Transcriptional signature in microglia isolated from an Alzheimer's disease mouse model treated with scanning ultrasound

Abstract Transcranial scanning ultrasound combined with intravenously injected microbubbles (SUS+MB) has been shown to transiently open the blood–brain barrier and reduce the amyloid‐β (Aβ) pathology in the APP23 mouse model of Alzheimer's disease (AD). This has been accomplished through the activation of microglial cells; however, their response to the SUS treatment is incompletely understood. Here, wild‐type (WT) and APP23 mice were subjected to SUS+MB, using nonsonicated mice as sham controls. After 48 h, the APP23 mice were injected with methoxy‐XO4 to label Aβ aggregates, followed by microglial isolation into XO4+ and XO4− populations using flow cytometry. Both XO4+ and XO4− cells were subjected to RNA sequencing and transcriptome profiling. The analysis of the microglial cells revealed a clear segregation depending on genotype (AD model vs. WT mice) and Aβ internalization (XO4+ vs. XO4− microglia), but interestingly, no differences were found between SUS+MB and sham in WT mice. Differential gene expression analysis in APP23 mice detected 278 genes that were significantly changed by SUS+MB in the XO4+ cells (248 up/30 down) and 242 in XO− cells (225 up/17 down). Pathway analysis highlighted differential expression of genes related to the phagosome pathway and marked upregulation of cell cycle‐related transcripts in XO4+ and XO4‐ microglia isolated from SUS+MB‐treated APP23 mice. Together, this highlights the complexity of the microglial response to transcranial ultrasound, with potential applications for the treatment of AD.


| INTRODUCTION
Alzheimer disease (AD) is the most common cause of dementia worldwide. The disease is characterized by progressive and irreversible neurodegeneration. However, given the complexity of the disease combined with a lack of knowledge on how to treat AD efficiently, there is an acute requirement to develop novel treatment strategies. 1 At a histopathological level, AD is characterized by the accumulation of extracellular amyloid-β (Aβ) plaques, intraneuronal tau deposits and increased microglial activation. 2 A broad range of studies have revealed how microglial cells assume both a protective role (through shielding, recognition, and removal of Aβ) and a detrimental role (through removal of synapses or the release of neurotoxic factors), driving the progression of AD. 3 Transcriptomic studies on microglia have advanced our understanding of the pathogenesis of AD at the level of transcriptional network dynamics, highlighting important molecular players depending on the different phases of the disease. [4][5][6] Microglia are known to phagocytose aggregated forms of Aβ, and it has been proposed that deficiencies in this process may contribute to late-onset AD 7 and metabolic labeling in humans indicated that clearance of Aβ is impaired in AD. 8 Recently, it has been shown that Aβcontaining microglia differ in their transcriptional signature in comparison to microglia that have not internalized the peptide. 9 An obstacle to treating AD is the blood-brain barrier (BBB), which prevents large molecules such as antibodies from entering the brain, with IgG having 0.1% transfer across the barrier. 10 Approaches to modify anti-Aβ antibodies to increase levels in the brain are in development, 11 along with other approaches to circumvent the BBB.  (a) Scanning ultrasound (SUS +MB ) or sham (no ultrasound) treatment was applied to APP23 transgenic and wild-type (WT) mice. Two days post-treatment, the mice received a single injection with methoxy-XO4 (that binds Aβ) 2 h before euthanasia and collection of brain tissue. The brains of the mice were harvested and homogenized to form a single-cell suspension, followed by FACS-based isolation of XO4 + and XO4 À microglial cells. (b) Inhouse prepared microbubbles were used for scanning ultrasound (SUS +MB ) and their size and concentration were measured using a Coulter Counter. (c) The gating strategy used to isolate microglial cells into XO4 + and XO4 À populations via FACS with CD11b and CD45 antibodies to isolate a pure population of microglia, and methoxy-XO4 fluorescence to isolate microglial cells that contain methoxy-XO4 bound to Aβ. (d) Methoxy-XO4 (blue) binds to Aβ plaques in the brains of APP23 mice, with Iba1-positive microglia in red. Scale bar: 50 μm Studies in animal models of AD have indicated that repeated transient BBB openings that are achieved throughout the entire brain using transcranial ultrasound in a scanning mode together with intravenously injected microbubbles (SUS +MB ) significantly clear amyloid plaques. One study reported that plaque reduction can occur as fast as 48 h after BBB opening, 12 and we have shown that this process occurs through microglial phagocytosis. 13 Ultrasound-mediated bioeffects (including microglial activation) have also been demonstrated by specifically targeting the hippocampus, 14,15 but the therapeutic benefit seems to be most pronounced when the brain is treated more globally. 13 Of note, this clearing process requires BBB opening 16 and is even effective at reducing Aβ pathology in 22-month-old senescent mice. 17 Combination treatments with ultrasound for delivery of anti-Aβ antibodies such as Aducanumab that has been recently approved by the Food and Drug Administration (FDA) 18 or an anti-pyroglutamylated Aβ antibody, 19 led to more effective plaque removal and behavioral improvements than in those observed in mice that were treated with either ultrasound alone or antibodies alone. 18 Ultrasound-mediated BBB opening has also been achieved in a small safety trial that revealed tolerability in patients with mild AD when a small region of the frontal cortex was targeted. 20 A subsequent clinical study found that the BBB could be opened in parts of the hippocampus, 21 with a modest reduction in the amyloid PET signal following three treatments with ultrasound over a 6-month period. 22 A recent clinical trial opened the BBB in the frontal lobes bilaterally and resulted in a modest reduction in the amyloid PET signal and significant improvement in neuropsychiatric symptoms. 23 In all these studies, BBB opening by ultrasound was shown to be safe and reversible in that the BBB was fully restored after 24 h.
Several mechanisms have been proposed to explain how BBB opening leads to amyloid plaque reduction, including the uptake of endogenous immunoglobulins 24 or albumin binding to amyloid, 13 followed by microglial phagocytosis of Aβ and lysosomal digestion.
Here, to gain a better understanding of how the combination of SUS

SUS
vs sham in XO4-microglia +MB F I G U R E 3 SUS +MB treatment leads to an increase in the number of differentially regulated genes in XO4 + microglia, when compared with sham-treated APP23 mice. (a) A Venn diagram depicting the number of genes up-regulated by SUS +MB distribute similarly between XO4 + and XO4 À cells, with many genes up-regulated in both. (b) A larger number of genes were down-regulated in the XO4 + cells compared with XO4 À cells following SUS +MB , with few genes down in both groups (adjusted p < 0.05). fluorescent dye to detect Aβ internalization within the microglia, we identified differences between the microglial cells from mice treated with or without ultrasound, as well as between cells that had internalized Aβ or not.
2 | RESULTS 2.1 | XO4 and FACS-based isolation of Aβ-positive and Aβ-negative microglia To understand the different effects of ultrasound-mediated BBB opening on plaque-phagocytic and non-phagocytic microglia in AD, we applied SUS +MB or sham (i.e., mice were anesthetized and injected with microbubbles but not exposed to ultrasound) to the brains of APP23 mice or WT littermate controls (Figure 1a,b). In addition, to be able to distinguish between microglial cells that had internalized Aβ and those that had not, we used the fluorescent Congo-red derivative methoxy-XO4 to stain Aβ within microglia when injected into live mice, as previously done. 9,25 This allowed us to use a fluorescence activated cell sorting (FACS)-based technique to separate and isolate XO4 + (Aβ phagocytic) and XO4 À (non-phagocytic) microglia following both SUS +MB and sham treatment paradigms (Figure 1c,d). by Aβ uptake (XO4 + vs. XO4 À cells), as well as treatment (SUS +MB vs. sham-treated animals), which were markedly accentuated in the APP23 samples. Of note, there was no effect of SUS +MB treatment in the microglial transcriptome of WT mice. Thus, we subsequently focused our analysis on the effects of ultrasound ± Aβ internalization in APP23-derived microglia only.

| SUS treatment induces an increased number of up-regulated genes in microglia
To gain insight into the response of APP23 microglia to the SUS treatment regime, we further analyzed the transcripts obtained from XO4 + and XO4 À cells. Our analysis identified 397 differentially enriched genes (FDR ≤ 0.05), with 155 genes being specific for XO4 + cells, 199 genes specific for XO4 À microglia, and 123 genes being independent of the Aβ signature. Analyzing the treatment-dependency patterns, we observed that most of the up-regulated genes were induced by SUS +MB , with a total of 353 enriched genes across all the Aβ internalization levels (Figure 3a), and only 44 genes that were down-  and "DNA metabolic processes" (Table 3). KEGG pathway analysis revealed that the most enriched pathways included "DNA replication" and "cell cycle," as well as established pathways in relation to the role of microglia in AD, such as "phagosome" and the "complement and coagulation cascade" ( cycle" and "phagosome" pathways in a treatment-(SUS +MB versus sham) and Aβ internalization (XO4 + versus XO4 À )-dependent manner revealed similar trends, with a stronger response found for the XO4 + microglia containing internalized Aβ (Figure 5a,b). More genes in the phagosome pathway are significantly altered by SUS +-MB in XO4 + microglia (seven genes up-regulated and three downregulated) than XO4 À microglia (five genes up-regulated) with two of these genes up-regulated in both (Figure 5a). For the cell cycle pathway, there were also more genes up-regulated in XO4 + microglia (18 genes up-regulated) than XO4 À microglia (11 genes up-regulated with 9 of these genes up-regulated in both; Figure 5b).

| The magnitude of BBB opening after SUS treatment does not differ significantly between APP23 and WT mice
While it was not the major focus of this work, it was surprising to find that there was no effect of SUS +MB treatment on the trans-

| DISCUSSION
In this study, we sought to investigate the changes to the microglial transcriptomic profile induced by the application of BBB opening achieved with therapeutic ultrasound in conjunction with intravenously injected microbubbles in a mouse model of AD. This profile Several previous studies have investigated the effect of ultrasound application to the brain by applying omics techniques to cell populations. One study investigating ultrasound-mediated delivery of plasmids to the brain of WT mice performed single-cell RNA sequencing and found an upregulation of lysosomal genes in microglia 48 h after ultrasound treatment. 31 In support of this, in a SWATH quantitative proteomics screen following a series of 6 weekly sessions of SUS +MB treatments in aged C57Bl/6 (WT) mice, 32 we identified an increase in two microglial proteins (LRBA and CAGP) that are involved in phagocytosis. 26 In our analysis performed a priori to our bioinformatic data mining, we evaluated whether Aβ load (XO4 + /XO4 À ) and treatment (SUS +MB /sham) are independent effects or interacting. Assessing the response of key pathways (phagocytosis and cell cycle), we conclude that the effects are independent and therefore can be analyzed in isolation. Our initial bioinformatics screen aimed to identify differences between microglia from WT and APP23 mice subjected to SUS +MB (with or without Aβ internalization) has revealed several interesting aspects related to the cellular response to the treatment. Thus, the WT microglia revealed the presence of a similar effect on the transcriptome in both sham-and SUS +MB -treated experimental groups, as revealed by both the PCA and heatmap analysis. This could be attributed to the fast resolution of microglial response to acute stimulation. 32 The WT transcriptome was found to cluster in the proximity of the transcriptome specific to XO4 À APP23 sham-treated microglia, reflecting a particular nonphagocytic cellular state, that is, most likely a nondisease associated microglial phenotype. We found microglia WT mice. 29 In addition, the response of microbubbles to ultrasound may differ between APP23 and WT mice because of differences in their cerebrovasculature, 33 or the fact that APP23 mice weigh less than their WT littermates. We performed recordings of acoustic emissions and found that APP23 mice had higher harmonics emissions than WT mice, but that ultraharmonic and broadband emissions were similar. Broadband emissions are associated with the largest magnitude and most violent cavitation activities, and these were mostly similar between WT and APP23 mice. The cause and significance of this difference in cavitation activity between WT and APP23 mice are unclear; however, it is conceivable that the increased cavitation recorded in APP23 mice might lead to an increased magnitude of transcriptomic changes at 48 h, which warrants further systematic studies, for instance, by using a cavitation controller. 27 If applied at an early stage of AD, boosting the Aβ phagocytic activity of microglia may present a promising therapeutic strategy by increasing the clearance of protein deposits. 34 A previous attempt to investigate the microglial response following ultrasound treatment focused on investigating transcripts related to the downstream effects of the NFκB pathway and damage-associated molecules (DAMs) in bulk lysates from WT rodent brains, with most transcript levels returning to baseline after 24 h. 30 A subsequent study, however, reported no significant changes in the expression of any of the NFκBrelated genes when using a lower, more clinically relevant dose of microbubbles. 35 These opposing effects could be attributed to the specific ultrasound parameters that elicit a cavitation-modulated inflammatory response through the microbubbles present in the blood circulation. 27 In addition, the transcriptomic response to ultrasoundinduced BBB opening was found to be dependent on the type of anesthesia used during the procedure. 31 Of note, we used ultrasound settings that we have previously demonstrated to increase microglial phagocytosis, 13 with no damage to neurons, 36  have been observed to remove synapses in AD through a mechanism involving members of the complement system. 37,38 In addition, it has been proposed that the metabolism of microglia is impaired in AD, an effect that can be ameliorated by enhancing the cellular energetic and biosynthetic metabolism. 39 Increased microglial numbers in the proximity of plaques are associated with more compact plaques and reduced axonal dystrophy, 40 and we have previously reported increased microglial numbers around plaques following SUS +MB treatment. 17 Higher numbers of microglia around plaques may result from an increased proliferation or metabolic activity, as hinted at in the present study. Reactivation of the cell-cycle machinery in microglia following ultrasound treatment is of particular interest, as it has been recently reported that repopulating microglial cells following ablation are neuroprotective in AD. 41

| Animals
In this study, we have used APP23 mice (harboring the AD Swedish

| Acute isolation of microglia and FACS
Two hours prior to brain harvest, mice were injected intraperitoneally with methoxy-X04 (2 mg/ml  (Figure 1d).