Treatment by low-dose brain radiation therapy improves memory performances without changes of the amyloid load in the TgF344-AD rat model

Alzheimer's disease (AD) is a neurodegenerative condition affecting memory performance. This pathology is characterized by intracerebral amyloid plaques and tau tangles coupled with neuroinflammation. During the last century, numerous therapeutic trials unfortunately failed highlighting the need to find new therapeutic approaches. Low-dose brain radiotherapy (LD-RT) showed efficacy to reduce amyloid load and inflammation in patients with peripheral diseases. In this study, the therapeutic potential of 2 LD-RT schedules was tested on the TgF344-AD rat model of AD. Fifteen-month-old rats were irradiated with 5 fractions of 2 Gy delivered either daily or weekly. The daily treatment induced an improvement of memory performance in the Y-maze. In contrast, the weekly treatment increased the microglial reactivity in the hippocampus. A lack of effect of both regimens on amyloid pathology was unexpectedly observed. The positive effect on cognition encourages to further evaluate the LD-RT therapeutic potential and highlights the impact of the design choice of the LD-RT regimen.


Introduction
Alzheimer's disease (AD) is the most common form of dementia and thus represents a major societal challenge. The pathological hallmarks of the disease are amyloid and tau deposits, accompanied by neuroinflammation. Amyloid pathology takes place after abnormal accumulation of amyloid beta (A β) peptides which aggregates to form fibrils, oligomers and finally plaques and progress throughout the disease ( Masters et al., 2015 ;Thal et al., 2002 ). The plaque morphology evolves with the pathology, with a shift in favor of fibrillar plaques vs. diffuse plaques at end-stage of the disease ( Dickson and Vickers, 2001 ), while the proportion of dense-core plaques (~20%) appears to be stable. The amyloid acidic protein (GFAP) and Vimentin, and a hypertrophy of soma and processes. These structural changes are accompanied by many transcriptomic and functional modifications which may positively or negatively influence the progression of the pathology ( Ben Haim et al., 2015 ;Escartin et al., 2019 ). Microglial cells also undergo important morphological, transcriptomic and functional changes when they become activated. Indeed, overexpression of proteins such as ionized calcium binding adaptor molecule 1 (IBA1) or Cluster of Differentiation 68 (CD68) is observed and the most extreme reactive state of microglial cells is characterized by an ameboid form. Functionally, reactive glial cells produce higher rate of pro-and anti-inflammatory molecules such as cytokines, chemokines and reactive oxygen species ( Bagyinszky et al., 2017 ).
The involvement of glial cells in A β phagocytosis and debris degradation is also well described ( Ries and Sastre, 2016 ). However, it is often postulated that neuroinflammation is firstly a mechanism of defense for the brain, but with disease progression, this mechanism is overload. It generates itself pro-inflammatory molecules and reactive oxygen species, harmful for neuronal functioning and survival, consequently also participating to disease damage ( De Strooper and Karran, 2016 ).
Despite multiple clinical trials, no disease-modifying treatment against AD has been validated yet ( Ceyzériat et al., 2020b ). It is consequently important to develop and test alternative strategies. Radiation therapy (RT) at low-dose (LD) offers promising applications as it has the potential to act on both amyloidosis and neuroinflammation. The ability to disaggregate amyloid pathology, possibly interacting directly with the beta-sheet structure, has been shown in multiple forms of systemic amyloidosis ( Ceyzériat et al., 2020a ). Low-dose brain radiotherapy (LD-RT) has also shown remarkable anti-inflammatory effects in benign and chronic degenerative inflammatory diseases (see ref Ceyzériat et al., 2020a for a detailed review). Only 4 studies have evaluated the LD-RT effect in the context of AD using AD mouse models ( Kim et al., 2020a( Kim et al., , 2020bMarples et al., 2016 ;Wilson et al., 2020 ). In a first report, authors observed a decrease of amyloid plaques, and accompanied by an improvement of memory performances in a pure amyloid model, more important with a fractionated protocol of LD-RT (5 fractions of 2 Grays [Gy], delivered daily) ( Marples et al., 2016 ). They hypothesized that proinflammatory mechanisms of RT could play a role in A β clearance, but a global downregulation of microglial activation through an increase of anti-inflammatory cytokines could also be postulated ( Garibotto et al., 2016 ). The same group confirmed their results on amyloid load in the triple transgenic model of AD (3xTg-AD mice) and described a slight decrease of NFT ( Wilson et al., 2020 ). In a third model, no effect on amyloid was observed 4 days post RT but LD-RT did reduce microglial activation and improved synaptic protein levels, suggesting that LD-RT protected neurons from cellular death ( Kim et al., 2020b ). Eight weeks post treatment, a significant decrease of amyloid plaques was measured, with a shift from a pro-inflammatory to an anti-inflammatory cytokine production, mainly characterized in vitro ( Kim et al., 2020a ). Given the discrepant results reported, the mechanisms involved in amyloid load reduction and the effects of LD-RT on the neuroinflammation in the AD brain need further investigation.
The major aim of this study is to evaluate the therapeutic potential of 2 different schedules (applied weekly or daily) of LD-RT. We choose to treat animals at an advanced stage to have clear cognitive and behavioral baseline alterations, in order to be able to better evaluate the therapeutic potential of LD-RT on cognition. For that, we investigated the impact of LD-RT on memory performances, amyloid load and neuroinflammation using a TgF344-AD (TgAD) rat model ( Cohen et al., 2013 ). As the majority of the RT schedules commonly used to treat benign degenerative inflamma-tory pathologies deliver total doses in the range of 5 to 10 Gy in weekly or daily fractions ≤2 Gy ( Kriz et al., 2018 ), in the present study we tested 2 different brain RT regimens: 5 fractions of 2 Gy, delivered weekly and 5 fractions of 2 Gy, delivered daily.

Animals
The TgF344-AD rats present human APP Swedish and PS1 delta E9 transgenes in the Fisher 344 background ( Cohen et al., 2013 ). The hemi-brain of TgAD female rats was treated by 2 LD-RT schedules (described below) and analyzed 4 months after the end of LD-RT. Sham treated TgAD and non-transgenic littermates (WT) were used as controls. Animals were housed in a 12-hours light-dark cycle, with food and water ad-libitum . All the experimentation procedures on our animals were conducted with respect of the ethical approval by the Ethics Committee for Animal Experimentation of the Canton of Geneva, Switzerland (GE9917).

Radiation therapy
TgAD rats were anesthetized (2% isoflurane) and treated with half brain irradiation (right hemisphere) using a Truebeam Linear Accelerator (Varian Medical Systems, Palo Alto, CA, US) with the technique previously described ( Koutsouvelis et al., 2020 ). Two fractionation schedules were used: a) 10 Gy in 5 weekly fractions of 2 Gy (5 weeks overall treatment time) or b) 10 Gy in 5 daily fractions of 2 Gy (5 days overall treatment time). Sham-treated groups underwent only anesthesia to induce the same stress due to handling and repeated anesthesia.

Behavior
Treated animals of the first cohort (2 Gy x 5 fractions weekly) underwent behavioral experiments before RT and 4 months post treatment at 19-month-old. Treated animals and sham RT rats of the second cohort (2 Gy x 5 fractions daily) performed experiments only 4 months after RT. Locomotor activity was assessed using an Open Field (90 × 90 × 40 cm). Rats were placed at the center and video-track for 30 min using the EthoVision software (Noldus). The alternative Y maze test was used to evaluate the spatial working memory. Animals were placed at the extremity of the start arm of the device (50 × 50 × 10 cm) and video-tracked for 5 min. The alternations were measured and analyzed using the Etho-Vision software. Rats performing less than 4 total entries were excluded from the analysis. All devices were washed with water and carefully dried between each rat.

Histology
Animals were killed under anesthesia (2% isoflurane) by intracardiac perfusion with a saline solution. Their brain was fixed in a 4% paraformaldehyde solution for 24 hours and cryoprotect with a sucrose gradient (5%-20%). Serial brain sections (35 μm) were realized with a cryostat and stored in an anti-freeze solution at -20 °C until used for immunostaining. Left and right hemispheres were identified by a cut in the left thalamus.

Image analysis
All images were acquired using an Axioscan.Z1 (Zeiss) at 10x and analyzed using ImageJ software. For the A4G8 staining only performed on the cohort treated daily, a manual segmentation of the cortex and the hippocampus was realized to measure the potential dose effect due to laterality from the region targeted by radiations ( Koutsouvelis et al., 2020 ). The number of amyloid plaques and their individual area were measured using a semi-automatic detection of amyloid plaques with size and intensity threshold. The same technic was used to quantify MXO4 + plaques in the hippocampus. A distribution analysis of plaque area was realized using STATISTICA software: for small plaques ( < 50 μm ²) a pitch of 10 μm ² was used and a pitch of 200-250 μm ² was used for larger plaques ( > 50 μm ²). For GFAP and IBA1 staining, the mean grey value, representing the mean intensity of pixels on 16 bits images, was measured in the entire hippocampus (3 fields per slices, 4 slices per animal). For Vimentin and CD68 staining, the percentage of the area positively stained was measured using an intensity threshold in the molecular layer and the hilus of the hippocampus, 2 separated regions of interest (ROI) manually delimited (2-5 slices per animal). The cellular layer of the subgranular zone of the dentate gyrus was quantified using an intensity threshold in a defined field (3-4 slices per animal) to measure the percentage of the ROI positively stained with DAPI or Cresyl violet.

Statistical analysis
A sample size analysis with the graphical Douglas Altman's nomogram approach ( Ashby, 1991 ) was performed and significant data were reported if p ≤ 0.05 and β < 0.2. Analysis was performed in blind conditions and normality of residues was assessed with the Shapiro-Wilks test. For data presenting left and right comparison, Two-way ANOVA (Group and Hemisphere as between factors) and LSD post hoc test were used to compare the groups. Paired two-tailed student's t test or unpaired two-tailed student's t-test were used to compare 2 groups. Chi-2 test was performed to analyze plaque distribution data. Statistical details, including t or F distributions, are indicated in the Supplemental Table 1. All analyses were performed on Prism 8 (GraphPad). Results are presented as Mean ± SEM.

Results
All TgAD rats analyzed were unilaterally treated with the 2 different radiation regimens as planned ( Fig. 1 A). As the hippocampus, one of the main source of neural stem cells, is known to be particularly sensitive to radiations even at low doses ( Gondi et al., 2012 ), we validated that both LD-RT regimens do not induced a neuronal toxicity, as shown by the absence of hippocampal atrophy and modulation of the dentate gyrus thickness (Supplemental file 1).

Daily LD-RT improves memory performances and restores locomotion
The spatial working memory of animals was evaluated using the alternative Y maze test before and after LD-RT. Animals treated with the weekly schedule did not perform better the test ( Fig. 1 B), however, we observed a significant decrease of locomotion ( p = 0.045 ; Fig. 1 C). Interestingly, animals treated with the daily schedule greatly improved their memory performances ( + 42% of success to the test, p = 0.032 ; Fig. 1 D), accompanied with a reduction of the total distance traveled in the open field after treatment compared to sham-RT animals ( p = 0.044 ; Fig. 1 E).

LD-RT does not impact amyloid plaques in the hippocampus of aged TgAD rats
The MXO4 + amyloid plaques, labelling fibrillar dense-core plaques ( Serrano-Pozo et al., 2011 ), were counted in the hippocampus of treated and sham-treated TgAD rats ( Fig. 2 A). LD-RT applied weekly did not impact the density of amyloid plaques (number of plaques per μm ²; Fig. 2 B). We also analyzed the intensity of the staining in order to evaluate if plaques present a more diffuse morphology after treatment but no difference was obtained ( Fig. 2 D). The mean plaque area ( Fig. 2 F) and the distribution of the number of plaques in different categories of plaque size were also unchanged by LD-RT, suggesting that LD-RT did not impact neither small ( < 50 μm ² corresponding to ~79.2 ± 0.9 % of the plaques) or larger ( > 50μm ²) plaques ( Fig. 2 H). The same results were obtained with the daily schedule (number of plaques per μm ², The pitch size chosen to analyze small and larger plaques did not influence the results, as we could replicate the same results using a constant pitch of 10 μm ² ( Chi-2 test, p > 0.999 for both schedules ; data not shown).

No lateral effect of LD-RT on amyloid plaques
To go further, we evaluated the impact of LD-RT on both diffuse and dense-core plaques, using the A4G8 antibody, in the daily treated group which presented significant memory improvements in the open field is significantly decreased in the treated group compared to sham-treated TgAD rats. * p < 0.05. Abbreviation: Gy = Gray, mo = month-old, RT = radiation therapy, fx = fraction. ( Fig. 3 A). The absence of effect of the daily treatment in the hippocampus was confirmed for all parameters studied: plaque density per μm ², A4G8 staining intensity, mean plaque area ( Fig. 3 B). Thus, LD-RT did not impact either diffuse or dense-core plaques. As the radiation dose is less homogenous at the midline in the hemibrain irradiation ( Koutsouvelis et al., 2020 ), we quantified, in addition to the hippocampus, the amyloid plaques in the cortex. No effect was observed in the cortex either (plaque density per μm ², A4G8 staining intensity, mean plaque area; Fig. 3 C). Results were confirmed by right/left ratio (Supplemental file 3). The plaque distribution in different area categories was also unchanged in both regions (Supplemental file 4). Consequently, the absence of impact on amyloid plaques in our rat model did not differ across the irradiated volume.

Astrocyte reactivity is unchanged by LD-RT treatments in TgAD rats
As expected in 19-month-old animals, we observed a significant overexpression of GFAP in TgAD groups compared to WT rats, characterizing an astrocyte reactivity in this model ( p < 0.0 0 01 ; Figs. 4 A and B). No difference was observed after LD-RT compared to the sham-treated group, whatever the regimen studied ( Figs. 4 B and C, Supplemental file 5A and B).
To go further, we quantified separately the molecular layer and the hilus of the hippocampus, the second region containing a clear larger number of amyloid plaques. The quantification of the percentage of the ROI positively stained with an anti-Vimentin antibody, reflecting the Vimentin + astrocyte density in the hippocampus, tended also to be increased in the molecular layer of TgAD rats ( p = 0.059 for the main effect of group; Fig. 4 D). The Vimentin + astrocyte density tended to increase or reached the significance in the hilus of sham-treated and weekly treated TgAD rats respectively, compared to WT, showing that astrocyte reactivity was mainly localized around amyloid plaques, highly present in the hilus ( p = 0.125 WT vs sham-treated rats, p = 0.006 WT vs RT rats, p = 0.288 sham-treated vs RT rats; Fig. 4 F). However, neither a weekly or a daily treatment influenced the Vimentin + astrocyte density compared to sham-treated animals ( Figs. 4 D-G, Supplemental file 5C-F).  Right hemispheres of TgAD rats were treated with 2 Gy x 5 fractions delivered daily. (A) A4G8 + amyloid plaques were quantified in the hippocampus (B) and cortex (C) to evaluate potential depth or laterality effects of LD-RT treatment from the targeted area. Scale bar = 500 μm. No differential effect on amyloid plaques (density of plaques, A4G8 intensity, mean plaque area) is observed in the quantified regions. Abbreviation: L = Left hemisphere, R = Right hemisphere, fx = fraction.

Microglial reactivity is increased by weekly LD-RT treatment in TgAD rats
We observed a differential expression of IBA1 between groups. The overexpression of IBA1 almost reached the significance in sham-treated TgAD groups compared to WT rats, suggesting a microglial reactivity in this model ( p = 0.056; Figs. 5 A and B). How-ever, no treatment effect was observed on IBA1 levels ( Figs. 5 B and C, Supplemental file 6A, B).
A second marker, the CD68, more specific of microglial reactivity than IBA1 expression was quantified in the molecular layer and in the hilus of the hippocampus. A clear overexpression of CD68 was observed in microglial cells associated with amyloid plaques ( Fig. 5 A). The quantification of the percentage of the ROI positively stained with an anti-CD68 antibody, reflecting the CD68 + microglia density in the hippocampus, revealed significant differences in the molecular layer ( Fig. 5 D) and in the hilus ( Fig. 5 F). Indeed, the CD68 + microglia density significantly increased in both regions in TgAD rats compared to WT rats ( p = 0.018 WT vs sham-treated rats, p = 0.003 WT vs RT rats in the molecular layer, Fig. 5 D; p = 0.068 WT vs sham-treated rats, p = 0.002 WT vs RT rats in the hilus, Fig. 5 F). Moreover, a significant increase of CD68 + microglial density was observed in the weekly treated hemisphere compared to the contralateral side in the molecular layer ( p = 0.001) . These results showed a slight increase of microglial reactivity in the hippocampus after a weekly but not daily exposure to LD-RT ( Figs. 5 E  Fig. 5. Weekly LD-RT increases microglial reactivity in the treated hemisphere. Right hemispheres of TgAD rats were treated with different fractionated protocols of low-dose radiation therapy: 2 Gy x 5 fractions delivered weekly (left column), 2 Gy x 5 fractions delivered daily (right column). (A) Confocal illustration (40x) of microglia with IBA1 staining (Green) or CD68 (White) in WT and TgAD rats. Scale bar = 25 μm. (B) IBA1 levels tend to be increased in sham-treated TgAD rats compared to WT, characterizing microglial reactivity. However, no difference of IBA1 levels is observed after treatments (B, C). (d) The % of CD68 + area significantly increases also in the molecular layer of both TgAD groups. LD-RT (2 Gy x 5 fractions weekly) increases the % of area positively stained with CD68 in the molecular layer (D) but not the daily treatment (E). In the hilus, the % of CD68 + area tends to increase in sham-treated TgAD rats compared to WT and reach significance in weekly treated rats (F). However, any treatment impacts reactive microglia in the hilus (F, G). * p < 0.05, * * p < 0.01, ## p < 0.01. Abbreviation: L = Left hemisphere, R = Right hemisphere, fx = fraction. and G). However, the absence of significative difference in right/left ratio, considering bilateral variability of each animals, suggests that this increase of CD68 + microglial density is not due to a proinflammatory effect of weekly LD-RT, but seems to be a global effect of inflammation in TgAD rats compared to WT animals (Supplemental file 6C-F).

Discussion
Our study shows that the same LD-RT dose delivered weekly or daily does not have the same therapeutic potential. Indeed, only the dose of 10 Gy delivered in 5 fractions daily presented interesting therapeutic effects at an advanced stage of the pathology as it improved efficiently the memory of TgAD rats ( + 42%) and restored their locomotor activity. However, this regimen did not impact amyloid plaques or neuroinflammation markers. A slight increase of astrocyte and microglia reactivity was instead observed in the treated hemisphere with the weekly schedule, without any influence on amyloid load.
The toxic side effects of high-dose RT for brain tumors are well described ( Makale et al., 2017 ;Monje et al., 2002 ). The doses used in this protocol are more than 3 times lower compared to those used in clinical practice for brain metastases or focal glioblastoma for example and the long-term impact of low-doses irradiation are not fully understood. The toxicity of RT, even at low doses, could be related to hippocampus exposure ( Gondi et al., 2012 ). Indeed, the 2 key niches of neural stem cells in the adult brain are present in the subventricular zone and the subgranular layer of the dentate gyrus in the hippocampus ( Bacigaluppi et al., 2020 ), making the hippocampus a particularly sensitive region to irradiation ( Kazda et al., 2014 ). We verified that both regimens did not induce hippocampal atrophy by measuring the total volume of the hippocampus in sham-treated and irradiated animals and also did not induce neuronal loss in the subgranular zone of the dentate gyrus. These results validate the safety of regimens delivered weekly and daily in the TgAD rat model.
The primary aim of this study was to evaluate the therapeutic potential of both radiation regimens when applied at a stage presenting a severe amyloid pathology. The main cognitive alteration in AD being memory deficits, we pay attention to the effect of both treatments on the memory alteration displayed by TgAD rats at 14month-old. LD-RT delivered weekly did not ameliorate the spatial working memory of treated rats in contrary to the regimen delivered daily which improved it significantly. Interestingly, both regimens reduced the locomotor activity of rats. As a hyperactivity is described in this model ( Cohen et al., 2013 ), this reduction suggests a restoration of the normal behavior of TgAD rats. It is also interesting to note that rats were unilaterally treated. Considering the beneficial effect observed on cognition with the hemi-brain irradiation, we can hypothesize that a whole brain irradiation could have an even stronger impact on memory. Further studies will be necessary to evaluate the therapeutic potential of the regimen 2 Gy x 5 fractions applied bilaterally.
A positive effect of LD-RT on amyloid load was previously reported in patients presenting peripheral amyloidosis ( Cooper et al., 2018 ;Copperman et al., 2019 ;Khaira et al., 2008 ;Leibovitch et al., 2006 ;Luo et al., 2016 ;Neuner et al., 2012 ;Ren and Ren, 2012 ) and recently in 3 AD mouse models ( Kim et al., 2020a ;Marples et al., 2016 ;Wilson et al., 2020 ). Nevertheless, this reduction of the number of amyloid plaques was not observed in our rat model. A difference of MXO4 + plaque density between the 2 cohorts was observed in the hippocampus. This difference could come from the number of amyloid transgenes as it is not possible to distinguish homozygous from heterozygous animals. Consequently, it is possible that the second cohort included more homozygous animals, which would increase the number of amyloid plaques detected. This emphasizes the necessity of including an untreated control group for each cohort, as in our experimental design. We went further by analyzing plaques with 2 different methods: Methoxy-XO4 (MXO4), a fluorescent derivative of Congo Red, which labels fibrillar/dense-core plaques ( Serrano-Pozo et al., 2011 ) and the A4G8 antibody which stains both diffuse and fibrillar/dense-core plaques ( Alafuzoff et al., 2008 ). No difference was observed for the 2 methods, showing that neither diffuse nor fibrillar/densecore plaques are affected by LD-RT, whatever the regimen. MXO4 and A4G8 staining intensities, which could also reflect the plaque morphology (diffuse vs fibrillar/dense-core plaques), and the size of each individual plaques were also analyzed. LD-RT did not im-pact plaque morphology, without any differential effect for small or larger plaques, as shown by size distribution analyses. The major difference between this study and the published results in AD mice, in addition to the species studied, is that we applied the treatment at an advanced amyloid stage, whereas studies in mice applied LD-RT at the beginning of amyloid deposit formation. Indeed, for example, our model presented 8326 ± 1562 plaques only in the hippocampus of sham-treated rats compared to 63 ± 6.64 plaques in the whole brain of sham-treated APP/PS1 mice used in the study of Marples and collaborators ( Marples et al., 2016 ). Consequently, it is possible that the pathology in our TgAD rats was already too severe to observe a reduction of amyloid plaques in the hippocampus or in the cortex. A recent publication supports this hypothesis as the decrease of amyloid plaques was not observed after the same LD-RT regimen in 5XFAD mice, treated when an aggressive amyloid pathology was already present ( Kim et al., 2020b ). Furthermore, a recent study described a decrease of amyloid load in the hippocampus of TgAD rats after an early treatment with scyllo-Inositol ( Morrone et al., 2020 ), comforting the idea that an early amyloid-targeting treatment of AD rats may be more efficient on amyloid deposits.
Moreover, we cannot exclude that LD-RT did not impact amyloid plaques but decreased the soluble forms of the A β such as A β40, A β42 and/or A β oligomers, known to be the most toxic forms ( Walsh and Selkoe, 2007 ). Further analyses are consequently necessary to understand the impact of LD-RT on those soluble forms. Another hypothesis is the importance of the delay post treatment. Indeed, in the publications mentioned above, the maximal delay post RT analyzed is 8 weeks ( Kim et al., 2020a ;Marples et al., 2016 ;Wilson et al., 2020 ). Thus, it is possible that 4 months post RT, as in our study, the beneficial effect of LD-RT on the amyloid is caught up by the disease progression. Furthermore, it is important to keep in mind that amyloid load is not directly related to cognitive deficits. In addition to amyloid, the neuroinflammation now appears as a key player in AD pathology. For example, it has been recently demonstrated a functional link between neuroinflammation and functional network and/or cognitive impairment using PET imaging and functional magnetic resonance imaging in AD patients ( Passamonti et al., 2019 ). An anti-inflammatory effect of LD-RT, as shown in benign and degenerative inflammatory diseases in periphery ( Arenas et al., 2012 ;Rödel et al., 2012Rödel et al., , 2007, could obviously be an interesting aspect for AD. Neuroinflammation partners in the brain are astrocyte and microglial cells. Both are known to react to alteration in the brain tissue homeostasis. This reaction is characterized by a reactive state, classically identified through the overexpression of markers such as GFAP and Vimentin for astrocytes and IBA1 and CD68 for microglia, as examples. These morphological changes are accompanied by transcriptomic and functional modifications, which clearly influence AD pathology progression ( Ben Haim et al., 2015 ;De Strooper and Karran, 2016 ;Escartin et al., 2019 ). The anti-inflammatory effect of LD-RT, associated with the decrease of amyloid plaques, was also observed in AD mice ( Kim et al., 2020b ;Marples et al., 2016 ). Unexpectedly, our study showed that a weekly treatment by LD-RT did not reduce astrocyte reactivity and microglial activation but on the contrary, seems to induce an increase of CD68 overexpression in the hippocampus. Nevertheless, this increase seems to be slight and is not accompanied by cognitive benefits. Thus, complementary analyses on functional changes induced by LD-RT are necessary. The daily treatment did not influence the expression of the markers studied, suggesting that the delay between each irradiation may be an important parameter in the neuroinflammatory response. Besides, as reactive astrocytes and microglial activation are mainly observed around amyloid plaques, it is possible that the amyloid pathology of our animals was too important to expect a reduction of neuroinflammation at this stage.
On the contrary, we may also hypothesize that daily LD-RT did not reduce amyloid load because it is not able to modulate neuroinflammation partners. The mechanisms by which LD-RT reduces efficiently the amyloid load in peripheral pathologies are not understood. One hypothesis suggests a direct effect of LD-RT on the β-sheet structure of amyloid ( Bistolfi, 2008 ). But it is also possible that this decrease of amyloid is due to a modulation of inflammation. Indeed, peripheral macrophages can internalize A β ( Lai and McLaurin, 2012 ) but this ability could be altered in presence of chronic inflammation, as it is the case for microglial cells at an advanced stage of AD ( Heneka et al., 2015 ). In our model, LD-RT does not reduce inflammation and therefore does not enhance A β clearance capacities of microglia and astrocytes. Consequently, it would be interesting to perform the same experiments in younger animals to validate this hypothesis.
Finally, protective effects of LD-RT have been described against neurodegeneration. Indeed, in 5XFAD mice, a restoration of the immunoreactivity of the neuronal marker NeuN and the pre-synaptic marker synaptophysin was described in the hippocampus (CA1, CA3 and in the molecular layer) after LD-RT ( Kim et al., 2020b ). An increase of the microtubule-associated protein 2 and the postsynaptic density protein 95 was also observed in the hippocampus of AD mice after chronic irradiation ( Kempf et al., 2016 ). These results supposed a restoration of synaptic and neuronal integrity. At the transcriptomic level ( Kempf et al., 2016 ), a modulation of pathways related to synaptic plasticity and a significant decrease of synaptic neurodegenerative processes was pointed out in the hippocampus after a cumulative dose of 6 Gy but not at lower dose (0.3 Gy). Interestingly, a reduction of phosphorylation pathways of microtubule-associated protein tau is also observed in the GO term analysis, suggesting a reduction of the hyperphosphorylation of the tau protein, known to be related with its disassembly with microtubules, their disintegration and the aggregation of tau in NFT ( Barbier et al., 2019 ;Chang et al., 2008 ;Derisbourg et al., 2015 ). Consequently, the restoration of synaptic and neuronal integrity, in addition to a modulation of hyperphosphorylated forms of tau could explain the memory improvement observed in our model.
One major limitation of this study is the well-advanced amyloid pathology. Indeed, as discussed, it is highly possible that LD-RT treatments were applied too late in the pathology to expect to have a reduction of amyloid plaques and neuroinflammatory markers. Nevertheless, it could be interesting to measure the soluble forms of amyloid peptides, known to be the more toxic forms in AD. As our tissue samples were prepared for immunohistology, it was not possible to process the biochemical experiments to realize these types of measurements. Consequently, further analyses are necessary to better understand the impacts of LD-RT in the brain at early and late stages.

Conclusions
In our study, we validated that LD-RT with a schedule delivering 5 daily fractions of 2 Gy improves memory abilities of TgAD rats unlike the same RT dose using 2 Gy delivered once a week. With this regimen, we were unable to replicate previously reported effects on amyloid load or neuroinflammatory mechanisms. A slight increase of microglial activation markers was observed only in the treated hemisphere of weekly treated rats, however without associated memory or behavioral changes. These results contribute to the evidence of a biological effect of LD-RT in AD, in favor of a daily scheme. The lack of effect on amyloid or neuroinflammation at an advanced stage of pathology highlights the importance to study the mechanisms of action of LD-RT at an earlier disease stage.

Disclosure statement
The authors report no conflicts of interest.

Acknowledgements
We are grateful to Maria Surini and Yesica Gloria for the technical assistance. We thank all the team of the radiation-oncology division, Geneva University Hospitals, for its help with the radiation treatments of animals. Authors also thank the Rat Resource and Research Center (RRRC, Columbia) for providing the rat model. VG was funded by the Swiss National Science Foundation (projects 320 030_169876 and 320 030_185028); TZ was funded by the Swiss National Science Foundation (project 320030_182366). This work was supported by the Velux foundation (grant number 1123 ).

Verification
1. The authors have no conflicts of interest to declare, and their institutions have no contracts relating to this research.
2. This work was supported by the Velux foundation (grant number 1123).
3. This manuscript has not been previously published and is not under consideration in the same or a substantially similar form in any other peer-reviewed media.
4. All experimental procedures were approved by the Ethics Committee for Animal Experimentation of the Canton of Geneva, Switzerland.
5. All authors approve the manuscript and the accuracy of the data.

Supplementary materials
Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.neurobiolaging.2021. 03.008 .