Glymphatic fluid transport is suppressed by the AQP4 inhibitor AER-271

Abstract

The glymphatic system transports cerebrospinal fluid (CSF) into the brain via arterial perivascular spaces and removes interstitial fluid from the brain along perivenous spaces and white matter tracts. This directional fluid flow supports the clearance of metabolic wastes produced by the brain. Glymphatic fluid transport is facilitated by aquaporin-4 (AQP4) water channels, which are enriched in the astrocytic vascular endfeet comprising the outer boundary of the perivascular space. Yet, prior studies of AQP4 function have relied on genetic models, or correlated altered AQP4 expression with glymphatic flow in disease states. Herein, we sought to pharmacologically manipulate AQP4 function with the inhibitor AER-271 to assess the contribution of AQP4 to glymphatic fluid transport in mouse brain. Administration of AER-271 inhibited glymphatic influx as measured by CSF tracer infused into the cisterna magna and inhibited increases in the interstitial fluid volume as measured by diffusion-weighted MRI. Furthermore, AER-271 inhibited glymphatic efflux as assessed by an in vivo clearance assay. Importantly, AER-271 did not affect AQP4 localization to the astrocytic endfeet, nor have any effect in AQP4 deficient mice. Since acute pharmacological inhibition of AQP4 directly decreased glymphatic flow in wild-type but not in AQP4 deficient mice, we foresee AER-271 as a new tool for manipulation of the glymphatic system in rodent brain.

Graphical Abstract

Summary Figure: AER-271 reduces glymphatic fluid flow dependent on AQP4 Arterial pulsation pumps cerebrospinal fluid (CSF) within the perivascular space (PVS), and AQP4 channels on astrocytic endfeet facilitate the influx of CSF into the brain parenchyma. Fluid and metabolic waste from the parenchyma are cleared by directional flow from arterial-side PVSs to venous PVSs, and along white matter tracts, finding ultimate egress via lymphatic vessels. Inhibiting AQP4 channels with AER-271 treatment decreases AQP4 clearance of interstitial water, thereby reducing turnover of fluid within the brain and decreasing clearance of interstitial solutes.

Introduction

The brain is endowed with a unique fluid transport network, the glymphatic system, which facilitates cerebrospinal fluid (CSF) influx into the brain and interstitial fluid (ISF) clearance from the brain ¹. CSF flows within the perivascular spaces along pial arteries and penetrating arterioles, thereby infiltrating the brain parenchyma to mix with ISF, and then exits the brain along perivenous spaces and white matter tracts, ultimately clearing to meningeal and cervical lymphatic vessels ^1–4. This unidirectional net flow transport system supports the elimination of endogenously produced metabolic wastes such as lactate ⁵ and toxic proteins such as amyloid beta¹, tau, and synuclein⁶ from the brain. As such, the glymphatic system is implicated in neurodegenerative diseases such as Alzheimer’s disease.

Importantly, the outer boundary of the perivascular space is composed of astrocytic endfeet densely expressing the water channel aquaporin-4 (AQP4) ⁷, and glymphatic flow is markedly reduced in AQP4 deficient mice ^1,8,9. AQP4 is normally expressed in a polarized pattern, with high levels at astrocytic vascular endfeet and lower expression in astrocytic soma and peri-synaptic processes facing the neuropil^7,10. Appropriate endfoot polarization of AQP4 is vital for glymphatic flow. Alpha-syntrophin deficient mice have normal levels of total AQP4 protein, but lack polarized expression at the astrocytic vascular endfeet ^11,12. These mice exhibit attenuated glymphatic flow, comparable to that in mice with full genetic deficiency of AQP4, which underscores the functional importance of the polarized AQP4 expression ^8,13. Further, loss of AQP4 polarization during aging^14,15, in neurodegenerative disease ^6,14, or inflammation ¹⁶ reduces glymphatic flow. Moreover, the extent of AQP4 polarization declines with progression of cognitive impairment in human Alzheimer’s disease ¹⁴. AQP4 polarization is not only related to pathology, it is also regulated by circadian rhythms; AQP4 polarization is greater during the rest phase in mice coincident with increased glymphatic flow¹⁷. Thus, AQP4 polarization to the astrocytic endfeet and AQP4 water permeability appear to present physiological targets for modulating glymphatic flow.

There have been few agents for pharmacological modulation of AQP4 water channels. A recent preclinical investigation identified IMD-0354, hereby referred to as AER-270, as an AQP4 inhibitor using a cell-based drug screening assay¹⁸. Treatment with AER-270 reduced brain edema in mouse models of ischemic stroke and water intoxication, both processes mediated by water flux through AQP4^12,18,19. AER-271 is a water soluble prodrug developed by addition of a phosphate group to the less soluble AER-270, improving administration in live animals¹⁸. Given that glymphatic flow is also dependent on AQP4^1,8, and that glymphatic influx of CSF is the primary source of edema during the acute phase of ischemic stroke²⁰, AER-271 presents itself as a potential tool for regulating glymphatic flow. Here we test the hypothesis that inhibition of AQP4 with AER-271 acutely inhibits glymphatic flow in mouse brain.

Methods

Animals

133 C57BL/6 mice (Charles River Laboratories) aged 2–5 months were used for the main experiments, along with 20 AQP4 deficient mice (APQ4KO), 5 Wild-Type littermates maintained on a C57BL/6 background at the vivaria of the University of Rochester Medical Center and Copenhagen University Faculty for Health Sciences ²¹. Animals were kept on a 12:12 light/dark cycle, and experiments were conducted in the middle of the light phase between 10am-4pm with experimental groups interleaved to minimize variation in data due to circadian rhythms. All procedures were approved by the University of Rochester Committee on Animal Resources and University of Copenhagen Animal Experiment Inspectorate.

Drugs

AER-271 (Sigma Aldrich, SML2737) was prepared at a concentration of 0.25 mg/mL in 0.9% normal saline solution, and administered to mice intraperitoneally at a dose of 5 mg/kg body weight (b.w.). For a subset of water intoxication experiments (Fig. S1), AER-271 was included in the water bolus at a concentration of 0.025 mg/mL, still administered intraperitoneally at a dose of 5 mg/kg b.w.. TGN-020 (Sigma Aldrich, SML0136) was prepared in same water bolus used for water intoxication at a concentration of 0.5 mg/mL and administered intraperitoneally at dose of 100 mg/kg b.w.. For cisterna magna infusions and solute clearance assay, mice were anesthetized with an intraperitoneal mixture of racemic ketamine (100 mg/kg) and xylazine (20 mg/kg). Depth of anesthesia was confirmed by toe pinch every 10 minutes, with addition of a one tenth dose of ketamine/xylazine if responsive. For MRI experiments, animals were anesthetized with ketamine (75 mg/kg) and dexmedetomidine (1 mg/kg) administered intraperitoneally.

Acute water intoxication

Mice in both experimental and control groups were injected intraperitoneally with 0.2 mL/g b.w. distilled water warmed to 37 °C, and then monitored for breathing impairment, motor impairment, and mortality for 120 minutes, as described by Manley and colleagues ¹⁹. Breathing impairment was scored as 1 (unimpaired), 2 (arrhythmic), 3 (arrhythmic and labored), and 4 (labored, slowed breathing). Motor impairment was scored as 1 (reduced spontaneous movement), 2 (uneven gait), 3 (poor gait and unable to navigate), 4 (stationary), and 5 (stationary with postural deficits and impaired righting reflex). Mice were evaluated by these scales every 10 minutes by two blinded investigators, and time of death was recorded for Kaplan-Meier survival analysis. Mean impairment in each group was plotted over time, and the sum of these values was taken as the area under the curve (AUC, Fig. S1).

Cisterna magna infusion

Cisterna magna (CM) tracer infusions were conducted as previously described ^1,17. In brief, fluorescent CSF tracer (bovine serum albumin AlexaFluor-647 conjugate, BSA-647; 66 kDa; Invitrogen, Life Technologies, Eugene, OR) was formulated in artificial CSF at a concentration of 0.5% weight by volume. Anesthetized mice were fixed in a stereotaxic frame, the CM surgically exposed, and a 30-gauge needle connected to PE10 tubing filled with the tracer was inserted into the CM. 10 μL of CSF tracer was infused at a rate of 2 μl/min for 5 min with a syringe pump (100 μL Hamilton Gastight 1700 series, Harvard Apparatus).

To visualize tracer movement from the cisternal compartment into the brain parenchyma, the animals were sacrificed by decapitation 30 min after the start of intracisternal infusion (T=0 at pump start, T=30 at sacrifice). The brains were extracted and fixed overnight by immersion in 4% paraformaldehyde in phosphate-buffered saline (PBS). Coronal vibratome slices (100 μm) were cut and mounted. Tracer influx into the brain was imaged ex vivo by macroscopic whole-brain and whole-slice conventional fluorescence microscopy (MVX10, Olympus; light: PRIOR Lumen 1600-LED; camera: Flash 4.0 digital, Hamamatsu).

For each animal, tracer influx was quantified by a blinded investigator using ImageJ software (v1.53t). To compare superficial tracer distribution, regions of interest (ROIs) were drawn on dorsal and ventral brain surfaces on un-sectioned whole brain images, and an ROI isolating the middle cerebral artery over the sensorimotor cortices. Next, images of each coronal slice were manually outlined, and additional ROIs were drawn over specific brain structures on images of sections extending from +1.2 mm to −1.8 mm anterior/posterior relative to bregma, for evaluating tracer penetration and distribution within the brain. For each animal, the area and mean fluorescence intensity within each ROI were measured, and the mean fluorescence intensity was calculated across six slices evenly spaced between +1.2mm and −1.8 mm anterior/posterior relative to bregma, resulting in a single biological replicate value. Total fluorescence was calculated by multiplying mean pixel intensity by number of pixels measured for each slice, then summing this value across all six slices.

Diffusion-weighted magnetic resonance imaging

Diffusion-weighted imaging (DWI) was performed at 9.4 Tesla magnetic resonance (MR) system (BioSpec 94/30USR, Bruker BioSpin, Ettlingen, Germany) in a group of 12 AQP4KO and 12 wildtype (WT) mice, including 5 WT littermates and 7 C57BL/6 mice. To assess genotype effects of AER-271 on the brain water mobility, echo-planar-imaging (EPI) sequence was performed using a volumetric Tx/Rx resonator (in. ø40 mm) and 1500 mT/m gradient coil (BFG6S, Bruker). All animals underwent DWI with 14 b-values measured in 6 directions of diffusion encoding gradients (EPI: TR/TE = 3500/30 ms; number of acquisitions = 1; voxel dimension = 0.15 × 0.15 × 0.5 mm; b-values = 40, 42, 44, 47, 90, 131, 159, 187, 227, 327, 627, 1027, 1527, 2027 s/mm²).

The imaging protocol consisted of 3 respiratory-gated EPI sequences acquired consecutively, and separated by two 15 minutes pauses. The end of the baseline DWI was followed by AER-271 (0.005 mg/20 μL/g b.w.) or 0.9% NaCl (20 μL/g b.w.) injection. Because each respiratory-gated EPI lasted between 22 and 27 minutes, the pause was applied so that the middle of the second DWI acquisition would depict the brain water changes at the peak of AER-271 plasma concentration after the injection (i.e., ~30 min). Subsequently, the third DWI was acquired after the same delay from the second DWI (i.e., middle of acquisition ~70–75 min after the injection). During the whole protocol, the animals remained in a head-first prone position inside the MR bore, with body temperature monitored and maintained at 37±1°C using a thermostatically-controlled waterbed and monitored, along with respiratory rate using an MR-compatible remote monitoring system (SA Instruments, NY, USA). Administration of AER-271 or 0.9% NaCl vehicle was made via a chronically implanted intraperitoneal catheter connected to a 1 ml plastic syringe.

DWI processing and analysis

For every animal, all acquired DWI volumes were subjected to rigid-body motion-correction (6 degrees of freedom) in AFNI²² to minimize the influence of random and subtle frame-to-frame image displacements. The motion-correction was performed four times or until no further improvement in reference to the first image acquired using the first b-value (b0), with visual inspection to confirm correctness. To assess the water molecules mobility within the brain parenchyma in the motion-corrected DWI images from each diffusion-encoding direction, we performed evaluation as presented previously²³. A standard mono-exponential model considering the apparent diffusion coefficient (ADC) was applied to estimate both ADC and DW image without influence of diffusion gradients (S₀), by means of voxel-wise fitting of the acquired DW intensities over 14 b-values. All DWI analysis was performed using in-house scripts and functions in MATLAB (r2021a, The Mathworks, Inc., Natick, MA.).

Regions of interest definition

To evaluate the brain-wide MR diffusion changes upon AER-271 or vehicle injection in animals of different genotypes, mean and standard deviation (SD) of ADC signal intensities were calculated within 19 ROIs for every diffusion direction separately in every animal. To enhance accuracy of the results, the ROIs were set automatically in a template volume, based on the Reference Space Model (RPM) from the Allen Brain Atlas (atlas.brain-map.org.), to which all acquired DWI volumes were co-registered (rigid-body) in ITK-SNAP²². The template volume was calculated as the mean DW image from all DWI volumes from a single WT animal, to which the RPM was realigned and resliced in ITK-SNAP to match spatial resolution. Due to limitations in the resolution of the acquired images, and to avoid partial volume effects and ghosting, we considered only 19 representative large parenchymal and cortical ROIs. Because of slight differences in the individual brain volumes, the coverage of cortical and ventral ROIs was further manually modified to match the corresponding brain regions in all animals. The defined ROIs comprised from 50 to 480 voxels (i.e. from ~0.6 to 5.4 mm³) each, defined in the following regions: olfactory area - OLF; cingulate gyrus - CA; visual cortex - VIS; somatosensory cortex - SS; motor cortex - M; auditory cortex - AUD; hippocampus - HIP; temporal cortex - TEMP; gustatory cortex - GUST; prelimibic - PRE; Agranular (insular) cortex - INS; amygdala - AMY; caudate nucleus - CP; pallidum - PAL; thalamus - TH; hypothalamus – HY; midbrain - MB; pons - PO; medulla oblongata - MED.

Brain-wide comparison

To compare the MR diffusion values from the baseline DWI, mean ADC values from all diffusion directions were calculated ROI-wise for each animal, and compared group-wise (4 groups, 5–6 animals/group). Subsequently, the real and relative to baseline ROI-wise differences between the aggregated mean diffusion values from the second vs. baseline, and third vs. baseline DWI acquisitions were computed to evaluate the brain water diffusivity changes after injecting AER-271 vs. vehicle in AQP4KO and WT animals. The baseline and the differences in MR diffusion values were both compared using a two-way ANOVA with Bonferroni’s post-hoc. We considered an unpaired comparison using nonparametric approach when comparing the baseline mean ADC between AQP4WT and C57BL/6 mice, due to the number of animals in each group. The results were considered significant for p<0.05.

Correlational analysis

To associate the DWI findings with glymphatic influx evaluated using BSA-647 in WT animals, we tested ROI-wise correlations between the mean pixel intensity of BSA-647 from the fluorescent CSF influx experiment and the mean and SD of aggregated group-wise ADC values as well as the differences from baseline in the mean ADC at ~30 min and ~75 min after injection. The correlations were calculated only for the subset of 9 parenchymal ROI that were assessed with both methods. Aiming to match the segmented areas from the ex vivo evaluation, we included several previously defined RPM ROIs according to their position relative to the BSA-647 segmentation. The final nine BSA-647 ROIs were as follows (original DWI template ROI): dorsal anterior (cingulate, motor, somatosensory cortex), dorsal posterior (somatosensory cortex), lateral anterior (agranular, gustatory, somatosensory cortex), lateral posterior (somatosensory, auditory, temporal cortex), ventral (amygdalar), caudate (caudate), hippocampus (hippocampus), thalamus (thalamus), hypothalamus (midbrain, hypothalamus). Correlations were considered significant for linear Pearson’s or Spearman’s correlation coefficient >0.5, p<0.05 significance and non-zero regression slope.

DB53 Clearance Assay

The DB53 (Direct Blue 53, also called Evans Blue) assay was conducted as previously described ²⁴. In brief, anesthetized mice were placed in a stereotaxic frame. The scalp was opened, and a burr hole was drilled at the following coordinates from bregma: Anterior/Posterior +0.6 mm; Medial/Lateral −2.0 mm (striatum). The dura was punctured with a 30G needle, then a guide cannula (26G, C315G SPC, 4.5 mm bellow pedestal) and a dummy cannula (33G, C315DC/SP, 0.1 mm projection) (PlasticsOne, Roanoke, VA) were placed at DV −3.3 mm (striatum) relative to bregma. The cannula was then secured in place with dental cement (Fisher Scientific, NC9991371) and the incision closed.

After 24 hr of recovery, animals were anesthetized and the dummy cannula was replaced by an internal cannula (33G, C315I/SP, 0.1 mm projection) connected by PE10 tubing to a 10 μL Hamilton syringe containing 4% wt/vol DB53 (Sigma Aldrich, E2129) in sterile aCSF. Mice were then placed under a fluorescent macroscope (MVX10, Olympus; light: PRIOR Lumen 1600-LED; camera: Flash 4.0 digital, Hamamatsu) in the lateral position. The left femoral vein was exposed via skin resection and baseline images collected. The start of pump infusion (1 μL total volume, 0.2 μL/min) was triggered simultaneously with the start of imaging of the vein (T=0), and images taken once every 15 min for 2 hr. Exposed tissue was kept hydrated with 0.9% saline. At the end of the experiment, animals were sacrificed by decapitation while still anesthetized, and their brains were harvested and fixed by immersion in 4% paraformaldehyde in PBS solution.

Extracted brains were sectioned coronally (100 μm), and 6 sections were taken per mouse evenly spaced between +1.2 mm and −1.8 mm relative to bregma. The sections were imaged on a fluorescence macroscope (Olympus). The DB53 signal in brain sections was analyzed in ImageJ software by drawing ROIs outlining each section, and then measuring the area fraction covered by DB53 fluoresence. Cannula placement in the striatum was confirmed by inspection of the brain sections, mice with incorrect cannula position were excluded. To quantify DB53 clearance from brain, we measured in vivo DB53 blood signals by taking the mean pixel intensity of a 0.011 mm² ROI within the femoral vein on each image.

Mathematical model

A mathematical compartment model was fitted to in vivo measurements of venous DB53 clearance as described in detail previously²⁴. The model assumes the DB53 in brain is well mixed and clears to blood through two routes $q_1$ and $q_2$ (in units of volume per time), where $m_i$ is DB53 mass in the brain, $v_i$ is injection site volume, and $c_i$ is concentration of DB53 in the brain:

$$\frac{{dm}_i}{dt}=v_i\frac{{dc}_i}{dt}=-\left(q_1+q_2\left(t\right)\right)c_i\left(t\right).$$

The two routes are distinguished in that the second route, $q_2\left(t\right)$, is zero until timepoint $t_0$, interpreted as the time at which tracer arrives at the second efflux route (see ²⁴). DB53 mass arrives in the blood pool at the same rate as it is cleared from the brain, as described below where $m_b$ is DB53 mass in the blood and $c_b$ is concentration of DB53 in the blood.

$$\frac{{dm}_b}{dt}=v_b\frac{{dc}_b}{dt}=\left(q_1+q_2\left(t\right)\right)c_i\left(t\right).$$

Combining these two equations algebraically with the requirement that DB53 mass is conserved gives the blood concentration $c_b$ as

$$c_b=\left\{\begin{array}{c}C\left(1-e^{{-k}_{1}t}\right)t<t_0\\ C\left(1-e^{{-k}_{1}t}{e}^{-(k_1+k_2)(t-t_0)}\right)\mathrm{else}.\end{array}\right.$$

Here $C$ is the total injected DB53 mass divided by blood volume, $k_1=q_1/v_i$, and $k_2=q_2/v_i$ is the non-zero rate of efflux through the second route. Note that when estimating $C$ from blood-concentration data, it indicates the long-run amount of tracer to be cleared from the brain.

Data for each mouse were normalized by subtracting the venous DB53 mean pixel intensity (MPI) at t=0 from all measurements and dividing by the mean DB53 MPI at t=120 within the saline group. Model parameters were then estimated for each animal to minimize the sum square distance between model prediction and normalized data. The fitted parameters were evaluated by ${\mathrm{R}}^2$, which was calculated as ${\mathrm{R}}^2=1-\mathrm{S}{\mathrm{S}}_{\mathrm{R}}/\mathrm{S}{\mathrm{S}}_{\mathrm{T}}$, where $\mathrm{S}{\mathrm{S}}_{\mathrm{T}}$ is the total variation in the data and $\mathrm{S}{\mathrm{S}}_{\mathrm{R}}$ is the variation in the residuals.

Physiological recordings

Mice were administered 2 doses of AER-271 or vehicle solutions at 75 and 15 minutes prior to recording just as in the CM infusion experiment (Fig. 2A). Ketamine/xylazine anesthesia was administered 15 minutes prior to physiological recording of ECG and respiration using a rodent physiology board (Harvard Apparatus, sampling frequency=1kHz) connected to a PC (Axoscope). Subsequently, mice were implanted with a femoral artery catheter filled with heparinized saline (50U/mL, 0.9%saline) connected to a blood pressure transducer (World Precision Instruments, BLPR2) and also connected to a PC (Axoscope). All recording lasted 30 minutes and afterwards heart and respiratory rates were calculated from the ECG and pressure pad signals using MATLAB. Mean of the heart rate, respiration rate, and blood pressure signals from each animal were calculated in 2-minute windows displaced by 1/3^rd of the window length. The calculated mean series were statistically compared between AER-271 and vehicle groups using 2-way repeated measurements mixed model ANOVA with Bonferroni’s post-hoc. The results were considered significant for p<0.05.

Figure 2: AER-271 reduces glymphatic influx.

(A) Experimental schema for AER-271 treatment and cisterna magna tracer infusions under ketamine/xylazine anesthesia. (B) Representative images of BSA-647 CSF tracer influx in whole brain of vehicle- and AER-271-treated mice. CSF tracer is visible surrounding the circle of Willis (bottom) and middle cerebral arteries (MCA, Top) (scale bar = 1 mm). (C) Representative coronal sections depicting BSA-647 CSF tracer influx. White arrows emphasize the reduced influx in ventral cortical areas. (scale bar = 1 mm) (D) Total fluorescence present in whole brain surfaces for ventral (left, p=.024), dorsal (center, p=.050), and MCA PVSs (right, p=.191). (E) Total fluorescence signal across 6 brain sections extending from 1.2 mm+ to −1.8 mm relative to bregma (p=.01) and mean pixel intensity (p=.033). (F) Anterior sections from each mouse brain were divided into subregions, and tracer signal intensity analyzed for each subregion, showing significantly less tracer in ventral cortex of AER-271 treated mice (p=.005). (G) Posterior sections divided into subregions and tracer intensity analyzed for each subregion, showing no significant differences within each subregion. (F-G) D CTX = dorsal cortex, L CTX = lateral cortex, V CTX = ventral cortex, CPu= caudate-putamen, HIP=hippocampus, TH=thalamus, HY=hypothalamus.

All statistical tests were unpaired two-tailed t-tests. *-p<.05, **-p<.01. All plots depict the mean and standard deviation of each group, where each point represents one animal (n=5 Vehicle, n=7 AER-271).

Evaluation of the physiological signals variability

To evaluate influence of AER-271 and vehicle administration on the acquired ECG, blood pressure and respiratory signals variability over the course of experiment, we have employed a nonlinear analysis method based on calculation of Higuchi fractal dimension (HFD)²⁵, using an in-house implementation in MATLAB. HFD measures dimensional complexity of the signals and attains values between 1 and 2, where 1 reflects a regular signal and 2 a white noise. All the acquired physiological signals were subjected to 60Hz notch (Butterworth filter, order=2, half power frequency between 59.5 and 60.5 Hz) with subsequent low-pass filtration to remove the influence amplitude current and high frequency fluctuations on final evaluation. Furthermore, the all signals underwent low-pass filtration at the level of 35 Hz for the blood pressure and heart rate signals, and 30Hz for respiratory signals. In order to reduce and standardize the length of the signals, all the signals were uniformly resampled with 50 Hz – outside the range of analyzed signals after filtration in frequency domain. Afterwards, 100 samples-long fragments were cut from the beginning and the end of each signal to remove the portions of the signals affected by filtration at the border conditions. Finally, HFD was calculated in 2-minute windows displaced by 1/3^rd of the window length for all the signals. Based on previous literature ^26–28 and empirical evaluation, the k_max=8 parameter was chosen for HFD calculation. All obtained HFD time series were statistically compared between AER-271 and vehicle groups using 2-way repeated measurements mixed model ANOVA with Bonferroni’s post-hoc, and the results were considered significant for p<0.05.

AQP4 Immunohistochemistry

Three 100 μm-thick brain sections originating between −1.2 mm to −1.8 mm posterior to bregma were selected from each mouse for AQP4 channel immunohistochemistry. Slices were permeabilized with 0.1% Triton-X-100 in PBS, blocked with 10% normal donkey serum (Jackson Immunoresearch) in PBS with 0.03% Triton-X-100 and incubated with primary antibody overnight at 4C, followed by three washes in PBS and incubation with the fluorophore-linked secondary antibodies for 2 hours. Stained slices were mounted with Fluormount G (Thermofisher Scientific). Primary rabbit anti-AQP4 antibody (AB3594, Millipore, 1:500 dilution) with secondary Cy3 donkey anti-rabbit antibody (711–165-152, Jackson ImmunoResearch 1:500 dilution) were used, and cell nuclei were identified by DAPI counterstaining (D1306, Invitrogen, 1:2000).

Images of dorsal cortex, lateral cortex, ventral cortex, were imaged for each slice (three images per slice) with a confocal microscope (FluoView, Objective: UPlanXApo 40x/numerical aperture 0.95, ∞/compatible cover glass thickness 0.17 mm/field number 26.5 mm, Olympus). For quantification of AQP4 polarization, 50 μm segments centered on blood vessels (as identified by vascular-shaped AQP4 localization) were analyzed using the plot profile tool in ImageJ. Selected vessels were perpendicular to the imaging plane so as to be visible longitudinally, and measured <6 μm wide and circa 30 μm long (range 20–55 μm).

For AQP4 polarity calculation, the baseline was defined as the average intensity of the line plot over the first 10 μm. MATLAB was used to locate the two highest peak intensities for each line segment, these peak values correspond to perivascular, astrocytic endfoot AQP4 on either side of the vessel. Baseline fluorescence was subtracted from the peak intensity in each segment, and individual polarization values were averaged for each animal to obtain one representative result (Fig. 4), and then aggregated from all animals as individual data points (Fig. S3).

Figure 4: AER-271 reduces glymphatic efflux.

(A) Schematic showing striatal cannula implantation, recovery, and in vivo DB53 measurement. (B) Representative coronal sections (+0.6 mm AP bregma) depicting DB53 infusion in striatum (scale bar = 1 mm). (C) DB53 infusion volume measured as total area covered by DB53. (Left) DB53 area covered in individual coronal sections. (Right) Mean area covered by DB53 across 6 coronal sections, where each data point is one animal. (D) Representative images of DB53 fluorescence in mouse femoral vein after DB53 striatal infusion (scale bar= 500 μm). (E) Fluorescent intensity of DB53 in femoral vein plotted over time (solid lines, mean ± SEM plotted for each group, dotted lines; mean temporal derivative of DB53 for each group). Results for the entire tracer circulation time (0–120 min;left), and inset focusing on 60–120 min (right). (F) Venous DB53 intensity at 0 minutes (left) and 120 minutes (right) after striatal infusion of DB53 tracer (120 min, p=.037; n=6 Vehicle, 6 AER-271). Legend: ns - not significant, * - p<.05, by means of unpaired two-tailed t-test.

A subset of WT and AQP4KO animals were immunohistochemically-stained to demonstrate lack of AQP4 channels in KO mice using GFAP (Primary antibody= MAB360 Millipore 1:500 dilution; Secondary antibody= 715–225-150 Jackson ImmunoResearch, 1:500 dilution) and AQP4 antibodies. (Primary antibody= AB3594 Millipore, 1:500 dilution; Secondary antibody= 711–165-152 Jackson ImmunoResearch, 1:500 dilution).

Statistical analysis

GraphPad Prism 9 was used to perform statistical tests. All p-values listed are derived from an unpaired, two tailed t-tests with α=0.05, unless otherwise specified.

Results

AER-271 is protective from hyponatremia induced brain swelling

To assess whether AER-271 inhibits in vivo AQP4 water permeability, AER-271 was first tested in a mouse model of acute severe hyponatremia, i.e., water intoxication ^21,29. Intraperitoneal injection of distilled water in mice results in plasma hyponatremia, which creates an osmotic gradient from brain to blood, with the brain having more osmolytes. Osmotic brain swelling develops as water exits the low osmolality blood compartment and moves into the higher osmolyte containing brain tissue. AQP4 water channels on astrocytic vascular endfeet facilitate this water movement (Fig. 1A) ^12,19,30. The tissue swelling elevates intracranial pressure, which in turn impairs motor function and breathing, ultimately causing death in cases of brainstem herniation ²⁹. AQP4 channels promote water intoxication-induced cerebral edema ^12,19,30, and previous work showed that the active drug AER-270 protected against edema formation and mortality in this model ¹⁸.

Figure 1: AER-271 protects from water intoxication.

(A) Schematic for hypo-osomotic induction of brain swelling via water intoxication (WI). WI induces an acute drop in plasma osmolality, resulting in increased water influx to the brain across astrocytic endfeet, facilitated by AQP4 water channels. (B) Experimental scheme. AER-271 is given to 5 months old wild-type (WT) mice at 5mg/kg administered intraperitoneally 20minutes prior to WI challenge. (C) Survival curves for mice treated with AER-271 or vehicle after acute WI (WT, males n=12 vehicle, 13 AER-271, p=.031 Mantel-Cox test). (D) Survival curves for mice treated with TGN-020 or vehicle control after acute WI, TGN-020 was included at T=0 because it is already its active form. (WT, males n=8 vehicle, 10 TGN-020, p=.034 Mantel-Cox test). (E) Survival curves for male and female mice after acute WI with no drug administration (p=.008 Mantel-Cox test, n= 13 female, 12 male)

AER-271 is a pro-drug bearing an additional phosphate group that is cleaved by endogenous phosphatases to the active drug AER-270. AER-270 peaks in blood and brain approximately 20 minutes after intraperitoneal injection of AER-271 ¹⁸. Given this conversion time, we pretreated mice aged five months with AER-271 at 20 minutes prior to water intoxication (Fig. 1B), finding protection from mortality after water intoxication (Fig. 1C, p=.031 Mantel-Cox test, n=12 vehicle, 13 AER-271). AER-271 included in the water bolus proved to be less protective, likely because the onset of brain edema preceded the conversion of prodrug to its active form (Fig. S1AB, p=.24 Mantel-Cox test, n=14 vehicle, 15 AER-271). In mice aged two months, there was no mortality, perhaps due to differences in body composition at younger age, but AER-271 treatment significantly improved breathing and motor function (Fig. S1CDE).

Next, we tested another AQP4 inhibitor, TGN-020, in the acute water intoxication paradigm described above. TGN-020 has inhibited AQP4 function in cell-based assays ^31,32 and inhibited glymphatic influx³³, but never been tested with water intoxication. Here, we found TGN-020 to protect from water intoxication mortality (Fig. 1D, p=.034 Mantel-Cox test, n=8 vehicle, 10 TGN-020). We tested both sexes with vehicle injection and WI, and found female mice were protected from the water intoxication without any drug administration (Fig. 1E, P=.008 Mantel-Cox test, n=13 female, 12 male), so we have only included male mice in the drug comparisons (Fig. 1CD). To our knowledge this paradigm has previously been used with only male mice, and it could be females have a higher capacity to buffer the osmotic challenge via estrogen signaling in the kidney³⁴

AER-271 reduces glymphatic influx

As expected, AER-271 was more effective when administered prior to the water intoxication. Therefore, we applied AER-271 pre-treatment in all subsequent studies of the glymphatic system (Fig. 2A). AER-271 or saline vehicle was administered to mice aged 8–12 weeks old that then received a cisterna magna infusion of tracer under ketamine/xylazine anesthesia. CSF was labeled with BSA-647 (66 kDa, 0.5% wt/vol in artificial CSF, 10μL infused 2μL/min), and the distribution of BSA-647 was analyzed ex vivo. We first collected macroscopic images of BSA-647 distribution on the brain surface (Fig. 2B), and then drew ROIs over the middle cerebral arteries. Quantitative analysis showed that there was a significant reduction in BSA-647 influx on the surface of the brain (Fig. 2D, ventral surface p=.024, dorsal surface p=.050, n=5 vehicle, 7 AER-271), and a trend towards decreased influx along the middle cerebral artery perivascular spaces (Fig. 2D, right, p=.19). We next prepared six coronal vibratome sections (100 μm-thick, 500 μm intervals, +1.2 mm to −1.8 mm anterior/posterior relative to bregma; Fig. 2C). The quantitative analysis showed less total fluorescence and lower surface area covered by the tracer in the sections prepared from mice pre-treated with AER-271 compared to vehicle controls (Fig. 2E). Coronal sections were segmented into ROIs based on anatomical boundaries referring to coordinates in the Allen Institute coronal mouse brain atlas (Fig. 2FG; atlas.brain-map.org.). Vehicle-treated mice exhibited highest tracer influx near the brain surface, particularly in the lateral and ventral cortices, and in hypothalamus (Fig. 2FG). Mice pre-treated with AER-271 exhibited a trend towards decreased CSF tracer influx across all brain regions, with the greatest differences noted in the lateral and ventral cortices, with a statistically significant decrease in the ventral cortex of anterior sections (Fig. 2F p=.005).

AER-271 prevents ADC changes after saline injection

Because ADC was shown to correlate to fluid movement within the brain parenchyma ^23,35, we sought to determine if AER-271 AQP4 inhibition altered this fluid movement. We performed diffusion-weighted MRI and measured baseline ADC values in WT and AQP4KO mice, with and without AER-271. AQP4KO mice showed 4.7–6.9% (mean 5.3%) higher baseline ADC values compared to their WT littermates (Fig. 3B, p<.0001; 2-way ANOVA with Bonferroni’s post-hoc), likely indicative of the enlarged extracellular space in AQP4KO mice, as described previously²³. In WT mice, there was an increase of ADC from baseline at 30 minutes and 75 minutes (Fig. 3C, p<.0001 for both time points; rel. to baseline ΔADC=1.82±0.46% at 30 min and rel. ΔADC=1.81±0.83 at 75 min; 2-way ANOVA with Bonferroni’s post-hoc). The increase of ADC relative to the initial baseline values in vehicle-treated WT mice could represent increased fluid influx resulting in increased ISF volume within the brain parenchyma. Unlike the WT vehicle-treated mice, vehicle-treated AQP4KOs exhibited no change in ADC from baseline both at 30 and 75 minutes, which could be due to reduced fluid exchange between brain parenchyma and CSF spaces thus preventing any change in the ISF volume (min. p>.7043 for both comparisons; AQP4KO rel. ΔADC=−0.28±0.91% at 30 min; rel. ΔADC=−0.43±0.85% at 75 min).

Figure 3: Diffusion-weighted MRI in vivo: AER-271 inhibits brain-wide ADC in WT animals.

(A) Experimental timeline for DWI experiment. (B) Box and whisker plots for the group–wise mean ADC values at the baseline and (C) the relative ADC changes from the baseline DWI at ~30 and ~75 minutes post AER-271 and saline (vehicle) injection, measured in 19 parenchymal ROIs in AQP4 KO (n=5 vehicle, 5 AER-271) and WT (n=6 vehicle and n=6 AER-271) mice, by means of DWI acquired in 6 diffusion-encoding directions. (D) ROI-wise correlation plots for the mean ΔADC vs. mean pixel intensity values of BSA-647 tracer, considering both saline and AER-271 injected mice jointly. Legend: ns - not significant, * - p<.05, ** - p<.01, *** - p<.001, **** - p<.0001 from 2-way ANOVA with Bonferroni’s post-hoc (B-C). Correlation plots show the respective regression lines along with semi-transparent areas marking 95% confidence intervals of the fitting. Correlations were considered significant for Person’s r or Spearman’s rho>0.5 with p<0.05, and non-zero regression slope.

Notably, when AER-271 was administered to WT mice, the increase of ADC from baseline was lost (Fig. 3C, p>.99 for both; rel. to baseline ΔADC=0.22±0.83% at 30 min and rel. ΔADC=0.54±1.15% at 75 min), and the relative ADC of WT AER-271 treated mice was similar to AQP4KO mice treated with vehicle (Fig 3C, p>.99 for WT AER-271 vs. AQP4KO Vehicle). There was no difference of ADC from baseline values in AER-271 treated AQP4KO mice at 30 minutes (p>.99; rel. ΔADC=0.03±0.58%), which was expected as the mice lack AQP4 channels. However, there was an increase of ADC at 75 minutes post in AER-271 treated AQP4KO mice, although the variability within the group is high (p<.0001; rel. ΔADC=1.33±1.66%).

We were interested to see whether the in vivo DWI results would correlate to CSF tracer influx, so we performed a regional correlation analysis of mean ΔADC with ex vivo BSA-647 CSF tracer influx (same regional influx data from Fig. 1FG). There was a low correlation between the average changes in mean ADC from baseline and the regional BSA-647 intensities only at 75 minutes after injection in aggregated group from AER-271 and vehicle WT mice (Fig. 3D; Spearman’s rho=0.55, p<0.02). When we divided into treatment groups a moderate correlation for the vehicle group was found (Spearman’s rho=0.733, p<0.05), but the slope of associated regression line was not different from zero. No other statistically significant correlations were found.

Altogether the observed changes in ADC support the conclusion that AER-271 acutely reduces brain fluid influx, as AER-271 prevents increases of ADC in WT animals.

AER-271 reduces glymphatic clearance

Next, we tested whether AER-271 inhibits glymphatic efflux as measured by a newly developed in vivo Direct Blue 53 (DB53) clearance assay ²⁴. In this assay, mice are implanted with an intrastriatal cannula 24 hours prior to the experiment, then DB53 dye is infused in the striatum and DB53 outflow to blood is measured subsequently. DB53 diffuses freely through the brain, and then clears from the brain via various efflux routes before ultimately reaching the blood, where it irreversibly binds to albumin. As a result, DB53 accumulation in blood reflects total DB53 clearance from brain, independent of the particular efflux paths (Fig. 4A). For this experiment, 8–12 weeks old mice maintained under ketamine/xylazine anesthesia were treated with AER-271 or vehicle 15 minutes prior to intrastriatal DB53 infusion. The similar areas covered by DB53 in coronal sections (Fig. 4BC) indicated that mice in the AER-271 and vehicle groups received comparable DB53 infusion volumes. At time 0 of the experiment venous signal was not different between the groups which is indicative of comparable levels of autofluorescence at baseline (Fig. 4F), and by 120 minutes post infusion the AER-271 treated mice exhibited less DB53 clearance to blood (Fig. 4DEF, p=.037 at 120 min).

We next fit the data to a mathematical compartment model with two components of DB53 clearance, namely a slow clearance component that is immediately active after DB53 infusion, and a faster clearance component that is recruited roughly 40 minutes after infusion ²⁴. The slow component may correspond to initial dispersion of the tracer from the infusion site and local transport across the blood brain barrier, and the fast component might reflect later clearance via glymphatic flow along perivenous spaces or cranial nerves. The model explained most of the variance in each recording (with coefficients of determination near 1, Table 2), however the variances in the predicted rates of DB53 clearance within the groups were too high to distinguish any meaningful differences (Table 2). Likewise, the model predicted a statistically insignificant reduction in total DB53 clearance to the blood circulation in the AER-271 treated group (Variable “C”, Table 2. p=.156), corresponding to the decreased cumulative clearance of DB53 to blood evident at 120 minutes (Fig. 4F). Thus, the effects AER-271 may have on the slow or fast clearance components are too subtle to detect with the current mathematical model given the sample size.

Table 2:

Descriptive statistics for fitted model coefficients

		C [−]	k1 [h−1]	k3 [h−1]	t0 [h]	R2
Saline n=6	Maximum	3.69	0.43	2.02	1.17	1.00
	Minimum	0.75	−0.05	0.20	0.10	0.97
	Range	2.94	0.48	1.82	1.07	0.03
	Mean	1.40	0.27	1.30	0.62	0.99
	SD	1.14	0.16	0.73	0.38	0.01
AER-271 n=6	Maximum	0.80	1.50	50.93	1.50	1.00
	Minimum	0.57	0.05	−1.39	0.41	0.97
	Range	0.23	1.45	52.32	1.09	0.02
	mean	0.69	0.57	10.14	0.80	0.99
	SD	0.10	0.53	20.05	0.38	0.01

Coefficients from the clearance curves are fitted to two-route clearance mathematical model. C is Long run blood concentration of DB53 (measured in normalized mean pixel intensities), k1 is the rate of clearance route one (per hour), k3 is the rate of clearance route two (per hour), t0 is the onset time of clearance route two (hours), and R² is the coefficient of determination.

AER-271 glymphatic reduction is AQP4 dependent

To test whether effects of AER-271 require the presence of AQP4 channels, we next repeated the glymphatic influx experiments in mice with genetic deletion of AQP4, AQP4KO mice ²¹ (Fig. 5A). The AQP4KO mice were pretreated with 2 doses of AER-271 (5mg/kg) or vehicle at 75 and 15 minutes prior to cisternal infusion of BSA-647, which was allowed to circulate for 30 minutes (Fig. 5B). Macroscopic images of whole brains showed no effect of AER-271 treatment on BSA-647 influx across the ventral brain surface, dorsal brain surface, or MCA perivascular spaces in the AQP4KO mice (Fig. 5CE). There was also no effect of AER-271 treatment on tracer influx in coronal sections from the AQP4KO mice (Fig. 5DF). As expected, CSF tracer influx was reduced in all groups of AQP4KO mice relative to WT mice (Fig. 5G), replicating earlier studies ^1,8.

Figure 5: AER-271 suppression of glymphatic influx is Aquaporin 4 dependent.

(A) Schematic showing AQP4 genetic knock-out strategy and representative immunohistochemistry images demonstrating the absence of AQP4 in AQP4 deficient mice (AQP4KO). (Green=glial fibrillary acid protein, Magenta = aquaporin 4, Scale bar = 25 μm) (B) Schematic showing AER-271 or vehicle treatment prior to cisterna magna tracer infusion in AQP4KO mice. (C) Whole mount brain images showing total BSA-647 CSF tracer distribution along the dorsal brain surface (Top) and ventral surface (Bottom). (Scale bar = 1 mm) (D) (Top) Representative coronal sections showing BSA-647 influx in AQP4KO mice. (Bottom) Representative images of BSA-647 influx in wild-type mice (WT, same mice as Figure 2C). Equivalent anterior sections shown (+0.6 mm bregma, Scale bar= 1 mm). (E) (Left) Total fluorescence intensity across the ventral brain surface, (Center) dorsal surface, and (Right) middle cerebral arteries in AQP4KO mice treated with either AER-271 or vehicle. (F) (Left) Total CSF tracer fluorescence in AQP4KO mouse coronal sections. (Right) Mean pixel intensity of CSF tracer in AQP4KO mice. (G) Total coronal section tracer fluorescence from WT mice (same data as Figure 2E; circles) plotted together with AQP4KO mice (same data as Panel F right; triangles). WT vehicle-treated mice had significantly more tracer than vehicle-treated AQP4KO mice (p=.001), WT AER-271 treated mice had more influx than vehicle (p=.017) and AER-271 treated (p=.026) AQP4KO mice. (H) Representative confocal images of AQP4 immunostaining from WT mice treated with vehicle or AER-271 (Same mice as Figure 2 B–G, Scale bar = 25 μm). (I) (Left) Intensity plots taken perpendicular to small vessels ( <6 μm), showing mean ± SD (Vehicle n= 5 mice, 123 vessels; AER-271 n=7 mice, 168 vessels). (Right) Mean AQP4 polarization, where each point represents mean AQP4 polarization measurement for all vessels of individual animals (p=.578). All plots depict mean ± SD, with each individual point representing one animal. All statistical tests were unpaired two-tailed t-tests, α=.05. Legend: ns - not significant, * - p<.05, ** - p<.01.

AER-271’s active metabolite AER-270 is also known as IMD-0354, an inhibitor of IκB kinase subunit beta (IKKβ). Inhibition of IKKβ reduces NF-κB nuclear translocation, where NF-κB can increase AQP4 expression by directly binding to the Aqp4 promoter ^36,37. Thus, we sought to test if AER-271 treatment would reduce AQP4 expression or localization via inhibition of NF-κB signaling. We performed AQP4 immunohistochemistry in brain sections from mice treated with two doses of AER-271 at 105 and 45 minutes prior to sacrifice, being the same mice used in the glymphatic influx experiment (Fig. 2). There was no change in total area of AQP4 expression after the AER-271 treatment in two successive doses (Fig. 4H). We also measured the AQP4 intensity profiles across small vessels (Fig. 5I, <6 μm diameter) in the dorsal, lateral, and ventral cortex, and then calculated the AQP4 polarization index by quantifying the peak AQP4 fluorescence signal around the vessel (i.e., astrocytic endfeet) with subtraction of the surrounding parenchymal AQP4 signal ¹⁷. The AER-271 treatment had no effect on AQP4 vessel localization in cortical brain areas (Fig. 5I, Fig. S3).

AER-271 does not alter heart rate, respiration or blood pressure

AQP4 is also expressed peripherally and arterial motion is a driver of glymphatic flow^38,39, thus we tested whether AER-271 altered the cardiovascular system. We recorded blood pressure via femoral artery catheter, electrocardiogram (ECG) via footpad electrodes, and respiration via pressure pad. We found no differences in the mean blood pressure, heart rate calculated from ECG, and respiration rate between the AER-271 and vehicle groups over the course of a 30-minute recording (Fig. S2A–C, left, p>.05 for AER-271 vs. vehicle; 2-way repeated measurements mixed model ANOVA). However a trend difference was observed for the HR (p=0.503). All the recordings showed slight changes appearing over their duration and similar in both vehicle and AER-271 groups (p<.0001). These slight changes were observed with decrease in blood pressure, and increase in respiratory and heart rates, likely attributed to the deepness of anesthesia over time.

Yet, the mean values of these signals are only one aspect of these physiological phenomena, so we applied Higuchi Fractal Dimension (HFD) analysis to quantify the variability of these signals across the recording time. We also found no differences in HFD values from the blood pressure respiration and heart rate signals between AER-271 or vehicle treatment groups (Fig. S2A–C, min. p=.3028, 2-way repeated measurements mixed model ANOVA). The variability assessed with HFD, was found similarly increased slightly over time in both groups for the blood pressure and respiratory rate but not for the heart rate signal (Fig. S2A–C, right). These analyses demonstrate AER-271 does not significantly affect the heart rate, respiration rate, or blood pressure over the duration of our experiment.

Discussion

We first validated AER-271’s ability to inhibit AQP4 in vivo using a mouse model of hyponatremia, and found AER-271 reduced mortality after acute WI. We found similar protection with another AQP4 inhibitor, TGN-020, which has not yet been tested with this assay. We also found that female mice were protected from mortality after WI which may be due to increased estrogen signaling in the kidney endowing female mice with a greater capacity to buffer the osmotic challenge³⁴. Importantly, AER-271 was only protective when given prior to the osmotic challenge, likely because the pro-drug needs time to be converted into the active form to be effective.

Next, we found AER-271 impaired glymphatic influx without affecting AQP4 localization in mouse brain. Administering AER-271 at 75 and 15 minutes prior to cisterna magna tracer infusion of BSA-647 CSF tracer inhibited its influx into the brain parenchyma. Interestingly, the roughly 15% pharmacological inhibition fell below the 45% reduction observed in AQP4KO mice compared to WT. These data are consistent with earlier studies reporting that deletion of AQP4 reduces glymphatic influx in the range of 40–60%, depending on the measuring technique used ⁸. Thus, it seems the present doses of AER-271 only partially blocked AQP4, or duration of AQP4 blockade may play a role in the magnitude of glymphatic reduction. Still, the effects of AER-271 were dependent on AQP4 expression, as AER-271 treated AQP4KO mice had no change in CSF tracer influx.

We then undertook in vivo DWI to determine whether AER-271 alters the brain fluid mobility as measured with apparent diffusion coefficient (ADC). Importantly, the DWI approach employed here allowed assessing the mean ADC changes from 6 diffusion encoding directions instead of assuming isotropic diffusion environment. Due to this, the observed effects at each timepoint are a superposition of 22–27 minutes lasting acquisition, and the magnitude of ADC changes is expected to be smaller. We found ADC increased brain-wide in WT mice by 2% relative to baseline at 30 minutes after vehicle treatment, with the effect remaining at 75 minutes. However, AER-271 completely abolished the temporal pattern in changes of ADC in WT mice, indicating inhibition of the physiological role of AQP4 function in supporting glymphatic influx. We also found no effect of AER-271 on ADC in AQP4KO mice at 30 minutes, while an increase was observed at 75 minutes after AER-271 administration. This delayed response may reflect that AER-271 impacts NF-κB signaling in AQP4KO mice to change ADC independent of AQP4. Overall, these data support the hypothesis that AER-271 significantly inhibits increases in ISF volume by acutely reducing the transmembrane fluid exchange in WT mice, whereas congenital lack of AQP4 channels reduces and delays the effect of AER-271 on CSF influx.

AQP4KO mice exhibit enlarged extracellular spaces, decreased potassium buffering capability, and reduced water exchange in the brain^23,40,41. Since these long-term structural and physiological brain changes might confound the existing findings of reduced glymphatic flow in AQP4KO mice, we emphasize that the present data show acute pharmacological blockage of AQP4 reduced glymphatic flow and prevented changes in ADC. AER-271 treatment reduced glymphatic influx in the same brain regions that showed the greatest glymphatic decrease in the AQP4KO mice, namely the lateral and ventral cortical areas ⁸. These brain regions exhibited the highest AQP4 expression in WT mice²³ which supports the notion that the drug is inhibiting AQP4, as areas with the most AQP4 were most affected by drug treatment. Our findings further validate that slow MR diffusion as a surrogate indicator of glymphatic function²³, and support the observation that AER-271 reduces glymphatic flow. Here we find the inhibitor reduced transmembrane water exchange leading to inhibition of ISF volume increases in WT mice, as measure using ADC. The AQP4KO mice had higher baseline ADC values than WT likely due to their enlarged brain extracellular space. Thus, pharmacological AQP4 inhibition emulates glymphatic decreases seen in AQP4KO mice, albeit with less potency, and apparently in relation to the local AQP4 expression.

We next tested effects of AER-271 using the in vivo glymphatic clearance assay ²⁴. There was a roughly 15% reduction in glymphatic clearance after AER-271 treatment, again falling below the 45–50% reduction of glymphatic clearance reported in AQP4KO mice ²⁴. This clearance pattern matches what has been previously observed in AQP4KO mice, albeit with lesser magnitude²⁴. Upon finding the reduction in total solute clearance, we applied the same mathematical model used in the previous study ²⁴ to our data to compare the kinetics between AER-271 and vehicle groups. The model predicted a reduced long-term clearance of DB53 to the blood after AER-271 treatment, matching the total solute clearance observed (Table 2). However, the transport rates (slow or fast clearance components) predicted by the model had high variability within experimental groups, and thus were not informative with our small sample size. This may call for refining the existing mathematical models of interstitial clearance by targeting infusions to different brain regions, thereby ascertaining the anatomical contributions to net solute clearance. In previous work, DWI slow MR diffusion parameters were reduced in AQP4KO mice, indicating an overall decrease of fluid turnover in the brain, which is synonymous with glymphatic influx and clearance²³. Our new analysis further confirmed those findings, insofar as the AQP4 inhibitor AER-271 acutely abolished typical changes in ADC observed in wildtype mice, as well as inhibited solute clearance, which supports the hypothesis that AQP4 facilitates glymphatic flow by promoting water clearance from the brain.

Acute manipulation of AQP4 has been historically challenging, as the features of a pharmacophore for the highly water selective AQP4 have yet to be defined. TGN-020 is one of the few existing small molecule inhibitors of AQP4, originally demonstrated to reduce edema following ischemic stroke ^31,42, and further developed as a radioligand for positron emission tomography used in both mice and humans ⁴³. There is some doubt about the specificity of TGN-020 for AQP4, as the drug was developed using computer models of the AQP4 monomer, not its endogenous homotetramer form, and the initial cell swelling assays had poor temporal resolution ⁴⁴. However, TGN-020 specifically inhibited AQP4 in a recent high temporal resolution cell swelling study comparing AQP channels ³². Indeed, we are the first to test TGN-020 in vivo with acute water intoxication and find it protective from mortality, and TGN-020 has been previously shown to inhibit glymphatic flow ⁴⁵. Yet, TGN-020 requires doses in the range of 100–300 mg/kg^6,31,43 and is poorly soluble in water which limits administration options and experimental design. Others have targeted AQP4 expression with shRNA, which indeed resulted in decreased glymphatic flow ^46,47. However, such experiments have a time-course of days, and require invasive intraparenchymal delivery of shRNAs to AQP4, which possibly induces inflammation that could reduce glymphatic flow¹⁶, and furthermore has off-target effects on coregulated RNAs such as Connexin 43⁴⁷. Given the limitations of current approaches to target AQP4, there remains considerable scope for developing more potent AQP4 inhibitors.

AER-271 is one such AQP4 inhibitor, it has a lower effective dose than previous AQP4 inhibitors (0.8–5 mg/kg in vivo and IC₅₀ 0.39 μM in vitro ¹⁸), efficient BBB permeability, and few off target interactions as tested by a ligand screening assay¹⁸. AER-271 is a modified version of IMD-0354, a phenylbenzamide that has previously been used as an IKKβ inhibitor ¹⁸. IMD-0354 proved to inhibit osmotic swelling through AQP4 blockade in a Chinese hamster ovary (CHO) cell model, leading to designing of the prodrug AER-271. AER-271 bears a phosphate group to improve water solubility; cleavage by endogenous phosphatases after injection yields its active, brain-penetrating homologue AER-270 (IMD-0354). The concentration of AER-270 in brain peaked at 150ng/g approximately 30 minutes after IP injection¹⁸, which corresponds to a brain concentration of 1.6μM concentration assuming a 20% extracellular space volume, so our two-dose administration paradigm should have at least this concentration which is above the IC₅₀ for the compound. AER-270 reduced edema following ischemic stroke, water intoxication, and global anoxic cerebral edema due to cardiac arrest, all of which are mediated by AQP4 ^12,19,48. These features make AER-271 an effective option for AQP4 inhibition, and the high solubility in water allows for more flexible administration options and experimental designs which may include repeated, chronic dosing.

Still, the exact mechanism of AQP4 inhibition by AER-271 is not fully understood. Signaling via the transcription factor NF-κB upregulates AQP4 expression ^36,37. NF-κB signaling is activated in the setting of ischemic stroke⁴⁹, which may thus mediate the AQP4 upregulation seen at 24 and 72 hours after ischemic stroke⁵⁰. Conversely, inhibition of NF-κB binding by curcumin reduced edema in a rat model of ischemic reperfusion stroke⁵¹. Given these findings, AER-271 may reduce cerebral edema by inhibiting NF-κB-mediated AQP4 expression. Yet, AER-271 has been shown to reduce osmotic swelling in AQP4 expressing CHO cells¹⁸, and here it inhibited edema during water intoxication, a setting in which there is no evidence of NF-κB activation. As such, we speculate that AER-271 blocks water passage through existing AQP4 channels rather than altering gene expression, although the mechanism is unclear. AER-271 might alter AQP4 function by altering phosphorylation of AQP4 or binding to an allosteric site on AQP4. It remains possible that AER-271 causes internalization or relocalization of AQP4 from the endfoot surface and our immunohistochemical approach was not sensitive enough to detect the difference; this could be more definitively tested with high resolution electron microscopy of AQP4¹².

Altogether AER-271 is a suitable alternative to the use of AQP4KO mice in future preclinical glymphatic studies, and is a promising candidate for pharmacological treatment of pathological fluid flow in the brain, such as occurs in the setting of ischemic stroke^18,20.

Table 1:

Descriptive statistics for diffusion-weighted MRI.

		Aqp4 (+/+)				AQP4 (−/−)
		Saline		AER-271		Saline		AER-271
		30 min	75 min	30 min	75 min	30 min	75 min	30 min	75 min
	No.+ sex	5 M/ 1 F		4 M / 2 F		3 M / 2 F		4 M / 1 F
Relative from baseline Δ(ADC)	Minimum	0.7983	0.3739	−1.521	−1.382	−1.431	−1.941	−1.128	−2.619
	Maximum	2.490	4.070	1.156	1.884	1.398	1.248	1.142	3.168
	Range	1.691	3.696	2.677	3.266	2.829	3.189	2.270	5.786
	Mean	1.823	1.810	−0.2171	0.5413	−0.2783	−0.4291	0.02635	1.330
	SD	0.4555	0.8283	0.8285	1.149	0.9120	0.8537	0.5836	1.662

Aqp4 (−/−) - Aquaporin-4 knockout mice, Aqp4 (−/−) - Aquaporin-4 wildtype mice, M - male, F - female.

Main points:

• The aquaporin-4 inhibitor AER-271 acutely decreases glymphatic flow similar to glymphatic decrease with genetic deletion of aquaporin-4

• AER-271 prevents ADC changes in WT mice

• Glymphatic inhibition by AER-271 requires aquaporin-4 expression

Data availability

The data used in this manuscript is available open access at the DANDI Archive https://dandiarchive.org/dandiset/000872