The ribosome-associated protein RACK1 represses Kir4.1 translation in astrocytes and influences neuronal activity

Summary

The regulation of translation in astrocytes, the main glial cells in the brain, remains poorly characterized. We developed a high-throughput proteomics screen for polysome-associated proteins in astrocytes and focused on ribosomal protein receptor of activated protein C kinase 1 (RACK1), a critical factor in translational regulation. In astrocyte somata and perisynaptic astrocytic processes (PAPs), RACK1 preferentially binds to a number of mRNAs, including Kcnj10, encoding the inward-rectifying potassium (K⁺) channel Kir4.1. By developing an astrocyte-specific, conditional RACK1 knockout mouse model, we show that RACK1 represses production of Kir4.1 in hippocampal astrocytes and PAPs. Upregulation of Kir4.1 in the absence of RACK1 increases astrocytic Kir4.1-mediated K⁺ currents and volume. It also modifies neuronal activity attenuating burst frequency and duration. Reporter-based assays reveal that RACK1 controls Kcnj10 translation through the transcript’s 5′ untranslated region. Hence, translational regulation by RACK1 in astrocytes represses Kir4.1 expression and influences neuronal activity.

Graphical abstract

Highlights

• •RACK1, a ribosome-associated protein, interacts with specific mRNAs in astrocytes
• •RACK1 represses translation of Kcnj10, encoding the potassium (K⁺) channel Kir4.1
• •Upregulation of Kir4.1 in absence of RACK1 increases astrocyte K⁺ current and volume
• •Upregulation of Kir4.1 in absence of RACK1 attenuates neuronal bursting activity

The present study focuses on mechanisms of translation in astrocytes. Oudart et al. show that, in astrocytes, the ribosomal protein RACK1 associates with specific mRNAs and represses translation of Kcnj10, encoding the inward-rectifying K⁺ channel Kir4.1, thus modifying astrocytic K⁺ currents and influencing neuronal activity.

Introduction

There are few data on the mechanisms that regulate translation in astrocytes, the main glial cells in the brain, which project long processes to neurons (perisynaptic astrocytic processes [PAPs]) and brain blood vessels (perivascular astrocytic processes [PvAPs]) and dynamically regulate synaptic and vascular functions.^1,2 Translation is known to be mediated by cis-acting elements, including RNA motifs and secondary structures that influence the binding of trans-acting proteins (also known as RNA-binding proteins [RBPs]).³ A few RBPs have been studied in astrocytes.^4,5 Fragile X mental retardation protein (FMRP) has been shown to bind and transport mRNAs encoding autism-related signaling proteins and cytoskeletal regulators in radial glial cells.⁶ Selective loss of FMRP in astrocytes has been shown to dysregulate protein synthesis in general and expression of the glutamate transporter GLT1 in particular.⁷ In the mouse, expression of a pathological form of FMRP in astrocytes has been found to impair motor performance.⁸ More recently, mRNAs enriched in PAPs have been shown to contain a larger number of Quaking-binding motifs,⁹ and inactivation of the cytoplasmic Quaking isoform QKI-6 in astrocytes alters the binding of a subset of mRNAs to ribosomes.⁹ Quaking has also been shown to regulate differentiation of neural stem cells into glial precursor cells by upregulating several genes involved in gliogenesis.¹⁰ Another general parameter in regulation of translation is the composition of the translation machinery itself, including ribosomal RNAs (rRNA) and proteins.^11,12 This aspect had not been studied previously in astrocytes. Last, RNA distribution and local translation are important, highly conserved mechanisms for translational regulation in most morphologically complex cells.¹³ We and others have demonstrated that local translation occurs in astrocyte PvAPs and PAPs and might sustain the cells’ molecular and functional polarity.^14,15,16

To advance our understanding of translation mechanisms in astrocytes, we identified a pool of polysome-associated proteins in astrocytes by combining translating ribosome affinity purification¹⁷ with mass spectrometry (TRAP-MS). We then focused on receptor of activated protein C kinase 1 (RACK1), a highly conserved eukaryotic protein that is involved in several aspects of translation. RACK1 is positioned at the head of the 40S subunit in the vicinity of the mRNA exit channel.^18,19 It regulates not only ribosome activities (frameshifting and quality control responses) but also polysome localization and mRNA stability.^20,21 In the brain, RACK1 has been mainly described in neurons²² and is involved in local translation and axonal guidance and growth.²³ The changes in RACK1 expression observed in several neuropathological contexts (such as bipolar disorder,²⁴ Alzheimer’s disease,^25,26 epilepsy,^27,28 addiction,²⁹ amyotrophic lateral sclerosis,³⁰ and Huntington’s disease³¹) indicate the importance of the protein’s physiological role in the brain.

In the present study, we demonstrate that, in astrocytes, RACK1 associates with specific mRNAs and represses translation of Kcnj10 (encoding the inward-rectifying K⁺ channel Kir4.1), modifying astrocytic K⁺ currents and influencing neuronal activity.

Results

Identification of polysome-associated proteins in astrocytes

We used TRAP-MS to identify polysome-associated proteins in astrocytes (Figure 1A). Enhanced green fluorescent protein (EGFP)-tagged polysomal complexes were immunopurified from whole-brain cytosolic extracts prepared from 2-month-old Aldh1l1:L10a-EGFP transgenic mice following a refined TRAP protocol.¹⁷ These animals express the EGFP-tagged ribosomal protein RPL10a specifically in astrocytes³² (Figure 1A). The same experiment was performed on brain samples from C57BL/6 mice as a control to identify the background nonspecific signal (Figure 1A). A western blot analysis of immunoprecipitated proteins showed that RPL10a-GFP and the ribosomal protein S6 (RPS6, a component of the 40S ribosomal subunit) were present in Aldh1l1:L10a-EGFP immunoprecipitates only, demonstrating the efficiency and specificity of TRAP-MS (Figure 1B). Extracted proteins were characterized using quantitative label-free tandem MS (liquid chromatography-tandem MS [LC-MS/MS]) (Figures 1A and 1C). Three proteins were found only in C57BL/6 extracts, and 139 were found only in Aldh1l1:L10a-EGFP extracts (fold change [FC]: – or +∞), 61 proteins were enriched in C57BL/6 extracts (p < 0.05; Log2 FC < −1), 106 proteins were detected in C57BL/6 and Aldh1l1:L10a-EGFP extracts (p < 0.05; −1 < Log2 FC < 1), and 110 proteins were enriched in Aldh1l1:L10a-EGFP extracts (p < 0.05; Log2 FC > 1) (Figure 1C; Tables S1 and S2). Most proteins identified in C57BL/6 extracts were immunoglobulins (provided by the TRAP columns), while proteins enriched or specifically identified in Aldh1l1:L10a-EGFP immunoprecipitates belonged to polysomal complexes, again demonstrating the efficiency and specificity of TRAP-MS (Table S2). Gene Ontology (GO) analysis of the 249 enriched immunoprecipitates or those specifically identified in Aldh1l1:L10a-EGFP indicated that most were ribosomal proteins (26%) or RBPs (39.1%) involved in ribosome biogenesis (22.7%) and gene expression (30.9%) (Figure 1D; Table S2). Thus, we were able to identify a set of polysome-associated proteins in astrocytes.

Figure 1.

Identification of polysome-associated proteins in astrocytes

(A) Flowchart of the TRAP-MS analysis on whole-brain extracts. Proteins extracted from whole brain in C57BL/6 mice or Aldh1l1:L10a-EGFP mice were immunoprecipitated by TRAP and analyzed by LC-MS/MS.

(B) Western blot detection of Rpl10a-EGFP and RPS6 in whole-brain extracts or TRAP immunoprecipitated proteins (immunoprecipitation [IP] GFP). C57BL/6 extracts were used as negative controls. The position of molecular weight markers is indicated on the right.

(C) Volcano plot of the TRAP-MS results. Each protein is represented by a dot. The dot size is proportional to the number of peptides identified by LC-MS/MS. Dots for proteins specific to or enriched in Aldh1l1:L10a-EGFP mice are represented with a color code: ribosomal proteins in light blue, ribosome-associated proteins in red, RBPs in purple, cytoskeleton-associated proteins in green, vesicle-ER-Golgi apparatus-lysosome-associated proteins in orange, and other proteins in black. Five independent replicates were analyzed (one brain per sample). The protein distribution is represented as the Log2 FC of the Aldh1l1:L10a-EGFP/C57BL/6 (x axis) versus −Log10-adjusted p value (y axis): proteins identified only in C57BL/6 extracts (3 proteins) or only in Aldh1l1:L10a-EGFP extracts (139 proteins) (FC: – or +∞); proteins enriched in C57BL/6 or Aldh1l1:L10a-EGFP extracts. The threshold for the enrichment in C57BL/6 or Aldh1l1:L10a-EGFP extracts is p < 0.05 (red line) and Log2 FC > 1 or < −1 (green lines).

(D) A GO analysis of the proteins enriched or detected solely in Aldh1l1:L10a-EGFP extracts (p <0.05; Log2 FC > 1) for biological processes (left) and molecular functions (right). The raw data are given in Table S2.

RACK1 associates with polysomes in astrocytes

Among the ribosome-associated proteins preferentially extracted with TRAP-MS, we focused on RACK1. This protein binds to the small ribosomal subunit 40S and has a key role in translation of capped polyadenylated mRNAs³³ (Figure 2A). RACK1’s role in astrocytes had not been assessed previously. However, Gnb2l1 mRNA (encoding RACK1) is highly translated in PAPs, suggesting that RACK1 has an important role at this cellular interface.¹⁶ A western blot analysis of TRAP-MS immunoprecipitated proteins showed that RACK1 was specifically detected in the Aldh1l1:L10a-EGFP condition and thus confirmed co-immunoprecipitation of RACK1 with astrocytic polysomes (Figure 2B). We next characterized RACK1 expression in astrocytes with a focus on the hippocampus. We used fluorescence in situ hybridization (FISH) and detected Gnb2l1 mRNAs in astrocytes, consistent with our previous translatome analysis¹⁶ and other transcriptomics analyses^34,35 (Figure 2C). Because Gnb2l1 is ubiquitous, we immunolabeled astrocytes for glial fibrillary acidic protein (GFAP). In line with our previous results, Gnb2l1 mRNAs were detected somewhat in astrocyte somata but mainly in astrocytic processes¹⁶ (Figure 2C). We next performed immunofluorescence imaging of RACK1 on hippocampal sections (Figure 2D). RACK1 was detected in neurons and GFAP-immunolabeled astrocytic somata and processes (Figure 2D). These results suggest that RACK1 is expressed in astrocytes and associates with astrocytic polysomes.

Figure 2.

RACK1 is associated with ribosomes in astrocytes

(A) Representation of the human 80S ribosome on the basis of the high-resolution cryoelectron microscopy (cryo-EM) structure.³⁶ RPL10a is shown in green and RACK1 in black.

(B) Western blot detection of RACK1 in whole-brain protein extracts and in TRAP-MS extracts (IP GFP) under Aldh1l1:L10a-EGFP conditions. C57BL/6 extracts were used as negative controls. The position of molecular weight markers is indicated on the right.

(C) FISH detection of Gnb2l1 mRNAs encoding RACK1 in hippocampal astrocytes immunolabeled for GFAP. From left to right: confocal image of a GFAP-immunolabeled astrocyte (green), FISH detection of Gnb2l1 mRNAs (red dots), and merged image. White arrows indicate Gnb2l1 mRNAs located on GFAP-positive processes.

(D) Confocal images of RACK1 immunofluorescence detection (red) in the hippocampus. Astrocytes (^∗) are co-immunolabeled for GFAP (green). A neuron (°) is also labeled for RACK1.

Nuclei are stained with DAPI in (C) and (D). Scale bars, 20 μm.

RACK1 associates with specific mRNAs in astrocytes and in PAPs

We next determined which mRNAs were associated with RACK1 in astrocytes via mRNA immunoprecipitation (using a RACK1-specific antibody) of whole-brain astrocytic cytoplasmic extracts prepared from 2-month-old mice. We first checked the efficiency of RACK1 immunoprecipitation on western blots. Increasing levels of anti-RACK1 antibody indeed immunoprecipitated higher levels of RACK1 and RPS6 (Figure 3A). We then repeated the experiment with the optimal quantity of RACK1 antibody, extracted the immunoprecipitated mRNAs, and analyzed them with qPCR (Figure 3B). Non-specific mouse immunoglobulin G (IgG) was used as a negative control (Figure 3B). Extracts prepared from whole Aldh1l1:L10a-EGFP brain were immunoprecipitated (Figure 3C). RACK1 was present in astrocyte processes (Figure 2D). We therefore also immunoprecipitated RACK1 in Aldh1l1:L10a-EGFP synaptogliosome preparations consisting of PAPs attached to synaptic neuronal membranes³⁷ (Figure 3C′). Because RACK1 is ubiquitously expressed in the brain, we limited our analysis to a selection of astrocyte-specific mRNAs, such as Kcnj10, encoding the inward-rectifying K⁺ channel Kir4.1; Slc1a2, encoding the glutamate transporter GLT1; Aqp4, encoding the water channel aquaporin 4; Slc1a3, encoding the L-Glutamate/L-Aspartate transporter GLAST; and Gja1 and Gjb6, encoding the gap junction proteins connexin 43 and 30, respectively. With the exception of Gjb6, all the tested mRNAs were immunoprecipitated more significantly by RACK1 than by IgG in whole-brain (Figure 3C) and synaptogliosome extracts (Figure 3C′). These results were probably influenced by the level of polysomal mRNAs in astrocytes and PAPs. We therefore determined the level of each polysomal mRNA in astrocytes and PAPs by performing TRAP and qPCR on whole-brain extracts from 2-month-old Aldh1l1:L10a-EGFP mice (Figures 3D and 3E) or on synaptogliosome extracts (Figures 3D and 3E′). The mean value for each mRNA was then used to normalize the quantity of RACK1-immunoprecipitated mRNA (Figure 3F) in whole brain (Figure 3G) and in synaptogliosomes (Figure 3G′). The results of these experiments suggested that Slc1a2 and Kcnj10 were preferentially associated with RACK1 in astrocytes (Figure 3G) and in PAPs (Figure 3G′).

Figure 3.

RACK1 associates with specific mRNAs in astrocytes and PAPs

(A) Western blot analysis of RACK1 IP in whole-brain extracts from 2-month-old C57BL/6 mice (arrow). Increasing quantities of RACK1 antibodies (0, 2, and 5 μg) were used. The higher band immunoprecipitated by RACK1 antibodies corresponds to a non-specific background noise of the IP. RPS6 is also detected in RACK1-immunoprecipitated proteins. The position of molecular weight markers is indicated on the right.

(B–C′) Flowchart (B) of RNA IP using anti-RACK1 antibodies (red) on whole-brain extracts (C) or synaptogliosome extracts (C′) prepared from 2-month-old Aldh1l1:L10a-EGFP mice. Red dots on ribosomes represent RACK1. Immunoprecipitated RNAs were purified and screened (in qPCR assays) for a selection of astrocyte-specific mRNAs. IgG-subtracted signals were normalized against rRNA 18S. The data are quoted as the mean ± SD (N = 5 or 6 samples; 1 mouse brain per sample); one-sample t test vs. 0 (except for the Gjb6 whole-brain experiment, one-sample Wilcoxon test).

(D–E′) Flowchart of polysomal IP using anti-GFP antibodies (green) on whole-brain (E) or synaptogliosomes extracts (E′) prepared from 2-month-old Aldh1l1:L10a-EGFP mice. Immunoprecipitated RNAs are purified and analyzed by qPCR for a selection of astrocyte-specific mRNAs. Signals were normalized against rRNA 18S. The data are quoted as the mean ± SD (N = 4 samples, 1 mouse brain per sample).

(F–G′) Flowchart of normalization of RACK1 IP against GFP IP mean values. Ratios were calculated for experiments on whole brain (G) or synaptogliosomes (G′). The data are quoted as the mean ± SD (N = 5 or 6 samples, 1 mouse brain per sample). The p values are indicated in green when Kncj10 results were the reference and in red when Slc1a2 results were the reference. Two-tailed unpaired t test or two-tailed Mann-Whitney test. ns, not significant (p > 0.05); ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001. The raw data are presented in Table S3.

These results suggest that RACK1 associates preferentially with specific mRNAs in astrocytes and in PAPs.

RACK1 represses expression of Kir4.1 in astrocytes and in PAPs

To gain insights into RACK1’s function in astrocytes, we generated a RACK1 conditional knockout (cKO) mouse model by crossing RACK1 fl/fl mice with Aldh1l1-CreERT2 mice (Figure 4A). Two-month-old Aldh1l1-CreERT2: RACK1 fl/fl mice were injected with tamoxifen to induce RACK1 KO in astrocytes (Figure 4A). KO in astrocytes was confirmed by PCRs on DNA extracted from whole brain (Figure 4B) and by immunofluorescence assays on hippocampal sections (Figure 4C). In cKO mice, RACK1 was detected in pyramidal layer neurons but not in astrocytes immunolabeled for GFAP (Figure 4C). We next sought to determine the impact of RACK1 KO in astrocytes on the levels of GLT1 and Kir4.1 in whole brain, hippocampus, whole brain, and hippocampal synaptogliosome protein extracts from RACK1 fl/fl and RACK1 cKO mice on western blots (Figures 4D and S2). It was also tested in PvAPs copurified with brain microvessels because Kcnj10 (Kir4.1) and Slc1a2 (GLT1) are translated at this cellular interface.¹⁴ Interestingly, the various extracts did not differ significantly with regard to the level of GLT1. However, the level of Kir4.1 was significantly higher in all extracts from RACK1 cKO mice except in brain microvessels (Figure 4D). Similar results were found in the cortex but with a stronger effect in astrocytes and PAPs (Figure S1A). Consistent with this, immunofluorescence of Kir4.1 in cortical astrocytes on brain sections was higher in RACK1 cKO compared with RACK1 fl/fl mice (Figure S1B). In this brain area, upregulation of Kir4.1 was also detected in PvAPs (Figures S1A and S2).

Figure 4.

RACK1 KO in astrocytes leads to higher levels of Kir4.1 in astrocyte somata and PAPs

(A) Generation of a mouse line with RACK1 KO in astrocytes (RACK1 cKO). Shown is a schematic of the RACK1 fl/fl and Aldh1l1-Cre/ERT2 alleles. Deletion of exon 2 in Gnb2l1 (the gene coding for RACK1) is induced in astrocytes by tamoxifen injection; this results in a frameshift and premature termination of Gnb2l1 translation. Primers are indicated by red arrows.

(B) PCR assays for Gnb2l1 KO in brain DNA from RACK1 fl/fl or Aldh1l1-CreERT2: RACK1 fl/fl tamoxifen-injected mice (RACK1 cKO). The 898-bp band corresponds to the floxed allele (Table S4). The 672-bp band corresponds to the exon 2-deleted allele.

(C) Confocal images of RACK1 immunofluorescence (red) in the hippocampus in RACK1 fl/fl and RACK1 cKO mice. Astrocytes are co-immunolabeled for GFAP (green). Some astrocytes are indicated by white arrowheads. Nuclei are stained with DAPI. The bottom panel gives a higher-magnification view of the boxed areas in the RACK1 fl/fl and RACK1 cKO images, which shows that RACK1 is specifically depleted in astrocytes (^∗) and is still expressed by neurons (°). Scale bars, 20 μm.

(D), Western blot detection and analysis of Kir4.1 and GLT-1 in protein extracts from whole brain, hippocampus, whole-brain synaptogliosomes, hippocampal synaptogliosomes, or whole-brain microvessels purified from RACK1 fl/fl or RACK1 cKO mice. The position of molecular weight markers is indicated on the right. Signals were normalized against that of stain-free membranes except for the experiment on purified microvessels, where histone 3 was used. The data are quoted as the mean ± SD (N = 5 samples per genotype, 1 mouse per sample); two-tailed unpaired t test or Mann-Whitney test. ns, p > 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001. The raw data are presented in Table S3 and Figure S2.

These results demonstrate that RACK1 deficiency in astrocytes leads to a higher level of Kir4.1 in astrocytes, PAPs, and cortical PvAPs.

Effect of Kir4.1 upregulation on astrocyte inward K⁺ currents and cell volume

Kir4.1 is a weakly inward-rectifying K⁺ channel that confers astrocytes with high K⁺ conductance. Elevation of Kir4.1 levels in RACK1 cKO mice might be associated with a greater K⁺ inward current. To investigate this, we performed patch-clamp whole-cell recordings of hippocampal cornu ammonis (CA1) stratum radiatum astrocytes (Figure 5A). RACK1 fl/fl and RACK1 cKO astrocytes displayed similar linear current-voltage (I-V) curves under control conditions and after treatment with the Kir4.1 K+ channel blocker VU0134992 (30 μM), which similarly decreased the evoked currents in both genotypes (Figures 5A and 5B). We then tested the ability of RACK1 cKO astrocytes to buffer K⁺ during high-frequency stimulation of Schaffer collaterals (SC).^38,39,40 SCs were stimulated at 10 Hz for 1 s, and astrocyte whole-cell currents were recorded under control conditions and in the presence of VU0134992 (30 μM). Astrocytic Kir4.1 currents were isolated by subtracting the VU0134992-insensitive component from the total current (Figure S3A). We found that, in response to repetitive high-frequency stimulation, RACK1 cKO astrocytes displayed an ∼2-fold increase in the amplitude of Kir4.1-mediated currents (after 8 pulses) compared with RACK1 fl/fl astrocytes (Figure 5C). These results indicate that the upregulation of Kir4.1 induced by RACK1 deficiency results in formation of functional channels, which are activated during high regimens of activity.

Figure 5.

Absence of RACK1 in astrocytes leads to higher Kir4.1-mediated astrocytic inward K⁺ currents and cell volumes

(A) Schematic of electrode positions used to record astrocyte whole-cell currents evoked by Schaffer collateral (SC) stimulation in the CA1 region of hippocampal slices.

(B) Left: representative traces of astrocytic whole-cell currents induced by 150-ms voltage steps (from −200 mV to +100 mV, 10-mV steps; black traces at the bottom) in RACK1 fl/fl and RACK1 cKO mice before (black and pink) and after (blue) application of a KIR 4.1 antagonist (VU). Scale bars, 50 ms, 5 nA. Right: current-voltage (I-V) plot in RACK1 fl/fl (white filled dots) and in RACK1 cKO (pink-filled dots) mice before (black) and after (blue) application of VU (N = 7 and 10 astrocytes for RACK1 fl/fl and RACK1 cKO mice, respectively; repeated-measures two-way ANOVA with Sidak’s correction for multiple comparisons) The data are quoted as the mean ± SD.

(C) Left: representative astrocytic Kir4.1 (VU-sensitive) currents induced by SC stimulation (10 Hz, 1 s) in RACK1 fl/fl (black) and RACK1 cKO (pink) mice. Scale bars, 200 ms, 50 pA. Right: quantification of astrocytic Kir4.1 current peak amplitude after each stimulus during SC stimulation (10 Hz, 1 s) in RACK1 fl/fl (white filled dots) and RACK1 cKO (pink filled dots) mice (N = 6 and 5 astrocytes for RACK1 fl/fl and RACK1 cKO mice, respectively; repeated-measures two-way ANOVA with Sidak’s correction for multiple comparisons). The data are quoted as the mean ± SD. ^∗p < 0.05, ^∗∗p < 0.01, ^∗∗∗p < 0.001.

(D) Illustration of the imaging method with a representative raw confocal image of an isolated RACK1 fl/fl CA1 astrocyte expressing tdTomato.

(E–G) Imaris analysis: filament tracing (E), convex hull volume (F), and a 3D Sholl analysis (G).

(H) Mean territory volume and filament length of RACK1 fl/fl and RACK1 cKO astrocytes. Shown is a histogram of the data, presented as the mean ± SD (N = 4 mice per genotype, 45 astrocytes); two-tailed t test. The data are quoted as the mean ± SD.

(I) A Sholl analysis of the ramification complexity of RACK1 fl/fl and RACK1 cKO astrocytes. Two-way analysis of variance. ^∗p < 0.05, ^∗∗p < 0.01. The data are quoted as the mean ± SD. The raw data are presented in Table S3.

K⁺ influx into astrocytes is thought to be coupled with water intake, leading to transient or prolonged swelling.^41,42 Thus, elevation of K⁺ inward currents in RACK1 cKO mice might be associated with a greater hippocampal astrocyte volume. We used an adeno-associated virus (AAV) bearing the gfaABC1D synthetic promoter (derived from Gfap⁴³) to drive expression of the fluorescent protein tdTomato in astrocytes. AAVs were injected into the CA1 region of the dorsal hippocampus of 2-month-old mice (Figure 5D). A three-dimensional (3D) analysis was performed on sparse labeled CA1 hippocampal astrocytes from RACK1 fl/fl and RACK1 cKO mice (Figures 5D–5I) to reconstruct the astrocytes’ processes (filaments; Figure 5E) and volume (Figure 5F). RACK1 cKO astrocytes had a larger territory volume with longer processes (filaments) (Figure 5H). Using a 3D Sholl analysis, we determined that RACK1 cKO astrocytes had longer distal processes (Figures 5G and 5I) than RACK1 fl/fl astrocytes.

These results indicate that astroglial RACK1 is required for correct astrocytic K⁺ homeostasis and territory volume.

Effect of Kir4.1 upregulation on neuronal activity

During their activity, neurons release large amounts of K⁺ at the synapses. The K⁺ is rapidly taken up by astrocytic Kir4.1 and is redistributed across the astrocytic network. This clearance mechanism maintains perisynaptic homeostasis and prevents neuronal hyperexcitability. Because RACK1 cKO astrocytes contain higher levels of Kir4.1 and display higher K⁺ inward currents upon high-frequency stimulation, we sought to determine whether these changes alter basal excitatory synaptic transmission. We stimulated CA1 SC synapses in acute hippocampal slices and thus evoked α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR)-mediated field excitatory postsynaptic potentials (fEPSPs) (Figure 6A). The size of the presynaptic fiber volley (the input) was compared with the slope of the fEPSP (the output). RACK1 cKO mice and control RACK1 fl/fl mice did not differ with regard to basal synaptic transmission (Figure 6B). We next investigated the effect of the Kir4.1K⁺ channel blocker VU0134992 (30 μM) and found a similar overall decrease in excitatory synaptic transmission in RACK1 fl/fl and RACK1 cKO animals after 20 min of application (Figure 6B). These results indicated that the elevated expression of astrocytic Kir4.1 in RACK1 cKO mice did not modify hippocampal basal excitatory synaptic transmission evoked in the hippocampal CA1 region.

Figure 6.

Absence of RACK1 in astrocytes alters network population activity and neuronal responses to intense stimulation

(A) Schematic of electrode positions used to record field excitatory postsynaptic potentials (fEPSPs) evoked by SC stimulation in the CA1 region of hippocampal slices.

(B) Input-output curves for basal synaptic transmission. Left: representative recordings in RACK1 fl/fl mice (black) and RACK1 cKO mice before (pink) and after (blue) application of a Kir 4.1 antagonist (VU). Scale bars, 10 ms, 0.5 mV. Right: quantification of the fEPSP slope for different fiber volley amplitudes after SC stimulation. The data are quoted as the mean ± SD. RACK1 fl/fl: n = 5 slices from 4 mice; p = 0.0087; RACK1 cKO: n = 5 slices from 5 mice; repeated-measures two-way ANOVA with Sidak’s correction for multiple comparisons.

(C) Top: a representative recording of fEPSPs evoked by repetitive stimulation (10 Hz, 30 s) of CA1 SCs in RACK1 fl/fl mice under control conditions. Scale bars, 5 s, 0.2 mV. Bottom: enlarged view of fEPSPs evoked by the first 10 stimuli. Scale bars, 200 ms, 0.2 mV.

(D) Changes in the fEPSP slope induced by 10-Hz stimulation relative to responses measured before the onset of stimulation (baseline responses) in RACK1 fl/fl mice (white filled dots) and in RACK1 cKO mice (pink-filled dots) before (black) and after (blue) application of VU. The data are quoted as the mean ± SD. RACK1 fl/fl: n = 5 from 5 mice; RACK1 cKO: n = 6 slices from 4 mice; repeated-measures two-way ANOVA with Sidak’s correction for multiple comparisons.

(E) Schematic (left) and picture (right) of a hippocampal slice placed on a multielectrode array (MEA). Scale bar, 200 μm.

(F) Representative MEA recordings of burst activity induced in hippocampal slices of RACK1 fl/fl (black) and RACK1 cKO (pink) mice by incubation in Mg²⁺-free ACSF containing 6 mM KCl. The expanded recordings of the bursts (surrounded by gray rectangles) are shown on the right. Scale bars, 10 s (left)/200 ms (right), 50 μV.

(G) Quantification of burst frequency (top) and burst duration (bottom) in RACK1 fl/fl (white) and RACK1 cKO (pink) hippocampal slices. The data are quoted as the mean ± SD. n = 15 slices from 5 mice for RACK1 fl/fl and n = 18 slices from 6 mice for RACK1 cKO; unpaired t test.

(H) Representative MEA recordings of hippocampal bursting activity in RACK1 fl/fl (top) and RACK1 cKO (bottom) slices in control (Ct) and during 25-min treatment with the Kir4.1 blocker VU0134992 (VU). The corresponding time-frequency plots are shown under the traces. Scale bar, 20 s/2 min, 50 μV.

(I) Quantification of VU’s effect on burst frequency (top) and duration (bottom) in RACK1 fl/fl (white) and RACK1 cKO (pink) hippocampal slices (RACK1 fl/fl: n = 15 slices from 5 mice for burst frequency and duration, respectively; RACK1 cKO: n = 18 slices from 6 mice for burst frequency and duration; paired t test. The data are quoted as the mean ± SD. ns, p > 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001. The raw data are presented in Table S3.

We reasoned that the elevated expression of Kir4.1 in RACK1 cKO astrocytes and the consequent enhancement in K⁺ buffering capacity might have major roles during intense neuronal activity. To test this hypothesis, we repeatedly stimulated SCs (10 Hz, 30 s) and analyzed the fEPSPs in the CA1 region of the hippocampus. This stimulation induced rapid synaptic facilitation and then depression (Figure 6C), which results from depletion of the presynaptic glutamate pool. Facilitation was greater and depression was slower in RACK1 cKO slices than in RACK1 fl/fl slices (Figure 6D). This finding indicates that the elevated expression of Kir4.1 in RACK1 cKO mice sustains repetitive excitatory synaptic activity. Accordingly, Kir4.1 K⁺ channel inhibition by VU0134992 (30 μM for 20 min) had more of an effect on synaptic facilitation and depression in RACK1 cKO mice than in RACK1 fl/fl mice (Figure 6D). In the presence of VU0134992, repetitive stimulation-induced facilitation and subsequent depression were similar in RACK1 fl/fl and RACK1 cKO mice (Figure 6D). These results indicate that astrocytic RACK1 influences neuronal activity in response to repetitive stimulation by controlling the expression of Kir4.1.

We next tested the impact of RACK1 on recurrent burst activity. This was induced in hippocampal slices by incubation in a pro-epileptic artificial cerebrospinal fluid (ACSF) (Mg²⁺ free with 6 mM KCl [0Mg6K ACSF]). We recorded neuronal bursts in all hippocampal regions by using the multielectrode array (MEA) technique (Figures 6E and S3B). We found that bursts were less frequent and lasted longer in RACK1 cKO mice than in RACK1 fl/fl mice (Figures 6F and 6G). Hence, the burst rate under pro-epileptic conditions appeared to be better controlled in astrocytic RACK1 cKO mice than in RACK1 fl/fl mice. To check whether this was due to more efficient buffering of extracellular K⁺ released during sustained activity, we recorded burst activity in RACK1 fl/fl and RACK1 cKO slices in the presence of VU0134992 (30 μM). In RACK1 fl/fl mice, inhibition of the Kir4.1 K⁺ channel by VU0134992 caused an increase in burst frequency that was likely due to increased neuronal excitability resulting from depolarization induced by extracellular K⁺ accumulation. This was accompanied by a decrease in burst complexity, as evidenced by the decreased duration and amplitude of individual bursts (Figures 6H and 6I). Interestingly, no initial increase in burst frequency after Kir4.1 channel inhibition was observed in RACK1 cKO mice. However, a reduction in burst complexity occurred during VU0134992 (VU) treatment, as evidenced by the decreased amplitude of individual bursts (Figures 6H and 6I). Indeed, in RACK1 cKO mice, the basal burst frequency was already decreased compared with RACK1fl/fl mice because of enhanced expression of Kir4.1 channels, which reduces extracellular potassium concentration ([K⁺]_e). Blockade of Kir4.1 channels with VU likely inhibited a higher number of Kir4.1 channels and resulted right away in a higher and excessive [K⁺]_e increase compared with RACK1fl/fl mice that rapidly leaded to significant neuronal depression.

Kir4.1 KO in astrocytes induces tonic-clonic seizure and premature death,^38,40 and pathophysiological alterations of Kir4.1 have been reported in various animal models of epilepsy.⁴⁴ Thus, enhancement of Kir4.1 expression and the associated decreased burst frequency induced by RACK1 cKO in astrocytes could ameliorate epileptogenic processes. We compared the epileptic activity of RACK1 fl/fl and RACK1 cKO mice upon injection of pentylenetetrazole (PTZ) or pilocarpine, which induce acute seizures^45,46,47 (Figures S4A and S4E). No significant difference in the dose for PTZ-inducing tonic seizures (Figure S4B), tonic-clonic seizures (Figure S4C), and death (Figure S4D) was observed. No difference in the latency before first seizure (Figure S4F), first seizure duration (Figure S4G), and time of survival (Figure S4H) was observed upon pilocarine injection in mice of both genotypes.

These results indicate that, by controlling Kir4.1 expression and the associated K⁺ buffering capacity in astrocytes, RACK1 helps to modulate the firing rate when neuronal activity is sustained. However, this effect is not sufficient to increase the seizure threshold in acute models.

RACK1 controls Kcnj10 translation through its 5′ untranslated region (5′ UTR)

We examined how RACK1 controls expression of Kir4.1. RACK1 is an essential factor in translation and ribosome quality control. It senses ribosome stalling on rare codons (such as CGA, coding for arginine, and AAA, coding for lysine) and can cause translation elongation to pause or abort. The absence of RACK1 results in more frequent translation of mRNAs with stalling sequences and, eventually, accumulation of peptides with frameshifts.²¹ Because we observed RACK1-dependent downregulation of Kir4.1, we first hypothesized that the Kcnj10 gene’s coding sequence (CDS) is subject to a stalling event that can only be resolved by RACK1. We generated human embryonic kidney 293T (HEK293T) cells in which RACK1 expression was disrupted through a CRISPR-Cas9-based strategy (RACK1^KO cells; Figure 7A). We then designed a dual-fluorescence reporter system in which GFP and mCherry fluorescent proteins were expressed in frame from a single mRNA and were separated by the Kcnj10 CDS (Figure 7B). The Kcnj10 CDS was insulated with viral P2A sequences, where ribosomes skip formation of a peptide bond without interrupting elongation.⁴⁸ Complete translation of this cassette generates three proteins (GFP, mKir4.1, and mCherry). The presence of any stall-inducing sequences in Kcnj10 CDS would modify the translation rate prior to mCherry synthesis and would result in a sub-stoichiometric mCherry:GFP ratio. A reporter without Kcnj10 CDS served as a negative control. The positive control was a reporter in which the Kcnj10 CDS had been replaced by a sequence containing a stretch of consecutive lysine AAA codons (termed K20) and that was known to induce ribosome stalling (Figure S5A).⁴⁹ These reporters were expressed in wild-type (WT) and RACK1^KO cells, and the mCherry:GFP ratio was measured at the single-cell level using fluorescence-activated cell sorting (Figures 7C and S5B). RACK1^KO cells displayed a robust elevation of the mCherry level expressed downstream of the K20 sequence, confirming that loss of RACK1 impairs ribosome stalling (Figures S5B and S5C). In contrast, the absence of RACK1 did not modify the mCherry/GFP ratio of the reporter cassette containing Kcnj10 CDS compared with WT cells (Figures 7C and 7D). These data indicate that the sensitivity of the Kcnj10 mRNA with regard to RACK1 is not mediated by a RACK1-modulated ribosomal event involving its CDS.

Figure 7.

RACK1 controls translation of Kcnj10 through its 5′ UTR

(A) CRISPR-Cas9-based KO of RACK1 in HEK293T cells. The western blot for the indicated proteins was performed using WT and RACK1^KO cell extracts. The position of molecular weight markers is indicated on the right.

(B) Topology of the reporters for flow cytometry analysis of Kcnj10 mRNA translation. The constructs contain GFP and mCherry separated by a multiple cloning site (into which the Kcnj10 CDS had been inserted) and two viral 2A sequences (where ribosomes skip formation of a peptide bond without interrupting chain elongation).

(C) Representative flow cytometry-based assay of the fluorescence ratio in WT or RACK1^KO HEK293T cells transfected by the constructs described in (B).

(D) Histogram of the data presented as the mean ± SD (N = 3); two-tailed t test with Welch’s correction.

(E) Schematic of the Renilla luciferase (RLuc) reporter constructs harboring the 5′ UTR and/or 3′ UTR of mouse Kcnj10 mRNA. Sequences were inserted into the psiCHECK2 vector, which also encodes the firefly luciferase (FLuc). The Ct RLuc was generated by transfecting the empty psiCHECK2 vector.

(F) WT and RACK1^KO HEK293T cells were transfected with the reporters described in (E). Luciferase activity was measured 24 h after transfection. RLuc values were normalized against FLuc levels, and a ratio was calculated for each Kcnj10 reporter relative to the empty psiCHECK2 reporter (value set to 1) for each population. shown is a histogram of the data, presented as the mean ± SD (N = 4); unpaired Mann-Whitney test or unpaired t test.

(G) Schematic of the truncated versions of the Kcnj10 5′ UTR inserted in the RLuc reporter. Blue boxes indicate GC-rich regions (Figure S5E). Plasmids were transfected in WT and RACK1^KO HEK293T cells. For each experiment, the ratio was calculated for each Kcnj10 reporter relative to the empty psiCHECK2 reporter (value set to 1).

(H) A histogram of the data, presented as the mean ± SD (N = 4); unpaired Mann-Whitney test. ns, p > 0.05; ^∗p < 0.05; ^∗∗p < 0.01; ^∗∗∗p < 0.001; ^∗∗∗∗p < 0.0001. The raw data are presented in Table S3.

We next sought to determine whether Kcnj10’s sensitivity to RACK1 was conferred by its 5′ UTR. Two distinct 5′ UTRs have been reported for Kcnj10 in the mouse (NM_001039484.1 and AB039879.1, hereafter referred to, respectively, as 5′ UTR#1 and 5′ UTR#2). 5′ UTR#1 is composed of a GC-rich first half (region 1–104, not found in other mammalian Kcnj10 orthologs) and a highly conserved second half (the 147–242 region, which is shared with 5′ UTR#2) (Figures S5D and S5E). 5′ UTR#1 and 5′ UTR#2 were inserted into the psiCHECK-2 luciferase reporter vector downstream of the Renilla luciferase (RLuc) CDS (Figure 7E). A reporter harboring the 3′ UTR of Kcnj10 upstream of RLuc was also constructed as a control. These reporters were transfected into WT and RACK1^KO HEK293T cells, and the effect of RACK1 loss on RLuc activity for each UTR construct was calculated relative to an empty psiCHECK2 reporter level (control RLuc). RLuc activity was normalized against the activity of the co-expressed firefly luciferase (FLuc) (Figure 7F). We detected significantly greater RLuc activity when the RLuc reporters harboring Kcnj10 5′ UTRs were expressed in RACK1^KO cells (relative to expression in WT cells) (Figure 7F). In contrast, no difference between WT and RACK1^KO cells was observed for the RLuc-Kcnj10 3′ UTR construct, indicating that control by RACK1 was mediated by Kcnj10 5′ UTRs (Figure 7F). A qPCR analysis did not show significant differences in RLuc mRNA levels under any conditions, which confirmed that the Kcnj10 5′ UTR-mediated effect on RLuc activity was only translational (Figure S5G). Because 5′ UTR#1 and 5′ UTR#2 share a common 96-nt region (Figure S5D), we hypothesized that this sequence confers RACK1-dependent translation control. To test this hypothesis, we truncated the 242-nt-long Kcnj10 5′ UTR#1 into five overlapping fragments, which were inserted upstream of the RLuc sequence and expressed in WT and RACK1^KO cells (Figure 7G). The 127–242 region and the shorter 181–242 region were sufficient to increase the RLuc activity in RACK1^KO cells (relative to WT cells), whereas the first half (region 1–146) did not confer RACK1 sensitivity (Figures 7G and 7H).

Thus, RACK1 controls Kcnj10 translation, and this regulation depends on a portion of the Kcnj10 5′ UTR rather than its CDS or 3′ UTR.

Collectively, our results show that RACK1 associates with specific mRNAs in astrocytes and, in particular, represses translation of Kcnj10 mRNA. This translational effect is mediated by the Kcnj10 gene’s 5′ UTR. RACK1 cKO in astrocytes is associated with higher Kir4.1 levels in astrocytes; in turn, this increases astrocyte K⁺ currents and volume and attenuates recurrent neuronal burst activity.

Discussion

The objective of the present study was to investigate the molecular mechanisms that regulate translation in astrocytes. We focused on the 40S-associated protein RACK1, a critical factor in translational regulation.^18,50 We demonstrated that RACK1 interacts with specific mRNAs in astrocytes and PAPs, represses translation of Kcnj10 (encoding Kir4.1), and impacts astrocyte K⁺ currents, volume, and neurotransmission.

The TRAP technique used to purify astrocyte polysomes was originally developed for analysis of polysomal mRNAs.⁵¹ Here, we demonstrated that TRAP was compatible with MS. TRAP was performed following a refined protocol that considerably reduces background noise.¹⁷ Importantly, the Aldh1l1 promoter used to activate the TRAP gene in astrocytes is active in neurogenic areas.⁵² Thus, neural stem cells might also contribute to our polysome-associated proteome. As expected, the most abundant immunopurified proteins were ribosomal or translation complex-associated proteins, although other proteins were also identified that might be candidates for regulation of translation in astrocytes. The most highly represented cytoskeleton-associated proteins in our screen included Ckap4 (CLIMP63), an endoplasmic reticulum (ER) integral membrane protein that binds to microtubules and promotes ER tubule elongation.⁵³ Ckap4 has a crucial role in dendritic organization of the ER in neurons.⁵⁴ The cytoplasmic linker-associated protein CLASP2 mediates asymmetric microtubule nucleation in the Golgi apparatus and is crucial for establishing the latter’s continuity and shape.⁵⁵ CLASP2 has been shown to underlie microtubule stabilization, neuronal polarity, and synapse formation and activity.⁵⁶ Our experiments did not pinpoint all known RBPs in astrocytes; for instance, we did not detect Qki, which has been shown recently to regulate translation in astrocytes.^9,57 Thus, the TRAP-MS technique probably does not give a comprehensive view of the ribosome-associated proteome in the astrocyte. However, it is cell specific and so might be a powerful approach for identifying some of the key post-transcriptional regulators in astrocytes.

Among the astrocytic ribosome-associated proteins identified in our screen, we focused on the highly enriched RACK1. Interestingly, use of a selection of astrocyte-specific mRNAs enabled us to determine the preferential association of RACK1 with Kcnj10 and Slc1a2 mRNAs. This finding indicated that RACK1-containing ribosomes associate with specific mRNAs in astrocytes and probably confer specific translational properties on the ribosomes. Along the same lines, it has been shown that ribosomes with different stoichiometries of RACK1 translate different subsets of mRNAs.⁵⁸ RACK1 has also been described recently as one of the ribosomal proteins translated in neurites and able to rapidly go on and off the ribosomes in neurons, suggesting that RACK1 is involved in ribosome filtering mechanisms¹¹ and might control translation in a dynamic mode.⁵⁹ It remains to be seen how the interaction between RACK1-containing ribosomes and specific mRNAs is achieved, but various elements might be involved in this process. RACK1 has been shown to discriminate between mRNAs according to their length and to promote translation of mRNAs with a short open reading frame.⁶⁰ In viruses, RACK1 might mediate translation of mRNAs with an internal ribosome entry site.⁶¹ Other studies demonstrated that RACK1 controls translation by sensing 5′ UTR sequences,⁶² as shown here for the 5′ UTR of Kcnj10.

Astrocyte-selective RACK1 KO led to higher levels of Kir4.1. Our ex vivo analysis demonstrated that this effect was linked to RACK1 repression of Kcnj10 translation. RACK1 has been shown to control important aspects of ribosome quality control by sensing stalled ribosomes on polyarginine or proline codons and contributing to degradation of nascent protein chains on stalled ribosomes.^21,63,64,65 We further confirmed this effect on a lysine AAA sequence. However, we showed that this mechanism does not operate on the Kcnj10 CDS. In contrast, we showed that the RACK1-mediated control of Kcnj10 translation relied on specific 5′ UTR sequences. It now remains to be determined how this regulatory mechanism operates. Several mechanisms might be involved. Indeed, RACK1 recruits and controls the activity of translation factors, like the elongation factor eIF6.⁶⁶ RACK1 interacts with components of the microRNA-induced gene silencing complex.⁶⁷ Baum et al.⁶⁸ suggested that recruitment of the mRNA-binding protein Scp160 to the yeast homolog (Asc1p) of RACK1 may influence the translation of specific mRNAs. Thus, changes to the translation machinery recruited on the Kcnj10 5′ UTR might occur in the absence of RACK1, which would change the ribosomal translational efficiency.

We previously demonstrated that the RACK1-encoding gene Gnb2l1 is highly translated in PAPs.¹⁶ We also determined that Kcnj10 polysomal mRNAs are present in PAPs.¹⁶ Here, we found that RACK1 was associated with Kcnj10 mRNAs in astrocytes in general but also in PAPs in particular. Moreover, we found that PAP levels of Kir4.1 were higher in the absence of RACK1. Upregulation of Kir4.1 in the absence of RACK1 was more pronounced in the cortex compared with the hippocampus and was also detected in cortical PvAPs. These observations suggest that regulation of Kcnj10 translation by RACK1 is not uniform in astrocytes.

Kir4.1 is a weakly inwardly rectifying K⁺ channel; in astrocytes, it helps to maintain the resting membrane potential, high K⁺ conductance, volume regulation, and glutamate uptake.^{38,39,49,69,70,71,72} During neuronal activity, neurons release large amounts of K⁺ at the synapse. K⁺ is rapidly taken up by astrocytic Kir4.1 and transported through the astrocytic network to regions with lower K⁺ levels. This clearance mechanism is vital for maintenance of K⁺ homeostasis and prevention of neuronal hyperexcitability. Here, we showed that Kir4.1 overexpression in RACK1 cKO mice led to bigger K⁺ currents in response to sustained stimulation and increased astrocytic volumes. Regarding neuronal activity, it did not modify basal excitatory synaptic transmission but was critical during sustained activity. We observed greater facilitation and slower depression after repetitive stimulation in RACK1 cKO mice. This effect was abolished upon addition of a specific Kir4.1 blocker, demonstrating that the greater facilitation and slower depression observed in RACK1 cKO were related to upregulation of Kir4.1. These results are consistent with previous reports of a role of Kir4.1 in the 3- to 10-Hz frequency band but not in the baseline activity (0.1 Hz).^38,39 Regarding the neuronal network burst activity under pro-epileptic conditions (with 0Mg6K ACSF), it was lower in RACK1 cKO than in RACK1 fl/fl mice. These results were in line with the fact that upregulation of Kir4.1 in RACK1 cKO leads to more efficient buffering of extracellular K⁺, allowing better control of burst frequency. Under WT conditions, when Kir4.1 channels were specifically blocked with VU, burst frequency increased as K⁺ accumulated in the extracellular space and depolarized neurons, thus increasing their excitability. In contrast, in RACK1 cKO, we did not observe the increase in burst frequency. Blockade of Kir4.1 channels with VU in RACK1 cKO slices rapidly caused a higher extracellular K⁺ increase, leading to neuronal depression. Altogether, our finding suggests that astrocytes more efficiently buffer extracellular K⁺ through higher levels of Kir4.1, which enhanced the RACK1 cKO mice’s ability to control extracellular K⁺ levels and bursting. These data indicate that, by modulating Kir4.1 expression, RACK1 influences neurotransmission.

Kir4.1 dysfunction has been illustrated in wide array of neurological diseases.^44,73,74,75 Some of these pathological contexts are also associated with RACK1 alteration, suggesting involvement of the RACK1/Kir4.1 pathway. In the amyloid precursor protein-Presenilin 1 (APP/PS1) Alzheimer’s disease mouse model, a localized increase in Kir4.1 has been reported recently in astrocytes proximal to β-amyloid (Aβ) deposits, while RACK1 was significantly decreased in the brain of aging animals and AD patients.^25,76 Traumatic brain injury causes RACK1 overexpression and Kir4.1 downregulation in the pericontusional cortex.^77,78 Similar observations were made in the cortex of aged rodents.^76,79 Reduction of Kir4.1 expression is commonly observed in murine and human epileptic tissues,^44,73 while RACK1 levels in the hippocampus are elevated after epileptic episodes in the lithium-pilocarpine rat model.²⁷ In a rat model of mesial temporal lobe epilepsy, RACK1 was upregulated in the granular layer dorsal dentate gyrus and downregulated in the ventral dentate gyrus.²⁸ Here, we tested whether elevation of Kir4.1 in RACK1 cKO could ameliorate epileptogenesis. Our results indicated that it was not able to rescue PTZ- or pilocarpine-induced epilepsy. This might be due to the fact that upregulation of Kir4.1 is not sufficient to counteract the huge hyperexcitability induced by PTZ or pilocarpine. On the other hand, Kir4.1 downregulation might not be a driving mechanism in epilepsy but, rather, a pathological phenomenon associated with astrocytic reactivity.^44,73 Considering the number of pathological situations in which Kir4.1 functions are altered, and given the fact that there are very few known expression enhancer of Kir4.1,^44,73 regulation of Kir4.1 translation by RACK1 might nevertheless be a valuable candidate mechanism for future therapy.

Limitations of the study

The TRAP MS method does not provide a complete view of the astrocytic ribosome-associated proteome. Additional strategies to stabilize the polysomal protein complex might help to identify more proteins. It is not yet known whether TRAP can distinguish between monosomes and polysomes in the brain. Thus, the pool of proteins identified in our study could be related to specific types of transcripts. Because RACK1 is ubiquitously expressed, we limited our study to a short list of astrocyte-specific mRNAs, whereas RACK1-associated ribosomes are likely to bind other mRNAs. One way to further explore this question would be to develop an astrocyte-specific labeled version of RACK1. Our in vitro analysis demonstrates that RACK1 is sensitive to a specific portion of the Kcnj10 5′ UTR. We did not determine whether this sensitivity is related to a specific motif or a 3D structure of the mRNA. One way to address this question would be to identify more RACK1-sensitive mRNAs in astrocytes and compare their sequences. We tested the possible beneficial effect of Kir4.1 upregulation in two acute models of epilepsy because Kir4.1 downregulation is a hallmark of epilepsy, but no effect was found. It is possible that Kir4.1 upregulation in RACK1cKO may rescue epileptogenesis in a more chronic context. Finally, we studied constitutive in vitro and in vivo deletion models of RACK1, which did not allow determination of whether RACK1 could dynamically regulate Kcnj10 translation.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Antibodies
Chicken polyclonal anti-GFP	Aves	Cat#GFP-1020; RRID: AB_10000240
Rabbit polyclonal anti-GLT1	Frontier Institute	Cat#Rb-Af670; RRID: AB_2571718
Rabbit polyclonal anti-KIR4.1	Alomone labs	Cat#APC-035; RRID: AB_2040120
Rabbit polyclonal anti-KIR4.1 extracellular	Alomone labs	Cat#APC-165; RRID: AB_2341043
Mouse monoclonal anti-RACK1 (B-3)	Santa Cruz Biotechnology	Cat#sc-17754; RRID: AB_2247471
Rabbit monoclonal anti-RPS6	Cell signaling Technology	Cat#2217; RRID: AB_331355
Mouse monoclonal anti-RACK1	BD transduction laboratories	Cat#610177; RRID: AB_397576
Rabbit polyclonal anti-GFAP	Sigma-Aldrich	Cat#G9269; RRID: AB_477035
Peroxidase AffiniPure Goat anti-chicken IgY (IgG) (H + L)	Jackson ImmunoResearch	Cat#103-035-155; RRID: AB_2337381
Goat anti-rabbit IgG (H + L) – HRP	CliniSciences	RRID: AB_CSA211
Goat anti-mouse IgG (H + L) – HRP	CliniSciences	Cat#RRID: AB_CSA2108
EasyBlot anti Mouse IgG (HRP)	GeneTex	Cat# GTX221667-01; RRID: AB_10728926
EasyBlot anti Rabbit IgG (HRP)	GeneTex	Cat#GTX221666-01; RRID: AB_10620421
Goat anti-Rabbit IgG (H + L) Alexa Fluor 488	Invitrogen	Cat#A-11034; RRID: AB_2576217
Goat anti-Mouse IgG (H + L) Alexa Fluor 555	Invitrogen	Cat#A-21424; RRID: AB_141780
Mouse IgG Isotype Control	Thermo Fisher Scientific	Cat#10400C; RRID: AB_2532980
Bacterial and virus strains
pAAV2/9-GFA-Bglob-TdTomato	This paper	N/A
Chemicals, peptides, and recombinant proteins
VU 0134992 KIR4.1 Blocker	Biotechne	Cat#6877
Pentylenetetrazole	Sigma-Aldrich	Cat#P6500
Pilocarpine hydrochloride	Sigma-Aldrich	Cat#P6503
Critical commercial assays
RNAscope® Multiplex Fluorescent Detection Kit v2	Biotechne	Cat#323110
Deposited data
Mass spectrometry data	This paper	http://www.ebi.ac.uk/pride; PXD033121
Experimental models: Cell lines
Human: HEK-293 cells	ATCC	Cat# PTA-4488; RRID: CVCL_0045)
Experimental models: Organisms/strains
Mouse: C57BL/6J	Janvier labs	SC-C57J-M
Mouse: Tg(Aldh1l1-eGFP/Rpl10a) JD130Htz	The Jackson Laboratory	MGI:5496674; RRID: IMSR_JAX:030247
Mouse: B6J.Cg-Rack1tm1.1Cart/Mmucd Rack1fl/fl	The Jackson Laboratory	MGI: 6343619
Mouse: B6N.FVB-Tg(Aldh1l1-cre/ERT2)1Khakh/J Aldh1l1-CreERT2	The Jackson Laboratory	MGI: 5806593; RRID: MGI:5911538
Oligonucleotides
Taqman probe Slc1a2	Thermofisher	Mm01275814_m1
Taqman probe Gjb6	Thermofisher	Mm00433661_s1
Taqman probe Gja1	Thermofisher	Mm01179639_s1
Taqman probe Slc1a3	Thermofisher	Mm00600697_m1
Taqman probe Aqp4	Thermofisher	Mm00802131_m1
Taqman probe Kcnj10	Thermofisher	Mm00445028_m1
Taqman probe Gnb2l1	Thermofisher	Mm01291968_g1
Taqman probe rRNA18S	Thermofisher	Mm03928990_g1
Software and algorithms
ImageJ	https://imagej.nih.gov/ij/	N/A
GraphPad Prism v8.0.3 (263)	GraphPad Software, Inc. version 8.0.3 (263)	N/A
IMARIS Bitplane	Oxford Instruments	N/A
MEA_Monitor v9.7.2	Multichannel Systems	N/A
MC_Rack v4.5.1	Multichannel Systems	N/A
Neuroexplorer v4.109	Nex Technologies	N/A
Other
FISH probe Mm-Gnb2l1-E1-E3	Bio-Techne	Cat#443621

Resource availability

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Martine Cohen-Salmon (martine.cohen-salmon@college-de-france.fr).

Materials availability

All plasmids generated for this manuscript are available from the lead contact upon request.

Experimental model and subject details

Animal

Tg (Aldh1l1-eGFP/Rpl10a) JD130Htz (MGI: 5496674) (Aldh1l1:L10a-eGFP) mice were obtained from Nathaniel Heintz’s laboratory (Rockefeller University, New York City, NY) and kept under pathogen-free conditions⁸⁰ (www.bactrap.org). Tg(Aldh1l1-cre/ERT2)1Khakh (MGI:5806568) (Aldh1l1-Cre/ERT2) mice⁸¹ were obtained from the Jackson laboratory (https://www.jax.org/) and B6J.Cg-Rack1^tm1.1Cart/Mmucd (RACK1 fl/fl) (MMRRC 044021-UCD) from the mutant mouse resource and research center (MMRRC) (https://www.mmrrc.org/).⁸² C57BL/6 mice were purchased from Janvier Labs (Le Genest-Saint-Isle, France). Mice were maintained on a C57BL/6 genetic background.

Animal care and use

Mice were kept in pathogen-free conditions. All animal experiments were carried out in compliance with (i) the European Directive 2010/63/EU on the protection of animals used for scientific purposes and (ii) the guidelines issued by the French National Animal Care and Use Committee (reference: 2013/118). The study was also approved by the French Ministry for Research and Higher Education’s institutional review board (reference #21817; Key-Obs SAS N° 27). All experiments were performed on 2-month-old mice. Both sexes were used for all experiments except epilepsy tests which were carried out on males only.

TRAP-MS experiments were performed on 5 Aldh1l1:L10a-eGFP mice and 5 C57BL/6 mice (control) (1 mouse per sample). RNA immunoprecipitation against RACK1 was performed on 5 Aldh1l1:L10a-eGFP mice (1 mouse per sample) for total brain and 12 Aldh1l1:L10a-eGFP mice (2 mice per sample) for synaptogliosomes. RNA immunoprecipitation against GFP was performed on 4 Aldh1l1:L10a-eGFP mice (1 mouse per sample) for total brain and on 8 Aldh1l1:L10a-eGFP mice (2 mice per sample) for synaptogliosomes. Western blots against KIR4.1 and GLT1 in RACK1 cKO model were performed on 5 RACK1 fl/fl mice and 4 RACK1 cKO mice (1 mouse per sample) for brain, on 5 RACK1 fl/fl mice and 5 RACK1 cKO mice (1 mouse per sample) for hippocampus, on 8 RACK1 fl/fl mice and 9 RACK1 cKO mice (1 mouse per sample) for brain synaptogliosomes, on 5 RACK1 fl/fl mice and 5 RACK1 cKO mice (1 mouse per sample) for hippocampus synaptogliosomes, on 5 RACK1 fl/fl mice and 5 RACK1 cKO mice (1 mouse per sample) for brain microvessels, on 4 RACK1 fl/fl mice and 4 RACK1 cKO mice (1 mouse per sample) for cortex, on 4 RACK1 fl/fl mice and 4 RACK1 cKO mice (1 mouse per sample) for cortex synaptogliosomes and on 4 RACK1 fl/fl mice and 5 RACK1 cKO mice (1 mouse per sample) for cortex microvessels. Immunofluorescence against KIR4.1 in the cortex was performed on 9 images from 3 RACK1 fl/fl mice and on 15 images from 3 RACK1 cKO mice. I/V curves in patch clamp recordings were performed on 7 astrocytes from 6 RACK1 fl/fl mice and 10 astrocytes from 6 RACK1 cKO mice. Current experiments in patch clamp recordings were performed on 6 astrocytes from 4 RACK1 fl/fl and 5 astrocytes from 3 RACK1 cKO mice. Volume and Sholl analyses were performed on 45 astrocytes from 4 RACK1 fl/fl mice and 45 astrocytes from 4 RACK1 cKO mice. Input/output curves in field recordings were performed on 5 slices from 4 RACK1 fl/fl mice and on 5 slices from 5 RACK1 cKO mice. 10Hz stimulations in field recordings were performed on 5 slices from 5 RACK1 fl/fl mice and 6 slices from 4 RACK1 cKO mice. MEA experiments were performed on 15 slices from 5 RACK1 fl/fl mice and 18 slices from 6 RACK1 cKO mice. Pentylenetetrazol experiments were performed on 11 RACK1 fl/fl mice and 9 RACK1 cKO mice. Pilocarpine experiments were performed on 9 RACK1 fl/fl mice 8 RACK1 cKO mice.

Cell use

In vitro experiments were performed on HEK-293T cells.⁸³ Flow cytometry analyses of GFP/RFP reporters were performed 3 times. Each WT and RACK1 KO experiments was performed on 6-well plates. Luciferase tests were performed 4 times. Each WT and RACK1 KO experiment was performed on 24-well plates.

Method details

Tamoxifen induction of RACK1 inactivation

Two-month-old mice received a daily intraperitoneal injection of 100 mg/kg tamoxifen solution in corn oil (10 mg/mL dissolved extemporaneously for 6-8h at 37°C) for 5 consecutive days and were analyzed 3 weeks later. For controls, RACK1 fl/fl received corn oil only (immunofluorescence; Western blot; qPCR), or tamoxifen (astrocyte volume study; electrophysiology). No specific symptom was observed in RACK1 cKO compared to RACK1 fl/fl mice.

TRAP-MS

Whole brain homogenates (one brain per sample) from 2-month-old C57BL/6 mice (negative control) and Aldh1l1:L10a-eGFP mice were submitted to TRAP by immunoprecipitating GFP-fused astrocytic polyribosomes with anti-GFP antibodies and protein-G-coupled magnetic beads.¹⁷ Here, 1 mg of proteins were used for the immunoprecipitation on 25 $\mu$ L G-protein-coupled magnetic Dynabeads coated with anti-GFP antibodies at 4°C. At the end of the procedure, immunoprecipitated proteins were eluted by boiling the beads in 20 μL of 0.35 M KCl buffer with 5X Laemmli buffer for 5 min. Samples were run on SDS-PAGE gels (Invitrogen) without separation as a clean-up step and then stained with colloidal blue staining (LabSafe GEL Blue G Biosciences). Gel slices were excised, and proteins were reduced with 10 mM DTT prior to alkylation with 55 mM iodoacetamide. After washing and shrinking the gel pieces with 100% acetonitrile, in-gel digestion was performed using 0.10 μg trypsin/Lys-C (Promega) overnight in 25 mM NH₄HCO₃ at 30°C. Peptides were then extracted (using 60/35/5 acetonitrile/H₂O/HCOOH) and vacuum concentrated to dryness. Peptides were reconstituted in injection buffer (0.3% TFA) before LC-MS/MS analysis. Five replicates per condition were prepared.

LC-MS/MS analysis

Online chromatography was performed with an RSLCnano system (Ultimate 3000, Thermo Scientific) coupled to a Q Exactive HF-X. Peptides were first trapped onto a C18 column (75 μm inner diameter × 2 cm; nanoViper Acclaim PepMapTM 100, Thermo Scientific) with buffer A (0.1% formic acid) at a flow rate of 2.5 μL/min over 4 min. The peptides were separated on a 50 cm × 75 μm C18 column (nanoViper C18, 3 μm, 100 Å, Acclaim PepMap RSLC, Thermo Scientific) at 50°C, with a linear gradient from 2% to 30% buffer B (100% acetonitrile, 0.1% formic acid) at a flow rate of 300 nL/min over 91 min. Full MS scans were performed with the ultrahigh-field Orbitrap mass analyzer in the range m/z 375–1500, with a resolution of 120,000 at m/z 200. The top 20 intense ions were subjected to Orbitrap for further fragmentation via high energy collision dissociation activation and a resolution of 15,000, with the intensity threshold kept at 1.3 × 10⁵. We selected ions with a charge from 2+ to 6+ for screening. The normalized collision energy was set to 27 and the dynamic exclusion was set to 40 s.

Data analysis

Data were searched against the Mus musculus UniProt canonical database (downloaded in August 2017 and containing 16888 sequences) using Sequest HT via proteome discoverer (version 2.0). The enzyme specificity was set to trypsin, and a maximum of two missed cleavage sites was allowed. Oxidized methionine, carbamidomethyled cysteine, and N-terminal acetylation were set as variable modifications. The maximum allowed mass deviation was set to 10 ppm for monoisotopic precursor ions and to 0.02 Da for MS/MS peaks. The resulting files were further processed using myProMS⁸⁴ version 3.9.3 (https://github.com/bioinfo-pf-curie/myproms). The false-discovery rate (FDR) was calculated using Percolator⁸⁵ and was set to 1% at the peptide level for the whole study. Label-free quantification was performed using peptide extracted ion chromatograms (XICs) computed with MassChroQ⁸⁶ v.2.2.1. For protein quantification, XICs from proteotypic peptides shared between compared conditions (TopN matching) with two-missed cleavages were used. Median and scale normalization at the peptide level was applied to the total signal, in order to correct the XICs in each biological replicate. To estimate the significance of the change in protein abundance, a linear model (adjusted for peptides and biological replicates) was used, and p values were adjusted using the Benjamini–Hochberg FDR procedure. Proteins with at least three total peptides in all replicates, a two-fold enrichment and an adjusted p value ≤0.05 were considered to be significantly enriched in sample comparisons. Proteins only found in one condition were also considered if they matched the peptide criteria. Proteins selected with these criteria were further analyzed and subjected to a GO functional enrichment analysis.

GO analysis

A GO analysis was performed for proteins with at least three peptides read by LC-MS/MS and found to be enriched in Aldh1l1:L10a-eGFP extracts (p value <0.05 and Log2 FC > 1), using UniProt bank annotations for the mouse (UniProt-GOA Mouse - Mus musculus). GO-term-associated p values were computed with the GOTermFinder module of myProMS.⁸⁴ We analyzed biological processes and molecular functions (p value threshold: 0.05). For each family, GO terms were classified manually according to the GO hierarchy, taking into account the number of genes from the study included in the highest GO. For instance, the in the “Gene Expression” category were included in the highest GO “Metabolic process”, and the proteins in the ‘Ribosomal proteins’ category were included in “Structural molecule activity”. The number of proteins in each category was expressed as a percentage of the total number of proteins. These data should be taken as illustrative because some proteins have more than one role and so the categories overlap.

RACK1 immunoprecipitation (IP)

IP was performed according to the TRAP-MS protocol (i.e. using an anti-RACK1 antibody) but with some changes, as follows. Columns (bead volumes: 100 $\mu$ L for the precleaning column, 25 $\mu$ L for precleaning + IgG column, 25 $\mu$ L for IP column) were prepared the day before. The IP column was first blocked 1 h with 2% bovine serum albumin and 0.1 mg/100 μL beads of yeast tRNA in 0.15 M KCl buffer, rinsed with 0.15 M KCl three times and coated with 5 $\mu$ g of anti-RACK1 antibodies or 5 $\mu$ g of non-specific immunoglobulins IgG (negative control). 500 $\mu$ g of protein extract was used. The precleaning steps are detailed in the STAR protocol.¹⁷ The precleaned extract was incubated with IP columns for 30 min at 4°C. The beads were rinsed three times with 0.35 M KCl and RNA were eluted in 300 $\mu$ L RLT buffer (Qiagen, Hilden, Germany) for 5 min at room temperature (RT) and kept at −80°C until extraction.

Quantitative RT-PCR

RNA was extracted using the Rneasy kit (Qiagen, Hilden, Germany). cDNA was then generated using the Superscript III Reverse Transcriptase kit (ThermoFisher). Differential levels of cDNA expression were measured using the droplet digital PCR (ddPCR) system (Bio-Rad) and TaqMan copy number assay probes or primers (Table S4). Briefly, cDNA and 6-carboxyfluorescein probes or primers were distributed into 10,000–20,000 droplets. The nucleic acids were then PCR-amplified in a thermal cycler and read (as the number of positive and negative droplets) with a QX200 ddPCR system. The results were normalized as follows: the IgG IP results were subtracted from the RACK1 RNA IP results for each gene. The results were then normalized against 18S rRNA gene expression. For GFP RNA IP (TRAP), results were normalized against the 18S rRNA.

FISH, immunofluorescence and confocal imaging

Mice were anesthetized with a mix of ketamine/xylazine (80/100 mg/kg i.p.) and killed by transcardiac perfusion with PBS/PFA 4%. The brain was removed, incubated in 30% sucrose overnight, and cut into 40-μm-thick sections using a cryomicrotome (HM 450, Thermo Scientific). For long-term storage, slices were kept at −20°C in a cryoprotectant solution (30% ethylene glycol, 30% glycerol in PBS). FISH was performed using the v2 Multiplex RNAscope technique (Advanced Cell Diagnostics, Inc., Newark, CA, USA) and a specific probe against Gnb2l1 mRNAs according to the manufacturer’s instructions. Following the FISH procedure, slides were incubated with a blocking solution (0.2% normal goat serum, 0.375% Triton X-100, and 1 mg mL⁻¹ bovine serum albumin in 1X PBS) for 1 h at RT, incubated with the primary antibody overnight at 4°C, rinsed three times with 1X PBS, and incubated with the secondary antibody for 2h at RT. Lastly, the slides were washed three times in 1X PBS and mounted in Fluor mount and DAPI (Southern Biotech).

Immunohistochemical labeling was performed on frozen brain sections (see above) rinsed in PBS and incubated for 2 h at RT in blocking solution (5% normal goat serum, 0.5% Triton X-100 in PBS). Sections were incubated with primary antibodies diluted in the blocking solution overnight at 4°C, rinsed for 5 min in PBS three times, incubated with secondary antibodies diluted in blocking solution for 2 h at RT, rinsed for 5 min in PBS three times, and mounted in Fluoromount (Southern Biotech, Birmingham, AL). Brain sections were imaged on X1 or W1 spinning-disk confocal microscopes (Yokogawa). Images were acquired with a 40X oil immersion objective (Zeiss). For the astrocyte morphology study, an LSM 980 confocal (Zeiss) and a 63X oil immersion objective (Zeiss) were used.

The antibodies and RNA probe used in the present study are listed in key resource table.

Preparation of synaptogliosomes

All steps were performed at 4°C. Hippocampi (two per extract; 1 mouse) were dissected and homogenized with a tight glass homogenizer (20 strokes) in buffer solution (0.32 M sucrose and 10 mM HEPES in DNAse/RNAse-free water, with 0.5 mM DTT, protease inhibitors (cOmplete, EDTA free, 1 minitablet/10 mL), ribonuclease inhibitor (1 μL/mL), cycloheximide (CHX) 100 μg/mL freshly prepared). The homogenate was centrifuged at 900 g for 15 min. The pellet was discarded, and the supernatant was centrifuged at 16,000 g for 15 min. The new supernatant was discarded, and the pellet (containing synaptogliosomes) was diluted in 600 μL of buffer solution and centrifuged again at 16,000 g for 15 min. The final pellet contained the synaptogliosomes.

Western blots

Brain tissues were crushed with a pestle and a mortar at −80°C. Proteins were extracted from tissue powder or synaptogliosome or microvessels pellets in 2% SDS (500, 200 or 50 μL per sample, respectively) with EDTA-free Complete Protease Inhibitor (Roche), sonicated twice for 5 min, once for 5 min or for 90 s, respectively (Bioruptor UCD 200, diagenode), and centrifuged at 20,000 g for 20 min at 4°C. Supernatants were heated at 56°C in Laemmli loading buffer containing DTT for 5 min. Protein content was measured using the Pierce 660 nm protein assay reagent (Thermo Scientific) and the Multiskan FC spectrophotometer (Thermo Scientific). Equal amounts of protein (whole immunoprecipitation extracts, 10 to 20 μg for whole tissue, synaptogliosomes and microvessels) were separated by denaturing electrophoresis in Mini-Protean TGX stain-free gels (Biorad) and then electrotransferred to nitrocellulose membranes using the Trans-blot Turbo Transfer System (Biorad). The antibodies used to hybridize the membranes are listed in key resource table. Horseradish peroxidase activity was visualized by enhanced chemiluminescence in a Western Lightning Plus system (PerkinElmer, Waltham, MA, USA). Chemiluminescent imaging was performed on a Fusion FX system (Vilber). The chemiluminescence signal intensity for each antibody was normalized against that of stain-free membranes, Actin or Histone3 accordingly.

Representative structure of the human 80S ribosome

The human 80S ribosome’s representative structure was depicted using the PyMol software (version 2.3.4, python 3.7, https://pymol.org/2/). A high-resolution cryo-electron microscopy (EM) structure of the human 80S ribosome (Natchiar et al., 2017) was obtained from the Protein DataBank in Europe (code 6EK0) because the mouse 80S ribosome is not available. The chain codes can be found on the Protein DataBank in Europe website. The Lz chain for RPL10a and the Sg chain for RACK1 were selected.

Preparation of microvessels

Brains were resuspended in HBSS/HEPES using an automated douncer at 4°C. After a first centrifugation at 2000 g for 10 min, the pellet was resuspended in HBSS/Dextran 18% and centrifuged at 4000 g for 15 min to separate myelin from the vessels. This new pellet contained the brain vessels on which PvAPs stay attached. It was resuspend in HBSS/BSA 1%. Filtration on a 20 μm-mesh filter allowed the retention of the brain microvessels.⁸⁷

Viral vectors and stereotaxic injection

Two-month-old mice were anesthetized with a mixture of ketamine (95 mg/kg; Merial) and xylazine (10 mg/kg; Bayer) in 0.9% NaCl and placed on a stereotaxic frame with constant body temperature monitoring. AAVs were diluted in PBS with 0.01% Pluronic F-68 at a concentration of 9 × 10¹² vg/ml and 1 μL of virus was injected bilaterally into the hippocampus at a rate of 0.1 μL/min, using a 29-gauge blunt-tip needle linked to a 2 μL Hamilton syringe (Phymep). The stereotaxic coordinates relative to the bregma were as follows: anteroposterior, ±2 mm; mediolateral: +1.5 mm; dorsoventral, −1.5 mm. The needle was left in place for 5 min and then removed slowly. The skin was glued back in place, and the animals’ recovery was checked regularly for the next 24 h. After 11 days, the mice were sacrificed and the tissues were processed for immunofluorescence assays.

Measurement of astrocyte volume

To drive expression in astrocytes, the transgene encoding cytosolic red fluorescent protein Td tomato was inserted under the control of the gfaABC1D⁴³ into an AAV shuttle plasmid containing the inverted terminal repeats of AAV2. Pseudotyped serotype 9 AAV particles were produced by transient co-transfection of HEK-293T cells.⁸³ Viral titers were determined by quantitative PCR amplification of the inverted terminal repeats on DNase-resistant particles and were expressed in vg per ml.

Astrocytes on 100 μm brain sections were reconstructed in 3D, using IMARIS software (Oxford Instruments, version 9.7.2). Filaments were created with a unique starting point in the astrocyte soma and with seeds defined with a manual threshold, according to the fluorescence intensity. Filaments outside the astrocyte were removed manually. An envelope of the astrocyte territory was created using the convex hull plugin (MATLAB). The following variables were computed and exported for analysis: astrocyte volume (corresponding to the envelope volume), the sum of the filament length and data for a 3D Sholl analysis (5 μm steps).

Cortical Kir4.1 immunofluorescence analysis

The immunofluorescence signals of Kir4.1 between RACK1 fl/fl and RACK1 cKO cortices was analyzed using ImageJ. A triangle thresholding method was applied to the stacks and the sum of intensity was performed to create a projection (removing the first and last slices). The integrated density (sum of pixel values in the image) was computed.

Electrophysiology

Electrophysiological recordings were performed in the hippocampus of 3-month-old RACK1 fl/fl (Control) and RACK1 cKO mice 3 weeks after tamoxifen injection, using ACSF in the presence or absence of a Kir4.1 blocker (30 μM VU0134992, Tocris, Biotechne).⁸⁸

Acute hippocampal slice preparation

Acute transverse hippocampal slices (400 μm) were prepared from 3-month-old RACK1 fl/fl or astrocytic RACK1 cKO mice.⁸⁹ Briefly, slices were cut at low speed (0.04 mm/s) and at a vibration frequency of 70 Hz in ice-cold oxygenated ACSF supplemented with sucrose (in mM: 87 NaCl, 2.5 KCl, 2.5 CaCl₂, 7 MgCl₂, 1 NaH₂PO₄, 25 NaHCO₃ and 10 glucose, saturated with 95% O₂ and 5% CO₂). Slices were then maintained at 32°C in a storage chamber containing standard ACSF (in mM: 119 NaCl, 2.5 KCl, 2.5 CaCl₂, 1.3 MgSO₄, 1 NaH2PO₄, 26.2 NaHCO₃ and 11 glucose, saturated with 95% O₂ and 5% CO₂), for at least 1 h prior to recording.

Patch-clamp recordings

To better visualize astrocytes for patch-clamp recordings, slices were loaded with sulforhodamine (1 μM, 20 min, 32°C, followed by 20 min wash in standard ACSF) before recording. Slices were then transferred to a submerged recording chamber mounted on an Olympus BX51WI microscope equipped for infrared-differential interference microscopy and were perfused with standard ACSF at a rate of 1–2 mL/min at 32°C. Whole-cell recordings were obtained from visually identified CA1 stratum radiatum astrocytes using 5–10 MΩ glass pipettes filled with 105 mM K-gluconate, 30 mM KCl, 10 mM HEPES, 10 mM phosphocreatine, 4 mM ATP-Mg, 0.3 mM GTP-Tris, and 0.3 mM EGTA (pH 7.4, 280 mOsm). CA1 astrocytes resting membrane potential, membrane and series resistance as well as membrane capacitance were monitored throughout the recordings. Evoked astroglial Kir4.1 currents were induced by stimulating Schaffer collaterals at 10 Hz for 1 s in the CA1 stratum radiatum.

Field recordings

Slices were transferred to a submerged recording chamber mounted on an Olympus BX51WI microscope equipped for infrared-differential interference microscopy and were perfused with standard ACSF at a rate of 1–2 mL/min at 32°C. Extracellular field recordings were performed with glass pipettes (2–5 MΩ) filled with ACSF and placed in the stratum radiatum. Stimulus artifacts were blanked in sample recordings. Basal excitatory synaptic transmission (input/output curves) was evaluated in presence of picrotoxin (100 μM), and the tissue was cut between CA1 and CA3 to prevent the propagation of epileptiform activity. Evoked postsynaptic responses were induced by stimulating SCs at 0.1 Hz in the CA1 stratum radiatum. Slices underwent prolonged, repetitive stimulation at 10 Hz for 30 s. Responses (neuronal fEPSP slope) were binned (bin size: 1.2 s) and normalized against the mean baseline response measured at 0.1 Hz prior to repetitive stimulation. Both basal excitatory synaptic transmission and responses to repetitive stimulation were evaluated before and after treatment with VU0134992.

Field potentials and patch-clamp recordings were acquired with Axopatch-1D amplifiers (Molecular Devices), digitized at 10 kHz, filtered at 2 kHz, and stored and analyzed on a computer using pCLAMP9 and Clampfit10 software (Molecular Devices).

MEA recordings

After a 20 min incubation in standard ACSF at 32°C, slices were stored for at least 1 h before recording in magnesium-free ACSF containing 6 mM KCl (0Mg6K ACSF) at 32°C. Hippocampal slices were then transferred onto planar MEA petri dishes (200–30 indium tin oxide electrodes, organized in a 12 × 12 matrix, with an internal reference, 30 μm diameter and 200 μm inter-electrode distance; Multichannel Systems), kept in place with a small platinum anchor, and continuously perfused at 1–2 mL/min with 0Mg6K ACSF at 32°C. Pictures of cortical slices on MEAs were acquired with a video microscope table (MEA-VMT1; Multichannel Systems). MEA_Monitor software (Multichannel Systems) was used to identify the location of the electrodes relative to the various regions of the hippocampal. Data were sampled at 10 kHz, and the slice activity was recorded at 32°C using a MEA2100-120 system (bandwidth: 1–3000 Hz; gain: 5x; Multichannel Systems) and MC_Rack 4.5.1 software (Multichannel Systems). The slices’ activity was recorded in 0Mg6K ACSF before and after treatment with 30 μM VU0134992. Raw data on 0Mg6K ACSF-induced network burst activity was analyzed with MC Rack software (Multichannel Systems). Bursts were detected with the Spike Sorter algorithm, which sets a threshold based on multiples of the standard deviation of the noise calculated over the first 500 ms of recording free of electrical activity. A 5-fold standard deviation threshold was used to automatically detect each event. If required (after a visual check), each event could be modified in real-time by the operator. Bursts were defined arbitrarily as discharges lasting less than 5 s. The bursts were characterized by fast voltage oscillations and then slow oscillations or negative shifts. The burst duration was measured using Neuroexplorer software (version 4.109, Nex Technologies, USA). Time-frequency plots were computed using Neuroexplorer (version 4.109, Nex Technologies, USA), setting the maximum frequency at 500 Hz and a time shift of 1 s. Values were normalised to the logarithm of the power spectral density (PSD) and expressed as dB. After the calculation the spectrum was smoothed with a Gaussian filter. In the figure, only frequencies between 1 and 100 Hz were displayed.

Induced seizure tests

The Penthylenetetrazole (PTZ) test: Mice were placed in a plastic cylinder (length, 12 cm, diameter 3.5 cm) where only limited movement is possible and leaving the tail of the mouse outside the cylinder. A needle was inserted into the lateral tail vein, fixed to the vein by a piece of adhesive. PTZ (pentylenetetrazole, Sigma-Aldrich, France, p6500) solution (10 mg/mL in 0.9% NaCl) was infused using a syringe pump at a concentration rate of 0.25 mL/min. The latency (in sec) of tonic seizure (startle), tonic-clonic seizure and death were recorded.^46,47

The pilocarpine test: Mice were intraperitoneally injected with a single dose of pilocarpine (MedChemExpress) solution (60 mg/kg in 0.9% NaCl) and seizure severity (latency before the first seizure: first seizure duration and time of death) was assessed immediately after injection.⁴⁵

Cell lines and culture conditions

HEK293T (Thermo Fisher Scientific, Waltham, MA) cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin/streptomycin (P/S) (Wisent Technologies). Control and RACK1 KO HEK293 cells were maintained in DMEM supplemented with 10% FBS, 1% P/S, 100 μg/mL zeocin (Thermo Fisher Scientific, R25001), and 15 μg/mL blasticidin (Thermo Fisher Scientific, R210-01). All cells were cultured at 37°C, in a humidified atmosphere with 5% CO₂.

Plasmid constructs

To generate the Kcnj10 CDS containing a fluorescent reporter, a control cassette was first created by replacing the BspEI/KpnI segment of the pmGFP-P2A-K0-P2A-RFP (Addgene plasmids 105686) with a linker containing a P2A site, a Flag coding sequence, and the EcoRI and NotI restriction sites. A gene block (Integrated DNA Technologies) encoding mKir4.1 (AAI41089.1) without a stop codon was then inserted at the EcoRI/NotI sites of this control cassette in frame with both the GFP and mCherry coding sequences. The psiCHECK-2 vector (Promega, C8021) was used to build the dual luciferase reporters with Kcnj10 UTRs. The UTR sequences of mouse Kcnj10 mRNA (NM_001039484.1 and AB039879.1) were synthesized as gBlocks and inserted at the NheI site of the psiCHECK-2 vector. The 3′UTR of the Kcnj10 mRNA (AB039879.1) was inserted as a gBlock into the XhoI and NotI restriction sites in the psiCHECK-2 vector downstream of the Renilla luciferase reporter gene. The truncated versions of the Kcnj10 5′ UTR (1–146; 127–242; 95–242; 95–191; 1–191; 181–242) were inserted as NheI/NheI PCR fragments into the psiCHECK-2 vector at the 5′ end of the Renilla luciferase gene. The sequences of the primers and gBlocks used for subcloning are listed in Table S4.

Flow cytometry analysis

Transient transfection of fluorescent reporter constructs was performed using Lipofectamine 2000 (Thermo Fisher Scientific), according to manufacturer’s instructions. In all experiments, WT and RACK1^KO HEK293T cells were plated in 6-well plates at a concentration of 500,000 cells per well, and transfected with 10 ng of plasmids on the following day. The cells were then trypsinized, washed once with PBS and pelleted at RT at 500 g for 5 min. The cells were resuspended in 500 μL PBS containing 10% FBS, passed through a 40 μm filter, and analyzed with a CytoFlex flow cytometer (Beckman Coulter). 10,000 fluorescent cells were selected for the analysis of GFP and mCherry signals. The data were analyzed using FlowJo software.

CRISPR/Cas9-mediated genome editing

CRISPR-Cas9-mediated genome editing of HEK293 cells was performed according to the protocol by Ran et al.⁹⁰ The DNA oligonucleotides (encoding a small guide RNAs (sgRNAs) cognate to the coding region of human Rack1/Gnb2l1 gene) are detailed in key resource table. These oligos contained BbsI restriction sites and were annealed to create overhangs for cloning of the guide sequence oligos into pSpCas9(BB)-2A-Puro (PX459) V2.0 (Addgene plasmid 62988) by BbsI digestion. To generate KO HEK293T cells, we transfected 500,000 cells with the guide sequence containing the pSpCas9(BB)−2A-puro plasmid. Twenty-four hours after transfection, puromycin was added to the cell medium. After 72 h, puromycin-resistant cells were isolated in 96-well plates and cultured until monoclonal colonies were obtained. Clonal cell populations were analyzed for protein depletion in Western blots.

Dual luciferase reporter assays

WT or RACK1^KO HEK293T cells were transfected with 20 ng per well of each psiCHECK2 construct or the empty psiCHECK2 in a 24-well plate by using Lipofectamine 2000 (Thermo Scientific, 11668019), according to the manufacturer’s instructions. Cells were lysed 24 h after transfection, and luciferase activities were measured with the Dual-Luciferase Reporter Assay System (Promega) in a GloMax 20/20 luminometer (Promega). The RLuc activity was normalized against the activity of co-expressed FLuc, and the normalized RLuc values were quoted relative to the corresponding control.

Quantification and statistical analysis

All statistical details, including the exact value of n, what n represents and which statistical test was performed are listed in Table S3 and can also be found in the figure legends or in the figures.All statistical analyses were performed using GraphPad Prism software (version 8.0.2, GraphPad Software, Inc.). The data are quoted as the mean ± SD. p values <0.05 were considered statistically significant.

Published: April 30, 2023

Data and code availability

• •MS proteomics data data have been deposited at ProteomeXchange Consortium PRIDE repository and are publicly available. Accession number is listed in the key resources table.
• •This paper does not report original code.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.