Genome-wide kinase-MAM interactome screening reveals the role of CK2A1 in MAM Ca2+ dynamics linked to DEE66

Significance The mitochondria-associated endoplasmic reticulum membrane (MAM) is a highly dynamic structure that serves as a signaling platform for a variety of cellular activities, including Ca2+ homeostasis. Using a kinome-wide screening for MAM structural alterations, we identify casein kinase 2 alpha 1 (CK2A1), a catalytic subunit of casein kinase 2, as a regulator of the MAM structure and MAM Ca2+ crosstalk via establishing the CK2A1–PACS2–PKD2 complex. PACS2 phosphorylation by CK2A1 affects the distribution of this complex at MAMs and PKD2-dependent Ca2+ homeostasis. Importantly, we demonstrate that PACS2 pathogenic mutations causing the developmental and epileptic encephalopathy-66 (DEE66) disorder are associated with the disruption of PACS2 phosphorylation by CK2A1 and dysregulation of MAM Ca2+ dynamics, suggesting a potential therapeutic route for DEE66-associated clinical conditions.

develop for 10-14 days. Immunoblotting was used to examine protein expression, and probably positive clones were validated using PCR cloning and sequencing of the targeted area from genomic DNA.

Human kinase-MAM interactome screening
HeLa cells were plated on 18 mm glass coverslips coated with poly-D-lysine (50 µg/mL) and allowed to grow to 50-70% confluency. Cells were transfected by lipofectamine 2000 (Invitrogen). Each kinase from the 408 kinase ORF library was transfected with two MAM-BiFC marker fragments and incubated for 48 h. Cells were fixed in 4% PFA (Sigma-Aldrich, #158127) in PBS for 10 min at room temperature before mounting in the fluorescence mounting medium (Agilent Technologies, #S302380-2). Four hundred-eight samples were blindly marked with the serial numbers and divided into 11 batches for imaging with 11 groups of accompanied negative controls. Cells in the negative control groups were treated identically, and transfected with empty vector and MAM-BiFC markers. Each batch contained at least 37 kinases and one negative control. At least 50 cells per sample were used to gather MAM-BiFC intensity. The positive candidates inducing MAM integrity were identified by positive log2 fold change (log2 FC), and the negative candidates causing MAM loosening were indicated by negative log2 FC.
The quality of samples was strictly controlled by checking the cell morphology and MAM-BiFC signals before imaging. Cell images were obtained by the FV3000 confocal laser scanning microscope (Olympus) using a UPLSAPO 60X/1.4 NA oil objective. All the images were changed to 8-bit images, subtracted background, and performed Otsu thresholding. Otsu automatic thresholding can highlight the area where pixel values are higher than the threshold. Then, the average pixel intensity over the threshold area without background was measured. Let us define as a log2-transformed intensity of sample in group ( = 1, … , , = 1, … , ). Each group corresponds to one of kinases in the context of this paper. We denote samples in control groups by , = 1, … , 0 , = 1, … , . is used to denote different control groups from different batches. For batch , let indicate the index set of kinases obtained in batch . The table below shows the data structure of our problem. Within each batch, we compared multiple kinases to a control group, which is often referred to many-to-one comparisons in statistics. Let us denote by 0 : 0 = the hypothesis for two-sample mean testing from control versus kinases . Then, the multiple comparison at batch is to test 0 ( ) = ⋂ 0 ∈ , based on the observations at the -th row. The testing can be performed by a modified Dunnett's test proposed by Herberich and colleagues (4).
Many-to-one comparisons and its hypothesis can be expressed as the general linear hypothesis under the one-way ANOVA model. Also, an asymptotic theory of mean estimators under the model is available even when group sizes are unequal, and heteroscedasticity exists. As a consequence, the virtue of this statistical procedure is being free from strong assumptions required for the classical Dunnett's test, i.e., Gaussianity, equal group sizes, and homogeneity across groups, which are usually non-realistic in the settings of biological experiments. The test provided (the number of kinases in batch ) P-values that are adjusted to control the family-wise error rate for 0 ( ) , not for the whole null hypothesis 0 = ⋂ 0 ( ) 1≤ ≤ . Therefore, we needed another adjustment of those P-values to satisfy the overall significance level of testing. In this additional adjustment, we simply relied on Bonferroni correction. In other words, we say the intensity of at least one of kinases in batch is significantly different from that of a control group (significance level at ) if P < / , where P is the individual P-value computed from batch . was set to be 0.05 in our study. The whole process was implemented using R multcomp package.

MAM proteome collection
A systemic search in PubMed was conducted to collect the studies that identified MAM proteome using the terms: "(mitochondria or mitochondrial) and (ER or endoplasmic reticulum) and (contact or contacts or mitochondria-associated membrane or mitochondria-associated membranes or MAM or MAMs or mitochondria-ER contacts or MERCs) and ("proteomic profiling" or "proteomic analysis" or "proteomic characterization" or proteome or proteomics or proteomic or labeling)". The literature search was finished on 1 June 2020. Then, the references to relevant articles were reviewed to identify relevant studies for further assessment. A study was qualified when it had the criteria below: (1) species (Homo sapiens, Mus musculus), (2) biospecimens (cell lines, tissues), (3) proteomics platform (mass spectrometry), and (4) well-validated MAM proteins confirmed by multiple experiments and reported by individual research articles. Data extraction was conducted according to the workflow as shown in Fig  S11. (1) Studies were separated by species (Homo sapiens, Mus musculus). (2) The majority of studies were conducted in Homo sapiens using mammalian cell lines (HEK293T, fibroblasts, PH5CH8, and Huh7). Another group included all studies using Mus musculus tissues (brain, retina, testis, liver). (3) Methods to enrich MAM proteins included Percoll gradient biochemical fractionation and proximity labeling (APEX, split APEX, split-BioID, contact-ID).

Subcellular fractionation of MAMs
Subcellular fractionation was carried out to isolate ER, MAM, or mitochondria as described previously (5,6). Briefly, harvested cells were rinsed with solution A (0.32 M sucrose, 1 mM NaHCO 3 , 1 mM MgCl 2 , 0.5 mM CaCl 2 , protease inhibitor cocktail). Cells were homogenized and centrifuged at 1400 g for 5 min at 4 °C. A part of the supernatant was preserved as a whole lysate fraction, while the remainder was centrifuged for 10 min at 13,800 g, 4 °C. The crude MAM pellet (MAM+Mito) was resuspended in solution B (0.32 M sucrose, 1 mM NaHCO 3 , protease inhibitor cocktail). The supernatant, on the other hand, was placed into a discontinuous sucrose gradient and centrifuged for 70 min at 152,000 g, 4 °C. The white band was collected using a syringe and centrifuged for 45 min at 126,000 g, 4 °C, with the pellet collected as the ER fraction. The preceding crude MAM pellet, resuspended in solution B, was put on a discontinuous sucrose gradient containing 1 mM NaHCO 3 and centrifuged for 2 h at 82,500 g, 4 °C. The third white band was collected with a syringe as the microsomal fraction, and the resultant pellet was reconstituted in an isolation medium (250 mM mannitol, 5 mM HEPES pH 7.4, 0.5 mM EGTA, 0.1% BSA) and put on top of Percoll gradient (225 mM mannitol, 25 mM HEPES pH 7.4, 1 mM EGTA, 30% Percoll). Upper and lower bands were collected as MAM and mitochondrial fraction, respectively.
MAM fractionation using mouse livers was conducted by differential ultracentrifugation using the Percoll gradient according to a previous report (6). All the fractions including whole liver lysate, cytosol, ER, pure mitochondria, and MAMs were collected, lysed in 1X modified RIPA lysis buffer, measured protein concentration using BCA assay (Thermo Fisher Scientific, # 23228), and subjected to immunoblotting experiments.

Immunoprecipitation assays
Cells were lysed in 1X Triton X lysis buffer (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% Triton X-100, 1 mM EDTA) supplemented with phosphatase inhibitor cocktail (Thermo Fisher Scientific, #A32957), and protease inhibitor cocktail (Thermo Fisher Scientific, #A32963). Lysates were incubated with 1 μg of antibody at 4 °C for 16 h with rotation. Protein A agarose beads (Roche, #5015979001) were rinsed three times with 1X Triton X lysis buffer and combined with immunoprecipitated lysates and rotated at 4 °C for at least 4 h. The beads were collected via centrifugation and washed three times with 1X Triton X lysis solution before being combined with the SDS sample buffer for carryout out the immunoblotting analysis.

Protein purification and in vitro kinase assay
Recombinant GST-fused PACS2 middle region (MR) fragments were expressed and purified from the BL21 strain of Escherichia coli using glutathione-sepharose affinity chromatography (GE Healthcare) as previously described (7). Each purified protein was incubated in the presence of purified recombinant CK2 purchased from NEB (Cat#P6010S). Reactions were carried out in a 1X NEBuffer™ for protein kinases (50 mM Tris-HCl, 10 mM MgCl2, 0.1 mM EDTA, 2 mM DTT, 0.01% Brij 35, pH 7.5) containing [γ-32 P] ATP (10 μCi) at 37 o C for 1 h and then terminated by adding 5X SDS sample buffer and boiling for 10 min. Samples were subjected to SDS-PAGE, stained by Coomassie Brilliant Blue, and dried, and then phosphorylated PACS2 fragments were detected by autoradiography.

Immunofluorescence and colocalization assay
Cells expressing Sec61β-mEmerald and TOM20-mScarlet or stained with Alexa 488 and Alexa 647-conjugated antibodies (Molecular Probes, dilution factor; 1:200) were analyzed using an FV3000 confocal laser scanning microscope (Olympus) with a UPLSAPO 60X/1.4 NA oil objective, 488-nm laser line and 543-nm / 647-nm laser line. Each channel's photomultiplier gain was adjusted to reduce background noise and saturated pixels. The 3D deconvolution was completed with Imaris software CUDA Deconvolution extension (Bitplain, Version 9.1.2). ImarisColoc was utilized to automate the selection of the thresholds, measure the number of co-localized voxels, and ROI that was co-localized, and calculate Manders' coefficient.

Single-molecule colocalization
Glass coverslips (18 mm) were prepared as previously mentioned (8). Then, coverslips were coated with poly-D-lysine (50 µg/mL) dissolved in PBS for 1 h before seeding SY5Y expressing MAM-BiFC. After 12 h, the cells were fixed with 4% paraformaldehyde in PBS for 10 min at room temperature, washed three times with PBS, permeabilized with 0.05% Triton X for 3 min, and then blocked with 4% BSA for 1 h. In order to identify endogenous CK2A1, cells were treated with primary antibodies at room temperature for 3 h, followed by indirect immunofluorescence labeling with Alexa Fluor 647-conjugated secondary antibodies. On a handmade objective-type total internal reflection fluorescence (TIRF) microscope constructed on an inverted microscope (IX-81, Olympus) with an XYZ automated stage, fluorescence imaging was performed (MS-2000, Applied Scientific Instrumentation). A 561-nm laser (YLK 6150T, Lasos) was aligned with an oil-immersion TIRF objective lens (APON 100X/1.49 NA, Olympus). An electron-multiplying charge-coupled device (EM-CCD) camera (iXon Ultra 897, Andor Technology) equipped with an adapter was used to capture images (TuCam, Andor Technology). To improve the magnification, a 1.6X amplifier, and a 1.43X tube lens were used. All instrument operations and data acquisitions were managed by MetaMorph (Molecular Devices) and MATLAB plug-ins that were developed specifically for this study (MathWorks). The eternity buffer was prepared as previously described (8) for inducing the photoblinking of Alexa Fluor 647, and its signal was excited using the 642nm laser (VFL-P-1000-642, MPB Communications) at 30 W/cm 2 with a frame rate of 0.05 Hz in the farred channel (654-870 nm). GFP signal was excited using a 488-nm laser (543-BS-A03, Melles Griot) at 10 W/cm 2 with a 0.05 Hz frame rate in the green channel (500-549 nm). Multiple particle detection method was performed based on U-track as previously described (9). To eliminate the false positives (MAM-BiFC with the random colocalized coordinate), we compared the distribution of the colocalization frequency of MAM-BiFC to their randomizing coordinates. The randomization function and quantitative analysis were written in Matlab (The MathWorks).

Electron microscopy
Cells were grown in 35 mm glass-bottomed culture dishes to 50%-60% confluency. Cells were fixed with 2 ml of the fixative solution containing 2% paraformaldehyde and 2.5% of glutaraldehyde diluted in sodium cacodylate buffer (pH 7.2) at 4 °C. After being washed, then post-fixed in 2% osmium tetroxide (OsO4) containing 1.5% potassium ferrocyanide for 1 h at 4 °C. The fixed cells were dehydrated using an ethanol series (50%, 60%, 70%, 80%, 90%, and 100%) for 20 min at each concentration and infiltrated with an embedding medium. Samples were cut horizontally to the plane of the block by 60 nm (Leica Microtome GmbH, Vienna, Austria) and were mounted on copper slot grids with a specimen support film and double-stained with lead nitrate and 2% uranyl acetate. The samples were then inspected using a transmission electron microscope Tecnai G2 (Thermo Fisher Scientific, Waltham, MA, USA). The 16 mosaic images per cell were analyzed using the lasso selection tool of Microscopy Image Browser (MIB) (10), ER, mitochondria, and MAM areas were selected and converted to masks corresponding to object identification. MAM was quantified using two different approaches as calculated the percentage of MAM length per total mitochondrial perimeter (11) and the distance between ER and mitochondrial membrane.

Proximity ligation assay
DuolinkTM in situ proximity ligation assay (PLA) was used to measure the interaction between VDAC1 and IP3R1 at the ER-mitochondria contacts, following a previously described protocol (12,13). Cells were fixed and blocked, then incubated overnight at 4 °C with primary antibodies against rabbit anti-VDAC1 (Proteintech, #55259-1-AP) and mouse anti-IP3R1 (Santa Cruz, #sc-271197). The PLA probes against the specific primary antibodies host were applied for 1 h at 37 °C. Then, ligase was added for 30 minutes at 37 °C, and amplification was completed. Lastly, cells were mounted after Hoechst staining. Images were obtained using an FV3000 confocal laser scanning microscope (Olympus) with a UPLSAPO 100X/1.35 NA oil objective. The PLA signals were detected by Texas red-labeled probes and indicated ER-mitochondrial contact points. The number of red dots as PLA plots per cell was quantified using Imaris software.

Calcium imaging
Cells were transfected with GCaMP6mt, GCaMP6-cyto, or RCEPIA1-er, along with respective expression constructs or shRNA plasmids for 72 h. Following transfection, the medium was replaced with extracellular buffer containing HEPES (pH 7.4; 25 mM), CaCl2 (2 mM), MgSO4 (1 mM), NaHCO3 (4 mM), and D-glucose (30 mM). Depending on the experiment's purpose, cells were treated with indicated reagents (250 μM histamine, 250 μM ATP, or 200 nM triptolide) and subsequently imaged at intervals of 2 s using FV3000 confocal laser scanning microscope (Olympus) with a UPLSAPO 20X/0.75 NA objective. For experiments involving permeabilized cells, cells were treated with 6 µM ionomycin in HBSS supplemented with HEPES (pH 7.4; 2.5 mM) for 3 min before imaging. After 10 s of imaging, cells were then applied 30 μM IP3 and imaged for another 145 s. Primary neurons transfected with GCaMP6mt or GCaMP6-cyto along with PACS2-related constructs were imaged in an extracellular buffer. Neurons were treated with 400 nM triptolide after baseline recordings. Images were analyzed using the Cellsense software (Olympus). The whole cell area or neuronal soma was selected by rectangle ROI by Cellsense, and average intensities were measured with background correction. After intensities were corrected for background subtraction, ΔF values were calculated from (F-Fo). F values displayed the intensities at peak while Fo values were defined by averaging 5 frames before stimulation and used for normalization.

TMRM imaging
Cells were incubated with 20 nM tetramethylrhodamine methyl ester (TMRM) (Invitrogen, # T668) in HEPES-buffered HBSS for 30 min at 37 °C. Measurements were obtained by using FV3000 confocal microscope with a UPLSAPO 20X/0.75 NA objective while keeping a bath of 5 nM TMRM in the imaging solution. TMRM was excited using the 543-nm laser line and fluorescence intensity was recorded. Basal mitochondrial membrane potential as well as dynamic changes upon application of oligomycin and FCCP were measured at a single focal plane. The intensity profile was collected by Cellsense software. TMRM was used in the redistribution mode, meaning that a reduction in TMRM fluorescence represents mitochondrial depolarization.

Mitochondrial ATP determination
Mito-AT1.03 probe-transfected cells were stimulated at 405 nm and CFP and YFP emissions were collected concurrently at 475 nm and 525 nm for FRET-based ATP determination in mitochondria. The YFP/CFP fluorescence intensity ratio was used to determine the mitochondrial ATP level in each cell.

SypHluorin imaging
To monitor presynaptic neurotransmitter vesicle release from axonal boutons, synaptophysin-pHluorin (sypHluorin) was co-expressed with PACS2 phospho-dead mutation (3A) or human DEE66associated PACS2 mutations (E209K, E211K) in primary glutamatergic neurons at DIV 15-16. KCl 50 mM was used to chemically induce synaptic activity. SypHluorin signals were analyzed by the Cellsense software (Olympus). After background subtraction, ΔF values were deduced from (F-Fo) where F displayed the intensities at peak following KCl stimulation while Fo indicated the baseline values.

Untargeted lipidomics analysis
SY5Y WT and CK2A1 KO cells were cultured as mentioned above, seeded into a 15 cm tissue culture dish at a density of 7 × 10 6 cells per plate, and harvested 72 h after seeding. Cells were collected and subjected to differential ultracentrifugation using the Percoll gradient to obtain the MAM fraction as described previously (6).
Approximately 20 mg of the fraction pellet was used for extracting lipids. In particular, lipids were extracted using the MTBE method (14,15). Firstly, 250 μL MeOH (-80 °C) was added to the sample and then homogenized using Precellys 24 at 5500 rpm x 15 s x 2, resting on dry ice for 2 min between each cycle. The sample was transferred to a new tube and 850 μL MTBE (-20 °C) was subsequently added and vortex for 60 s. The sample was shaken at 1500 rpm, 4 °C for 1 h. Next, 210 μL H2O was added, vortex for 60 s before additionally incubating on a shaker for 15 min at 4 °C. Next, samples were centrifuged at 16000 rcf x 10 min at 4 oC. The upper layer (400 μL x 2) was transferred to another tube and dried under nitrogen purge at room temperature until completely dry. Lipid extract was stored at -80 °C until analysis. For untargeted lipidomics analysis, the sample was suspended in 80 μL of MeOH:toluene (9:1 v/v), and a post-extraction pooled QC sample was created by mixing 20 μL of each sample.
Annotated lipid was used to further data analysis and visualization. Data were first filtered to remain only the peaks with a relative standard deviation (RSD) in QC less than 20% and then processed with median normalization and log transformation before statistical analysis. Principle component analysis (PCA) and heatmap were constructed using MetaboAnalyst 5.0 (18). Lipid class analysis was conducted on Lipidsig in two separated ion modes with the log transformation setting (19).

Statistical analysis
Quantitative data were presented as Mean and SD unless otherwise stated. The sample size varied depending on the properties of the experiments. Statistical analyses were conducted by either R 4.2 statistics or GraphPad Prism 9.3.1. A two-tailed independent-samples t-test was employed to determine whether the means of two independent groups were significantly different. One-way or two-way analysis of variance (ANOVA) and Tukey's post hoc test were applied to compare the means of more than two separate groups. As the level of statistical significance, a criterion of 0.05 was adopted unless otherwise stated. For lipidomic analysis, significant lipids were considered with an adjusted P-value (i.e., False-Discovery Rate) < 0.05.  Data presented in line graphs indicate mean ± SEM. Scatter plots show mean ± SD. Each experiment was repeated at least three times. One-way ANOVA with Tukey's post hoc test for multiple comparisons was used to determine statistical significance. ns, not significant; **P<0.01; ****P<0.0001.    immunofluorescence. Scale bar = 10 µm and 1 µm for higher magnification. Quantification of Mander's coefficient is shown in scatter plot. Total cells of each group: CK2A1 KO = 124, CK2A1 WT rescue = 123, CK2A1 K68M rescue = 126. One-way ANOVA with Tukey's post hoc test for multiple comparisons was used to determine statistical significance. ns, not significant; ****P<0.0001. (C-E) The indicated cells in panel A were transfected with organelle calcium sensors (GCaMP6mt, GCaMP6cyto, and R-CEIPIA1er, respectively) and subjected to monitoring for changes in the Ca 2+ sensor signal (ΔF) in mitochondria (C), cytosol (D), and ER (E) before and after triptolide stimulation. Total cells analyzed: GCaMP6mt (CK2A1 KO = 510; CK2A1 WT rescue = 516; CK2A1 K68M rescue = 527); GCaMP6cyto (CK2A1 KO = 416; CK2A1 WT rescue = 454; CK2A1 K68M rescue = 429); R-CEPIA1er (CK2A1 KO = 336; CK2A1 WT rescue = 348; CK2A1 K68M rescue = 337). (F-H) SY5Y cells expressing GCaMP6mt, or GCaMP6cyto, or R-CEPIA1er were pre-exposed to either DMSO or CX-4945 12 µM for 6 h to observe the changes of Ca 2+ flux in mitochondria (panel F), cytosol (panel G), and ER (panel H), respectively, after triptolide stimulation. More than 800 cells from each group were collected and analyzed. Statistical significance: ns, not significant; *P<0.05; **P<0.01; ****P<0.0001 by two-tailed unpaired Student's t-tests. In panel C-G, line graphs display mean ± SEM and scatter plots show the peak amplitude of ΔF/Fo as mean ± SD. Each experiment was performed at least three times.  . Line graphs present mean ± SEM. Data in scatter plots are displayed as mean ± SD. Each experiment was performed independently at least three times. Statistical significance was determined by one-way ANOVA with Tukey's post hoc test for multiple comparisons. ns, not significant; **P<0.01; ***P<0.001; ****P<0.0001.