Vulnerability to shear stress caused by altered peri-endothelial matrix is a key feature of Moyamoya disease

Moyamoya disease (MMD) is characterized by progressive bilateral stenotic changes in the terminal portion of the internal carotid arteries. Although RNF213 was identified as a susceptibility gene for MMD, the exact pathogenesis remains unknown. Immunohistochemical analysis of autopsy specimens from a patient with MMD revealed marked accumulation of hyaluronan and chondroitin sulfate (CS) in the thickened intima of occlusive lesions of MMD. Hyaluronan synthase 2 was strongly expressed in endothelial progenitor cells in the thickened intima. Furthermore, MMD lesions showed minimal staining for CS and hyaluronan in the endothelium, in contrast to control endothelium showing positive staining for both. Glycosaminoglycans of endothelial cells derived from MMD and control induced pluripotent stem cells demonstrated a decreased amount of CS, especially sulfated CS, in MMD. A computational fluid dynamics model showed highest wall shear stress values in the terminal portion of the internal carotid artery, which is the predisposing region in MMD. Because the peri-endothelial extracellular matrix plays an important role in protection, cell adhesion and migration, an altered peri-endothelial matrix in MMD may contribute to endothelial vulnerability to wall shear stress. Invading endothelial progenitor cells repairing endothelial injury would produce excessive hyaluronan and CS in the intima, and cause vascular stenosis.


Results
Immunohistochemistry of MMD lesions. Staining of samples for HA with HA-binding protein revealed marked accumulation of HA in the thickened intima of a specimen from a patient with MMD, whereas in the control specimens, HA staining was detected in the endothelium and in the outside margins of the internal elastic lamina, with only a small amount of HA detected within the intima. In contrast with the control specimens, there was no staining for HA in the endothelium of the specimen from the patient with MMD ( Fig. 1A-D). There was strong staining for HA synthase 2 (HAS2) in the infiltrated cells in the thickened intima, the endothelium, and in vascular smooth muscle cells (VSMCs) in the specimen from the patient with MMD. There was also strong staining for HAS2 in the VSMCs of controls. There was strong HAS2 staining in the endothelium of both control specimens (Fig. 1E-H). There was weak staining for cyclooxygenase 2 (COX2) in the infiltrated cells in the thickened intima and VSMCs of the specimen from the patient with MMD, and comparable COX2 staining in the VSMCs in both control specimens ( Fig. 1I-L). There was strong staining for chondroitin sulfate (CS) with C6 sulfation (chondroitin 6-sulfate [CS-C6]) and relatively weak staining for CS with C4 sulfation (chondroitin 4-sulfate [CS-C4]) in the thickened intima and VSMCs of both MMD and control specimens. There was slight, minimal staining for CS-C6 and CS-C4 in the peri-endothelial cells of MMD, whereas in the control specimen there was relative stronger staining for both CS-C6 and CS-C4 in the endothelial cells (Fig. 2). There was strong staining for HA in the thickened intima of the middle cerebral artery from the patient with MMD (A,C). Specimens from Control 1 (B) and Control 2 (D) exhibited HA staining on vascular endothelium, which was not stained in the MMD sample (arrows). There was weak staining for HA in the thickened intima of the carotid artery from Control 2 (D). There was strong HAS2 staining in the VSMCs in the MMD sample (E,G) and both Control 1 (F) and Control 2 (H), as well as in the infiltrating cells within the thickened intima of the MMD specimen (E,G). VSMCs stained weakly for COX2 in specimens from the patient with MMD (I,K), Control 1 (J), and Control 2 (L). In addition, infiltrating cells within the thickened intima of the MMD specimen (I,K) stained positive for COX2. Scale bar: 500 μm for (A,B,E,F,I,J); 100 μm for (C,D,G,H,K,L). www.nature.com/scientificreports/  www.nature.com/scientificreports/ Endothelial differentiation of iPSCs derived from MMD and control. All three patients with MMD had the common R4810K mutation of RNF213, while control individuals had no R4810K mutation in RNF213 ( Figure S1) . After endothelial differentiation, we sorted 17.5%-64.5% of CD31 + CD144 + cells from every iPSC line (Fig. 3A). The purified cells had the endothelial cell markers, VE-cadherin, von Willebrand factor, CD31 and CD105, and the smooth muscle cell marker, α-smooth muscle actin (Fig. 3B, C, S4). We regarded these cells as iPSC-derived endothelial cells (iPSECs).
Disaccharide analysis of GAGs from iPSECs. No significant differences were noted in HA between control-and MMD-derived iPSECs (Fig. 4A), whereas the amount of CS was significantly decreased in MMD (Fig. 4B). Furthermore, sulfated CS, especially in the C6 position, was decreased in MMD (    www.nature.com/scientificreports/

Discussion
The results of the immunohistochemical analysis showed marked accumulation of HA within the thickened intima, accompanied by increased HAS2 and COX2 expression in the infiltrated EPCs, in the occlusive lesion of the patient with MMD. Prostaglandin E 2 plays an important role in the closure of the ductus arteriosus. Prostaglandin E 2 dilates vessels by relaxing vascular smooth muscle and forms an intimal cushion by promoting HA synthesis through the induction of HAS2 in the VSMCs migrating into the intima 9 . In a previous study, it was shown that the infiltrating cells in the thickened intima of patients with MMD are bone marrow-derived EPCs, which are involved in vascular repair and remodeling 7 . In the present study, HAS2 expression in VSMCs did not differ significantly between the patient with MMD and the controls, and this HA derived from VSMCs would not be able to enter the intima because of the internal elastic lamina. Thus, in MMD, the HA responsible for intimal thickening is most likely produced by infiltrating EPCs. HAS2 can be induced by certain proinflammatory cytokines (e.g. interleukin-1 and tumor necrosis factor-α), mechanical stress, and COX2-derived prostaglandins 10,11 . Because the terminal portion of the ICA is subjected to very large hemodynamic stress, as shown in our computational simulation, HAS2 could be strongly induced in this region. Once stenosis occurs by intimal thickening, hemodynamic stress become much greater, and a vicious circle is formed. Furthermore, cerebral infarctions associated with MMD are sometimes triggered by   www.nature.com/scientificreports/ infection (e.g. influenza). Inflammatory cytokines and prostaglandins may also contribute to intimal thickening by inducing HAS2. Interestingly, overexpression of HAS2 has been reported in Down syndrome, which is one of the underling diseases of quasi-MMD 12 . Immunohistochemical analysis also demonstrated decreased staining of HA and CS in the endothelium of MMD. Especially, both CS-C4 and CS-C6 were decreased in the endothelium of MMD. Quantitative analysis using iPSECs derived from control and MMD patients showed decreased amounts of sulfated CS in MMD, which was confirmed by immunofluorescent staining. Enzyme expression related to CS synthesis and sulfation also tended to be decreased in the iPSECs of patients with MMD.
Decreased expression of C4ST-1 and its relationship to the pathogenesis of Costello syndrome has been reported 13 . Costello syndrome is caused by mutations of HRAS, a proto-oncogene, and is one of the underlying diseases of quasi-MMD 14 . Increased HRAS signaling causes down-regulation of C4ST-1 expression, and results in cell proliferation and defects in elastic fiber formation in Costello syndrome. Interestingly, other diseases which increase RAS signaling, such as Noonan syndrome and neurofibromatosis type 1, are also sometimes complicated with quasi-MMD 15 . Thus, decreased C4ST-1 and hyposulfation could be involved in the pathogenesis of MMD.
The endothelium synthesizes and secretes GAGs, such as heparan sulfate, CS, and HA 16 . Some of these GAGs are deposited into the peri-endothelial extracellular matrix, where they are involved in various functions, such as cell proliferation, adhesion, migration, protection, and signal transduction 16,17 . Furthermore, changes in the sulfation of CS profoundly influence its function, such as inflammation and angiogenesis 18,19 . Downregulation of extracellular matrix receptor-related genes was demonstrated by DNA microarray analysis using iPSECs of MMD 20 . These changes in the extracellular matrix could cause the arterial endothelium to become vulnerable to WSS, and cause invasion of EPCs into the intima to repair the damaged endothelium. Our computational simulation has indicated that the target region for MMD matched the location of high WSS.
Although it is unknown how RNF213 contributes to the pathogenesis of MMD, RNF213 is possible to interact with sulfation of CS according to the computational network analysis GeneMANIA (https ://genem ania. org/, Fig. S5). In this study, we have shown decreased sulfation of CS in iPSECs derived from three genetically independent MMD patients with variant RNF213. Thus, in MMD, variant RNF213 causes changes in CS, and the endothelium becomes susceptible to shear stress, making it easier for EPCs to invade the vascular intima, where they would produce HA, resulting in intimal thickening and vascular stenosis. According to the results using a stenotic carotid artery model, WSS would be further increased in the stenotic lesion 21 . Because shear stress induces HAS2 expression, the EPCs that have invaded the intima may be induced to produce even more HA, resulting in a vicious cycle (Fig. 8).
In conclusion, the altered peri-endothelial matrix in patients with MMD may contribute to injury of the vascular endothelium and invasion of EPCs into the intima. HA produced by the infiltrated EPCs cause intimal thickening and vascular obstruction. Further studies would be needed to confirm our hypothesis. www.nature.com/scientificreports/ Methods Patients. A 19-year-old female with MMD died as a result of cerebral hemorrhage. An autopsy was performed and specimens of the supraclinoid ICA were obtained. As controls, autopsy specimens were obtained from a 73-year old male with subarachnoid hemorrhage (Control 1) and an 18-year-old male with malignant lymphoma (Control 2). Informed consents from legally authorized representatives/next of kin were obtained for all autopsy samples.

Immunohistochemistry.
To identify HA, specimens were treated with proteinase K (Dako), endogenous peroxidase was blocked, and sections were stained using a biotinylated HA-binding protein (1:50 dilution; Hokudo) at room temperature for 1 h. Sections were then treated with avitin-biotin peroxidase complex solution (Nichirei Biosciences Inc.) and 3,3′-diaminobenzidine (Dako) according to the manufacturer's instructions.
After microwave antigen retrieval and blockade of endogenous peroxidase, immunostaining was performed with the appropriate primary antibody at room temperature for 2 h. Sections were then treated with EnVision (Dako), according to the manufacturer's instructions, as well as 3,3′-diaminobenzidine. Sections were counterstained with hematoxylin.
iPSCs. MMD-specific iPSCs were derived from three unrelated patients with MMD. The gender and age of the patients with MMD were as follows: MMD 1 (A182), 6-year-old girl; MMD2 (A205), 3-year-old girl; and MMD3 (A206), 37-year-old woman. Their diagnoses were based on criteria from the Japanese Research Committee on MMD (Ministry of Health, Labour and Welfare, Japan). The disease staging for all three patients were www.nature.com/scientificreports/ stage 2. Informed consent was obtained from a parent and /or legal guardian as minors (6 year and 3 year). For control iPSCs, the following two cell lines were used: C1 (409B2), purchased from Riken BRC Cell Bank; and C2 (N1), derived from a 48-year-old man.
iPSC generation. Mononuclear cells were isolated by gradient centrifugation with Ficoll-Paque, and then activated and expanded in KBM502 medium (Kohjin Bio Co.) on anti-CD3 antibody-coated dishes (eBioScience). iPSCs were generated from activated mononuclear cells as described previously 22 . In brief, 5 × 10 5 mononuclear cells were infected with Sendai virus carrying OCT3/4, SOX2, KLF4, and c-MYC at a multiplicity of infection of 10. Sendai virus was prepared as described previously 23 . After 2 days of culture, the infected cells were seeded at 2 × 10 4 cells per 10 cm dish on mitomycin C-treated mouse embryonic fibroblasts. On the next day, the medium was replaced with iPSC medium. From 15 to 17 days after infection, the colonies were selected and expanded on mouse embryonic fibroblasts with iPSC medium.

RT-PCR analysis.
RT-PCR analysis was performed as described previously 20 . In brief, total RNA was purified with Isogen (Nippon Gene Co.) and transcribed to DNA with Superscript III (Invitrogen) and random primers (Invitrogen). RT-PCR was conducted using QuickTaq (Toyobo), according to the manufacturer's instructions. The sequences of primers and amplification conditions for the detection of pluripotent markers were designed as described previously. The primers used for OCT3/4, SOX2, KLF4, and c-MYC were designed to detect the expression of endogenous genes but not of transgenes.
To detect the Sendai virus genome, nested RT-PCR was performed. The primers used for RT-PCR to analyze the results of endothelial differentiation are shown in supplementary Table 1. DNA isolation and Sanger sequencing. DNA isolation and Sanger sequences were performed as described previously 20 . In brief, cells were treated with Lysis buffer (10 mmol/L Tris-HCL pH 7.5, 10 mmol/L EDTA, 10 mmol/L NaCl, 1 mg/mL proteinase K, 0.5% SDS). Genomic DNA was precipitated with ice-cold 75 mM NaCl in ethanol and suspended with 50 μL of TE buffer (10 mmol/L Tris-HCl, 1 mmol/L EDTA, pH 8.0). Mutation of RNF213 (p.R4810K: rs112735431, G > A) was analyzed by direct sequencing. The sequence reactions were performed using a Big Dye Terminator cycle sequencing kit (Life Technologies) and analyzed using an ABI PRISM 310 Genetic Analyzer (Applied Biosystems). The sequences of primers for detecting R4810K in RNF213 were as follows: forward, 5′-AAA GTT CCT GCC TGA GAT TTTG-3′, reverse, 5′-AAA TGC GGG ACA GTC CTG GT-3' .

Disaccharide analysis of GAGs from endothelial cells derived from MMD and control
iPSCs. Endothelial cells (3 × 10 4 ) derived from MMD and control iPSCs were seeded onto 35 mm dishes and cultured for 24 h, and cells were harvested with a scraper into a tube. After centrifugation, supernatant medium was carefully removed, and the residual cells were weighed.
GAGs were isolated and purified from the cells as described previously 25 . Briefly, cells were homogenized and extracted with acetone three times, and air-dried thoroughly. The dried materials were digested with heatactivated actinase E (10% by weight of dried materials) in 0.1 M borate-sodium, pH 8.0, containing 10 mM CaCl 2 at 55 °C for 48 h. The samples were adjusted to 5% trichloroacetic acid and centrifuged. The resultant supernatants were extracted with diethyl ether three times to remove trichloroacetic acid, and then neutralized using 20% NH 4  Computational fluid dynamics model. Two 3-dimensional models of arteries were made from computed tomography data of a 66-year-old female patient, using a slice thickness of 0.5 mm so that blood flow in the model lumens could be simulated by computational fluid dynamics. As shown in supplemental Figure S2 The incompressible Navier-Stokes equations were solved at each computational cell by the ANSYS Fluent Ver.19.0 software. Coupling of the velocity and pressure fields was performed by the SIMPLE algorithm 26 . Blood was treated as a Newtonian fluid with a viscosity of 0.004 Pa s and a density of 1050 kg/m 3 . Steady-state calculation was carried out with the arterial walls assumed to be rigid. Although actual blood flow in the vasculature is pulsatile, steady-state calculation was considered to be sufficient for comparison of the magnitudes of WSS near the terminals of the ICA and CCA. Calculation was terminated after the dimensionless residual of each governing equation had reduced to 10 −6 .
At the proximal opening of model B, the Poiseuille (parabolic) velocity profile with a maximum velocity of 60 cm/s was imposed as the inflow boundary condition. This condition corresponded to an inlet WSS of 1.27 Pa. As the outflow boundary condition, flow rates at the two distal openings were specified with the assumption that flow division at the terminal bifurcation of the CCA followed Murray's law 27 . In model A, the inflow rate at the ICA was the same as at the terminal of the ICA in model B, because blood passing through the ICA in model B was considered to reach the ICA in model A. Flow division at the bifurcation in model A was also assumed to follow Murray's law.
Statistical analysis. Every result is presented as the mean ± SEM. The Wilcoxon test was used to analyze the difference between the means in each group. Statistical significance was accepted at the 95% confidential level (P < 0.05). All experiments were repeated at least three times, and representative data are shown. Study approval. All patients and control individuals were recruited with written informed consent which was approved by the institutional review board of Saga University Hospital (approval numbers: 22-39). All experimental procedures using human samples were approved by the following ethics committees: Ethics Committee of Saga University Hospital; Ethics Committee for Epidemiological and General Research at the Faculty of Life Science, Kumamoto University; Ethics Committee for Human Genome and Gene Analysis Research at the Faculty of Life Sciences, Kumamoto University; and Ethics Committee for Clinical Research and Advanced Medical Technology, Kumamoto University (approval numbers: 012-0317; 318; 153; and 1018, respectively). All methods were performed in accordance with the relevant guidelines and regulations.

Data availability
All data generated or analysed during this study are included in this article and its supplementary information files.