Epidermal growth factor receptor inhibition prevents vascular calcifying extracellular vesicle biogenesis

Abstract

Chronic kidney disease (CKD) increases the risk of cardiovascular disease, including vascular calcification, leading to higher mortality. The release of calcifying extracellular vesicles (EVs) by vascular smooth muscle cells (VSMCs) promotes ectopic mineralization of vessel walls. Caveolin-1 (CAV1), a structural protein in the plasma membrane, plays a major role in calcifying EV biogenesis in VSMCs. Epidermal growth factor receptor (EGFR) colocalizes with and influences the intracellular trafficking of CAV1. Using a diet-induced mouse model of CKD followed by a high-phosphate diet to promote vascular calcification, we assessed the potential of EGFR inhibition to prevent vascular calcification. Furthermore, we computationally analyzed 7,651 individuals in the Multi-Ethnic Study of Atherosclerosis (MESA) and Framingham cohorts to assess potential correlations between coronary artery calcium and single-nucleotide polymorphisms (SNPs) associated with elevated serum levels of EGFR. Mice with CKD developed widespread vascular calcification, associated with increased serum levels of EGFR. In both the CKD mice and human VSMC culture, EGFR inhibition significantly reduced vascular calcification by mitigating the release of CAV1-positive calcifying EVs. EGFR inhibition also increased bone mineral density in CKD mice. Individuals in the MESA and Framingham cohorts with SNPs associated with increased serum EGFR exhibit elevated coronary artery calcium. Given that EGFR inhibitors exhibit clinical safety and efficacy in other pathologies, the current data suggest that EGFR may represent an ideal target to prevent pathological vascular calcification in CKD.

NEW & NOTEWORTHY Here, we investigate the potential of epidermal growth factor receptor (EGFR) inhibition to prevent vascular calcification, a leading indicator of and contributor to cardiovascular morbidity and mortality. EGFR interacts and affects the trafficking of the plasma membrane scaffolding protein caveolin-1. Previous studies reported a key role for caveolin-1 in the development of specialized extracellular vesicles that mediate vascular calcification; however, no role of EGFR has been reported. We demonstrated that EGFR inhibition modulates caveolin-1 trafficking and hinders calcifying extracellular vesicle formation, which prevents vascular calcification. Given that EGFR inhibitors are clinically approved for other indications, this may represent a novel therapeutic strategy for vascular calcification.

INTRODUCTION

Medial calcinosis manifests as the formation of calcium phosphate mineral in the media layer of arterial walls, leading to vascular stiffening, dysfunction, and cardiac overload (1, 2). Medial calcinosis highly correlates with cardiovascular morbidity and mortality (3), and calcification of arterial media commonly occurs in patients with chronic kidney disease (CKD) (2, 4). Patients with CKD and no detectable vascular calcification have 8-yr all-cause survival rates of ∼90% compared with 50% survivability in age-matched patients with medial calcification (5). Imbalanced serum calcium and phosphorous levels elevate the risk of medial calcinosis in patients with CKD. Impaired renal excretion of phosphorous also leads to abnormal bone remodeling and mediates osteogenic differentiation of vascular smooth muscle cells (VSMCs) in the arterial walls (6).

Osteogenic differentiation of resident VSMCs and release of calcifying extracellular vesicles (EVs) mediate nucleation and growth of ectopic vascular calcification (3, 7). This process mimics aspects of the physiological mineralization of osteoblasts and chondrocytes in bone via the release of matrix vesicles (8). Although calcifying EVs released into the vascular wall and bone matrix vesicles contribute to similar end points of mineralization, they originate through different pathways (9, 10). The development of pharmaceuticals that target mechanisms specific to the formation of vascular calcifying EVs could avoid deleterious off-target effects on bone. Production of a specific subset of calcifying EVs from VSMCs requires caveolin-1 (CAV1), a scaffolding membrane protein (11). CAV1 resides in caveolar domains, small invaginations (50–100 nm) on the plasma membrane, which consists of the caveolin protein family, cholesterol, sphingolipids, and receptors (12, 13). Caveolar functions include intra-/extracellular lipid transfer, endocytosis, mechanotransduction, and signaling mediation (13, 14). Calcifying VSMCs release CAV1-enriched EVs, and CAV1 knockdown abrogates calcification in these cells (11).

Epidermal growth factor receptor (EGFR) is a tyrosine kinase transmembrane glycoprotein (15), which localizes abundantly in caveolar domains. EGFR interacts with and modulates CAV1 trafficking (16) and recruits signaling proteins to caveolar domains (17). EGFR actively participates in human cancer progression, and EGFR tyrosine kinase inhibition has become a widely used strategy in cancer therapies (18). Both CAV1 and EGFR are elevated during breast cancer progression (19), and clinical studies indicate that overexpression of an EGFR family member in breast cancer associates with increased ectopic calcification (20). In cardiovascular pathogenesis, elevated EGFR activity correlates with oxidative stress and chronic inflammation (21). EGFR inhibition in apolipoprotein E-deficient mice fed a high-fat diet prevented atherosclerotic plaque development (21). In vascular calcification, the EGFR/ERK signaling axis plays a role downstream of thrombomodulin, a plasma membrane glycoprotein, by Gas6 regulation and promotes VSMC apoptosis and calcification (22). However, the direct targeting of EGFR in VSMC-mediated calcification has not been reported.

Given these associations and the known interactions between CAV1 and EGFR, we hypothesized that EGFR inhibition would prevent vascular calcification by mitigating the biogenesis of calcifying EVs. We showed that EGFR inhibition reduces the release of procalcific CAV1-positive EVs and prevents calcification in osteogenic VSMC cultures and in CKD mice fed a high-phosphate diet. Furthermore, we computationally analyzed 7,651 individuals in the Multi-Ethnic Study of Atherosclerosis (MESA) and Framingham cohorts, revealing a positive correlation between predicted serum EGFR and coronary artery calcification (CAC) measured by computed tomography. Interestingly, EGFR inhibitor treatment also significantly reversed bone mineral loss in the CKD mice. Given the demonstrated clinical safety, our data suggest that EGFR inhibition could represent a viable therapeutic strategy to prevent vascular calcification without deleterious bone effects in patients with CKD.

METHODS

Chronic Kidney Disease and Vascular Calcification Mouse Model

The in vivo study was approved by the Institutional Animal Care and Use Committee (IACUC) at Florida International University under Protocol AN20-006 and conformed to current NIH guidelines. The experimental design was based on procedures established in a previous study to induce CKD and vascular calcification in mice (23). Wild-type C57BL/6J mice (8 wk old; n = 38, 19 per biological sex) were fed an adenine-supplemented diet (0.2%, TestDiet, Richmond, IN) for 6 wk to induce severe kidney injury. The mice then received a diet containing 1.8% phosphate (TestDiet, Richmond, IN) and 0.2% adenine for an additional 2 wk to induce medial calcinosis. Along with this calcifying diet, a group of mice (n = 19) received daily tyrphostin AG1478 (10 mg/kg mouse, Millipore Sigma, T4182) via oral gavage. The remaining mice (n = 19) received vehicle treatment (1% wt/vol, carboxymethylcellulose sodium salt, Sigma, C5678). For nondiseased controls, a third group of mice (n = 12, 6 per biological sex) were fed a regular chow diet and received the vehicle for the final 2 wk. During the oral gavage, animals were partially anesthetized using isoflurane (1%, Patterson Veterinary, 07-893-1389, in 2 L·min⁻¹ oxygen flow). All animals received a tail vein injection with the calcium tracer OsteoSense 680EX (80 nmol·kg⁻¹ mouse, PerkinElmer, NEV10020EX) 48 h before euthanasia. At the study end point, mice were anesthetized with isoflurane (1%, in 2 L·min⁻¹ oxygen flow) followed by retro-orbital bleeding for blood collection. Mice were then immediately euthanized by laceration of the diaphragm before tissue collection. After resection, the aortas were imaged using a near-infrared scanner (LI-COR Odyssey) to visualize the vascular calcification burden. A custom MATLAB script quantified the total area of the calcium tracer, which was normalized to the total scanned aorta area.

Immediately after scanning, the tissue was incubated in a digestive solution (24) of 0.25 M sucrose (Sigma, S7903), 0.12 M NaCl (Fisher Chemical, BP358), 0.01 M KCl (Fisher Chemical, P217), 0.02 M Tris hydrochloride (Fisher Chemical, BP153), and 600 U/mL collagenase (Worthington Biochemical, LS004174) for 2 h at 37°C. The solution was then centrifuged at 1,000 g for 15 min to remove cell debris and at 33,000 g for 30 min to remove microvesicles. Finally, the supernatants were ultracentrifuged (Beckman Coulter, Optima MAX-TL) at 100,000 g for 1 h to isolate the EVs of interest. The pellet was suspended in RIPA lysis and extraction buffer (G Biosciences, 786-489) supplemented with pierce protease inhibitor (Thermo Scientific, A32963). To yield sufficient protein concentration for analysis, EVs isolated from two to three aortas were pooled.

Osteogenic Stimulation, In Vitro Calcification, and Extracellular Vesicle Isolation

Primary human coronary artery vascular smooth muscle cells (VSMCs, ATCC, PCS-100-021) were cultured using vascular smooth muscle cell media and growth kit (ATCC, PCS-100-042). VSMCs (passages 4–6) were harvested using 0.05% trypsin-EDTA solution (Caisson Laboratories, TRL04) and seeded with a density of 26,320 cells·cm⁻² and incubated for 72 h at 37°C, 5% CO₂ with controlled humidity before treatment. VSMCs were treated with either control media, consisting of DMEM (HyClone, SH30022.01), 10% vol/vol bovine calf serum (iron supplemented, R&D Systems, S11950), and 1% vol/vol penicillin-streptomycin (Gibco, 15070-063) or with an osteogenic media (OS) optimized to induce calcification (25, 26). OS media were supplemented with 10 mM β-glycerophosphate (Sigma, 13408-09-8), 0.1 mM l-ascorbic acid (Sigma, 113170-55-1), and 10 nM dexamethasone (Sigma, 50-02-2). To assess the role of EGFR inhibition, tyrphostin AG1478 (Millipore Sigma, T4182) was dissolved in the vehicle (DMSO-methanol, 1:1) and added to OS media to a final concentration of 2.5 µM. An equal volume of the vehicle was added to the control and OS groups. To confirm the specificity of our EGFR inhibitor (AG1478), we also used another EGFR inhibitor, PD153035 (Selleck Chemicals, S6546, 2.5 µM). We found that 28 days in OS culture media led to robust calcification by VSMCs; therefore, all cultures (n = 3, independent donors, male and female) were treated for 28 days and media were replaced every 3 days. On days 6, 13, 20, and 27, the media were replaced by an extracellular vesicle-free (EV free) media (ultracentrifuged for 15 h at 100,000 g at 4°C to remove background EVs common in the serum). After 24 h, conditioned media were collected on days 7, 14, 21, and 28. Collected media were centrifuged at 1,000 g for 5 min to remove cell debris. EV isolation was performed using ultracentrifugation at 100,000 g for 1 h.

Osteoblasts (from human fetus, hFOB 1.19, ATCC, CRL-11372) were cultured and grown in DMEM containing 10% vol/vol bovine calf serum and 1% vol/vol penicillin-streptomycin. Osteoblasts (passages 4–6) were harvested using 0.25% trypsin-EDTA solution (Caisson Laboratories, TRL01), seeded with a density of 5,200 cell·cm⁻², and incubated for 24 h at 37°C and 5% CO₂ with controlled humidity. The cells were treated in three groups of control, OS, and OS supplemented with tyrphostin AG1478 (2.5 μM) for 21 days and media were changed every 3 days. When these were compared with VSMCs, we observed more rapid mineralization in osteoblasts cultured in OS with full matrix mineralization apparent after 21 days. Similar to the VSMC experiments, EV-free media were added to the cultures on days 6, 13, and 20 and collected 24 h later on days 7, 14, and 21. Collected media were centrifuged at 1,000 g for 5 min to remove cell debris. Matrix vesicles were isolated using the ultracentrifugation at 100,000 g for 1 h.

Alizarin Red S Staining and Quantification

At the end of experiments (28 and 21 days of treatment for VSMCs and osteoblasts, respectively), media were removed, and the cells were fixed using formalin (10%, Fisher Chemical, SF100) for 15 min. To visualize in vitro calcification, Alizarin Red S stain (ARS, Ricca, 500-32) was added to the wells and incubated for 30 min at room temperature. The stain was then removed, and the cells were washed three times with milliQ water. To quantify the in vitro calcification, ARS stain was extracted using acetic acid (1.67 M, Fisher Chemical, A38S) on a shaker. After 30 min, the supernatants were collected, briefly vortexed, and heated at 85°C for 10 min. The samples were then cooled on ice for 5 min and centrifuged at 20,000 g for 15 min to remove background particles. Sample absorbance of 405-nm light was measured using a multimode reader (BioTek, Synergy HTX). For tissue ARS staining, aortic samples were cryosectioned with a thickness of 18 μm to preserve mineral. After removal of excess Tissue-Plus optimal cutting temperature (OCT) compound with PBS, the samples were incubated with ARS stain for 5 min, followed by one wash with PBS and one wash with milliQ water (27), shown in Supplemental Fig. S1A (all Supplemental material is available at https://doi.org/10.6084/m9.figshare.22001615).

Kidney Histological Analysis

To assess histological changes in kidneys due to renal injury, hematoxylin and eosin (H&E) staining was performed. The kidneys resected from the mice were fixed using formalin (10%) for 3 h. Tissues were embedded using Tissue-Plus OCT (Fisher Scientific, 23-730-571). The samples were cryosectioned with a thickness of 12 μm and stained using rapid chrome hematoxylin-and-eosin staining kit (Thermo Scientific, 9990001), shown in Supplemental Fig. S1B.

Quantitative Real-Time Polymerase Chain Reaction

Following 7 or 14 days in control, OS, or OS plus EGFR inhibitor media, VSMCs and osteoblasts were lysed in 1 mL of TRIzol solution (Invitrogen, 15596018). Total RNA was isolated according to the manufacturer’s protocol. To perform the quantitative real-time polymerase chain reaction (qRT-PCR), Power SYBR Green RNA-to-CT 1-Step Kit (Applied Biosystems, 4391178) was used. Isolated template RNA (50 ng) was added to each reaction for qRT-PCR. The results were normalized to glyceraldehyde 3-phosphate dehydrogenase (GAPDH) expression level as the housekeeping control. The relative gene expression levels were calculated using comparative CT method, considering control groups as the reference. The following human primers were purchased from Integrated DNA Technologies (IDT); GAPDH, forward: CTTCGCTCTCTGCTCCTCCTGTTCG, and reverse, ACCAGGCGCCCAATACGACCAAAT; RUNX2, forward, GCTCTCTAACCACAGTCTATGC, and reverse, AGGCTGTTTGATGCCATAGT; ALPL, forward, GGAGTATGAGAGTGACGAGAAAG, and reverse, GAAGTGGGAGTGCTTGTATCT; Osteocalcin (BGLAP), forward, TCACACTCCTCGCCCTATT, and reverse, CCTCCTGCTTGGACACAAAA.

To isolate RNA from the resected kidneys, the tissues were homogenized using a grinder (Sigma, Z529672) and lysed in 1 mL of TRIzol solution. After 10 min incubation at room temperature, the samples were centrifuged at 12,000 g for 10 min at 4°C. The supernatants were collected and 200 μL of chloroform (Sigma-Aldrich, C2432) were added to each sample. The samples were vortexed, incubated at room temperature for 10 min, and centrifuged for 15 min at 12,000 g at 4°C. The aqueous phase was collected from each sample, and 500 μL of isopropanol were added; the samples were vortexed, incubated for 15 min at room temperature followed by 15 min on ice, and centrifuged at 21,000 g for 15 min. The supernatants were discarded, and pellets were washed twice with 500 μL cold ethanol (75% vol/vol) and centrifuged at 21,000 g for 5 min (28–30). The isolated RNA templates were heated at 65°C for 15 min, and the concentrations were measured using a spectrophotometer (NanoDrop Lite, Thermo Scientific). Power SYBR Green RNA-to-CT 1-Step Kit with 100 ng isolated template RNA per reaction was used. The following mouse primers were purchased from Eurofins Scientific; Gapdh, forward, AACGACCCCTTCATTGAC, and reverse, TCCACGACATACTCAGCAC; Col1a1, forward, CCTCAGGGTATTGCTGGACAAC, and reverse, ACCACTTGATCCAGAAGGACCTT; Tgfb1, forward, TGGAGCAACATGTGGAACTC, and reverse, CAGCAGCCGGTTACCAAG.

Alkaline Phosphatase Activity Assay

To assess the activity of cellular tissue nonspecific alkaline phosphatase (TNAP), a colorimetric assay kit (BioVision, K412) was used. VSMCs (n = 3) after 14 days and osteoblasts (n = 3) after 7 days were lysed in 120 μL of assay buffer. Each sample (80 μL) were mixed with 50 μL of 5 mM p-nitrophenyl phosphate (pNPP) solution and incubated for 60 min at 25°C. The colorimetric change resulting from the reaction was detected using a plate reader to measure absorbance at 405 nm. The results were normalized to the total protein for associated samples measured by a bicinchoninic acid (BCA) protein assay (BioVision, K813). For EV or matrix vesicle TNAP activity measurement, after ultracentrifugation at 100,000 g for 1 h, the pellets were resuspended in 120 μL of assay buffer. The assessment was performed using the same assay protocol described for cellular TNAP activity and the results were normalized to the total protein for each sample. For mouse serum TNAP activity, the samples were diluted 1:20 and assessed according to the manufacturer’s protocol.

Serum Creatinine and Urea Nitrogen Assessment

To measure serum creatinine, a colorimetric assay (Cayman Chemicals, 700460) was used; 15 µL of each collected serum were added to 200 µL of a solution of assay reaction buffer and color reagent (1:1), incubated for 1 min at room temperature, and measured at 495 nm using a multimode reader. The absorbance was measured at 495 nm for the second time after 7 min. The changes in optical density (ΔOD) for each sample were associated to the creatinine concentration according to the manufacturer’s protocol.

To assess serum urea nitrogen, the serum samples were diluted 1:10; a colorimetric assay (Invitrogen, EIABUN) measured the serum urea nitrogen. Briefly, collected serum (50 µL) was mixed with 150 µL of assay color solution (reagent A-reagent B, 1:1) and incubated at room temperature for 30 min. The colorimetric changes were measured at 450 nm.

Extracellular Collagen Assessment

After 28 days of treatment, soluble collagen was extracted from the cultures using acetic acid (0.5 M) through overnight incubation at 4°C. A colorimetric assay, Sircol soluble collagen assay (Biocolor, S1000), measured the total soluble extracellular matrix (ECM) collagen in each group. Samples were prepared and assessed according to the manufacturer’s protocol. Results were then normalized to the total protein measured using BCA assay.

Subcellular Protein Fractionation for VSMCs and Aortas

Wild-type C57BL/6J mice (8 wk old; n = 20, female) received the adenine-supplemented diet for 6 wk to induce CKD, followed by an additional 2 wk of the diet containing 1.8% phosphate and 0.2% adenine to induce medial calcinosis. Mice were split into two groups (10 per group). The first group received daily tyrphostin AG1478 (10 mg/kg mouse), whereas the other group received vehicle (1% wt/vol, carboxymethylcellulose sodium salt). At the study end point, the animals were euthanized, and the aortas were resected. A subcellular protein fractionation kit for tissue (Thermo Fisher, 87790) was used to isolate cellular cytosolic fraction from the resected aortas, using the manufacturer’s protocol. Briefly, the tissues were minced and homogenized using a grinder. The samples were then incubated in a cytoplasmic extraction buffer for 10 min at 4°C, followed by centrifugation at 1,000 g for 5 min. The supernatants yielded the cytosolic fraction. To obtain sufficient protein for analyses, two aortas were pooled per data point.

VSMCs were treated with control, OS, and OS supplemented with tyrphostin AG1478 (2.5 μM) for 14 days. At the experiment end point, using a subcellular protein fraction kit for cultured cells (Thermo Fisher, 78840), cytosolic fraction was isolated according to the manufacturer’s protocol. Briefly, VSMCs were harvested using 0.25% trypsin solution and resuspended in cytoplasmic extraction buffer. After 10 min incubation at 4°C, the samples were centrifuged at 1,000 g for 5 min, and the supernatants were collected as cytosolic fractions. The protein concentration for aortic tissue and VSMC fractions were quantified using a BCA assay and samples were prepared for protein immunoblotting.

Immunoprecipitation and Lipid Raft Isolation

Following 14 days of treatment, VSMCs (n = 3) were lysed using Pierce immunoprecipitation lysis buffer (Thermo Scientific, 87788) supplemented with protease inhibitor. A Dynabeads Protein G immunoprecipitation kit (Invitrogen, 10007D) was used to precipitate CAV1 from the cell lysates (11). Briefly, Dynabeads coated with protein G were incubated with either CAV1 antibody (5 μg, Abcam, ab17052) or IgG mouse control antibody (5 μg, Proteintech, B900620) by rotation for 3 h at 4°C. After removal of supernatants using a magnet, 100 μg of protein were loaded to the beads and incubated for 5 h at 4°C while rotating. After removal of supernatants and three washes with washing buffer, 40 μL of elution buffer (from the kit) and 20 μL of 1:1 NuPAGE sample reducing agent (Thermo Scientific, NP0009) and NuPAGE LDS sample buffer (Thermo Fisher Scientific, NP0007) were added to the beads. The samples were incubated by rotation at 4°C for 30 min, and denatured at 70°C for 10 min. After removal of beads by a magnet, the samples were ready for protein immunoblotting.

To isolate lipid rafts, VSMCs (n = 3) were treated under control, control plus EGFR inhibitor, OS, or OS plus EGFR inhibitor for 14 days. Then, the cells were lysed in a buffer containing 25 mM HEPES (Fisher Scientific, BP310-100), 150 mM NaCl, 1 mM PMSF (Boston Bioproducts, PI120), 1 mM EDTA (Invitrogen, AM9260G), 1% vol/vol Triton-X100 (Fisher Scientific, BP151-100), and protease inhibitor. Five gradient layers were prepared using OptiPrep density gradient medium (Sigma-Aldrich, D1556), including 35% (with cell lysates) and 30%, 25%, 20%, and 0% (with lysis buffer), and loaded to ultracentrifuge tubes (Optiseal bell, Beckman Coulter, 361621), respectively (total volume of 4.5 mL). Samples were ultracentrifuged at 4°C and 110,000 g for 4 h; nine fractions (500 μL per fraction) were isolated for each group and used for protein immunoblotting.

Gel Electrophoresis and Protein Immunoblotting

VSMCs, osteoblasts, isolated EVs (either from cells or mouse aortas), and matrix vesicles (from osteoblasts) were lysed in RIPA lysis and extraction buffer supplemented with protease inhibitor. After Laemmli SDS-sample buffer (1:4 vol/vol, Boston BioProducts, BP-110R) was added to each lysate, the samples were denatured at 100°C for 10 min, loaded into 7.5%–12% 1-mm SDS-PAGE gel (15 to 20 μg protein per lane), and run at 170 V. The proteins were then transferred to Trans-Blot turbo nitrocellulose membranes (Bio-Rad, 1704158) at 25 V for 7 min. To quantify the total protein, the membranes were stained using 2% wt/vol Ponceau stain (Alfa Aesar, AAJ6074409) for 20 min, followed by one wash with 5% acetic acid and milliQ water for 5 min. After imaging, the intensity of each lane was measured in ImageJ for total protein normalization. Membranes were blocked with 5% wt/vol bovine serum albumin (HyClone, SH30574.01) in TBS-Tween (1×) for 1 h. The membranes were incubated with primary antibodies of interest, including CAV1 (1:200, Abcam, ab2910), EGFR (1:00, EMD Millipore, 06-874), CD63 (1:200, Abcam, ab231975), BMP2 (1:200, Abcam, ab214821), RUNX2 (1:200, Abcam, ab114133), Syntenin (1:200, Abcam, ab19903), GAPDH (1:100, Abcam, ab181602, as common cytosolic marker), and annexin V (1:200, proteintech, 11060-1-AP) overnight at 4°C. After three washes with TBS-Tween (1×), the membranes were incubated with secondary antibody (1:1,000, Li-Cor) for 1 h, followed by three washes with TBS-Tween (1×). The protein bands were visualized with Odyssey CLx scanner (Li-Cor) and quantified using Image Studio Lite software (Li-Cor). All Western blot images and corresponding Ponceau stains used for normalization are provided in Supplemental Figs. S3 and S4.

Immunofluorescence Staining and Imaging

VSMCs were fixed after 14 days of culture using formalin (10%) for 15 min and washed with PBS. A solution of PBS and Triton-X (0.1% vol/vol) permeabilized the plasma membrane for 10 min at room temperature. To avoid nonspecific antibody binding, the cells were incubated with a blocking buffer solution, consisting of BSA (1% wt/vol) and glycine (22.5 mg/mL) in PBS for 30 min at room temperature. The cells were next incubated for 2 h with primary antibody against CAV1 (1:200) and washed three times with PBS. Cells were then incubated with a secondary antibody, Alexa Fluor 594 (1:500, Abcam, ab150080), for 1 h at room temperature, followed by three washes with PBS. To visualize actin filaments, samples were incubated for 20 min with Phalloidin-iFluor 488 conjugate (1:50, Cayman Chemical, 20549) followed by three washes with PBS.

Resected mouse aortas were fixed in formalin (10%) for 2 h. The tissues were rinsed with PBS and embedded in OCT. The samples were cryosectioned with a thickness of 7 μm. The samples were incubated with a blocking buffer containing donkey serum (10% vol/vol), Triton-X (0.3% vol/vol), and BSA (1% wt/vol) in PBS for 1 h at room temperature. After blocking buffer removal, a solution of donkey serum (1% vol/vol), Triton-X (0.3% vol/vol), BSA (1% wt/vol) in PBS, with primary antibody against either CAV1 (1:200), EGFR (1:100), or TNAP (1:200) was added to the samples. After 1-h incubation at room temperature, the primary antibody solution was removed, and the samples were washed with PBS. Secondary antibody, Alexa Fluor 594 (1:500, Invitrogen, A21207) was added to the samples and incubated for 1 h at room temperature. After the samples were washed with PBS, they were stained with DAPI (0.2 μg/mL, Cayman Chemical, 14285) for 10 min and washed with PBS. The samples were mounted using Flouromount (Sigma Millipore, F4680). A confocal microscopy system (Eclipse Ti, Nikon) was used to image both cellular and tissue samples.

X-Ray Computed Tomography

Femurs were dissected from mice, wrapped in parafilm, and imaged directly in a Nikon XT H 225 scanner (macro-CT, Nikon Metrology, Tring, UK). The raw transmission images were reconstructed using commercial image reconstruction software package (CT Pro 3D, Nikon Metrology, Tring, UK), which uses a filtered back-projection algorithm. The scan was performed using 80-kV beam energy, 70-μA beam current, and a power of 5.6 W. A PerkinElmer 1620 flat panel detector was used, with 200-μm pixel size. The resulting effective pixel size was 5 μm. The exposure time per projection was 0.5 s, and a total of 1,601 projections were acquired, resulting in a scanning time of ∼13 min per sample. Bone structural parameters, including thickness and volume fraction [the ratio of bone volume (BV) to total volume (TV)], for both cortical and trabecular regions were assessed using a plug-in module, BoneJ, in ImageJ (NIH) (31–33).

Identification of Instrumental Variables for Mendelian Randomization

Instrumental variables (IVs) were selected using an agnostic P value threshold, P < 5 × 10⁻⁶, as advised by the methodological literature on Mendelian randomization (MR) (34). Single-nucleotide polymorphisms (SNPs) associated with significantly elevated serum EGFR concentration (P < 5 × 10⁻⁶) from a previous proteomics study were compared against genotyped SNPs in the Multi-Ethnic Study of Atherosclerosis (MESA) SNP Health Association Resource (SHARe), and all SNPs presented in both the proteomics study and MESA genotyping data associated beyond this P value threshold were included as IVs for the MR analysis (35). In total, three SNPs of rs12666347, rs2371816, and rs7806938 were included. The same three IVs and measure of CAC were used to replicate the significance of the MR analysis and validate results in the Offspring Cohort of the Framingham Heart Study (FHS).

Calculation of SNP-EGFR and SNP-CAC Association in the MESA and FHS Cohorts

Effect sizes of each SNP on EGFR concentration, as well as their standard errors, were extracted from the publicly available summary statistics (35). To calculate the effect sizes of each SNP on calcification levels, we identified 1,896 individuals from the FHS Offspring cohort, and 5,755 individuals who completed MESA Exam 1 who had available genotyping information. For each of these individuals, genotyping information, age, sex, study site, race, and Agatston score were extracted. Agatston scores are a measure of CAC determined through cardiac imaging, with an increasing Agatston score representing increased CAC. Associations between each IV SNP and CAC is calculated using logistic regression, treating Agatston scores as a binary variable (=0 vs. >0) and including age, sex, study site, and race as covariates in the model. All analyses were conducted using the R programming language.

Mendelian Randomization

Following the identification of SNP-CAC and SNP-EGFR association and standard error values, MR analysis was performed to determine the presence and estimate the magnitude of the causal effect that elevated serum EGFR has on CAC. Eleven different regressions were included in the MR analysis to correct for possible pleiotropic effects, a possible source of confounding. Included regressions were simple median, weighted median, penalized median, inverse-variance weighted (IVW), penalized IVW, robust IVW, penalized-robust IVW, MR-Egger, penalized MR-Egger, robust MR-Egger, and penalized-robust MR-Egger. MR analysis was performed using the Mendelian Randomization package in R (36, 37). We accounted for multiple testing errors using a Bonferroni-adjusted 0.05 significance level of 0.0045 (0.05/11).

Statistics

Data are presented as means of independent replications, and error bars represent standard error of the mean. The reported n values represent independent biological replicates. Statistical significance between groups was calculated using one-way ANOVA with Tukey’s post hoc test in GraphPad Prism 8. A P value < 0.05 was considered statistically significant. In case of comparison between two groups, the statistical significance was calculated using t test with P values < 0.05.

RESULTS

EGFR Inhibition Reduces Vascular Calcification in a CKD Mouse Model

Visualization of the calcium tracer, OsteoSense, showed widespread vascular calcification in CKD mice compared with the chow-fed control group. Daily EGFR inhibitor gavage (10 mg/kg/mouse) for 2 wk dramatically reduced vascular calcification in CKD animals (Fig. 1A, also confirmed with ARS staining, representative images shown in Supplemental Fig. S1A). Quantification of the OsteoSense intensity revealed a significant reduction in vascular calcification in the EGFR inhibited group (P < 0.0001), as shown in Fig. 1B. The level of serum EGFR was elevated in the CKD group compared with chow-fed animals (P = 0.038), with no significant difference between CKD and EGFR inhibited groups (P = 0.78) (Fig. 1C). Serum TNAP activity, urea nitrogen, and creatinine (Fig. 1, D–F) in CKD animals were significantly elevated compared with the control group (P < 0.0001). EGFR inhibition did not reduce serum TNAP activity (P = 0.06), urea nitrogen (P = 0.82), and creatinine (P = 0.94). Gene expression of common renal fibrosis markers, Tgfb1 and Col1a1 (Fig. 1, G and H), were significantly increased in both CKD mice (P = 0.047 and P = 0.04 for Tgfb1 and Col1a1, respectively) and CKD mice treated with EGFR inhibitor (P = 0.04 and P = 0.046 for Tgfb1 and Col1a1, respectively) when compared with chow-fed controls, with no significant differences between the CKD groups (P = 0.92 and P = 0.99 for Tgfb1 and Col1a1, respectively). Qualitative assessment of histological sections of resected kidney tissues showed enlarged tubular structures in both CKD and EGFR inhibitor treated CKD groups, compared with the chow-fed control (Supplemental Fig. S1B). These results indicate that EGFR inhibition reduces vascular calcification in CKD animals independent of effects on renal injury.

Figure 1.

Epidermal growth factor receptor (EGFR) inhibition prevents vascular calcification in vivo and in vitro. A: visualization of vascular calcification using calcium tracer OsteoSense. B: quantification of the OsteoSense to correlate with vascular calcification burden (n = 50). C: serum EGFR level collected from mouse groups. D: serum tissue nonspecific alkaline phosphatase (TNAP) activity collected from mouse groups. E: serum urea nitrogen level collected from mouse groups. F: serum creatinine level collected from mouse groups. G and H: gene expression of renal fibrotic markers, Tgfb1 and Col1a1. I: in vitro calcification visualization using Alizarin Red S staining and quantification. J and K: gene expression of osteogenic markers, RUNX2 and ALPL in vascular smooth muscle cells (VSMCs) following 14 days of treatment. L: extracellular matrix collagen accumulation in VSMC cultures. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, ANOVA with Tukey’s post hoc test. ▴, ●, and ■, female, male, and cell culture replications, respectively. CKD, chronic kidney disease; OS, osteogenic media.

EGFR Inhibition Attenuates In Vitro Vascular Smooth Muscle Cell Calcification

VSMCs calcified following 28 days of culture in OS media, as shown by ARS staining (Fig. 1I, representative image). Treatment of OS cultures with both EGFR inhibitors (AG1478 or PD153035) abrogated in vitro calcification of the VSMCs (Fig. 1I, and Supplemental Fig. S1D, respectively). Protein levels of RUNX2 and bone matrix protein 2 (BMP2) significantly decreased in EGFR inhibited VSMC cultures (Supplemental Fig. S2, A and B). However, gene expression analysis of the common osteogenic markers, RUNX2 and ALPL, revealed that VSMCs cultured in both OS (P = 0.023 and P = 0.012 for RUNX2 and ALPL, respectively) and OS treated with EGFR inhibitor (P = 0.006 and P = 0.002 for RUNX2 and ALPL, respectively) acquired an osteogenic phenotype after 14 days of culture (Fig. 1, J and K), with no significant differences between the groups (P = 0.46 and P = 0.20 for RUNX2 and ALPL, respectively). Moreover, OS media promoted the accumulation of ECM collagen in vitro, which creates a platform for calcifying EVs to initiate calcification (26) (Fig. 1L); EGFR inhibition did not affect the ECM collagen accumulation (P = 0.99). These data suggest that EGFR inhibition attenuates VSMC calcification downstream of changes in VSMC phenotype.

EGFR Inhibition Alters CAV1/TNAP Cellular Trafficking

Both OS cultured VSCMs and OS cultured VSMCs treated with EGFR inhibitor significantly increased the total level of cellular CAV1 protein in VSMCs compared with the control group (P < 0.0001) (Fig. 2A). OS media also increased cellular EGFR in VSMCs compared with the control group (P = 0.019, Fig. 2B). EGFR inhibition prevented the OS-induced increase in EGFR protein (P = 0.038). Parallel to the gene expression data (Fig. 1K), both OS cultured VSMCs and OS cultured VSMCs treated with EGFR inhibitor exhibited elevated cellular TNAP activity (P = 0.025 and P = 0.02, respectively, compared with control) (Fig. 2C). Confocal micrographs of VSMCs (Fig. 2D, and Supplemental Fig. S2C) showed alignment of CAV1 protein along actin filaments in VSMCs cultured in OS media. In the OS cultured VSMCs treated with EGFR inhibitor, larger clusters of CAV1 were observed between filaments. Subcellular protein fractionation of VSMCs revealed that cytosolic fractions of both CAV1 and TNAP were elevated in EGFR inhibited cultures compared with control (P = 0.021 and P = 0.002, respectively) and OS groups (P = 0.047 and P = 0.004, respectively, Fig. 2, E and F).

Figure 2.

Epidermal growth factor receptor (EGFR) inhibition modulates caveolin-1 (CAV1) trafficking in vascular smooth muscle cells (VSMCs). Cellular level of CAV1 (A), EGFR (B), and tissue nonspecific alkaline phosphatase (TNAP; C) activity in VSMCs after 14 days of culture. D: confocal micrographs of CAV1 distribution in VSMCs following 14 days of treatment (×1,200, scale bar = 0.5 µm). Cytosolic level of CAV1 (E) and TNAP (F) protein following 14 days of treatment. G: CAV1 level on extracellular vesicles (EVs) isolated from VSMC cultures after 14, 21, and 28 days. H: TNAP activity of the EVs isolated from VSMC cultures after 14, 21, and 28 days. EV level of EGFR (I), annexin V (J), and CD63 (K) liberated from VSMCs on day 28 of treatment. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, ANOVA with Tukey’s post hoc test. ■, cell culture replications. CKD, chronic kidney disease; OS, osteogenic media.

To compare the in vitro observations to the in vivo studies, qualitative analysis of confocal micrographs of CAV1, EGFR, and TNAP immunofluorescence in the aorta of mice indicated elevation of all three proteins in CKD mice and CKD mice treated with EGFR inhibitor, compared with chow-fed controls (Fig. 3A). Subcellular protein fractionation of aorta indicated higher cytosolic CAV1 and TNAP proteins in EGFR inhibited CKD animals compared with the CKD group (P = 0.041 and P = 0.0001, and P = 0.018, respectively), shown in Fig. 3, B–D, analogous to in vitro data. Both in vitro and in vivo analyses suggest that EGFR inhibition alters CAV1 subcellular distribution.

Figure 3.

Epidermal growth factor receptor (EGFR) inhibition redistributes cavolin-1 (CAV1) and tissue nonspecific alkaline phosphatase (TNAP) in vivo. A: immunofluorescence staining of CAV1, TNAP, and EGFR in aortic tissue. Cytosolic protein levels of CAV1 (B), TNAP (C), and EGFR (D) from aortic tissue. Extracellular vesicle (EV) levels of CAV1 protein (E) and TNAP (F) activity isolated from mouse aortas. Scale bar for ×10 and ×100, 200 and 20 µm, respectively. *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001, ANOVA with Tukey’s post hoc test. ▴ and ●, female and male replications, respectively. CKD, chronic kidney disease; OS, osteogenic media.

EGFR Inhibition Reduces the Release of CAV1-Positive EVs with High TNAP Activity In Vitro and In Vivo

EVs isolated from the aortas of CKD mice exhibited significantly elevated CAV1 protein and TNAP activity compared with chow-fed controls (P < 0.0001 and P = 0.03 for CAV1 and TNAP activity, respectively, Fig. 3, E and F). The EVs isolated from the CKD mice treated with EGFR inhibitor had significantly lower CAV1 protein and TNAP activity (P < 0.0001 and P = 0.007 for CAV1 and TNAP activity, respectively, Fig. 3, E and F). These data suggest that EGFR inhibition decreased formation of calcifying EVs in the CKD mouse aorta in vivo.

The EGFR inhibition led to similar outcomes in vitro. EVs obtained from VSMCs cultured in OS media contained significantly elevated CAV1 after 14, 21, and 28 days compared with controls (Fig. 2G). EV TNAP activity increased in OS VSMC cultures over time (Fig. 2H). EGFR inhibition significantly reduced the release of EV CAV1 (Fig. 2G) and EV TNAP activity (Fig. 2H). Furthermore, EVs isolated from VSMCs cultured in OS media were enriched with EGFR and annexin V, a calcium-binding protein, (Fig. 2, I and J); EGFR inhibited groups showed reduced levels of these proteins in the EVs. Of note, the level of CD63, a common exosomal marker, was preserved across the in vitro groups following 28 days of culture (P = 0.9 between the groups), as shown in Fig. 2K. AG1478 also did not alter the levels of EV syntenin 1, a common marker of secreted EVs (38), whereas PD153035 significantly decreased EV syntenin 1 (Supplemental Fig. S2E). The activity of TNAP was significantly reduced by EGFR inhibition (Supplemental Fig. S2, F and G). These data suggest that EGFR inhibition prevents the release of calcifying EVs independently of alterations to traditional exosome secretion.

EGFR Inhibition Attenuates the Interaction between CAV1 and EGFR and Retains CAV1 in Lipid Rafts

CAV1 immunoprecipitation from VSMC lysates and Western blot analysis for EGFR showed that in both control and OS cultures, CAV1 interacts with EGFR in VSMCs (Fig. 4A). EGFR inhibition significantly reduced the interaction between CAV1 and EGFR in both control (P = 0.0015) and OS (P < 0.0001) cultures (Fig. 4B). Using ultracentrifugation to perform a density-gradient based separation of VSMC lysates, we showed that CAV1 redistributes from lighter fractions associated with lipid rafts to more dense fractions in OS cultures with enrichment in fraction 7 compared with control cultures (Fig. 4, C and D). EGFR inhibition prevented the redistribution of CAV1, leading to a subcellular fractionation profile analogous to control cultures (Fig. 4D).

Figure 4.

Epidermal growth factor receptor (EGFR) inhibition attenuates caveolin-1 (CAV1) and EGFR interaction. A: EGFR and CAV1 immunoblotting after CAV1 immunoprecipitation from vascular smooth muscle cells (VSMCs) following 14 days of treatment. B: densitometry and quantification of the EGFR level. C: CAV1 immunoblotting on isolated lipid rafts. D: densitometry and quantification of CAV1 in isolated lipid rafts. **P < 0.01 and ****P < 0.0001, ANOVA with Tukey’s post hoc test. ■, cell culture replications. OS, osteogenic media.

EGFR Inhibition Does Not Cause Deleterious Effects on Physiological Bone Mineralization

To determine whether the anticalcification effects of EGFR inhibition in VSMCs would cause deleterious effects on physiological bone mineralization, we performed in vitro studies with human osteoblasts and assessed bone density from the treated mice. Both OS and OS cultured osteoblasts treated with EGFR inhibitor committed to osteogenic transition by downregulation of RUNX2 (39, 40) (Fig. 5A) and increased the expression of ALPL and Osteocalcin (BGLAP) (39), after 7 days (Fig. 5, B and C), with no significant differences between the groups (P = 0.9 and 0.9 for ALPL and BGLAP, respectively). Cellular levels of RUNX2 and BMP2 proteins in osteoblast OS cultures remained elevated upon EGFR inhibition (either AG1478 or PD153035, Supplemental Fig. S2, C and D). Parallel to ALPL expression, the osteoblasts demonstrated significantly increased cellular TNAP activity after 7 days in both cultures (Fig. 5D). Alizarin red staining demonstrated in vitro calcification in both groups and quantification of the in vitro calcification showed no significant difference between the groups (P = 0.86, Fig. 5F and Supplemental Fig. S1D).

Figure 5.

Epidermal growth factor receptor (EGFR) inhibition does not prevent osteoblast in vitro calcification. A–C: gene expression of common osteogenic markers, RUNX2, ALPL, and BGLAP, in osteoblasts following 7 days of treatment. D: osteoblast cellular tissue nonspecific alkaline phosphatase (TNAP) activity following 7 days of treatment. E: Alizarin Red S staining and quantification of osteoblast cultures after 21 days. F: osteoblast cellular caveolin-1 (CAV1) following 7 days of treatment. G: CAV1 level on matrix vesicles liberated from osteoblasts on days 7, 14, and 21 of culture. H: TNAP activity of matrix vesicles isolated from osteoblast cultures on days 7, 14, and 21. *P < 0.05, **P < 0.01, and ***P < 0.001, ANOVA with Tukey’s post hoc test. ■, cell culture replications. OS, osteogenic media.

In both OS and OS cultured osteoblasts treated with EGFR inhibitor, cellular CAV1 protein was significantly increased compared with the control group (P = 0.016 and 0.03 for the OS and OS with EGFR inhibitor groups, respectively, Fig. 5E). Matrix vesicles (MV) released by osteoblasts in both OS and OS treated with EGFR inhibitor groups had significantly increased TNAP activity; however, the MVs from these cells had lower levels of CAV1 protein compared with controls on days 14 and 21 in culture (Fig. 5, G and H, and Supplemental Fig. S2, I and J). EGFR inhibition (using AG1478 or PD153035) did not affect the level of syntenin 1 in MVs compared with OS cultures (Supplemental Fig. S2H).

We next assessed the femurs resected from murine groups to analyze the effects of EGFR inhibition on bone mineralization (Fig. 6, A–C). The thickness and bone volume fraction of both trabecular (epiphyseal and metaphysical regions) and cortical bone was significantly reduced in CKD animals compared with chow-fed controls. EGFR inhibition increased the thickness of both trabecular and cortical bone significantly in the CKD mice (P = 0.049 and 0.022 for epiphyseal and metaphysical regions, respectively, and P = 0.004 for cortical bone; Fig. 6, D–F). Interestingly, EGFR inhibition increased the bone volume fraction in trabecular bone, both epiphyseal (P = 0.009) and metaphysical (P = 0.001) regions, compared with CKD animals. Bone volume fraction did not significantly change in cortical bone (P = 0.25; Fig. 6, G–I). Detailed quantification of the bone structural parameters can be found in Supplemental Table S1.

Figure 6.

Epidermal growth factor receptor (EGFR) inhibition does not have deleterious effects on physiological bone mineralization. Three-dimensional (3-D) reconstructions of femoral head (A), cancellous bone (B), and cortical bone (C) resected from mouse groups (scale bar = 0.5 mm). Bone thickness at cortical (D), metaphyseal trabecular (E), and epiphyseal trabecular (F) regions. G–I: bone volume fraction (%) at cortical (G), metaphyseal trabecular (H), and epiphyseal trabecular (I) regions. Numbers represent P values, ANOVA with Tukey’s post hoc test. ▴ and ●, female and male replications, respectively. CKD, chronic kidney disease.

Mendelian Randomization Shows Positive Correlation between Serum EGFR and CAC

Of the 11 MR regressions performed in the MESA cohort, all regressions predicted positive correlation between serum EGFR concentration and CAC (i.e., elevated EGFR concentration predicts increased incidence of elevated CAC). Two of the MR regressions reached statistical significance beyond the Bonferroni-adjusted significance threshold: robust MR-Egger and penalized robust MR-Egger. The intercept tests for the MR-Egger estimates are statistically significant at P = 1.9 × 10⁻⁵, suggesting the presence of vertical pleiotropy among the IV SNPs accounted for in the MR-Egger type regressions (Fig. 7A). The causal estimates of the effect of EGFR concentration on increased CAC are associated with P values < 1 × 10⁻¹⁰ (Supplemental Table S2).

Figure 7.

Clinical data indicate positive correlation between serum epidermal growth factor receptor (EGFR) and coronary artery calcification. A: forest plot summarizing effect estimates of each Mendelian randomization (MR) regression along with their 95% confidence intervals for Multi-Ethnic Study of Atherosclerosis (MESA) cohort. B: offspring cohort of Framingham Heart Study (FHS). Robust MR-Egger and penalized robust MR-Egger estimates of effect are statistically significant and highlighted in red. IVW, inverse-variance weighted.

Replication of the 11 MR regressions in the FHS cohort also yielded significant estimates for the robust MR-Egger and penalized robust MR-Egger regression estimates with P values of 1.17 × 10⁻⁶ for both cohorts (Fig. 7B). However, the intercept test for vertical pleiotropy was not statistically significant (P = 0.06), possibly trending toward significance due to insufficient sample size. However, both regressions suggest a positive causal relation between serum EGFR concentration and CAC (Supplemental Table S2).

DISCUSSION

Despite the recognition that a particular population of EVs participate in vascular calcification, significant knowledge gaps exist into how these specialized structures form and to what extent they mediate mineral deposition. Here, we present new insight into the role of CAV1 trafficking in the formation of calcifying EVs and demonstrate that EGFR inhibition can alter CAV1 trafficking to prevent vascular calcification both in vitro and in vivo downstream of osteogenic changes in cellular phenotype.

Caveolae have a low buoyant density. Translocation of CAV1 to more dense regions in the gradient-based VSMC fractionation analyses indicates trafficking to noncaveolar domains (41) in procalcific conditions (Fig. 4D). The data presented in the current study suggest that physical interactions between EGFR and caveolin-1 are required for the cellular, noncaveolar trafficking mechanisms that lead to calcifying EV biogenesis. EGFR tyrosine kinase inhibition reduces EGFR-CAV1 coimmunoprecipitation (Fig. 4B) and retains CAV1 within less dense membrane fractions associated with caveolae lipid rafts (Fig. 4D). The EGFR inhibition, however, does not alter transition of VSMCs to a procalcifying phenotype (Fig. 1, J–L). By preventing the formation of calcifying EVs, factors expressed during this phenotypic transition accumulate within the VSMCs both in vitro and in vivo (Fig. 2, E and F and Fig. 3, B–D). Blocking the release of the procalcific factors in EVs resulted in reduced calcification in vitro (Fig. 1I) and in vivo (Fig. 1A), demonstrating the relevance of calcifying EVs in mineral formation.

Given the myriad of upstream vascular calcification initiators, altering calcifying EV biogenesis may represent a point of convergence that can be targeted therapeutically (42). Patients with CKD are particularly prone to develop widespread vascular calcification, increasing from 25% of patients in stages 3 and 4 to 50%–80% of the population in stage 5 (43). Previous studies showed that EGFR facilitates tyrosine kinase-mediated phosphorylation of CAV1 and modulates CAV1 trafficking (44–47). Therefore, we hypothesized that EGFR tyrosine kinase inhibition may prevent the CAV1-dependent formation of calcifying EVs in CKD.

We show that inhibiting EGFR tyrosine kinase activity prevents vascular calcification in a CKD mouse model, with 100% survival rate. The in vivo results showed reduced calcium burden in the aorta of CKD mice treated with EGFR tyrosine kinase inhibitor, AG1478. This effect was independent of kidney remodeling as AG1478 treatment did not reduce the expression of common markers of renal injury (Fig. 1, D–H; Supplemental Fig. S1B). Certainly, efforts should be (and are made clinically) to preserve and improve kidney function in patients with CKD. However, the presence of vascular calcification significantly predicts morbidity and mortality in these patients. A therapeutic to prevent and/or reduce vascular calcification could improve morbidity and mortality as other strategies are implemented to improve kidney function. Our results suggest that the reduction of vascular calcification in EGFR inhibited group is independent from worsening or improving kidney damage.

Demonstrating the relevance of the in vitro mechanistic studies to our in vivo analyses, we observed elevated CAV1-positive EVs in the aortae of CKD mice, which was reduced by EGFR inhibition. Similarly, TNAP activity was elevated in EVs isolated from the aortae of CKD mice, whereas EGFR inhibition reduced the activity of this enzyme in the EVs. Calcifying EVs are enriched in annexin V, a collagen-binding Ca²⁺ channel (3, 48). We found that annexin V was elevated in VSMC EVs, which was also reduced by EGFR inhibition. Taken together, these results support our hypothesis and suggest that targeting the CAV1-dependent formation of calcifying EVs by EGFR inhibition reduced vascular calcification in the CKD mouse model. Future studies with additional EGFR inhibitors, both monoclonal antibodies and tyrosine kinase inhibitors, and genetic deletion of EGFR are needed to assess the specificity of the anticalcific response.

Since our data indicate that EGFR inhibition disrupts calcifying EV formation, we also set out to determine whether the treatment alters other types of EV formation. We blotted for CD63 and syntenin 1, widely used markers enriched in exosomes and other EV subtypes, including high phosphate-induced VSMC calcification (49). The data demonstrate no differences in CD63 protein within EVs from VSMCs cultured in control, OS, or OS media samples treated with AG1478. The AG1478 treatment also did not alter EV syntenin 1; however, the PD153035 treatment significantly decreased EV syntenin 1. It is unclear whether calcifying EVs considered in our study derive from an exosomal population that is loaded with procalcific components, or whether they derive from a distinct population of EVs. Though these data do not show changes in CD63, it is possible that CD63-positive vesicles acquire procalcific properties in pathological conditions. The current data, however, suggest that CD63-positive EV release is not altered by EGFR inhibition. Future studies are needed to investigate potential changes in syntenin-1 release and clarify the intracellular trafficking mechanisms and specific cargo, and alterations due to EGFR inhibition, associated with calcifying EV formation in VSMCs.

Our data also suggest that osteogenic function of osteoblasts was not affected by EGFR inhibition. Culturing osteoblasts in OS media resulted in the release of TNAP-positive EVs and robust mineralization, neither of which was altered by EGFR inhibition. Interestingly, we showed reduced CAV1 levels in matrix vesicles released by osteoblasts cultured in OS media. These observations further suggest that despite many commonalities, bone matrix vesicles and vascular calcifying EVs originate through different mechanisms. Patients with CKD often exhibit bone disorders, including decreased bone mass density (50). Previous reports demonstrated that trabecular and cortical bone mass density increased in CAV1-deficient mice (51, 52). We demonstrated that EGFR inhibition significantly reversed reductions in trabecular and cortical thickness in the CKD mice; bone volume fraction in trabecular regions significantly increased by the treatment. At the least, these results suggest that EGFR inhibition does not induce deleterious bone remodeling and, at best, may improve CKD-induced bone pathologies. The calcification paradox, the observation that bone and vascular mineral are often negatively correlated (53), is poorly understood. Future studies that further explore the role of CAV1 and EGFR in calcification may provide new mechanistic insight into physiological and pathological mineralization differences.

The cardioinformatics analyses performed also suggest a potential underappreciated link between EGFR and vascular calcification. The association between increased serum EGFR and vascular calcification was also observed in the CKD mouse model; however, the link between these observations and the mechanistic analyses involving intracellular trafficking and calcifying EV formation remains unclear. Mendelian randomization is a causal inference technique used for the in silico identification of novel drug targets, potential drug-drug interactions/synergies, and for the estimation of magnitudes of effect for each of these (54). MR uses genetic variants as IVs, where the unique genetic composition of each individual is used to “randomize” individuals into different treatment groups, mimicking a randomized control trial since genetic composition is randomized at birth (55). The intuition in MR is that if genetic variants are correlated with the exposure variable and if the exposure variable is causal for the outcome variable, then the genetic variants should also explain variance in the outcome variable (56). The IV assumptions must be fulfilled for MR to yield valid results, though in practice, they are often violated because of pleiotropic effects of genetic variants (34). Therefore, a wide set of MR regression techniques have been developed, each with unique merits in accounting for minimizing and resisting potential violations of IV assumptions or other confounding factors (34). As the number of tools created to support in silico target discovery continues increasing, in particular database tools such as HeartBioPortal, OpenGWAS, and MRBase, MR becomes an increasingly attractive tool for exploration of novel pharmaceutical interventions (57–60).

The direction of effect was qualitatively replicated in each of the MR regressions in the FHS cohort. The significant causal estimates of the robust MR-Egger and penalized robust MR-Egger estimates were recreated, though the intercept test P value for the regressions fell short of reaching statistical significance in the FHS cohort. The intercept P values for the two MR-Egger regressions with significant causal estimates were P = 0.06, which likely did not reach statistical significance because of lower sample size of our FHS replication cohort (n = 1,896) relative to our MESA discovery cohort (n = 5,755). We interpreted the highly significant positive causal estimates and intercepts in the MESA cohort (P < 1 × 10⁻¹⁰ and 1.9 × 10⁻⁵) along with the significant positive causal estimates and nonsignificant intercept (P = 1.17 × 10⁻⁶ and 0.06) in the FHS cohort as suggestive that increased serum EGFR is causal for increased CAC. This cardioinformatics workflow (61) highlights the importance of bridging not only the bench-to-bedside but also the informatics-to-medicine divide that still exists in modern precision cardiology research. This approach can connect basic science to population-level data and enable computationally derived therapeutics.

Conclusions

Cardiovascular disease is the leading cause of death in patients with CKD, and the risk of mortality is directly associated with the presence of vascular calcification. Therefore, the development of a therapeutic strategy to prevent vascular mineralization in these patients would represent a breakthrough in CKD management. Other therapeutic strategies in promising clinical trials slow CKD-mediated vascular calcification by interacting directly with mineral (62). Other proposed preclinical strategies include targeting mechanisms that lead to a procalcific SMC phenotype. However, a myriad of initiators results in vascular calcification. Our data suggest a unique therapeutic strategy to modulate calcifying EV formation independent of cell phenotype. EGFR inhibitors have demonstrated clinical safety and efficacy in cancer treatments (63). The accessibility of EGFR has led to the suggestion that it may represent a therapeutic target worth exploring for cardiovascular diseases (64). Patients with CKD represent an identifiable population in need of therapeutics for vascular calcification. The confluence of an accessible target with approved therapeutics and a clear patient population that lack therapeutic options could accelerate the start of clinical trials.

DATA AVAILABILITY

Data will be made available upon reasonable request.

SUPPLEMENTAL DATA

Supplemental Tables S1 and S2 and Supplemental Figs. S1–S4: https://doi.org/10.6084/m9.figshare.22001615.

GRANTS

This work was supported by Florida International University (Dissertation Year Fellowship to A. B. Nik); and Florida Heart Research Foundation (Stop Heart Disease Researcher of the Year Award to J. D. Hutcheson); and the American Heart Association and the Herbert Wertheim College of Medicine Pilot Project Grant 19POST34380255 and FIUF 2400160 (to H. H. Ng). Research reported in this publication was supported by National Institutes of Health Grants 1R01HL160740 (to J. D. Hutcheson) and R01DK132090 (to B. B. Khomtchouk).

DISCLAIMERS

The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

DISCLOSURES

B.B.K. is a founder of Dock Therapeutics, Inc. None of the other authors has any conflicts of interest, financial or otherwise, to disclose.

AUTHOR CONTRIBUTIONS

A.B.N., H.H.N., B.B.K., and J.D.H., conceived and designed research; A.B.N., H.H.N., S.K.A., F.I., P.R.S., and S.B. performed experiments; A.B.N., H.H.N., S.K.A., P.S., F.I., P.R.S., S.B., D.M., B.B.K., and J.D.H. analyzed data; A.B.N., H.H.N., S.K.A., P.S., B.B.K., and J.D.H. interpreted results of experiments; A.BN. and B.B.K. prepared figures; A.B.N. drafted manuscript; A.B.N., H.H.N., S.K.A., P.S., F.I., P.R.S., S.B., D.M., B.B.K., and J.D.H. edited and revised manuscript; A.B.N., H.H.N., S.K.A., P.S., F.I., P.R.S., S.B., D.M., B.B.K., and J.D.H. approved final version of manuscript.