Bumetanide induces post-traumatic microglia–interneuron contact to promote neurogenesis and recovery

Abstract Although the Na-K-Cl cotransporter (NKCC1) inhibitor bumetanide has prominent positive effects on the pathophysiology of many neurological disorders, the mechanism of action is obscure. Attention paid to elucidating the role of Nkcc1 has mainly been focused on neurons, but recent single cell mRNA sequencing analysis has demonstrated that the major cellular populations expressing NKCC1 in the cortex are non-neuronal. We used a combination of conditional transgenic animals, in vivo electrophysiology, two-photon imaging, cognitive behavioural tests and flow cytometry to investigate the role of Nkcc1 inhibition by bumetanide in a mouse model of controlled cortical impact (CCI). Here, we found that bumetanide rescues parvalbumin-positive interneurons by increasing interneuron-microglia contacts shortly after injury. The longitudinal phenotypic changes in microglia were significantly modified by bumetanide, including an increase in the expression of microglial-derived BDNF. These effects were accompanied by the prevention of CCI-induced decrease in hippocampal neurogenesis. Treatment with bumetanide during the first week post-CCI resulted in significant recovery of working and episodic memory as well as changes in theta band oscillations 1 month later. These results disclose a novel mechanism for the neuroprotective action of bumetanide mediated by an acceleration of microglial activation dynamics that leads to an increase in parvalbumin interneuron survival following CCI, possibly resulting from increased microglial BDNF expression and contact with interneurons. Salvage of interneurons may normalize ambient GABA, resulting in the preservation of adult neurogenesis processes as well as contributing to bumetanide-mediated improvement of cognitive performance.

and maintained at 37 ± 2°C with a heating pad (Harvard Apparatus®).Animals received subcutaneous injections of Buprenorphine (0,03 mg/kg) and carprofen (5mg/ml).A metallic head-plate (Neurotar®), was attached to the skull using dental cement (Opti-Bond de Kerr) at the level of the posterior parietal cortex and the recording coordinates 2 mm posterior to and 1.5mm left of the bregma were marked using an indelible marker.Animals were monitored during the following week to make sure the recovery was successful.After a 72 hours recovery, animals were trained in the Mobile HomeCage® (MHC V5, Neurotar®) for the next 7 days.
Recordings were visually inspected using NeuroExplorer (version 5.305), and artifacts (e.g., disconnections, movement artifact) were manually rejected from all subsequent analyses.Raw data were then down sampled to 1,000 Hz and filtered for 50 Hz line noise using a Butterworth band stop IIR filter.Animal temperature as well as movement were used to determine active and non-active time windows.Delta, Theta and Gamma band were operationally defined as1-4 Hz, 4-12 Hz ;30-80 Hz 3 , and spectral power estimates across frequency bands (2-80 Hz, 80 logarithmically spaced points) were calculated using Morlet wavelets [cycles = 6; (56)] and first normalized using the ratio of power in the first minute to the last minute within each recording session.The mean normalized power of active time windows was subtracted from non-active for each recording.
Individual units were isolated from multisite LFP recordings using Spyking-Circus (https://github.com/spyking-circus) 4 .Clusters with >1% of ISIs violating the refractory period (<2 ms) or appearing otherwise contaminated were manually removed from the dataset.Pairs of units with similar waveforms and coordinated refractory periods in the cross-correlogram were combined into single clusters.Unit position with respect to electrode sites was characterized as the average of all electrode site positions weighted by the wave amplitude on each electrode.
The Single-unit activity were further classified into wide-spiking (WS) and narrow-spiking (NS) units 5 based on the spike waveform, auto-correlogram, and cross-correlogram.Units with short half-amplitude width, short trough-to-peak time, were classified as NS units; the rest of the units were classified into WS units (Fig S4).NS and WS presumably correspond to putative inhibitory and excitatory units, respectively 6,7 .

Immunohistochemistry
Mice were transcardially perfused with cold phosphate buffer saline (PBS 1X) prior to a 3% paraformaldehyde solution (AntigenFix, Diapath®).Brains were post-fixed overnight in 3% paraformaldehyde at 4°C and cells were fixed in 3% paraformaldehyde at RT for 5 mins.

PV microglia contacts quantification
Confocal stacks with double immunofluorescent labeling (Parvalbumin and Iba1) were acquired using a LSM-800 Zeiss confocal microscope Plan Flour 40x/1,30 NA oil immersion objective.Optical sections with a step size of 0.3 mm to a total of 15 mm were collected from the ipsi-and contralesional hippocampi of mice treated and non-treated with bumetanide, at 3 and 7 dpCCI.Image stacks were 3D-deconvoluted using the Autoquant X software (Media Cybernetics Inc).Images were then reconstructed in 3D with the Imaris Software.3D surfaces were rendered from each of the corresponding channels for microglia (green channel) and interneurons (red channel) and the contact points were defined as surfaces within 0.5 µm from each other.Then, contacts between microglia and PV interneurons were quantified with the ImageJ® plugin SynapCountJ from microglial processes of at least 0.5 long segments and visually inspected.
A metallic head-plate was attached to the coverslip forming a frame for the window.Animals were monitored during the following weeks ensure a successful recovery.
1 month after surgery, animals were trained for 5 days before the imaging session on a Mobile HomeCage® (MHC V5, Neurotar®).
Trained awake animals were head fixed on a Mobile HomeCage® platform (MHC V5, Neurotar®) under a custom-built two-photon microscope (Femtonics®) and imaged longitudinally using a Nikon Apo LWD 25X objective with 1.10 numerical aperture for 5 days.
The imaging protocol consisted of excitation at 920 nm using a Mai Tai® Ti:Sapphire Ultrafast Laser (Spectra-Physics) and emission collected at 530/30.An image stack with a 5 mm step size covering 300 mm was acquired.Each animal had a 30 min baseline recording.Injury was then achieved by focal excitation of 1 cell for 20 s at 75 µs/pixel at 100% laser intensity.After the injury, animals were IP injected with either the vehicle or bumetanide and z-stacks were collected covering the entire injury area (around 300 µm) with 5 min intervals for 1 h.Animals were IP injected twice per day with the vehicle or bumetanide and imaged using the protocol described daily from day 1 to day 5 post injury.
To compute microglia cell dynamics, we designed a semi-automated procedure using Python.
First, the 4D image stacks (3D + time) were loaded into the Napari viewer for a manual selection of the cells of interest.Sub-volumes were then cropped around each cell before undergoing a series of spatial registrations using pystackreg (G.Lichtner, pystackreg, (2022), https://github.com/glichtner/pystackreg)ported from the ImageJ TurboReg plugin 8 .The aim of these registrations was to correct translational, rotational and shearing drift that could be observed during acquisition.After maximum projection along the z-axis, the user was prompted to define regions of interest (ROIs) around each microglia and signal outside of these ROIs was discarded.To segment the microglia, an intensity threshold was determined and applied according to Li's method 9 and the cell dynamics was evaluated by calculating the percentage of binary pixels changing in value from one time point to another 8 .

Behavioral tests
1 month post-CCI, mice were tested on different tasks using the object recognition paradigm to test the individual components of episodic-like memory, namely the novel object recognition (NOR) and the object displacement task (ODT) 10 .Animals were habituated to the testing room 24 h before testing.For NOR, mice were placed in the center of a maze (Noldus apparatus®, 38.5 cm x 38.5 cm) and allowed to freely explore the space for 10 min.Then, two identical objects were added for a 3 min exploration time.Finally, after a 3 min retention time, one of the objects was replaced and the time of exploration was measured for a 3 min period.This test was used to assess short-term memory.The same experiment was also performed with a 1-hour retention time to test long-term memory.
For the ODT, mice were placed in the center of the same maze for 3 min, then 2 identical objects were added for 3 min.After a 3 min retention time, one of the objects was moved and the time of exploration of each object was recorded.
Mice were also tested for pure spatial memory learning using a Barnes maze (BM) test.Briefly, the Barnes maze is a black, metallic, circular table with a diameter of 100 cm elevated 60 cm from the ground.In the periphery there are 20 evenly spaced holes, where 19 of these holes are "false escapes" in which the mice cannot effectively hide.One hole contains the escape box (target hole), which leads to a dark recess below the table.1-month post-CCI, mice were trained twice daily for a duration of 4 days.On trial probe day (day 5), the target box was replaced with a false escape.The distance travelled, the number of errors, the time spent in the target zone and the latency to reach the target hole were measured.Finally, mice were tested for anxiety using the elevated plus maze (EPM) test.The maze from Ugo Basile company® is composed of 2 open arms and 2 closed arms that extended from a central platform (5 cm x 5 cm).Each arm is 50 cm long and are at a height of 50 cm above the floor, with the closed arms being surrounded by a 15 cm height wall.For each test, mice were placed in the center square facing an open arm and allowed to move freely for 5 mins.Entries and time spent in the open arms, closed arms and center were monitored.Recording and analyses was done using the Ethovision software (Noldus®).

Western blotting
Western blotting was performed as described in Goubert et al. 6 .Membranes were exposed overnight at 4°C to an NKCC1-specific antibody (DSHB Hybridoma Product T4) diluted 1:2000 in blocking solution.Chemiluminescent detection was performed using the ECL-plus kit (Pierce Biotech®).We measured signal intensities with the image analysis software G box (Syngene®).Then, membranes were stripped and probed with rabbit anti-α-tubulin (Sigma®, 1:10 000).Quantification was performed using the Gel Plot Analyzer plugin (Fiji®).

Cell culture -Cell lines:
As described previously 20 , BV2 cells were cultured in Gibco Dulbecco's Modified Eagle Medium (DMEM) media containing 10% fetal bovine serum, 100 IU/mL penicillin, and 100 mg/mL streptomycin for amplification.After seeding, we cultured BV2 cells in DMEM media containing only 1% fetal bovine serum, 100 IU/mL penicillin, and 100 mg/mL streptomycin since serum deprivation triggers cellular differentiation.Plates were incubated at 5% CO 2 and 37 °C.BLPS (50 ng/mL), bumetanide (40 nM) or LPS + Bumetanide were added to the culture media for either 24 or 72 h.
-Primary culture: Microglia were isolated from cultures of newborn CX3CR1-CreERT2 mice brain and the tissue was dissociated by trituration with trypsin (Gibco, United States) as described previously 21 .In brief, isolated cerebral cortices from newborn CX3CR1-CreERT2 mice were taken off the meninges, minced in DMEM complete medium (Gibco, United States) with 10% fetal bovine serum (Gibco, United States) and 1% antibiotics before being dissociated by trituration in 0.25% trypsin/EDTA.Cells were plated in 75 cm 2 plastic culture flasks containing 15 ml complete medium with 10% fetal bovine serum at a density of 2 × 10 6 cells/ml.Culture flasks were vigorously agitated on a rotary shaker twice for 10 minutes after 10 days of culture.Here, GFAP+ astroglia remain adherent to the flasks and the resulting cell suspension, rich in microglia, was placed in new plastic flasks (10 5 cells/ml) and allowed to adhere at 37°C.Cells were then plated into LAB-TEK II (Thermo Fisher Scientific) at a density of 2.5 × 10 4 cells/well for the phagocytosis assay and morphology analysis.

-Phagocytosis Assay
Microglia were seeded into Labtech plates at a density of 2.5 × 10 4 cells/well and cultured for 3 days under different conditions: co-transfected with pSico-shNKCC1 and pCAG-RFP and treated with bumetanide and tamoxifen, transfected with pCAG-RFP and treated with tamoxifen, or co-transfected with pSico-shNKCC1 and pCAG-RFP and treated with tamoxifen.Latex beads-Rabbit IgG-DyLight 633 Complex (Cayman chemical®, No.601490) were directly added to the pre-warmed culture medium (final dilution of 1:250).After incubating the DMEM medium containing latex beads with the treated microglia for 2 h, cells were fixed, stained with Hoechst and incubated in an anti-Iba1 antibody for subsequent fluorescence microscopy.

Deletion of Nkcc1 in Gfap-positive astrocytes does not show similar effect than bumetanide.
Although CCI induced robust changes at 3, 5 and 7 days in GFAP positive astrocytes we did not observe a significant effect of bumetanide on CCI induced changes in astrocyte morphology on either ipsi and contralateral side at 3 and 5 days after treatment (Fig S1A , B, C and D).We only did find however significant effects of bumetanide on the contralesional DG at 7 days post-CCI where bumetanide decreased the length of astrocyte processes (CCI Veh 290 µm ± 10.91 vs CCI Bum 147 µm ± 22.17 To investigate whether the effect of bumetanide on adult neurogenesis depends on Nkcc1 in astrocytes, we monitored the effect of CCI in transgenic mice where Nkcc1 was specifically deleted in astrocytes (GFAP-Nkcc1 KO), thus mimicking the effect of bumetanide on astrocytes (Fig S2 A and B).Interestingly, there was no impact on adult neurogenesis in hGFAP-Nkcc1 KO mice, neither in the contralesional DG (Sham +/+ 100 % ± 16.03 vs Sham Cre/+ 100 % ± There is a complex interplay between astrocyte and microglia activation during pathophysiological conditions.For this reason, we also investigated the effect of GFAP-Nkcc1 KO on microglia.After 7 days of bumetanide treatment, the number of microglial processes in WT animals did not differ between sham or contralesional DG of CCI animals (Sham +/+ 55.53 µm ± 12.56 vs Sham +/+ Bum 49.11 µm ± 24,52 and Contra +/+ 38.82 µm ± 6.241 vs Contra +/+ Bum 34.01µm ± 9.768) (Fig. S2H).However, it led to an increase in the number of microglial processes on the ipsilesional side (Ipsi +/+ 36.50 ± 6.025 vs Ipsi +/+ Bum 55.57 ± 12.40).In hGFAP-Nkcc1 KO mice, there was no change in the number of processes in either the sham condition (Sham +/+ 55.53 ± 12.56 vs Sham Cre/+ 50.92 ± 16.25) or on the ipsilesional side (Ipsi +/+ 36.50 ± 6.025 vs Ipsi Cre/+ 41.65 ± 4.984) of CCI animals.However, a significant increase was observed on the contralesional side of CCI animals (Contra +/+ 38.82 ± 6.241 vs Contra Cre/+ 52.78 ± 10.42).The average size of microglial soma was significantly reduced These results indicate that the depletion of Nkcc1 in astrocytes produced significantly different effects on astrocyte morphology than bumetanide treatment.In addition, the positive effects that occur in neurogenesis and interneuron survival after bumetanide treatment might not be mediated by astrocytes.

Bumetanide induces Nkcc1 expression and modify the morphology of a microglial cell line.
To investigate how bumetanide can influence microglia morphology and their trophic actions, we used a BV2 murine microglial cell line.24 hours of treatment with bumetanide did not show significant changes in Nkcc1 expression (normalized value on control: Ctrl 1 ± 0.17 vs Bum 0.82 ± 0.29, n= 12 wells per conditions, Fig. S3A).However, after 72 hours we detected a significant increase (Ctrl 1 ± 0.18 vs Bum 2.24 ± 0.99, Fig S3A).We then assessed the morphology of BV2 cells.We observed that 24h of bumetanide treatment induced a significant increase in average cell size (Bum 3161 µm² ± 371.

Supplementary Table 1 | Descriptive statistics of all data sets
Descriptive statistics of data sets per figure including the normality of the data sets, the test used, the groups compared, the difference between the group means, the SE of diff, the number of values per group, the DF, the statistically significant difference and the adjusted p-value.Descriptive statistics of data sets per figure including the normality of the data sets, the test used, the groups compared, the difference between the group means, the SE of diff, the number of values per group, the DF, the statistically significant difference and the adjusted p-value.
3 vs Ctrl 1993 µm² ± 164.1, Fig S3B and D).At 72h this effect was not significant (Bum 3100 ± 414.1 vs Ctrl 1974 ± 231.5, Fig S2E and G).The effect of bumetanide on the number of processes was the opposite.While no effect was found at 24 h (Bum 15.90 ± 4.73 vs Ctrl 22.08 ± 6.96 Fig S3B and C), bumetanide induced a significant decrease after 72 h (Bum 14.59 ± 5.205 vs Ctrl 35.16 ± 10.84, Fig S3F and G).These results further indicate that bumetanide can regulate microglial morphology by modulating chloride co-transporter Nkcc1 expression.

FIG
FIG S1 | Astrocyte morphological changes after CCI in the contralesional side 7 dpCCI and effect FIG S1 | Astrocyte morphological changes after CCI in the contralesional side 7 dpCCI and effect of bumetanide on major inflammatory cells.