, which component of sunlight is most bene cial in reducing AD-related neuropathology and cognitive dysfunction and whether they affect the microglia for Aβ clearance remains unclear.
We hypothesize that a component of sunlight improves AD-related cognitive dysfunction and that this bene cial effect may be via Aβ clearance by microglia. To test these hypotheses, we investigated the effects of the visible light (λ = 500 nm), near infrared light (λ = 800 nm) and far infrared light (λ = 3-25 µm) on the cognitive ability of AD-like mice. We also determined whether the light with a potential to treat AD could affect microglial Aβ clearance.

Methods
Animals. The APP swe /PSEN1dE9 double-transgenic (APP/PS1) and their littermate wild-type (WT) mice were obtained from Nanjing Biomedical Research Institute of Nanjing University (Nanjing, China). These mice were housed in speci c pathogen free (SPF) environment with a 12-hour light/12-hour dark cycle.
Mouse was caged singly and allowed to have food and water ad libitum. All experimental animal procedures were approved by the Institutional Animal Care and Use Committee (Approval No.: IACUC-20180918-010) and the Laboratory Animal Ethics Committee of Jinan University.
Light irradiation treatment. At the age of 8.5 months, APP/PS1 mice were randomly distributed to the following groups: sham treatment group, the visible light (λ = 500 nm) treatment group (VIS light-treated APP/PS1 mice), the near infrared light (λ = 800 nm) treatment group (NIR light-treated APP/PS1 mice), and the far infrared light (the wavelength range was from 3 to 25 µm) treatment group (FIR light-treated APP/PS1 mice). A WT mouse group was also included. Mice in all groups were allowed free feeding during light irradiation that was 60 min per day for 1.5 months.
Morris water maze. After 5-week treatment with different light (VIS, NIR and FIR), all mice began to be subjected to Morris water maze behavior test under SPF environment. The Morris water maze experiment apparatus consisted of a circular pool with a diameter of 120 cm and a height of 50 cm, and water was injected into the pool. The camera was located just above the tank (connected to the computer), which was used to record the movement of mice. An appropriate amount of non-toxic protein powder was added into water to facilitate this recording. The pool temperature was maintained at 20-22 °C, and the water surface was divided into four quadrants (from one to four) and twelve zones (from one to twelve).
The water maze experiment was divided into two parts: the rst is the training trial section (navigation stage of water maze), and the second is the spatial probe test (space memory stage). In the training trial, the target platform (8 cm in diameter, 20 cm in height) was located in the center between zone 2 and zone 6 in the second quadrant, and the water surface was 1 cm higher than the target platform, so that the mice could not see the platform. The experimental parameters were as follows: swimming time (60 s), residence time on target platform (3 s). During the water maze training trials, mice were placed into the water in accordance with the east and west directions. After mouse facing the pool wall into the water, the time required to nd the target platform was recorded, namely escape latency. Mice were allowed to stand on the target platform for 10 s when climbing up to the target platform. However, if mice were unable to nd the target platform within 60 s, they were manually guided to nd the target platform, and allowed to stay on it for 15 s. Each training cycle was done for 4 times and for 7 continuous days. In the spatial probe test that was 24 h after the last training trial, the target platform was removed. Mice were placed into the water again and observed for a period of time (60 s). The number of entry into effective zones to nd the target platform was recorded to re ect the spatial memory of mice.
Immunostaining. After the behavioral tests, the brain tissues of the mice were isolated and bisected longitudinally. The left hemisphere was frozen at -80 °C for further biochemical study. The right hemisphere was soaked in 4% paraformaldehyde solution for 24 h. Subsequently, they were immersed in 10%, 20% and 30% sucrose solutions each for 24 h and then stored in Tissue-Tek OCT compound at -20 °C. Coronal 30-µm thick sections from the right hemisphere were cut by a cryostat. The sections were washed with phosphate buffered saline (PBS).
For immunohistochemistry, the activity of endoperoxidase was blocked by 3% H 2 O 2 . Antigen retrieval with citric acid buffer (pH 6.0) was performed at 95 °C for 10 min. After washed with PBS, sections were incubated with 0.3% triton X-100 and 5% donkey serum for 60 min at room temperature. The sections were then incubated with mouse anti-Aβ antibody (1:10000, Sig-39300, Biolegend) overnight at 4 °C.
Sections were then washed with PBS containing 0.3% triton x-100 and incubated with goat antimouse/rabbit IgG conjugated with horseradish peroxidase (Universal kit, PV-600, Zhongshan Jinqiao, Beijing, China) at room temperature for 30 min. After washed with PBS, sections were incubated with hydrogen peroxide (DAB kit, Zhongshan Jinqiao) and co-stained with hematoxylin [E803FA0003, Sangon Biotech (Shanghai) Co., Ltd, China] at room temperature for 5 min. Images were captured using a Nikon microscope (Nikon Ni-U) For immuno uorescent staining, after blocking with 0.3% triton X-100 and 5% donkey serum at room temperature for 60 min, sections were then incubated at 4 °C overnight with the following primary antibodies: mouse monoclonal anti-Aβ (6E10) antibody (1:10000, Sig-39300, Biolegend), rabbit polyclonal anti-ionized calcium binding adapter molecule 1 (Iba1) (1:1000, 019-19741, Wako) or rat polyclonal anti-cluster of differentiation 68 (CD68) (1:2000, ab53444, Abcam). Sections were rinsed with PBS containing 0.3% Triton X-100. The sections were incubated at room temperature for 2 h with uorescent secondary antibodies as follows: donkey anti-mouse IgG antibody conjugated with Alexa Aβ plaque-associated microglial analysis. The quanti cation of Aβ plaque-associated microglia was similar to the previously reported method with minor modi cation [20] . Within 20 µm range from the edge of an Aβ plaque, the number of Iba1 staining positive cells was manually counted. At least 100 plaques in the cerebral cortex region and 40 plaques in the hippocampus per group were quanti ed. For further analysis, plaques were divided into several groups according to their sizes (< 300, 300 to 600, 600 to 1200 and > 1200 µm 2 ). The percentage of microglial CD68 + area was expressed as the co-staining area of Iba1 + and CD68 + divided by Iba1 + area. To further analyze the proximal interaction of microglia with Aβ plaque, the percentage of the co-staining area of 6E10 + and Iba1 + in total 6E10 + area was calculated. The co-staining area of 6E10 + , CD68 + and Iba1 + divided by the total 6E10 + area was calculated to analyze the proximal interaction of CD68 + cells with Aβ plaque.
Primary microglial culture. Primary microglial cultures were performed as previously described with slight modi cations [21,22] . Brie y, C57BL/6 mouse at postnatal 0 to 3 days was quickly immersed in 75% ethanol, and then the entire brain tissue was carefully removed with a sterile spatula. The olfactory bulbs, cerebellum and hind brain were dissected away, and the meninges were also removed. Subsequently, the cortical tissues were harvested, transferred to a 35-mm petri dish containing 2 mL Hanks' Balanced Salt Solution (HBSS) and minced into approximately 1 mm pieces by using sterile surgical scissors. The minced tissues were mechanically dissociated into suspension with a 1 mL pipette, and then ltered into a 50 mL conical tube by using a 40-µm mesh lter. The ltrated cell suspension was then transferred into a poly-L-lysine-coated T75 ask and cultured with Dulbecco's modi ed Eagle's medium (DMEM) (C1995500BT, Gibco) supplemented with 10% fetal bovine serum (FBS) (SFBE, Natocor). After 3 days, medium was changed to that composed of 25 ng/mL granulocyte-macrophage colony-stimulating factor (GM-CSF) (415-ML-010, R&D system) and 10% FBS. After cultured for 12-14 days, microglial cells on top of the mixed glial cell layer were harvested by shaking the culture ask and were used for further experiment when their purity evaluated by Iba1 staining was over 95%.
Phagocytosis assay. The microglial phagocytosis of Aβ 1−42 was analyzed similarly to a previously reported method with slight modi cations [23] . Brie y, uorescein amidite-labeled Aβ 1−42 (FAM-Aβ 1−42 ) (AS-23525, Anaspec) was rst aggregated at 37 °C for 24 h with agitation. Primary microglial cultures were seeded in the poly-L-lysine-coated 6-well plates at a density of 1 × 10 5 cells/cm 2 and then cultured overnight. After the cells were treated with FIR light for 1 h, the aggregated FAM-Aβ 1−42 was added into the culture medium with a nal concentration of 0.8 µg/mL. The primary microglial cultures were continuously treated with FIR light for additional 2 h. FAM-Aβ 1−42 -containing cultured medium was removed. Cultures were washed three times with PBS. Subsequently, the uorescent images were randomly captured in three elds for each well by using an OLYMPUS microscope (OLYMPUS, IX71). For uorescent intensity detection by ow cytometry, primary microglial cultures were digested with 0.25% trypsin-EDTA (25200056, Gibco), and then were centrifuged at 1500 g for 5 min at 4 °C and washed with PBS for three times. For the microscopic images, the green uorescent intensity of internalized FAM-Aβ 1−42 in each well, which was captured under a green uorescent protein channel, was quanti ed based on the mean of uorescent intensity from three random elds by using ImageJ 1.52n. The results of three images were averaged to re ect the phagocytosis activity of the cells in the well. In the case of detection by ow cytometry, the green uorescent intensity of internalized FAM-Aβ 1−42 from 10000 cells each sample was recorded under a FITC channel and then was expressed as the Geo mean for each sample. These measurements were performed by a person who was blind to group assignments. Finally, the green uorescent intensity of treated samples was normalized to the mean uorescent intensity of the control group without FIR light treatment.
Elisa assay. The protein amount of interleukin (IL)-1β and IL-6 in the cerebral cortex were determined by using ELISA kits (catalog No: E-EL-M0037c for IL-1β, Catalog No: E-EL-M0044c, Elabscience) according to the manufacture's instruction. Brie y, small pieces of the cerebral cortex of mice were weighted. The tissue was then homogenized in PBS on ice with protease inhibitor cocktail for general use (P1005, Beyotime) [tissue weight (g): PBS (mL) volume = 1:9]. Subsequently, the homogenates were centrifuged at 5000 g for 8 min at 4 °C. The amount of IL-1β and IL-6 in the supernatant was then determined. The nal level of IL-1β and IL-6 was then normalized to its protein content determined by a BCA protein assay kit (P0010, Beyotime).
Western blot. The level of synaptic proteins postsynaptic density-95 protein (PSD-95) and synaptophysin in the hippocampus were determined by using Western blotting as previously described with slight modi cations [24,25] . The hippocampal tissues were homogenized in the RIPA lysis buffer (CW23335, CWBIOTECH) by using 1 mL insulin syringe on ice. Homogenates were centrifuged at 12000 g at 4 °C for 15 min. The supernatant was used for Western blotting. The protein content of each sample was determined by the enhanced BCA protein assay kit (P0010, Beyotime). Twenty microgram protein per lane was electrophoresed on a 10% polyacrylamide gel and then transferred to a polyvinylidene uoride (PVDF) membrane. The membrane was blocked by 5% nonfat milk at room temperature for 60 min and then incubated at 4 °C over night with the following primary antibodies: mouse monoclonal anti-PSD-95 antibody (1:1000, 75 − 028, NeuroMab), mouse monoclonal anti-synaptophysin (1:1000, MABN1193, EMD Millipore), or rabbit monoclonal anti-β-actin (1:1000, 8457s, Cell Signaling). After washed with Tris Buffered Saline Tween (Catalog No: T1081, Solarbio), the membranes were then incubated with corresponding secondary antibodies at room temperature for 2 h. The signal of protein band was visualized using NcmECL Ultra (Catalog No: P10200, Ncmbiotech). The protein band intensity of PSD-95 and synaptophysin was normalized to the corresponding band intensity of β-actin from the same sample.
Statistical analysis. All data were expressed as means ± S.E.M. The statistical analysis of results was performed by using GraphPad Prism version 7.0. The signi cant difference was assessed by unpaired t test, one-way or two-way ANOVA followed by Turkey's or Dunett multiple comparisons test. The data of training sessions in the Morris water maze test was analyzed by two-way repeated measures ANOVA. A p value < 0.05 was considered statistically signi cant difference.

Results
FIR light ameliorated cognitive dysfunction of AD-like mice. The time-course of the in vivo experiments is shown in Fig. 1A. The Morris water maze test consisted of two stages: training trials and spatial probe test (Fig. 1B). WT mice took shorter time than APP/PS1 mice to nd the platform in the training trials of the test (Fig. 1C). This impaired learning in APP/PS1 mice was improved by the treatment of various wavelengths of light [F(4, 203) = 4.113, p = 0.0032]. Compared to APP/PS1 mice, FIR light-treated but not VIS light-or NIR light-treated APP/PS1 mice had more entries into the zones around the platform in the spatial probe test (Fig. 1D). Notably, there was no difference in swimming speed among WT mice and APP/PS1 mice treated with or without various lights (Fig. 1E). These results suggest that FIR light treatment improves spatial memory of AD-like mice.
FIR light enhanced the expression of synaptic protein in AD-like mice. The impairment of synaptic plasticity including the decreased expression of proteins in synapse is a common phenomenon in the progression of AD [26] . PSD-95 and synaptophysin are a pre-and post-synaptic protein, respectively, and play a pivotal role in regulating synaptic plasticity, which closely correlates with cognitive functions [27,28] .
Since the cognitive dysfunction of APP/PS1 mice was attenuated by FIR light, we next determined whether FIR light could in uence the level of PSD-95 and synaptophysin in the hippocampus. The level of PSD-95 did not exhibit signi cant changes in APP/PS1 mice compared with WT mice. However, the level of synaptophysin was decreased in the APP/PS1 mice compared to the WT mice but this decrease was attenuated in APP/PS1 mice treated with FIR light (Figs. 2A to 2C). Therefore, these results suggest that FIR light treatment is bene cial to restore synaptic protein expression in AD-like mice.
FIR light promoted microglial recruitment around the Aβ plaque and enhanced Aβ phagocytosis in AD-like mice. As the resident immune cells in the brain, microglia participate in the clearance of Aβ [29] . In the present study, numerous microglial cells were recruited around the Aβ plaques (Fig. 3A). In the cortex region, FIR light treatment enhanced the recruitment of microglia surrounding the Aβ plaques (Figs. 3B and 3C). The number of microglial cells surrounding the Aβ plaques was increased with increased sizes of Aβ plaques. FIR light increased the number of microglial cells no matter which size of Aβ plaques was considered (Fig. 3D). Similar results were observed in the hippocampus (Figs. 3E to 3G).
To investigate whether the microglia surrounding the plaques could engulf Aβ to achieve Aβ clearance, co-immunostaining of Iba1, CD68 and Aβ were carried out and the CD68 + microglial phagosomes and internalized Aβ were quanti ed. The results showed that the microglia around the Aβ plaque expressed the phagocytosis marker CD68 (Fig. 4A). Compared to the microglia in APP/PS1 mice, the microglia in the APP/PS1 mice with FIR light treatment signi cantly had increased CD68 expression (Fig. 4B). In addition, the interaction between microglial cells and Aβ was enhanced due to the increase in their contact area and phagocytic area in the These results were in agreement with the ndings that FIR light was able to signi cantly increase Aβ phagocytosis by microglia in the APP/PS1 mice.
Together, these results suggest that FIR light treatment increases microglial recruitment to the plaques and enhances microglial phagocytosis, which is bene cial to clean Aβ in the AD-like mice.
FIR light alleviated Aβ burden in the brain of AD-like mice. The Aβ accumulation and aggregation are mainly resulted from the imbalance of Aβ production and clearance and result in a series of subsequent pathological events for AD progression [5,6] . Since there was a signi cant enhancement of microglial phagocytosis of Aβ in the APP/PS1 mice treated with FIR light, we next determined whether FIR light could reduce the Aβ load in the AD-like mice. Compared to WT mouse, APP/PS1 mouse exhibits abundant Aβ plaque in both cerebral cortex and hippocampus (Fig. 6A). A signi cant reduction of Aβ plaque in these brain areas of the FIR light-treated APP/PS1-mice was found compared to the APP/PS1 mice (Figs. 6B to 6G). These results indicate that FIR light treatment signi cantly reduces Aβ load in the brain of APP/PS1 transgenic mice.
FIR light decreased neuroin ammation in AD-like mice. Neuroin ammation mainly resulted from abnormal activation of microglia and astrocytes has been observed in the AD-like mouse brains and patients with AD [30] . Neuroin ammation can cause neuronal damage [30] . Thus, we determined whether FIR light treatment could alleviate neuroin ammation in the AD-like mice. The pro-in ammatory factors including IL-1β and IL-6 were signi cantly higher in the APP/PS1 mice compared to WT mice (Figs. 7A and 7B). However, they were markedly reduced in the FIR light-treated APP/PS1 mice. These results suggest that FIR light treatment is able to alleviate neuroin ammation in the AD-like mice.

Discussion
With the increased life expectancy, much attention had been paid to the treatment of AD. During the progression of AD, the most common symptom is the decline in cognitive function including poor memory and learning skill. Accordingly, improving cognitive dysfunction has been a goal of AD treatment.
Interestingly, it has been reported that sunlight exposure could decrease the risk of developing AD [10,11] .
Sunlight mainly contains three lights, namely ultraviolet light, visible light and infrared radiation [19] . In the present study, VIS light (λ = 500 nm), NIR light (λ = 800 nm) and FIR light (λ = 3-25 nm) were tested as a treatment strategy. APP/PS1 mice with different light treatments showed a trend of improvement in learning compared to those without light treatment during training stage. Notably, FIR light-treated APP/PS1 mice had better spatial memory than APP/PS1 mice in the probe test, indicating that FIR light treatment may have potential to be a therapeutic strategy for AD.
The potential of using light as an AD treatment had been investigated. VIS light was found to inhibit the aggregation of Aβ under in vitro condition [31] . Photobiomodulation with NIR light attenuated AD-related pathology in K3 mice (aged 5 months) and APP/PS1 mice (aged 7 months) [12] . In the present study, we compared the effects of VIS, NIR and FIR light on the cognition function of the 8.5-month-old APP/PS1 mice that start to have cognition de cits by 6 months of age [32] . Only FIR light improved the spatial memory in APP/PS1 mice. FIR light irradiation that can penetrate up to 1.5 inches beneath the skin had been reported to have many positive biological effects in animals or humans, including improving blood circulation, ameliorating endothelial dysfunction, relieving fatigue and pain, lowering blood pressure, and promoting capillary dilation [33,34] . The current study shows for the rst time the bene cial effects of FIR light on the AD-like mice, such as improved cognitive function, restored the expression of the synaptic protein synaptophysin, reduced Aβ plaque burden in the cerebral cortex and hippocampus and decreased neuroin am-mation.
Signi cant Aβ accumulation is a character of AD brain. Excessive Aβ has been hypothesized to drive the neuropathogenic cascades of AD including neuroin ammation, oxidative stress and synaptic failure [35,36] . Reducing excessive Aβ in the brain has been proposed to be a potential treatment strategy for AD.
Previous studies also showed that NIR light could reduce Aβ burden in the brain of AD mice [12,13] . However, the mechanisms underlying these effects remain to be fully elucidated. As the resident immune cells in brain, microglia may be the rst responder to clean Aβ through phagocytosis [29] . Accordingly, we try to focus on the involvement of microglia in the underlying mechanism for FIR light to decrease Aβ burden in the brain of AD mice. It is known that the phagocytosis ability of microglia decreases with age, leading to the reduction of Aβ clearance during AD development [37] . In the current study, FIR light enhanced microglial phagocytosis of Aβ. FIR light promoted the recruitment of microglia to Aβ plaque with increased CD68 + phagosomes in the APP/PS1 mice and enhanced microglial Aβ engulfment in mouse brain and primary microglial cultures.
Therefore, this nding is in agreement with the result that FIR light was able to reduce Aβ plaque burden in the cerebral cortex and hippocampus of APP/PS1 mice. Namely, FIR light treatment may act on microglia to reduce the deposition of Aβ in the brain of AD-like mice.
Although the microglial recruitment to Aβ and phagocytosis of Aβ in APP/PS1 mice were enhanced after FIR light treatment, the exact mechanism of these FIR light biological effects on microglia remains poorly understood. Previous study showed that NIR light can be absorbed by mitochondrial chromophores, such as cytochrome c oxidase (COX) [38] . This photon absorption possibly promotes the dissociation of the inhibitory molecule, NO, from the copper B site, resulting in the activation of the enzyme [39] . The loss of NO will rapidly lead to the increase in electron transport, oxygen consumption, and adenosine triphosphate (ATP) production [34,39] . Consequently, signaling pathways will be activated and involved in the activation of transcription factors, such as nuclear factor-κB [40] . However, the principle chromophore at FIR light wavelengths is not COX but rather water [34] . Possible concept is that the nanometer water layers exist in the hydrophobic surfaces, such as cellular and mitochondrial membranes, and FIR light can perturb the structure of membrane underlying the nanoscopic water layers, resulting in the open of ion channels in the membrane and then possibly enhancing the mitochondrial respiration activity [34,41] .
Nevertheless, it cannot be excluded that FIR light may directly in uence the COX activity [42,43] .
Microglial phagocytosis requires dynamic reorganization of the actin cytoskeleton, for which a substantial amount of ATP are necessary [44] . However, microglia in response to cerebral Aβ deposition will had a metabolic switch to glycolysis [45] , which produces much lower amounts of ATP per glucose molecule than the oxidative metabolism modulated by mitochondrial respiratory activity [46] . Accordingly, we speculate that FIR light treatment promotes the microglial recruitment to Aβ and phagocytosis of Aβ through enhancing mitochondrial respiratory activity, which needs further systemic experiments to support this speculation. In addition, it is noteworthy that abnormally active microglia will release proin ammatory cytokines to cause neuronal damage [47] . However, glycolytic switch of microglia in response to cerebral Aβ deposition would promote pro-in ammatory state change, which could be blocked by inhibition of glycolysis [48] . In the present study, we observed that FIR light treatment signi cantly decreased pro-in ammatory cytokines, such as IL-1β and IL-6, in APP/PS1 mice. This phenomenon may be explained that FIR light treatment inhibits glycolysis switch and enhances mitochondrial oxidative phosphorylation, which would be consistent with the above speculation that FIR light treatment could enhance mitochondrial respiratory activity.
Previous studies suggest that proper neuronal function is the underlying mechanism for cognitive ability [49][50][51] . However, the excessive Aβ and the pro-in ammatory cytokines will cause neuronal damage [35,36,47] , leading to synaptic function impairment. Synaptic plasticity de cit with decreased expression of synaptic protein can compromise normal neuronal functions, resulting in cognitive dysfunction, which often occurs in the AD brain [27,52] . In the present study, we found that APP/PS1 mice treated with FIR light had higher expression of synaptophysin than those without FIR light treatment. This result may possibly be a subsequence of the decrease in Aβ and proin ammatory cytokines after FIR light treatment. Together, our results suggest that FIR light treatment may normalize synaptic integrity, which may be a mechanism for the improved cognitive function of AD-like mice.

Limitations
Firstly, the design that wavelength of 500 nm, 800 nm and 3-25 µm represents as VIS, NIR and FIR light, respectively, is an over-simpli cation, while their wavelengths all cover extremely broad range [34] . So it needs tremendous workload to investigate the effect of different wavelength of light on cognitive function of AD-like mice. Secondly, the complexity of illumination parameters such as wavelength, uence, power density, pulse structure and treatment timing also challenge the rational choose, which may lead to the publication of both negative and positive studies [53] . Therefore, we have not studied in detail the effects of different wavelength of light on cognitive function of AD-like mice. Thirdly, the APP/PS1 two transgenic mouse occur increased Aβ and amyloid plaque by the age of 4 months and cognitive de cits by the age of 6 months [32,54] . However, the APP/PS1 mice acted as AD-like mice employed in the current study began to be subjected to FIR light treatment by 8.5 months of age. At this point in time, they were on the progressive stage of disease. In order to explore the effectiveness of FIR light treatment in more depth, further studies should consider the FIR light treatment strategy of AD-like mice in different age phase, such as 4-month-old and 6-month-old.
In the present study, we mainly focused on the effect of FIR light on the microglia, which would enhance the phagocytosis and clearance of Aβ. In addition to microglia, astrocyte similarly engaged in the process of AD development [55] . Activated astrocytes surrounding the Aβ neprilysin, an amyloid-degrading enzyme [56] . Functional experiment also suggested that astrocytes were able to phagocyte Aβ [57] .
Simultaneously, astrocytes in conditions of AD-like pathology were observed to express β-site APPcleaving enzyme 1, an important endoprotease for production of Aβ, thereby acquiring the ability to produce Aβ [58] . It was known that Aβ could remodel the astroglial glucose metabolism that involved in the ATP production and release [59,60] , which would mediate the chemotactic response of microglia [61] . In light of the pivotal role of astrocyte in the Aβ-induced condition, the further studies on functional changes of astrocyte under FIR light irradiation should be taken into consideration. It should be pointed out that FIR light might directly act on the neuron and in uence Aβ-induced synaptic dysfunction. Accordingly, the functional changes of neuron and synaptic plasticity after FIR light treatment should not be ignored in the future studies.
Beyond the limitation mentioned above, the mechanism for how FIR light ameliorated cognitive dysfunction of AD-like mice has not been clearly de ned although FIR light could enhance microglial Aβ phagocytosis and reduce Aβ burden in the brain of AD-like mice, subsequently contributing to the bene cial effect on synaptic plasticity and cognitive function. However, it could not be ruled out that FIR probably had peripheral effects after FIR light irradiations of whole-body without making a distinction between head and body. Previous work by others showed that moderate UV light exposure could elevate blood urocanic acid (UAC) [14] . After crossing the blood-brain barrier (BBB), UAC was able to be converted to glutamate through intra-neuronal metabolic pathway. Such UV-triggered glutamate synthesis could contribute to some of neurobehavioral changes. Therefore, further studies should assess the potential FIR light-induced peripheral molecule that could across BBB and then directly or indirectly regulate the neurobehavioral function.
Lastly, the molecules and pathways mediating the biological effects of FIR light on microglia, especially for the mechanism underlying enhanced microglial phagocytosis of Aβ after FIR light treatment, should be well studied.

Conlusion
In summary, we found that FIR light at wavelengths of 3-25 µm signi cantly promoted recruitment of microglia to the Aβ plaque, enhanced the phagocytosis of Aβ by microglia, reduced Aβ burden, decreased neuroin ammatory cytokines and restored the expression of synaptic protein in AD-like mice. FIR light treatment also ameliorated the learning and memory impairment of these mice. These results suggest the therapeutic potential of FIR light for AD. Further study is needed to determine this potential and the molecular mechanisms for the protective effects of FIR light. Authors' contributions QL conceived and conducted most of the experiments, and wrote most of the manuscript. ZW conceived the idea for the project, coordinated the research and modi ed the manuscript with ZZ and WW. JR did much work to improved quality of our work such as revising the paper meticulously, providing a professional editing. HZ analyzed the results. YL and JW collected the data.      staining area normalized to the total 6E10+ area. Data were means ± S.E.M. (n = 49 -52 per group). *P < 0.05 and ****P < 0.0001 compared to corresponding values of APP/PS1 mice without FIR light treatment.