Neurogranin-MYH9 interaction regulates cytoskeletal remodeling in cerebral vasculature

Abstract

Neurogranin (Ng), a known regulator of neuronal Ca²⁺-calmodulin (CaM) signaling, is linked to Alzheimer’s disease. Though well-studied in neurons, Ng is also expressed in brain vasculature, where its function remains unclear. To investigate Ng’s role in brain microvascular endothelial cells, we defined its interactome using immunoprecipitation-mass spectrometry (IP-MS) under high- and low-Ca²⁺ conditions. Among 119 Ng-binding proteins, we discovered a novel interaction between Ng and MYH9, a key regulator of cytoskeletal remodeling. Ng-MYH9 binding was prominent in high Ca²⁺ and validated via CaM affinity pulldown and proximity ligation assays. Ng knockdown reduced F-actin levels, while MYH9 knockdown decreased both Ng and F-actin. Loss of Ng-MYH9 also impaired AKT-GSK3β signaling and elevated the endothelial activation marker VCAM1. Ng-null mice exhibited disrupted brain microvascular architecture and reduced MYH9 expression in endothelial cells. These findings reveal a novel Ng pathway promoting MYH9-dependent cytoskeletal remodeling and a potential role in maintaining blood-brain barrier integrity, a previously unrecognized function for Ng in brain health and Alzheimer’s disease.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12987-025-00709-x.

Introduction

Neurogranin (Ng) is a Ca²⁺ sensing molecule in neurons [1], and its genetic variant significantly contributes to the development of cognitive decline and neurological disease [2]. Ng, known as calmodulin (CaM) binding protein in neurons [3], regulates synaptic plasticity and long-term potentiation through Ca²⁺-dependent signaling pathways [4]. Clinically, a loss of Ng expression in the brain has been associated with schizophrenia [2] and Alzheimer’s disease [5]. Moreover, Ng is also expressed in peripheral tissue, but its function and underlying molecular mechanisms remain unclear. In endothelial cells, Ng contributes to vascular remodeling by regulating nitric oxide (NO) production through Ca²⁺-independent protein kinase B (AKT)-mediated endothelial nitric-oxide synthase (eNOS) regulation [6]. In skeletal muscle cells, Ng expression regulates CaN-mediated myoblast fusion, which involves skeletal maintenance, growth, and remodeling [7]. A recent study also reported that Ng is expressed in brain endothelial cells, regulating blood-brain barrier (BBB) integrity. A loss of Ng expression in endothelial cells decreases AKT activation and increases BBB permeability [8].

Among possible Ng mechanisms, the most critical Ca²⁺ binding protein, CaM, is well known as a regulator of Ng and an amplifier of the Ca²⁺ signal. In mammalian cells, CaM is involved in numerous cellular processes, including cell growth, motility, proliferation, and apoptosis [9]. Notably, CaM modulates numerous binding partners by interacting with proteins structurally adapted to its different conformations, which change in response to Ca²⁺ levels [10, 11]. For example, at high intracellular Ca²⁺ concentrations, the Ca²⁺-CaM complex regulates cell morphology through interactions with Ca²⁺/CaM-dependent protein kinase II (CaMKII) [12] and activates eNOS in endothelial cells [13]. Conversely, in low-Ca²⁺ conditions, Apo-CaM (Ca²⁺-free calmodulin) exhibits a strong affinity for Ng, effectively inhibiting CaM function. Therefore, it is necessary to understand how changes in Ca²⁺ concentration can induce conformational changes in Ng-CaM interactions.

To investigate the mechanism of Ng in the brain vasculature, we assessed the Ca²⁺-dependent interactomes of Ng and CaM in the human cerebral microvascular endothelial cell line (hCMEC/D3). The hCMEC/D3 cell line serves as a model of the human blood-brain barrier (BBB) that can be used to study drug delivery and permeability mechanisms in the central nervous system (CNS). The BBB, formed by tightly packed endothelial cells, maintains the optimal environment required for neuronal function [14] and regulates brain homeostasis by controlling the influx and efflux of circulating biological substances, including energy metabolites, oxygen, metabolic waste products, blood-derived cells, pathogens, and toxins [15, 16]. BBB breakdown, characterized by a reduction in tight and adherens junction proteins essential for BBB integrity [17], permits the influx of neurotoxic molecules into the brain [18] and contributes to neurodegenerative disorders such as Alzheimer’s disease [19]. Given the critical role of Ng in Alzheimer’s disease pathophysiology, its expression at the BBB may influence Ca²⁺ dysregulation and BBB-mediated clearance of amyloid-β (Aβ) through transporter mechanisms.

To elucidate the role of Ng in hCMEC/D3 cell structure and function, we analyzed the Ng interactome after transfecting HA-tagged Ng into hCMEC/D3 cells. Immunoprecipitation-mass spectrometry (IP-MS) assays were performed under high- Ca²⁺ and low-Ca²⁺ conditions using an anti-HA-tag antibody. CaM pull-down assay was also performed to identify CaM interactions in hCMEC/D3 cells. Label-free proteomics analysis identified novel binding proteins of Ng. Heat-map investigation and ingenuity pathway analysis (IPA) suggested new signaling pathways. These Ng interactomes and pathways were validated using proximity ligation assay (PLA), Ng overexpression, Ng-siRNA, and Ng null mice. Our new finding provides a novel Ng-mediated cytoskeletal remodeling pathway in the brain endothelial cells that may ultimately contribute to BBB permeability and CNS regulation that contributes to the pathophysiology of Alzheimer’s disease.

Materials and methods

Cell culture

We purchased human Cerebral Microvascular Endothelial Cells/D3 (hCMEC/D3) from Millipore (RRID: CVCL_U985; catalog no. SCC066; EMD Millipore, Darmstadt, Germany). The cells were cultured in EndoGRO-LS complete culture media consisting of EndoGRO basal medium supplemented with a kit containing 2% fetal bovine serum, 0.2% EndoGRO-LS supplement, 5 ng/ml recombinant human epidermal growth factor, 50 µg/ml ascorbic acid, 10 mM L-glutamine, 1 µg/ml hydrocortisone hemisuccinate, and 0.75 U/ml heparin sulfate at 37 °C in a humidified 95% air and 5% CO₂. We authenticated the hCMEC/D3 cell line by confirming the expression of typical endothelial markers (CD31 and VE-cadherin) as well as the expression of tight junction proteins (claudin5, occludin and zona occludens) with western blot analysis.

Transfection

To overexpress Ng, the mammalian gene expression vector containing human Ng (GenBank accession no. NM_001126181.2) with an HA-tag at the N-terminus was designed and constructed by VectorBuilder (VectorBuilder Inc., Chicago, IL, USA). The cells at 70% confluency were transfected with 1.0 µg/ml expression vectors (pRP-CMV-HA-hNRGN) encoding HA-tagged Ng using 50 nM Lipofectamine^® RNAiMAX reagent (catalog no. 13778; Invitrogen, Carlsbad, CA, USA) for 5 h. Subsequently, the cells were washed and cultured at 37 °C in a humidified 95% air and 5% CO₂. To induce the knockdown of Ng or MYH9, validated siRNA targeting Ng (ID: s9723) or MYH9 (ID: 4390824) was purchased from Ambion. The hCMEC/D3 cells at 70% confluency were transfected with 20 nmol/ml of Ng siRNA or 40/60 nmol/ml of MYH9 siRNA using 25 nM Lipofectamine^® RNAiMAX reagent for 6 h. Subsequently, the cells were washed and cultured at 37 °C in a humidified 95% air and 5% CO₂. The scRNA (Negative Control siRNA, ID: AM4611) group refers to a group transfected with a non-targeting siRNA.

Ng-immunoprecipitation assay

The cells overexpressing HA-tagged Ng were harvested at 40 h Post-transfection and lysed using a homogenization buffer of 150 mM NaCl, 20 mM Tris pH 8.0, 5 mM MgCl₂, and 1% Triton X-100 for 1 h on ice. Then, the lysates were collected after centrifugation at 15,000 x g for 20 min. As shown in Supplemental Fig. 1, the cell lysates were precleared with 40 µl of Protein A/G PLUS-Agarose beads (agarose beads; catalog no. sc-2003; Santa Cruz Biotechnology, Dallas, TX, USA) at 4 °C on a shaker for 30 min to remove non-specific binding proteins that bind to agarose beads. Then, the precleared lysates (supernatant) were collected after centrifugation at 1000 x g for 5 min. To isolate Ng-binding proteins under high-Ca²⁺ or low-Ca²⁺ conditions, CaCl₂ or EDTA (Ca²⁺ chelator) was added into the precleared lysates at 2 mM concentration for each. Simultaneously, the anti-HA-tag antibody was added to the precleared lysates at 1:50 dilution. Then, the lysates were incubated for 16 h at 4 °C. After that, the lysates were incubated with 40 µl of Protein A/G PLUS-Agarose beads, which were washed with homogenization buffer containing either 2 mM CaCl₂ or EDTA at 4 °C on a shaker for 3 h. As unbound protein, supernatants were carefully collected after centrifugation at 1000 x g for 5 min. Then, agarose beads were washed three times using 1 ml of the homogenization buffer (150 mM NaCl, 20 mM Tris pH 8.0, 5 mM MgCl₂, and 1% Triton X-100) containing either 2 mM CaCl₂ or EDTA. Bound proteins from the washed agarose beads were eluted by boiling with 2X sample buffer at 95 °C for 10 min with 1050 rpm shaking in Eppendorf ThermoMixer F1.5. Next, the heated agarose beads were cooled on ice, and the eluted bound proteins (supernatant) were collected after centrifugation at 1000 x g for 5 min.

Calmodulin pull-down assay

As shown in Supplemental Fig. 2, the CaM pull-down assay was conducted using a method described previously [20]. First, cell lysates were extracted using the homogenization buffer for 1 h on ice. The cell lysates were collected after centrifugation at 15,000 x g for 20 min and precleared with 40 µl of Sepharose^® 4B beads (catalog no. 4B200; Sigma-Aldrich, Milan, Italy) at 4 °C on a shaker for 3 h to remove non-specific binding proteins that bind to Sepharose beads, and then the precleared lysates (supernatant) were collected after centrifugation at 1500 x g for 3 min. To isolate CaM-binding proteins under high-Ca²⁺ or low-Ca²⁺ conditions, CaCl₂ or EDTA (Ca²⁺ chelator) was added into the precleared lysates up to 2 mM concentration for each, and then 1 mg of the precleared lysates were incubated with 40 µl of Calmodulin Sepharose™ 4B beads (CaM-beads; catalog no. 17-0529-01; GE Healthcare, Uppsala, Sweden), which is washed with homogenization buffer containing either 2 mM CaCl₂ or EDTA, at 4 °C on a shaker for 3 h. Subsequently, supernatants as unbound proteins were carefully collected after centrifugation at 1500 x g for 3 min. Then, CaM beads were washed three times using 1 ml of the homogenization buffer (150 mM NaCl, 20 mM Tris pH 8.0, 5 mM MgCl₂, and 1% Triton X-100) containing either 2 mM CaCl₂ or EDTA. Bound proteins from the washed CaM-beads were eluted by boiling with 2X sample buffer at 95 °C for 10 min with 1500 rpm shaking in Eppendorf ThermoMixer F1.5. The heated CaM beads were cooled on ice, and the eluted bound proteins (supernatant) were collected after centrifugation at 1500 x g for 3 min.

Label-free proteomics using orbitrap fusion

To minimize the effect of highly abundant Immunoglobulin G (IgG) and albumin in immunoaffinity approaches, we separated Ng or CaM pulldown samples using SDS PAGE. We confirm the quality of immunoprecipitation using western blots to detect the baits and known interactome. Based on Coomassie Brilliant Blue Dyes staining, each lane was cut into 4 gel pieces identically based on the molecular weight (top of 100 kDa marker, bottom of 50 kDa marker, and top of 25 kDa marker). Each gel piece was subjected to in-gel trypsin digestion as follows. Gel segments were destained in 50% methanol and 50 mM ammonium bicarbonate, followed by a reduction in 10 mM Tris[2-carboxyethyl] phosphine and alkylation in 50 mM iodoacetamide. Gel slices were then dehydrated in acetonitrile, followed by adding 100 ng porcine sequencing grade modified trypsin in 50 mM ammonium bicarbonate and incubating at 37 °C for 12–16 h. Peptide products were then acidified in 0.1% formic acid. Tryptic peptides were separated by reverse-phase XSelect CSH C18 2.5 μm resin on an in-line 150 × 0.075 mm column using a nanocavity UPLC system. Peptides were eluted using a 60 min gradient from 98:2 to 65:35 buffer A: B ratio [Buffer A = 0.1% formic acid, 0.5% acetonitrile; buffer B = 0.1% formic acid, 99.9% acetonitrile]. Eluted peptides were ionized by electrospray (2.4 kV) followed by MS/MS analysis using higher-energy collisional dissociation (HCD) on an Orbitrap Fusion Tribrid mass spectrometer in top-speed data-dependent mode. MS data were acquired using the FTMS analyzer in profile mode at a resolution of 240,000 over 375 to 1500 m/z. Following HCD activation, MS/MS data were obtained using the ion trap analyzer in centroid mode and normal mass range with precursor mass-dependent normalized collision energy between 28.0 and 31.0.

Results

Immunoprecipitation-mass spectrometry (IP-MS) for Ng protein interactomes

To investigate the interactome of Ng depending on Ca²⁺ concentration, HA-tagged Ng was overexpressed efficiently in hCMEC/D3 cells (Fig. 1A), and immunoprecipitation-mass spectrometry (IP-MS) were performed using an anti-HA-tag antibody under high-Ca²⁺ and low-Ca²⁺ buffer conditions (Fig. 1B and Supplemental Fig. 1). We carried out 4 independent biological replicate experiments to ensure IP quality by measuring HA-Ng expression in input lysate, pre-cleared lysate, binding fraction, and flow-through fraction. Then, we selected the most efficient sample based on the HA-tag assessment and ran proteomics to identify the bound proteins. Immunoprecipitated samples were separated on a 12% SDS-PAGE gel and stained with Coomassie blue. Different Ng protein binding profiles were observed under high-Ca²⁺ buffer conditions (2mM CaCl₂) and low-Ca²⁺ buffer conditions (2mM EDTA) (Fig. 1C). Each lane was cut into 4 fractions based on molecular weight. Each fraction was digested by trypsin and analyzed by label-free quantitative proteomics, which assesses the relative protein quantification using MS/MS spectra count [23, 24] (Fig. 1D and Supplemental Table S2). The bait protein (HA-Ng) was detected in both high-Ca²⁺ and low-Ca²⁺ (Fig. 1E). 116 and 34 binding proteins were detected in samples isolated from conditions of high-Ca²⁺ and low-Ca²⁺, respectively. 85 proteins, including UBB (Ubiquitin B), TPM1 (Tropomyosin 1), and MPRIP (Myosin Phosphatase Rho Interacting Protein), were detected only in high-Ca²⁺, while 3 proteins were detected only in the low-Ca²⁺condition. Interestingly, 31 proteins bound to Ng in both high-Ca²⁺ and low-Ca²⁺ conditions. MS/MS spectra count indicates that MYH9 (Myosin Heavy Chain 9; 227 kDa) binds to Ng with a higher affinity in the high-Ca²⁺ than in the low-Ca²⁺ condition. Compared to the input, we validated immunoprecipitated proteins using western blotting to ensure Ca²⁺-dependent Ng interaction (Fig. 1E). Consistent with the proteomics results, MYH9 and MPRIP exclusively bind to Ng under the high-Ca²⁺ condition. At the same time, the HA-tag was observed at the same levels in high-Ca²⁺ and low-Ca²⁺ conditions. GAPDH does not directly bind to Ng in both high-Ca²⁺ and low-Ca² conditions. These findings suggest that the Ca²⁺ levels play an essential role in Ng function by affecting protein-protein interactions in ECs.

Fig. 1.

Proteomic identification of Neurogranin (Ng) interactome under high-Ca²⁺ and low-Ca²⁺ conditions using immunoprecipitation-mass spectrometry (IP-MS) analysis. A, confirmation of overexpression of Ng in hCMEC/D3 transfected with an expression vector encoding HA-tagged Ng by immunocytochemistry (Ng, green; nucleus, blue). Negative control was stained using an Alexa Fluor 488-conjugated secondary antibody without a primary antibody. Scale bar, 50 μm. HA-Ng and endogenous Ng expression were confirmed by western blotting (n = 2). B, schematic representation of Ng IP in the presence of CaCl₂ or EDTA using an anti-HA-tag antibody. C, Coomassie blue SDS-PAGE gel image for lysate, bead bounds protein, Ng-bound proteins under high-Ca²⁺ and low-Ca²⁺ conditions. Each lane was cut into 4 gel pieces (dotted lines) for trypsin digestion (n = 3 per group). D, heat map and Venn diagram of Ng-bound proteins detected in high-Ca²⁺ or low-Ca²⁺ conditions. Numbers indicate the MS/MS spectra counts for each protein. E, validation of Ng-bound proteins by western blotting (n = 3 per group). “Input” means precleared cell lysates using Protein A/G PLUS-Agarose beads. HA-Ng is present in both high-Ca²⁺ and low-Ca²⁺ conditions. MYH9 and MPRIP significantly bind to Ng under high-Ca²⁺ conditions. *Ng-bound proteins validated by western blotting

Pathway analysis of Ng binding protein profiles based on calcium concentration

To determine the role of Ng in hCMEC/D3 cells, Ingenuity Pathway Analysis (IPA) was performed by analyzing the Ng-binding proteins. The IPA networks were algorithmically generated based on the functional connectivity of these proteins. Under high Ca²⁺ conditions, Ng interacts with proteins involved in canonical pathways such as epithelial adherens junction signaling, eukaryotic initiation factor 2 (EIF2) signaling, epithelial adherens junction remodeling, and the regulation of actin-based motility by Rho (Fig. 2A). Under low-Ca²⁺ conditions, Ng also binds to proteins associated with canonical pathways involving the Bcl2-associated athanogene 2 (BAG2) signaling pathway, the role of protein kinase R (PKR) in interferon induction pathway, and antiviral response, eNOS signaling, and aldosterone signaling (Fig. 2B). Overall, Ng-binding proteins in hCMEC/D3 cells play essential roles in developing the brain vascular system and organ morphology.

Fig. 2.

Ingenuity pathway analysis (IPA) of the Neurogranin (Ng) binding proteins. A, IPA results of 116 Ng-bound proteins detected under the high-Ca²⁺ condition. B, IPA results of 34 Ng-bound proteins that were simultaneously detected in both high-Ca²⁺ and low-Ca²⁺ conditions

Proteomic identification of the proteins that interact with cam

Since Ca²⁺ levels are essential for neuronal Ng-CaM binding [25, 26], we assessed Ca²⁺-dependent CaM interaction with Ng in the BBB using hCMEC/D3 cells. To identify the CaM-Ng interaction in hCMEC/D3 cells, we purified CaM-binding proteins using CaM-Sepharose beads under high-Ca²⁺ buffer conditions (2mM CaCl₂), representing Ca²⁺-CaM complex and low-Ca²⁺ buffer conditions (2mM EDTA), representing Apo-CaM (Fig. 3A and Supplemental Fig. 2) [27]. We carried out 4 independent biological replicates to ensure the quality of CaM affinity purification by assessing CaMKII binding because CaMKII is a CaM-binding enzyme activated by Ca²⁺-CaM binding [28, 29]. Finally, we selected the most efficient experimental condition and ran proteomics for binding-protein identification. The gel staining image demonstrated that Ca²⁺ concentrations significantly alter CaM-binding protein profiles (Fig. 3B). Western blot analysis showed that MYH9, CaMKII, and phosphorylated-protein kinase Cg (p-PKCg) interact with CaM under high-Ca²⁺ conditions. On the other hand, Ng only binds to CaM under low-Ca²⁺ conditions (Fig. 3C). This finding is consistent with previous findings that Ca²⁺-CaM binds to CaMKII while Apo-CaM dominantly binds to Ng in neurons [20].

Fig. 3.

Proteomic identification of Calmodulin (CaM) interactome under high-Ca²⁺ and low-Ca²⁺ conditions by affinity pull-down and label-free proteomics analysis. A, schematic representation of pull-down of CaM in the presence of CaCl₂ or EDTA using CaM-Sepharose beads. B, SDS-PAGE gel image of CaM-bound proteins stained with 0.1% (w/v) Coomassie blue. Each lane was cut into 4 gel pieces (dotted lines) for trypsin digestion (n = 3 per group). C, western blot analysis of CaM bound proteins. “Input” indicates precleared cell lysates using Sepharose 4B beads. MYH9, CaMKII, and p-PKCg only bind to CaM under high-Ca²⁺ conditions, while CaM binds to Ng under low-Ca²⁺ conditions (n = 3 per group). D, heat map and Venn diagram of CaM bound proteins detected in high-Ca²⁺ or low-Ca²⁺ conditions. Numbers indicate the MS/MS spectra count for each protein. *Indicates CaM-bound proteins reported in previous studies. E. intracellular Ng-CaM interaction measured by proximity ligation assay (PLA) (n = 5). Red spots (arrowhead) demonstrate PLA signals by the interaction of CaM-Ng in the proximity of 40 nm or less. The nucleus is stained blue (DAPI). Negative control was stained using a Duolink^® In Situ Red Kit without primary antibody treatment. The number of PLA signals per cell was quantified using ImageJ software. *p < 0.05, P values were determined by Student’s t-test. All data are presented as mean ± SEM. Scale bar, 10 μm

To identify the Ca²⁺-dependent CaM interactome in hCMEC/D3 cells, both lanes were cut into 4 fractions based on molecular weight (Fig. 3B). Each fraction was digested with trypsin and analyzed by label-free quantitative proteomics using LC-MS/MS (Fig. 3D and Supplemental Table S3). Proteomics results indicate that 858 proteins in hCMEC/D3 cells bind to CaM. Of these, 571 were bound to Ca²⁺-CaM when the Ca²⁺ level was high. 280 proteins bound to both Ca²⁺-CaM and Apo-CaM conditions. To determine CaM-binding affinity between high-Ca²⁺ and low-Ca²⁺ conditions, the MS/MS spectra number of these proteins was used for relative quantification [23, 24] (Fig. 3D). Notably, the IQ-motif containing GTPase-activating protein 1 (IQGAP1), which is a regulator of actin dynamics and assembly, and MYH9, which plays a vital role in cytoskeleton reorganization and focal contact formation, bound to CaM with higher affinity in the high-Ca²⁺ condition than in the low-Ca²⁺ state. CaMKIIs, well-known Ca²⁺-CaM binding proteins, were detected in both conditions, but MS/MS spectra count results showed a significantly increased number under high-Ca²⁺ conditions. Only 7 proteins bound to Apo-CaM, including phosphorylase kinase (PHK) and Ng (NRGN). Overall, CaM interacts with Ng under the low-Ca²⁺ condition. However, the affinity of Ng for Apo-CaM is weaker than that of other Apo-CaM-binding proteins.

To confirm the direct CaM-Ng interaction, we employed a proximity ligation assay (PLA), which demonstrates the in situ detection of endogenous protein binding between CaM and Ng (Fig. 3E). PLA signals (red spots) indicate CaM-Ng interactions within a proximity of 40 nm or less. To confirm that the PLA signals were not due to non-specific binding of secondary antibodies, a negative control was stained using a Duolink^® In Situ Red Kit without primary antibody treatment. This result supports the finding that Ng interacts with CaM in hCMEC/D3 cells, although its affinity in the BBB is relatively weak compared to that in neurons.

Cluster analysis identifies Ng-mediated cytoskeletal remodeling

To further investigate the interaction between Ng and CaM in the BBB, we assessed the co-binding proteins from the protein list from Ng IP protein list (119 proteins) and the CaM affinity pull-down assay (859 proteins) (Fig. 4A). Interestingly, 46 proteins were found to only bind to Ng, including T cell receptor alpha joining 56 (TRAJ56) piccolo presynaptic cytomatrix protein (PCLO), granulin (GRN), and myosin phosphatase Rho-interacting protein (MPRIP) (Fig. 4B and Supplemental Table S4). Moreover, albumin (ALB), immunoglobulin heavy constant gamma 1 (IGHG1), might have been eluted from the HA antibody-based IP. A total of 73 Ng-binding proteins also bound to CaM, suggesting novel molecular mechanisms involving Ng-CaM interaction. Among these Ng-binding proteins, MYH9 bound to Ng with a higher affinity than to CaM, while tropomyosin 1 (TPM1) showed a similar binding affinity for both Ng and CaM.

Fig. 4.

Cluster analysis of the Neurogranin-Calmodulin (Ng-CaM) binding proteins in the Blood Brain Barrier (BBB). A, schematic of Ng and CaM bound proteins from Ng-IP and CaM affinity pull-down assay. B, heat map and Venn diagram of Ng and CaM binding proteins. 73 proteins bind both Ng and CaM. C, Ingenuity pathway analysis (IPA) results of 73 Ng and CaM binding proteins

To elucidate the biological functions related to Ng-CaM interaction, we performed IPA using 73 proteins that bound to both Ng and CaM. IPA analysis demonstrated that the Hsp90, Akt, and PI3K pathways play an essential role in Ng-CaM interactions, consistent with the findings in Fig. 2A and B. In addition, canonical pathways significantly enriched in the dataset, included actin cytoskeleton signaling, regulation of actin-based motility by Rho, and epithelial adherens junction signaling (Fig. 4C). These results indicate that proteins interacting with both Ng and CaM play critical roles in vascular system development and organ morphology and organ development.

Intracellular Ng interactions with MYH9

Since our Ng IP-MS and CaM pull-down assays demonstrated significant Ng and MYH9 interactions in the lysates, we assessed intracellular protein-protein interactions using immunofluorescence. To assess intracellular Ng interaction with CaM and MYH9, we utilized proximity ligation assay (PLA), which detects in situ protein-protein interactions occurring within a distance of 40 nm or less (Fig. 5A). We also included a negative control experiment using Duolink^® In Situ Red Kit without any primary antibody treatment. PLA revealed red dots that indicate Ng binding to both CaM and MYH9, while there was no PLA signal in the negative control (Fig. 5B). Interestingly, PLA revealed significantly more Ng-MYH9 interactions than Ng-CaM interactions in hCMEC/D3 cells (p < 0.05) (Fig. 5C). This supports our IP-MS findings and suggests that Ng interacts with MYH9 intracellularly, potentially forming the basis of Ng’s function in brain endothelial cells.

Fig. 5.

Neurogranin (Ng) overexpression and knockdown decreases F-actin expression in the blood-brain barrier (BBB). A, schematic diagram for the proximity ligation assay (PLA) B, PLA detects endogenous Ng-MYH9 and CaM-MYH9 interactions. Red spots demonstrate PLA signals by protein-protein interactions in the proximity of 40 nm or less, a negative control was stained using a Duolink^® In Situ Red Kit without any primary antibody treatment. Red spots (arrowhead) demonstrate PLA signals by the interaction of Ng. The blue color indicates the nucleus (DAPI). Scale bar, 10 μm. C, quantification of PLA signal measured as a ratio of a number of PLA signal puncta to number of cells. Ng-MYH9 shows significantly more interactions in hCMEC/D3 cells compared to CaM-MYH9 interactions measured by Student’s t-test. D, Ng-mediated MYH9 and F-actin expression change in the hCMEC/D3 cells. There is no change in MYH9 expression. E, Both Ng-overexpression and Ng-siRNA showed a significant reduction of F-actin expression in hCMEC/D3 cells. scRNA is the control group treated with a non-targeting siRNA (n = 4 per group). *p < 0.05, **p < 0.01, P values were determined by one-way ANOVA with a Tukey’s post-hoc test. F and G, F-actin expression in scRNA and Ng-siRNA were observed with FITC-phalloidin (green). The distribution of F-actin was assessed by measuring the green fluorescence intensity profiles across individual cells, as indicated by the orange line. Ng-knockdown cells showed a significant reduction of F-actin (n = 4 per group). White scale bar, 25 μm. H, the intensity of F-actin significantly decreased Ng-siRNA. *p < 0.05, P values were determined by Student’s t-test. All data are presented as mean ± SEM

Ng expression modulates F-actin but not MyH9 expression levels in cultured brain endothelial cells

Our IP-MS results demonstrate that cellular Ca²⁺ levels play an important role in Ng binding with other proteins, ultimately contributing to cytoskeletal remodeling [30]. For example, Ng binds with cytoskeletal regulators in a Ca²⁺-dependent (e.g. MPRIP) or Ca²⁺-independent (e.g. MYH9) manner. Since MPRIP is described as a cytoskeletal protein that regulates stress fibers, which retains the F-actin binding capacity [31, 32], we assessed the impact of Ng expression levels on cytoskeletal protein expression. First, we evaluated how Ng knockdown and overexpression affect MYH9 and F-actin expression compared to scrambled controls. MYH9 expression remained unchanged despite alterations in Ng expression (Fig. 5D). Notably, both Ng knockdown and overexpression led to a significant decrease in F-actin expression [F(2,9) = 9.439, p = 0.0062] (Fig. 5E and Supplemental Fig. 4). To validate this finding, we assessed the F-actin expression using phalloidin staining (Fig. 5F, G). Using ImageJ, we measured fluorescence intensity profiles across the individual cells (indicated by the orange line). In scrambled control cells, the F-actin signal was highest in the cortical/cytoplasmic regions. In contrast, in Ng-siRNA cells, this peripheral enrichment was diminished and replaced by a more diffuse intracellular distribution. The intensity of F-actin filament staining was significantly decreased in the Ng-siRNA-treated cells compared to the scrambled controls (Fig. 5H) (n = 4, t = 3.031, df = 6, p = 0.0231). Overall, Ng expression is critical for F-actin expression in hCMEC/D3 cells but does not affect MYH9 expression levels. Since MYH9 is an F-actin-binding molecular motor that regulates cell adhesion, cell migration, cytokinesis, and polarization [34], Ng alone or its binding to MYH9 may suppress F-actin expression. These findings highlight the crucial role of the Ng-MYH9 interaction in regulating cytoskeletal remodeling in the hCMEC/D3 cells.

MYH9 expression regulates Ng expression and AKT activation in hCMEC/D3 cells

Then, we examined whether MYH9 expression levels affect Ng expression and subsequent F-actin expression in hCMEC/D3 cells using Western blot analysis (Fig. 6A). MYH9-siRNA (40 nM and 60 nM) conditions significantly decreased Ng expression [F(3,8) = 27.83, p = 0.0001] (Fig. 6B). MYH9 knockdown also significantly decreased F-actin expression, suggesting that MYH9 or the Ng-MYH9 interaction influences actin dynamics [F(3,8) = 12.94, p = 0.0020] (Fig. 6C). To further validate this, we assessed F-actin levels using phalloidin staining in hCMEC/D3 cells (Fig. 6D, E). Across individual cells, fluorescence intensity profiles showed robust cytoplasmic F-actin signals in the control, while MYH9-siRNA cells exhibited a diminished intracellular network. Overall, cells treated with MYH9-siRNA showed a significant reduction in F-actin expression compared to the control, similar to the observation in Ng-siRNA-treated cells (n = 10–12, t = 2.659, df = 20, p = 0.0151) (Fig. 6F).

Fig. 6.

Depletion of MYH9 suppresses Ng and F-actin expression in hCMEC/D3 cells. A, MYH9-siRNA decreases Ng and F-actin expression in the hCMEC/D3 cells. B, decreased Ng expression. C, decreased F-actin expression. n = 3 per group, *p < 0.05, **p < 0.01. P values were determined by a one-way ANOVA with Tukey post hoc test. D and E, phalloidin staining of control (scRNA) and MYH9-siRNA in hCMEC/D3 cells. The distribution of F-actin was assessed by measuring the green fluorescence intensity profiles across individual cells, as indicated by the orange Line. White scale bar, 25 μm. F, decreased intensity of phalloidin staining in MYH9-siRNA transfected hCMEC/D3 cells. (n = 10,12). p < 0.05. P value was determined by an unpaired student t-test

To explore the causal relationship between MYH9-Ng interaction and cytoskeletal remodeling, we tested Ng expression and downstream signaling under the MYH9-siRNA treatment (Fig. 7A). Since Ng knockdown in endothelial cells has suppressed the AKT pathway [6, 8], we measured AKT and GSK3b expression and activity changes under MYH9-siRNA. Consistent with results from Ng-siRNA, MYH9-siRNA decreased AKT [F(3,8) = 1.915, p = 0.0018] and GSK3β [F(3,8) = 9.049, p = 0.0060] phosphorylation, while there was no change in AKT and GSK3β expression (Fig. 7B-D). Additionally, we measured adhesion molecule changes, including VCAM1 and ICAM1, both of which are significantly increased by Ng-siRNA treatment [6, 8]. The MYH9-siRNA group exhibited elevated VCAM1 expression [F(3,8) = 12.37, p = 0.0023], indicating similar endothelial activation mediated by Ng-siRNA (Fig. 7E). However, ICAM1 was not increased by MYH9-siRNA, in contrast to the increase observed with Ng-siRNA. Because MYH9 contributes to maintaining the integrity of the BBB [33], and Ng regulates the BBB permeability as well [8], we measured whether MYH9 knockdown influences BBB integrity by measuring junctional protein expression change. Consistent with Ng-siRNA, we also observed that MYH9-siRNA significantly decreased the expression of the junction proteins claudin-1 [F(3,8) = 17.38, p = 0.0007] and ZO-2 [F(3,8) = 9.049, p = 0.0060] (Fig. 7F and G). These results support that Ng-MYH9 interaction plays a critical role in cytoskeletal remodeling, endothelial activation, and BBB integrity in hCMEC/D3 cells.

Fig. 7.

Loss of MYH9 expression alters AKT signaling, tight junction protein levels, and adhesion molecule expression in hCMEC/D3 cells. A, representative western blot for AKT signaling proteins, adhesion molecules, and tight junction proteins. B, MYH9-siRNA decreases expression of p-Akt. C, no change in total AKT expression D, MYH9-siRNA decreases expression of p-GSK3β. E, MYH9-siRNA increases expression of VCAM1 F, MYH9-siRNA decreases expression of Cldn-1 and G, ZO2. (n = 3 per group). *p < 0.05, **p < 0.01 P values were determined by one-way ANOVA with a Tukey pos-hoc test. H, immunofluorescent staining of cortical blood vessels in wildtype and I, Ng knockout mice. Green staining represents CD31, an endothelial cell marker. Red indicates MYH9 expression. J, Ng knockout (−/−) mice showed a substantial decrease in MYH9 expression in brain microvessels (CD31) compared to wild-type (+/+) mice. (Ng +/+: n = 4 and Ng -/-: n = 5). Scale bar, 50 μm. *p < 0.05. P values were determined by a Student’s t-test. All data are presented as mean ± SEM

Finally, we measured the in vivo change in MYH9 expression in the brain vasculature of Ng null (-/-) mice. MYH9 (green) is dominantly expressed within blood vessels (CD31 (red); endothelial cell marker) of the cortical tissues (Fig. 7H, I). Interestingly, Ng -/- mice demonstrated significantly decreased MYH9 expression (red) in the CD31 compared to that of wild-type mice (Fig. 7J). (n = 4, t = 4.026, df = 6, p < 0.01). Although Ng knockdown had no effect on MYH9 expression in hCMECs, the decreased Ng-mediated MYH9 expression in mice suggests that Ng expression is critical for maintaining MYH9-mediated neurovascular integrity through cytoskeletal remodeling.

Discussion

In the neuron, Ng binds with CaM when Ca²⁺ levels are low, playing a key role in Ca²⁺-CaM dynamics and Ca²⁺-CaM kinase (CaMKII) signaling [20, 35]. However, the effect of Ng knockout in the endothelial cells demonstrated Ca²⁺-CaM independent signaling [6, 8]. Therefore, we hypothesize that Ng has a novel function in brain endothelial cells, distinct from CaM binding observed in neurons. To test this, we assessed Ng interactomes using IP-MS, protein pulldown assay, label-free proteomics, and bioinformatics. Our pulldown assay results indicate that Apo-CaM binds Ng when Ca²⁺ levels are low. However, CaM is not detected in the HA-Ng pulldown. This result may be induced by biological variations during the pulldown assay, the possibility that Ng does not bind to CaM directly or the HA tag at the N-terminus interfering with CaM binding. Moreover, our proteomic results provided new Ng-binding proteins, including MYH9, which regulates cytoskeletal remodeling in the BBB. Finally, we validated intracellular Ng-MHY9 binding and signaling using intracellular PLA assay, Ng-siRNA, MYH9-siRNA, and Ng null mice.

Since CaM structural dynamics are altered by the number of Ca²⁺ ion bound, which in turn depends on the intracellular Ca²⁺ concentrations [36], we examined CaM-Ng binding under low Ca²⁺ lysate or high Ca²⁺ conditions. As an exploratory research platform, label-free proteomics identified possible Ng or CaM interactomes in response to different Ca²⁺ levels in the lysate. Molecular and cellular biology experiments validated our proteomics findings. Consequently, our results indicated that Ng significantly binds to MYH9, regulating cytoskeletal remodeling in hCMEC/D3 cells. Consistently, CaM also binds with MYH9, but its binding affinity is weaker than Ng-MYH9 (Fig. 4). Overall, our novel approach provides new mechanistic insight into Ng-mediated cytoskeletal remodeling regardless of Ca²⁺ levels and CaM signaling.

MYH9, known as non-muscle myosin heavy chain IIa (NMMIIA), is expressed mainly in endothelial cells and glandular cells and is involved in several cellular processes, including cell motility, polarization, phagocytosis, and morphology [37]. Furthermore, deletion of NMMIIA in mice induces abnormalities in the visceral endoderm, resulting in embryonic lethality via the reduction of cell-cell adhesion mediated by the epithelial-cadherin [38]. Therefore, we anticipate that MYH9 may play an essential role in endothelial cell function. MYH9 participates in many cellular processes, including cytokinesis [39], cell migration [40], adhesion [41], thrombosis [42], and morphogenesis [43]. In mice, mutation of MYH9 shows abnormal placenta development and induces defects in angiogenesis during embryonic development [44]. In addition, MYH9 mutation attenuates cell motility by inducing the formation of abnormal stress fibers and fewer focal adhesions in mouse embryonic fibroblast cells. MYH9 knockout reduces vascular sprout multicellularity in the human umbilical vein endothelial cells (HUVECs) [45]. Accordingly, MYH9 is well known to play an essential role in angiogenesis for new blood vessel formation. MYH9 regulates the contraction of F-actin observed in cytoskeletal structures such as stress fibers and contractile rings [46]. Actin facilitates cell migration and sprouting via cytoskeletal remodeling in non-muscle cells, including endothelial cells [47]. This actin-myosin complex, called actomyosin, enables cell motility by implicating translocation of the F-actin [48].

We also observed that both Ng knockdown and overexpression significantly decreased F-actin expression in hCMEC/D3 cells. Therefore, maintaining Ng expression in the BBB may be critical to F-actin-mediated cytoskeletal remodeling. In addition, MYH9 knockdown suppressed Ng expression and decreased F-actin expression, which is the consistent outcome of the effect of Ng knockdown in endothelial cells. Furthermore, we demonstrated that MYH9-siRNA significantly reduced AKT and GSK3b phosphorylation, outcomes similar to those observed with Ng-siRNA. This led us to hypothesize that MYH9 might be upstream of Ng. However, changes in ICAM1 and VCAM1 expression in response to Ng-siRNA did not align with those observed with MYH9-siRNA. Furthermore, Ng-null mice exhibited a significant decrease in MYH9 expression in the brain vasculature. Thus, we anticipate that Ng interaction with MYH9 is essential in Ng-mediated F-actin regulation and cytoskeletal remodeling in the brain vasculature. Further study using MYH9 in vivo model is required to explain the causal relationship between Ng and MYH9 expression in hCMEC/D3 cells and how Ng expression impacts vascular sprouting through cytoskeletal remodeling by interacting with MYH9 in the brain endothelial cells.

Proteins with a specific three-dimensional (3D) structure play essential roles in cellular functions and biological processes by interacting with other biological macromolecules, including protein, DNA, and RNA. Therefore, discovering protein-protein interactions depending on tCa²⁺ concentration, which modulates signaling cascades, is essential for understanding fundamental cellular and molecular processes and exploring therapeutic targets. This study used Ng immunoprecipitation and CaM affinity purification against different Ca²⁺ concentrations. Cell culture media contains 2mM of CaCl₂, which provides sufficient Ca²⁺ for cell growth, attachment, and differentiation [49]. Extracellular Ca²⁺ levels are known to reach the 2 mM range, while cytosolic Ca²⁺ shows 100 nM levels, so the endothelial cells maintain a 20,000-fold Ca²⁺ gradient [12]. To minimize the effect of extracellular Ca²⁺, our cell lysis was conducted under Ca²⁺-free lysis buffer so that our “standard condition” might be close to the 100–200 nM Ca²⁺ levels. In addition, 2 mM EDTA treatment during the pulldown completely depleted Ca²⁺ from the intracellular area and generated Apo-CaM [20, 50]. Therefore, we anticipate that our CaCl₂ or EDTA treatment induces Ca²⁺-dependent protein conformational changes. Protein conformational flexibility can generate interactomes within the cell lysate, enabling structural change of proteins with a specific 3D conformation [51]. In addition, the conformational change of Ca²⁺ sensing protein is regulated by Ca²⁺ dynamics and protein-protein environments [52]. We anticipated that the 3D conformational change induced by competitive elution using CaCl₂ or EDTA is suitable for elucidating Ca²⁺-dependent Ng or CaM interactions. Interestingly, our western blotting results indicate that competitive elution was not strong enough to elute both bait and its binding proteins. Therefore, we had to boil the beads for the binding protein elution, although it may increase the detection of non-specific interactions of Ng or CaM proteins.

Although, both immunoprecipitation and affinity precipitation can carry over non-specific proteins from beads, antibodies, or detergent-insoluble proteins, we utilized pre-clearing and control experiments to diminish the nonspecific binding. Pulldown conditions were first optimized using Coomassie staining and western blotting. While on-bead tryptic digestion with a high-resolution LC-MS/MS approach offers a simple and robust approach without pre-fractionation [53, 54], we employed SDS-PAGE to enhance the detection of lower-abundance proteins by separating out highly abundant immunoglobulin G (IgG) and albumin from the immunoprecipitated fraction. Furthermore, because Ca²⁺-dependent conformational change can impact the multimerization of binding proteins, we anticipated a molecular weight shift of proteins from their theoretical molecular weight [55]. Indeed, we observed that several proteins, including Ng, were detected in a higher molecular weight gel fraction under high Ca²⁺ conditions. In contrast, Ng was detected in the gel fraction of its theoretical molecular weight under low Ca²⁺ conditions. This finding supports the idea that Ng or CaM binding occurs selectively in response to distinct Ca²⁺-dependent conformational states [56]. To validate the Ng and CaM interactomes identified by proteomics, we performed PLA to confirm Ca²⁺-dependent intracellular protein–protein interactions.

Since the numbers of MS/MS spectra from label-free proteomics can be used for predicting protein amount under pair-wise sample comparison, we assessed the binding affinity of Ng or CaM using a heat-map analysis [22, 57]. Since our study tests the binding affinity between high Ca²⁺ and low Ca²⁺ conditions, we expect a “binding” or “non-binding” status. We agree with the precision of intensity-based absolute quantification (iBAQ) values in protein quantification. However, it has been demonstrated that among all the factors of identification, only spectral count showed a strong Linear correlation with relative protein abundance with a dynamic range over 2 orders of magnitude [58, 59]. Therefore, we employed spectral count to represent the affinity or relative abundance of the protein-protein interactions.

So far, most of the literature demonstrates the role of Ng in low Ca²⁺ cellular conditions due to its binding to Ng-Apo CaM bindings. In contrast, our study is the first to report that Ng interacts with distinct signaling networks depending on Ca²⁺ conditions. For example, network and IPA canonical pathway analysis of 85 proteins that bound to Ng at high-Ca²⁺ revealed enrichment in TCR, ERK1/2, and PI3K signaling pathways (Fig. 2A). The top canonical pathways included epithelial adherens junction signaling, EIF2 signaling, remodeling of epithelial adherens junctions, and regulation of actin-based motility by Rho. Notably, proteins such as TPM1 and MPRIP, which are involved in cytoskeletal remodeling, were among the Ng-binding partners under high Ca²⁺ conditions. Although the IPA software identified “epithelial adherens junction signaling,” the software does not distinguish between epithelial and endothelial pathways. While epithelial and endothelial adherens junction signaling share similarities, they differ in protein composition, localization and regulation. This limitation in the software algorithm and analysis should be considered when interpreting our pathway analysis results in the context of endothelial cells. The Rho family of small GTPases, a vital regulator of the actin cytoskeleton, are involved in angiogenesis [60]. Also, TPM1 regulates F-actin function and actin-myosin interaction [61], while myosin phosphatase Rho-interacting protein (MPRIP) modulates actin stress fibers formation as a cytoskeletal protein [62].

Interestingly, our study revealed that Ng overexpression and knockdown decreases F-actin, supporting a role for Ng in cytoskeletal remodeling. One possible explanation is that both excessive and insufficient Ng expression disrupt the finely balanced MYH9 signaling pathway, leading to impaired actin filament organization. Similar bidirectional effects of Ng expression on protein function have been reported in recent studies implicating Ng in Ca²⁺-CaM or redox homeostasis in endothelial cells [63]. It is possible that Ng acts as a regulatory unit for MYH9, and its overexpression may competitively interfere with F-actin interactions. To further investigate cytoskeletal remodeling, we reanalyzed phalloidin-stained hCMEC/d3 cells to assess F-actin distribution in addition to total signal. Although hCMEC/D3 cells do not exhibit a dense subcortical actin rim typical of fully differentiated BBB endothelium [64], the observed shift in F-actin localization upon Ng or MYH9 knockdown suggests structural remodeling that may be relevant to barrier formation in confluent monolayers. It is important to note that hCMEC/D3 cells in this study were not grown to confluency, and they were not validated to exhibit a BBB phenotype. However, most of our functional experiments, including pull-down assays, immunoprecipitation, and western blotting, were performed on confluent cultures. Still, because of this incongruence in confluency and the limitations of in vitro models, any inferences to BBB function should be interpreted with caution. We postulate that Ng may regulate F-actin through a distinct signaling mechanism that remains to be fully elucidated.

We also identified 31 proteins that bind to Ng in both high-Ca²⁺ and low-Ca²⁺ conditions. These proteins were linked to Akt, PI3K, and ERK1/2 signaling pathways (Fig. 2B). Key canonical pathways included BAG2 signaling pathway, the role of PKR in interferon induction and antiviral response, eNOS signaling, and aldosterone signaling in epithelial cells. Among the proteins, GRN and annexin A2 (ANXA2) were notable. GRN promotes NO synthesis via Akt/eNOS signaling pathway and inhibits the atherosclerotic inflammatory reaction [65]. While ANXA2 is an endogenous antioxidant as the thioredoxin substrate [66] and enhances angiogenesis via Akt/ERK pathway [67]. The eNOS signaling activated by Akt also plays a role in angiogenesis by producing NO [68]. The ERK1/2 [69] and PI3K [70] pathways are well known as crucial for angiogenesis and support endothelial cell migration and proliferation. Given that cytoskeletal components [71] and NO [72] are essential for maintaining BBB integrity, our findings suggest that Ng may contribute to both barrier function and angiogenesis through interactions with proteins such as TPM1, MPRIP, GRN, and ANXA2.

We reported 571 proteins binding to CaM under high-Ca²⁺ conditions and 280 proteins binding to CaM regardless of Ca²⁺ concentration. The binding proteins of CaM affinity pulldown are well established [27]. Consistent with the finding from the others, we also found CaM binding proteins, including IQGAP1 [73], IQGAP3 [74], CAMK2G [75], CAMK2D [75], Ng [76], MYO1E [77], MYO1C [78], MYO1B [77] MYO5 [79], MYO6 [80], MYO9B [81], MYH9 [82], ATP2B1 [83], CALD1 [84], WFS1 [85], SPTAN1 [86], CDC5L [87], AKAP12 [88], TBC1D4 [89], PHKB [90], PHKA2 [90], PHKA1 [90], PHKG2 [90], PLEC [91], PFKP [92], and PFKM [92]. Although there is batch-to-batch variability between antibodies and resins, it was well known that CAVN1, WSF1, SPTAN1, and AKAP12 are Ca²⁺-dependent CaM-binding proteins, while IQGAP, MYH9, CAMKII, unconventional myosins, and ATP binding proteins are Ca²⁺-independent CaM-binding proteins [93]. For example, CaMKII is well known as a CaM-binding protein in the neuron. CaMKII plays a structural role by interacting with actin in the dendritic spine [94]. CaMKII interacts with CaM in the BBB. Moreover, among the 281 proteins binding to CaM regardless of Ca²⁺ concentration, IQGAP1, MYH9, and plectin (PLEC) showed a stronger affinity with CaM. IQGAP1, with several binding domains containing F-actin and CaM, mainly participates in diverse cellular processes such as cell migration and adhesion by regulating actin dynamics and the actin-microtubule complex [95]. PLEC enhances vascular integrity by reinforcing adherens junction via crosstalk between vimentin and actin networks as a significant cytoskeletal linker protein [96]. Hence, CaM may regulate endothelial barrier function by interacting with the proteins, including CaMKII, IQGAP1, MYH9, and PLEC. Moreover, we found 7 proteins binding to Apo-CaM under the low-Ca²⁺ condition in the BBB. Phosphorylase kinase (PHK) subunits and Ng mainly interacted with CaM under low-Ca²⁺ conditions. PHK is a substrate-specific CaM kinase for stimulating glycogenolysis, breaking down glycogen into glucose, an energy substrate [97]. Reduction of PHK leads to dorsal longitudinal anastomotic angiogenesis failure in ECs [98]. Ng regulates synaptic plasticity and long-term potentiation. Ng is also well known as a CaM binding protein under the low-Ca²⁺ condition in the neurons. Overall, we observed that MYH9 is the most potent binding partner of Ng and CaM. In addition, 73 proteins simultaneously bind to both Ng and CaM, forming a network cluster with PI3K and Akt pathways (Fig. 4C). Based on IPA results, actin cytoskeleton signaling, regulation of actin-based motility by Rho, and epithelial adherens junction signaling were identified as the top 3 canonical pathways. Ultimately, these pathways impact the development and function of the cardiovascular system in the BBB.

Our data indicate that MYH9 knockdown recapitulates several effects of Ng depletion, including reduced phosphorylation of AKT and GSK3β, increased VCAM1 expression, and decreased junctional proteins claudin-1 and ZO-2, suggesting a shared pathway in regulating BBB integrity [8]. The absence of ICAM1 upregulation following MYH9 knockdown, however, points to partial divergence in downstream signaling. These findings support the hypothesis that MYH9 acts upstream or in parallel with Ng to mediate endothelial remodeling and barrier function. While claudin-5 is widely recognized as essential for BBB formation, we chose to include claudin-1 in our analysis due to its emerging, though somewhat debated, role in BBB regulation. Some studies report that increased claudin-1 expression disrupts BBB integrity following injury [99, 100], while others demonstrate that claudin-1 may contribute to tight junction sealing and BBB maintenance under both pathological and physiological conditions [101–103]. Although claudin-1 overexpression has been linked to barrier disruption, it remains unclear whether a loss of basal claudin-1 expression is detrimental to tight junction stability. Therefore, we evaluated claudin-1 expression as a potentially relevant contributor to BBB maintenance. These results support that Ng-MYH9 interaction plays a critical role in cytoskeletal remodeling, endothelial activation, and BBB integrity in hCMEC/D3 cells.

We have also noted the dysregulated CD31 staining in Ng -/- mice. While we currently have no explanation for this pattern, the disorganization of CD31 localization was consistent across all our immunofluorescent staining of mouse brain tissue. Notably, a western blot analysis of CD31 in whole-brain lysates did not reveal significant differences in overall CD31 expression between genotypes, suggesting that the observed phenotype reflects changes in localization rather than total protein levels. This disorganization may indicate structural alterations in vessel patterning or endothelial organization, which warrants further investigation.

Ultimately, our findings have revealed a novel role of Ng in hCMEC/D3 cells. We have described a new binding partner for Ng in cultured brain endothelial cells that may influence cytoskeletal remodeling and barrier function. Further studies would be required to determine if this novel interaction between Ng and MYH9 regulates brain endothelial function. Furthermore, brain Ng depletion has been observed in Alzheimer’s disease. Future studies can also focus on elucidating the role of endothelial Ng expression on BBB permeability and integrity in Alzheimer’s disease.

Data availability

The proteomics raw datasets supporting the conclusions of this article are available in the MassIVE proteomics data repository [MSV000092629 https://massive.ucsd.edu].

Database searching

Proteins were identified by database search using Mascot Distiller (version 2.5.1, Matrix Science) with a parent ion tolerance of 3 ppm and a daughter ion tolerance of 0.5 Da. All MS/MS samples were analyzed using Mascot (Matrix Science, London, UK; version 2.6.2). The mascot was set up to search the UniProt_2022_01_Homo_sapiens_UP000005640 database (79052 entries), assuming the digestion enzyme trypsin. Mascot was searched with a fragment ion mass tolerance of 0.5 Da and a parent ion tolerance of 3.0 ppm. Carbamidomethyl of cysteine was specified in Mascot as a fixed modification. Oxidation of methionine and acetyl of the N-terminus were specified in Mascot as variable modifications. A maximum of two missed cleavages were allowed. The minimum number of peptides for positive protein identification was set to at least one unique peptide sequence. Protein identifications were confirmed by duplicated sample analysis.

Criteria for protein identification/quantification and IPA analysis

Mascot peptide sequencing results were imported into Scaffold software to verify MS/MS-based peptide and protein identifications. The Scaffold (version Scaffold_5.1.2, Proteome Software Inc., Portland, OR) was used to validate the MS/MS-based peptide and protein identifications of all samples. Individual peptide identifications were accepted if they could be established at greater than 99.0% probability to achieve an FDR less than 1.0% by the Peptide Prophet al.gorithm [21] with Scaffold delta-mass correction. Protein identifications based on individual peptide sequencing were accepted if they could be established at greater than 99.0% probability to achieve an FDR less than 1.0% and contained at least 1 unique peptide sequence. Proteins that contained similar peptides and could not be differentiated based on MS/MS analysis alone were grouped to satisfy the principles of parsimony. Proteins, which share significant peptide evidence, were grouped into clusters, and the list of identified proteins is shown in the Supplemental Tables. Each protein’s total MS/MS spectra count has been used for a label-free quantification [22]. Then, we employed Ingenuity Pathway Analysis (IPA) to classify the protein functionally, where we entered the genes and the number of identified unique peptides. Proteins were clustered and classified by the IPA Network generation algorithm, and the network score ranked the networks. Solid and dashed lines indicate direct and indirect network interactions, respectively. Direct protein interactions require the two molecules to make direct physical contact with each other; there is no intermediate step. Indirect interactions do not need a physical connection between the two molecules, such as a signaling cascade, but instead involve the two molecules not making physical contact.

Duolink^® proximity ligation assay (PLA)

To detect protein-protein interactions, cells were fixed with 4% (v/v) formaldehyde for 30 min at 4 °C, permeabilized with 0.1% Triton X-100 for 10 min at room temperature (RT) and blocked with blocking solution included in Duolink^® In Situ Red Kit (catalog no. DUO92101; Sigma-Aldrich) for 1 h at 37 °C in humidified air. After rinsing with PBS, cells were incubated with primary antibodies against each of the two proteins of interest for 16 h at 4 °C, washed, and then incubated with secondary antibodies conjugated with oligonucleotides (species-specific PLUS and MINUS PLA probes) for 1 h at 37 °C in a humidified air. Subsequently, the oligonucleotides were hybridized and linked to a closed circle nearby (< 40 nm) by Ligation solution for 30 min at 37 °C in humidified air. As a primer, the oligonucleotide arm of PLA probes induced a rolling-circle amplification reaction (RCA) using the Ligated circle as a template by amplification solution for 100 min at 37 °C in humidified air. Fluorescently labeled oligonucleotides in the amplification solution were hybridized into RCA products, and the visible fluorescent spot signals were observed under a confocal laser scanning microscope (Nikon A1R). The information on the primary antibodies is listed in Supplemental Table S1.

F-actin visualization

Cells were fixed with 4% (v/v) formaldehyde for 30 min at 4 °C. The fixed cells were permeabilized with 0.1% (v/v) Triton X-100 diluted in PBS for 15 min and washed with PBS. F-actin expression was detected by incubating the cells with Alexa Fluor 488-conjugated phalloidin (catalog no. A12379; Invitrogen) diluted in PBS for 2 h at 4 °C. After rinsing with PBS, the stained cells were counterstained with DAPI solution and observed under a confocal laser scanning microscope (Nikon A1R). ImageJ was used to analyze immunofluorescence images by first splitting them into individual channels. For DAPI (nuclei) and FITC (phalloidin), thresholds were adjusted to accurately capture the stained regions. Integrated density was measured as a proxy for phalloidin expression, and mean intensity was calculated by dividing the integrated density by the stained area. Phalloidin intensity was then normalized to DAPI intensity to account for differences in cell number. The information on the phalloidin is listed in Supplemental Table S1.

Animals

Male Ng +/+ and Ng -/- mice (C57BL/6J background, Jackson Laboratories, Bar Harbor, ME) aged 12 weeks were used. Mice were group-housed in standard Plexiglas cages under a 12 h light/ dark cycle (Lights on at 6:00 AM) at a constant temperature (24 ± 0.5 °C) and humidity (60 ± 2%) with food and water available ad libitum. The animal brain tissue isolation and handling procedures were approved by the LSUHSC-Shreveport Institutional Animal Care and Use Committees following NIH guidelines.

Immunofluorescent staining

Mice (n = 4 ~ 5 per genotype) were anesthetized with isoflurane and fixed during transcardial perfusion with 10% Formalin. Brains were isolated, Post-fixed overnight, and incubated in 30% sucrose at room temperature until they sank. Using a cryostat, brains were sectioned at 30 μm thickness. Immunohistochemistry was performed using a free-floating method. Sections were incubated with primary antibody overnight at 4 °C. The localization of primary antibodies was detected by incubating the brain sections with secondary fluorescent antibodies purchased from Cell Signaling Technology in 1:400 dilution at room temperature for 2 h. Sections were mounted using the ProLong^® Gold Antifade Reagent with DAPI Immunofluorescence. Fluorescently labeled MYH9 and CD31 were detected on a Nikon A1R confocal microscope and were analyzed using ImageJ software. For vessel analysis, 5–7 vessels per image were quantified for each genotype. Cultured cells were fixed in 2% (w/v) paraformaldehyde (PFA) for 15 min. The fixed cells were stained with unconjugated primary antibodies diluted in PBS supplemented with 1% (w/v) bovine serum albumin (BSA) at 4 °C overnight. Subsequently, the proteins labeled with primary antibodies were detected by incubating the stained cells with fluorescence-conjugated secondary antibodies diluted in PBS supplemented with 1% (w/v) BSA for 2 h at 4 °C. After rinsing with PBS, the stained cells were counterstained with DAPI (catalog no. 8961; Cell Signaling Technology, Beverly, MA, USA) solution and observed under a confocal laser scanning microscope (Nikon A1R, Nikon Instruments Inc., Melville, NY, USA). The information on the primary and secondary antibodies is listed in Supplemental Table S1.

Western blot analysis

Proteins were loaded into 12% Mini-PROTEAN TGX gels, separated at 125 V for 1 h, and transferred onto Immuno-Blot^® PVDF membranes at 25 V for 7 min using the Trans-Blot^® Turbo™ System (serial no. 690BR025712; Bio-Rad Laboratories, inc.). The membranes were blocked with 5% (w/v) nonfat dry milk diluted in 1X Tris-buffered saline with 0.05% Tween 20 (TBST) for 1 h at RT and incubated with primary antibodies diluted in 1X TBST containing 2% (w/v) nonfat dry milk for 16 h at 4 °C. Subsequently, the membranes were incubated with horseradish peroxidase conjugated-secondary antibodies diluted in 1X TBST containing 2% (w/v) nonfat dry milk for 2 h at 4 °C. Protein signals were developed using the Clarity Max™ Western ECL Substrate. Images of membranes were taken using a Bio-Rad ChemiDoc™ Imaging System. Exposure time was adjusted where necessary to avoid signal saturation and signal intensities remained within the linear dynamic range of detection. The intensity of each protein band was quantified by the Image J software (version 1.48v) using the gel analysis plugin. Rectangular ROIs were drawn around each band using the rectangular tracing tool, and background intensity was subtracted before analysis of signal intensity. The information on primary and secondary antibodies is listed in Supplemental Table S1.

Experimental design and statistical rationale

We used the in vitro hCMEC/D3 cell culture model and an Ng knockout (Ng-/-) mice model to study brain vascular signaling that may contribute to BBB integrity and homeostasis. To determine the role of Ng hCMEC/D3 cells, we established gene transfection models for Ng overexpression using a vector (pRP-CMV-HA-hNRGN) encoding HA-tagged Ng or Ng knockdown using siRNA targeting Ng. Those gene expression methods were compared with three different gene constructs and two different concentration conditions, and the best conditions were selected for subsequent experiments. Lipofectamine treatment without the Ng construct or cells transfected with the non-targeting siRNA were used as a control. To conduct immunoprecipitation (IP) and affinity pull down, each cell lysate was pre-cleared with control beads (Protein A/G agarose or Sepharose 4B beads). Then, the lysates were randomly selected for CaCl₂ or EDTA treatment. We carried out 4 biological replicates to validate IP quality by assessing Ng expression of input lysate, pre-cleared lysate, binding fraction, and flow-through fraction. To identify the binding proteins to the HA-tag or CaM-Sepharose 4B bead, lysates were analyzed by label-free proteomics using nanoLC-ESI-MS/MS experiments. Each sample running was duplicated. Label-free proteomics was achieved using MS/MS spectra numbers using Scaffold 5 software. Ingenuity Pathway Analysis (IPA) bioinformatic analysis has elucidated relevant pathways related to Ng or CaM binding proteins. Then, the Ng-bound proteins were validated using western blotting (n = 4). Since immunoprecipitation and pull-down assay may include unspecific protein binding from resin and antibody, we validated intracellular Ng binding using proximity ligation assay (PLA; n = 3). The data are shown as the mean ± standard error of the mean (SEM). To detect statistical differences, we performed a two-tailed student’s t-test or a one-way ANOVA with a Tukey’s post-hoc test (Prism, GraphPad Software, La Jolla, CA). The criterion for statistical significance was p < 0.05.

Abbreviations

AKT

protein kinase B

BBB

Blood-brain barrier

Ca²⁺

calcium

CaM

calmodulin

CaMKII

calcium-calmodulin dependent protein kinase II

Cldn-1

Claudin 1

ECs

endothelial cells

F-actin

filamentous actin

GSK3β

Glycogen synthase kinase 3 beta

hCMEC/D3

human cerebral microvascular endothelial cells

ICAM

intercellular adhesion molecule 1

IP-MS

immunoprecipitation and mass spectrometry

IPA

ingenuity pathway analysis

MPRIP

myosin phosphatase Rho-interacting protein

MYH9

myosin heavy chain-9

Ng

neurogranin

PLA

proximity ligation assay

VCAM1

vascular cell adhesion molecule

ZO2

zonula occludens 2

Author contributions

Conceptualization and Supervision: HN; Methodology: AA, JP, RS, JA, and HN; Formal analysis and investigation: AA, JP, and HN; Writing - original draft preparation: AA and HN; Writing - review and editing: AA, RS, JA and HN, Resources: RS, JA, and HN. All authors reviewed the manuscript.

Data availability

The proteomics raw datasets supporting the conclusions of this article are available in the MassIVE proteomics data repository [MSV000092629 https://massive.ucsd.edu].

Declarations

Ethics approval and consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.