Transcriptomic response of brain tissue to focused ultrasound‐mediated blood–brain barrier disruption depends strongly on anesthesia

Abstract Focused ultrasound (FUS) mediated blood–brain barrier disruption (BBBD) targets the delivery of systemically‐administered therapeutics to the central nervous system. Preclinical investigations of BBBD have been performed on different anesthetic backgrounds; however, the influence of the choice of anesthetic on the molecular response to BBBD is unknown, despite its potential to critically affect interpretation of experimental therapeutic outcomes. Here, using bulk RNA sequencing, we comprehensively examined the transcriptomic response of both normal brain tissue and brain tissue exposed to FUS‐induced BBBD in mice anesthetized with either isoflurane with medical air (Iso) or ketamine/dexmedetomidine (KD). In normal murine brain tissue, Iso alone elicited minimal differential gene expression (DGE) and repressed pathways associated with neuronal signaling. KD alone, however, led to massive DGE and enrichment of pathways associated with protein synthesis. In brain tissue exposed to BBBD (1 MHz, 0.5 Hz pulse repetition frequency, 0.4 MPa peak‐negative pressure), we systematically evaluated the relative effects of anesthesia, microbubbles, and FUS on the transcriptome. Of particular interest, we observed that gene sets associated with sterile inflammatory responses and cell–cell junctional activity were induced by BBBD, regardless of the choice of anesthesia. Meanwhile, gene sets associated with metabolism, platelet activity, tissue repair, and signaling pathways, were differentially affected by BBBD, with a strong dependence on the anesthetic. We conclude that the underlying transcriptomic response to FUS‐mediated BBBD may be powerfully influenced by anesthesia. These findings raise considerations for the translation of FUS‐BBBD delivery approaches that impact, in particular, metabolism, tissue repair, and intracellular signaling.

influenced by anesthesia. These findings raise considerations for the translation of FUS-BBBD delivery approaches that impact, in particular, metabolism, tissue repair, and intracellular signaling.

K E Y W O R D S
anesthesia, blood-brain barrier, focused ultrasound, RNA sequencing

| INTRODUCTION
The blood-brain barrier (BBB) is essential to maintaining homeostasis in the central nervous system (CNS). The BBB describes a specialized vasculature, consisting of nonfenestrated endothelium, pericytes, astrocytic processes, microglia, and basement membrane working in concert to precisely permit nutrient transport while protecting against toxins and pathogens. However, the BBB also presents a significant neuropharmacological obstacle, preventing 98% of small-molecule therapeutics and nearly 100% of large-molecule therapeutics from accessing the CNS. 1 Significant efforts have focused on strategies to bypass or disrupt the BBB. Methods to bypass the BBB, including intracranial injection and intracerebroventricular infusion, require surgical intervention and thus carry significant risk. Chemical methods to disrupt the BBB, such as mannitol, cause global BBB disruption (BBBD), and lead to considerable neurotoxicity.
Focused ultrasound (FUS) following IV infusion of microbubbles (MB) is a promising approach for BBBD. [2][3][4] In this technique, ultrasound waves produced extracorporeally pass through the skull and cause MB circulating in a targeted region of the brain to oscillate. These oscillations disrupt BBB tight junctions and enhance transport of molecules into the brain parenchyma. FUS induced BBBD is an attractive alternative to surgical and chemical methods as it is targeted, noninvasive, and repeatable.
Many therapies normally restricted by the BBB have been successfully delivered with FUS + MB, including antibodies, 5-7 chemotherapeutics, [8][9][10] neural stem cells, 11,12 and genes. [13][14][15] BBBD with FUS is reversible and may be applied in a manner that yields little to no histological damage after repeated treatment. 3,16,17 However, recent molecular profiling studies have demonstrated that, under certain conditions, FUS induced BBBD can lead to increased expression of pro-inflammatory cytokines, homing receptors, and damage associated molecular patterns (DAMPs) as well as increased systemic macrophage accumulation. 18 These findings are consistent with sterile inflammation (SI), an innate immune response. The potential for FUS to induce local SI has sparked discussion of the cellular implications of FUS, both where additional inflammation may be desirable (such as cancer or Alzheimer's) or undesirable (such as multiple sclerosis or stroke). [19][20][21][22] Transcriptomic studies have shown that FUS induced SI is proportional to both microbubble dose and FUS acoustic pressure. 23,24 At pressures capable of reliably opening the BBB, as measured by MR contrast enhancement, we previously observed upregulation of proinflammatory transcripts (such as Ccl3, Ccl12, Ccl4, and GFAP) and pathways at 6 h post-FUS, trending toward resolution at 24 h post-FUS, consistent with previous studies. 18 ,24,25 Recent work has demonstrated the extent of post-FUS SI can be modulated by administration of dexamethasone. 26 Still, knowledge of the contributions of FUS experimental parameters to the SI response as well as noninflammatory effects on the brain parenchyma remain limited.
One such parameter is general anesthesia. Anesthetic protocols, ubiquitous in preclinical FUS BBBD studies, distinctly impact the circulation time of MB and the extent of FUS-induced vascular damage. 27,28 Common anesthetics vary widely in their effects on the CNS, differentially affecting cerebral vasculature, neuronal signaling, inflammation, and metabolism. [29][30][31] Indeed, a review of the FUS BBBD literature (Table S1) highlights considerable diversity in anesthetic protocols used in preclinical studies of experimental therapeutic efficacy, with isoflurane and ketamine being the most commonly chosen agents. We hypothesize that anesthetics differentially alter the underlying reactivity of the brain parenchyma when FUS is applied, which may produce anesthesia-dependent synergies and conflicts with respect to SI, drug metabolism, or neuronal damage. Herein, we test this hypothesis by detailing the cumulative transcriptome level and pathway level impacts of anesthesia, MB, and FUS on the brain parenchyma.

| Characterization of FUS-induced BBBD and passive cavitation analysis
Mice were anesthetized with either isoflurane in medical air (Iso) or ketamine/dexmedetomidine (KD) and treated with Magnetic Resonance-guided Focused Ultrasound (MRgFUS) targeted to the right or left striatum (n = 4 per group). Contrast-enhanced MRIs ( Figure 1(a)), collected before and after treatment, revealed enhanced signal in mice anesthetized with Iso compared to KD (Figure 1(b)). To evaluate MB activity, we analyzed acoustic emissions data obtained from a listening hydrophone embedded in the therapeutic transducer. No significant differences in harmonic emissions (i.e., 2nd, 3rd, and 4th harmonics) or broadband emissions (<10 MHz) were found between Iso and KD (Figure 1(c)).

| Transcriptomic variation is driven primarily by KD and secondarily by FUS BBBD
Bulk RNA sequencing was performed on mRNA extracted 6 h post-FUS from the treated region of each brain treated with MRgFUS shown in Figure 1. Brains extracted from naïve mice, mice treated with each anesthetic alone, and mice treated with each anesthetic and MB were also sequenced 6 h after treatment (n = 3 per group). After read alignment and QC, principal components analysis (PCA) was performed on transformed transcript counts from each sample to assess global differences between treatment conditions (Figure 2(a)). Interestingly, the first principal component segregated samples by whether they received KD, with Iso-treated mice clustering more closely to the naïve controls. FUS-treated mice formed a distinct cluster only in the KD treated mice. Similar results were obtained when hierarchical clustering was performed on inter-sample Euclidian distances computed between samples based on their transcript counts (Figure 2(b)). With the exception of one sample, the first branch point of the dendrogram separated samples by KD status, while the second and third branch points distinguished samples by FUS treatment.

| Overview of differential gene expression and gene set enrichment analyses
To evaluate relative transcriptomic differences between conditions, differential gene expression contrasts were computed for all 21 unique combinations of the 7 conditions evaluated (Figure 2(c)). KD alone produced the most profound effect on the transcriptome, with over 3000 genes significantly differentially regulated (p-adjusted <0.05) compared to naïve brain. Regardless of the anesthetic background, FUS and MB produced moderate (on the order of hundreds of differentially expressed genes) and negligible (<9 differentially expressed genes) effects on gene expression, respectively. Iso alone had a marginal effect on the transcriptome, only significantly changing the expression of 26 genes. Next, we performed gene set enrichment analysis (GSEA) to identify biological processes consistent with genes differentially expressed within each contrast (Figure 2(d)). GSEA was performed using the Gene Ontology (GO) Biological Pathways database, wherein each "GO" term represents a collection of genes associated with a particular biological phenomenon. Surprisingly, Iso alone affected more biological pathways than KD, despite KD affecting considerably more genes. The addition of MB changed relatively few biological pathways. FUS had the strongest effect on biological pathways on both anesthetic backgrounds, inducing more pathways than it repressed.

| Anesthetics differentially affect the transcriptome of normal brain tissue
The relative transcriptional impact of Iso and KD on the mouse striatum was marked, with Iso significantly changing expression of 26 genes compared to the 3291 significantly changed by KD (Figure 3 2.5 | Anesthetics differentially affect the transcriptome of brain tissue exposed to FUS BBBD

| Tissue damage elicited by FUS BBBD is minimal and not affected by anesthetic
Given the anesthesia-dependence of BBBD and FUS-induced gene expression, we next tested whether anesthesia affected the extent of damage in the brain parenchyma after treatment with the same FUS pressure. We performed histological analysis of murine brains treated with combinations of Iso, KD, and FUS (Figures 6(a-d)).   (Table S1), a factor that could affect the interpretation of how experimental therapeutic outcomes will translate to human applications, wherein such anesthetics are not utilized. Our study systematically addressed how choice of general anesthetic shapes acute transcriptomic responses to FUS with respect to SI, endothelial activity, metabolism, platelet activity, repair, molecular signaling, and BBB-associated genes (see summary in Table 1). Ultimately, we conclude that the underlying transcriptomic response to FUS-mediated BBBD may be strongly influenced by the choice of anesthetic. Such responses may synergize and/or conflict with responses generated by the therapeutic approach itself. Thus, our results provide a framework for rational anesthesia selection for preclinical BBBD studies and will likely find utility when comparing clinical outcomes to preclinical results for FUS mediated BBBD drug and gene delivery approaches.
PCA and hierarchical clustering performed on variance-stabilizing transformed RNA-seq counts data revealed the relative contributions of Iso, KD, MB, and FUS to intersample variability with respect to CNS gene expression (Figure 2(a),(b)). The most striking of these was KD, inducing DGE (p-adjusted <0.05) of 3291 genes when compared to naïve controls (Figure 3(a)). Whether this profound change in gene expression is attributable to ketamine, dexmedetomidine, or both is unclear. Microarray studies of developing rat brain have shown a similar magnitude of acute differential gene expression from ketamine alone. 33 More specifically, investigators reported 819 differentially expressed genes with fold change >1.4, p-adj <0.05 compared to the 1182 meeting these criteria in our study at an identical timepoint.
Though ketamine's mechanism of action is still unclear, recent studies into its rapid anti-depressant action suggest ketamine indirectly suppresses eukaryotic elongation factor 2 kinase (eEf2K), leading to increased protein translation. 34 This mechanism is in agreement with our pathway level findings (Figure 3(c)). Though fewer transcriptomic level studies exist for dexmedetomidine, it is known to acutely augment transcriptional programs associated with inflammation and F I G U R E 5 Anesthetics differentially affect transcripts associated with BBB structure and function. (a-d) Heatmaps of significance of upregulation (red) or downregulation (blue) for selected genes (rows) across multiple contrasts (columns), separated by anesthetic for transcripts associated with BBB structure and function. Selected categories include (a) leukocyte adhesion, (b) BBB tight junctions, (c) transporters, and (d) transcytosis/ miscellaneous. Contrast identities are shown by the color at the bottom of the column, corresponding to the key. Full opacity corresponds to an adjusted-p-value of 0, while full transparency corresponds to an adjusted p-value ≥0. 10. Each color in the key corresponds to a specific pairwise comparison of Anesthesia (An), An + MB, and An + MB + FUS for either Iso or KD, specifying the numerator (above the black line), and denominator (below the black line). For example, pink corresponds to the ratio of gene expression for mice treated with An + MB + FUS to those treated with just An + MB circadian rhythm. 35,36 In stark contrast to KD, we found Iso had a negligible impact on gene expression, only significantly altering the expression of 26 genes. This finding is in close agreement with existing acute transcriptomic studies of inhalable anesthetics in rats, which report between 0 and 20 differentially expressed genes. 37,38 Interestingly, despite weak changes in expression magnitude, Iso changed regulation of significantly more pathways than KD (Figure 2(d)). We thus hypothesize that, while Iso influences more targeted transcriptional programs, the combination of ketamine and dexmedetomidine elicits wide-ranging, complex transcription thereby preventing GSEA from detecting discrete pathway enrichment.
We observed increases in inflammatory signatures elicited by both anesthetics (Figure 3(d)). Of the few genes upregulated by Iso alone, a surprising number were immune-associated. Some examples include upregulation of T-cell associated markers Ly6a and Ctla2a, upregulation of adhesion markers Pecam1 and CD93, and downregulation of Nfkbia, the protein product of which inhibits NF-κB.
Indeed, activation of NF-κB has been proposed as a mechanism by which volatile anesthetics elicit neuroinflammation 39,40 . Several rodent studies have demonstrated volatile anesthetics can also acutely induce expression of IL-6, IL-1β, and activated caspase-3. [41][42][43][44] Under conditions of CNS stress, including ischemia or LPS exposure, volatile anesthetics attenuate inflammation, suggesting that these drugs may contribute to maintaining homeostasis in the brain, rather than being strictly pro-or anti-inflammatory. [45][46][47][48] KD also induced signatures associated with inflammation, though to a lesser extent Transcriptional impact -""" "" "" Endothelial activity --""" """ Platelet activity --""" " Signaling pathways # -""" -and with a less clear mechanism than Iso. At the chemokine level, for example, KD significantly upregulated Ccl17, Ccl2, Ccl3, and Ccl6 with minor but significant downregulation of Cxcl12 and Cx3cl1. These mixed effects may be caused by contrasting neuroinflammatory effects produced by ketamine and dexmedetomidine. Ketamine has been shown to be acutely inflammatory in naïve mice, increasing levels of IL-6, IL-1β, and TNF-α, 49 while Dexmedetomidine tends to protect against neuroinflammation. 50  whose tight junction protein product is essential to BBB integrity, in both FUS groups. This may indicate initiation of transcriptional programs to repair the disrupted barrier ( Figure 5(b)). In contrast, a microarray study of brain microvessels did not detect significant differences in claudin-5 post-FUS. 25 This discrepancy could be due to differences in species (i.e., mouse vs. rat), the source of the analyzed tissue in the brain, anesthesia protocol, and several FUS and microbubble parameters. Downregulation of multiple metabolic pathways in Iso-FUS contrasts further suggests Iso may prime the BBB for more significant alteration than KD.
Despite differential responses at the transcriptional level and in MRI signal enhancement (Figure 1), FUS applied under both anesthetics led to little to no generation of petechiae by H&E ( Figure 6).
With respect to coagulation signatures by RNA-seq, only Iso-FUS led to increased platelet activity despite no significant difference in RBC extravasation compared to KD-FUS (Figure 4(c)). While Iso has minimal effect on platelet activity, 58-60 both ketamine and dexmedetomidine reduce coagulability. [61][62][63][64] Thus, KD may minimize the inflammatory response resulting from blood products in the brain parenchyma compared to Iso upon FUS application.
Transient SI can provide beneficial effects in certain disease contexts with respect to clearance and regeneration. 65 Indeed, this may be the primary mechanism by which FUS promotes Aβ plaque clearance in Alzheimer's disease. 66 Similarly, neurogenesis observed after FUS may be attributable to tissue repair mechanisms preceded SI. 67,68 We observed activation of repair mechanisms by FUS, though to dif-  voltage-pressure calibration is provided in Figure S2. Immediately following the FUS treatment, mice received an intravenous injection of gadolinium-based contrast agent (0.05 ml of 105.8 mg/ml preparation; Multihance; Bracco Diagnostics), and contrast-enhanced images were acquired to assess BBBD using the same T1-weighted pulse sequence mentioned above.

| Passive cavitation detection
Acoustic emissions were detected with a 2.5 mm wideband unfocused hydrophone mounted in the center of the transducer. Acoustic signal was captured using a scope card (ATS460, Alazar, Pointe-Claire, Canada) and processed using an in-house MATLAB (MathWorks) algorithm. Acoustic emissions at the fundamental frequency, harmonics (2f, 3f, 4f), sub harmonic (0.5f), and ultra-harmonics (1.5f, 2.5f, 3.5f) were assessed by first taking the root mean square of the peak spectral amplitude (Vrms) in each frequency band after applying a 200 Hz bandwidth filter, and then summing the product of Vrms and individual sonication duration over the entire treatment period.
Broadband emissions were assessed by summing the product of Vrms and individual sonication duration for all remaining emissions over the entire treatment period.

| Bulk RNA sequencing and analysis
Six hours after treatment, mice were euthanized via an overdose of pentobarbital sodium and phenytoin sodium. Immediately following euthanasia, the mouse brains were harvested and the anterior right quadrants (~100 mg) were excised (with the exception of 1 mouse, which had FUS treatment on the left). For FUS-treated mice, contrast MR images were referenced to confirm extraction of the full volume of sonicated brain. A representative 3D image of the harvested tissue is provided in Figure S1.  Figure S2.

| Histological processing and analysis
Sixty minutes after Evans Blue injection, mice were euthanized via an overdose of pentobarbital sodium and phenytoin sodium. A macroscopic image was taken immediately after whole brain harvest. Brains were then placed in 10% NBF, embedded in paraffin, and sectioned 400 μm apart. H&E stained sections were imaged with 4× and 20× objectives on an Axioskop light microscope (Zeiss, Germany) equipped with a PROGRES GRYPAX microscope camera (Jenoptik, Germany).
Ten 20× images from the region of the right striatum with maximal Evans Blue extravasation were taken per section and 2-6 sections were imaged per brain. A researcher blinded to treatment condition assigned a semi-quantative score of 0 (none-complete absence of RBC extravasation/vacuolation), 1 (mild-sparse small sites of RBC extravasation/vacuolation), 2 (moderate-singular large OR multiple small sites of RBC extravasation/vacuolation), or 3 (severe-multiple large sites of RBC extravasation/vacuolation) to each 20× image for RBC extravasation and vacuolation using a custom MATLAB (MathWorks) script.

| Statistical methods
For contrast enhancement and acoustic emissions analyses, data are presented as mean ± SEM. Statistical significance was assessed by