Blood vessel organoids generated by base editing and harboring single nucleotide variation in Notch3 effectively recapitulate CADASIL-related pathogenesis

Human blood vessel organoids (hBVOs) offer a promising platform for investigating vascular diseases and identifying therapeutic targets. In this study, we focused on in vitro modeling and therapeutic target �nding of cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL), the most common form of hereditary stroke disorder caused by mutations in the Notch3 gene. Despite the identi�cation of these mutations, the underlying pathological mechanism is elusive, and effective therapeutic approaches are lacking. CADASIL primarily affects the blood vessels in the brain, leading to ischemic strokes, migraines, and dementia. By employing CRISPR/Cas9 base-editing technology, we generated human induced pluripotent stem cells (hiPSCs) carrying Notch3 mutations. These mutant hiPSCs were differentiated into hBVOs. The Notch3 mutated hBVOs exhibited CADASIL-like pathology, characterized by a reduced vessel diameter and degeneration of mural cells. Furthermore, we observed an accumulation of notch3 extracellular domain (notch3ECD), increased apoptosis, and cytoskeletal alterations in the Notch3 mutant hBVOs. Notably, treatment with ROCK inhibitors partially restored the disconnection between vascular cells in the mutant hBVOs. These �ndings shed light on the pathogenesis of CADASIL and highlight the potential of hBVOs for studying and developing therapeutic interventions for this debilitating human vascular disorder.


Introduction
Cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy (CADASIL) is the most prevalent hereditary stroke disorder [1,2].The condition arises from mutations in the NOTCH3 gene on chromosome 19p13.1 [3,4].The main symptoms are migraines, psychiatric disorders, recurrent strokes, and dementia [5].Notch3 mutation is associated with molecular pathologies such as arterial smooth muscle degeneration and the occurrence of granular osmiophilic material (GOM) [6].However, the mechanism of pathogenesis is unclear, and there is no speci c therapy for CADASIL.
The NOTCH3 gene encodes the Notch3 receptor, a single-pass transmembrane receptor predominantly expressed in mural cells, including vascular smooth muscle cells and pericytes [7,8].The Notch3 receptor consists of 29-36 epidermal growth factor repeats (EGFRs) in its extracellular domain [9].Variants that alter the number of cysteine residues in the EGFRs contribute to the CADASIL pathophysiology [2].An odd number of cysteine residues leads to the formation of incomplete disul de bridges, increasing the multimerization potential between Notch3 ectodomains (Notch3ECD) [10].Moreover, mutations in the Notch3 receptor result in mural cell degeneration, accumulation of notch3ECD, and the deposition of GOM [11,10,12].However, the relationship between NOTCH3 mutations and the pathophysiological symptoms is poorly understood.
In this study, we generated a novel CADASIL model using human blood vessel organoids (hBVOs), which represent human blood vessel-like structures and enable the investigation of vascular cell interactions at both the cellular and tissue levels [13,14].Using CRISPR/Cas9 base-editing technology, we generated human induced pluripotent stem cells (hiPSCs) carrying Notch3 mutations, free from other environmental events or background in uences.These mutant hiPSCs were differentiated into hBVOs.The Notch3mutated hBVOs exhibited decreased vessel diameter and loss of mural cells.Additionally, the levels of notch3ECD and apoptosis were increased in Notch3 mutant BVOs.Furthermore, cytoskeleton alterations were detected in Notch3 mutant BVOs.To identify factors capable of restoring connections between vascular cells, we conducted an inhibitor treatment study.The results revealed that ROCK inhibitors were key factors protecting mural cells from degeneration.This platform holds promise for elucidating pathophysiological mechanisms and identifying novel target drugs for human vascular disease, especially for CADASIL.

Generation and characterization of Notch3 mutant hiPSCs
The missense Notch3 mutant responsible for CADASIL is characterized by a single nucleotide change that results in the alteration of cysteine [15].We targeted the R153C or R182C Notch3 mutation, which induces a cytosine-to-thymine (C-T) change, leading to an amino acid substitution from arginine to cysteine (Fig. 1A).Speci c sgRNA candidates were selected to recognize the target region of the R153C or R182C Notch3 gene and direct the nickase Cas9 (nCas9) to induce single-strand breaks (Supplementary Table 1).The conversion of targeted single bases is enabled by CRISPR base editing [16].The base editor AncBE4max was used to induce a C-T change in the target region.The EF1α-AncBE4max-nCas9-2UGI-EGFP-U6-sgRNA plasmid vector was used for transfection into HEK293 cells (Fig. 1B).Following transfection, the plasmids containing sgRNAs (R153C-1, R182C-1, and R182C-2) exhibited target C-T base substitution e ciencies of 0.5%, 0.2%, and 0.5%, respectively (Fig. 1C).
We assessed the pluripotency and genomic stability of the gene-edited Notch3 mutant hiPSCs.The cell lines expressed pluripotency markers, including SOX2, Tra-1-60, Nanog, SSEA3, SSEA4, and Oct3/4, as con rmed by ow cytometry and immunostaining (Fig. 2A and 2B).Additionally, a normal karyotype of (46, XX) was observed in the analyzed cells (Fig. 2C).To identify any unintended mutations resulting from gene editing, an off-target analysis was conducted.The top 10 off-target sites were predicted using the benchling web tool (Supplementary Table 2 and Table 3).Sanger sequencing was performed for the top 10 predicted off-target sites around each sgRNA (R153C-1 and R182C-1) (Supplementary Fig. 1A and   1B).Notably, no mutation was detected in any of the off-target sequences, suggesting that off-target cleavage is unlikely to have contributed to CADASIL-like pathology.

Generation of and structural changes in Notch3 mutant hBVOs
To investigate the effects of NOTCH3 mutations on vascular structure, NOTCH3 mutant hiPSCs were differentiated into NOTCH3 hBVOs using a mesoderm induction and vascular differentiation protocol for 15 days (Fig. 3A).Visual inspection con rmed the presence of well-established blood vessel-like structures in both WT and NOTCH3 mutant hiPSCs (Fig. 3B).
To examine the differences in vasculature between WT and Notch3 mutant hBVOs, immunolabeling was performed for CD31 + endothelial cells and SMA + smooth muscle cells in each hBVO.The diameter of CD31 + endothelial vessels was decreased in Notch3 mutant hBVOs (Fig. 3C and 3D).Furthermore, PDGFRβ + mural cells were decreased, with heterozygous CADASIL BVOs exhibiting a 2.7-fold reduction and homozygous CADASIL BVOs a 6-fold reduction in mural cells compared to WT BVOs (Fig. 3E and  3F).These ndings suggest that Notch3 mutant hBVOs have a phenotype characterized by reduced blood vessel diameter and loss of mural cells, reminiscent of the pathology of CADASIL.

Increased apoptosis and altered cytoskeleton in Notch3 mutant hBVOs
We next identi ed the key characteristics speci c to CADASIL, including the accumulation of Notch3ECD and the presence of granular osmiophilic material (GOM).To examine the alteration in Notch3ECD expression, WT and Notch3 mutant hBVOs were immunolabeled using anti-Notch3ECD and anti-PDGFRβ antibodies.The ratio of Notch3ECD expression per total mural cells was signi cantly higher in Notch3 R153C homozygous mutant hBVOs (Fig. 4A).This indicates the accumulation of Notch3ECD in Notch3 mutant hBVOs despite a decrease in Notch3 receptor-expressing mural cells.
Another characteristic of CADASIL is the accumulation of GOM surrounding mural cells.The detection of GOM typically requires transmission electron microscopy (TEM).However, due to the low density of vascular cells in hBVOs (Supplementary Fig. 2), GOM was challenging to observe.Furthermore, GOM deposits are most abundant in arterioles, whereas hBVOs primarily consist of capillary-like blood vessels.
Next, we investigated the reasons for the degeneration of mural cells in Notch3 mutant hBVOs.We examined the alterations in the differentiation and proliferation of Notch3 mutant mural cells derived from Notch3 mutant hiPSCs (Supplementary Fig. 3A).Interestingly, there was no difference in the proportion of cells differentiating into mural cells between WT and Notch3 mutant hiPSCs (Supplementary Fig. 3B).Additionally, there was no signi cant disparity in the cell cycle phase distribution between WT and Notch3 R153C homozygous mural cells, as determined by propidium iodide (PI) staining (Supplementary Fig. 3C and 3D).Therefore, the degeneration of Notch3 mutant mural cells is not attributable to differences in their differentiation or proliferation rates.
Apoptotic cells were signi cantly increased in Notch3 R153C homozygous hBVOs (Fig. 4B).Moreover, the morphology of the cytoskeleton was altered in Notch3 R153C homozygous hBVOs and mural cells (Fig. 4C).In Notch3 R153C homozygous hBVOs, lamentous actin formed bundled structures, whereas in Notch3 R153C homozygous mural cells, lamentous actin was elongated.In addition, expression of the ACTA2 gene, which is one of the cytoskeleton-associated genes, was downregulated in Notch3 R153C homozygous hBVOs (Fig. 4D).The increase in apoptosis and the altered cytoskeleton may contribute to the degeneration of mural cells.
In summary, differentiation and proliferation are not important in the loss of mural cells in Notch3 mutant BVOs.Instead, Notch3 mutant mural cells may die after differentiation and proliferation.Furthermore, we observed increased apoptosis and an altered cytoskeleton in Notch3 mutant hBVOs and mural cells, implicating them in the degeneration of mural cells.

Restoration of mural cell interactions by a ROCK inhibitor
In search of potential therapeutic targets for CADASIL, we explored the effects of inhibitors and activators.From days 13 to 15 of hBVO differentiation, Notch3 R153C homozygous hBVOs were treated with inhibitors targeting p38, JNK, ERK, AKT, MEK1/2, GSK3β, mTOR, ROCK, γ-secretase, SHH activation, and TGF-β (Fig. 5A).In Notch3 mutant hBVOs, the co-localization of endothelial cells and mural cells was impaired.However, treatment with the ROCK inhibitor restored the interaction between mural cells and endothelial cells (Fig. 5B).Although the restoration did not reach the level in WT samples, the mural cells reestablished connections with the endothelial cells.These ndings highlight the potential of Notch3 mutant hBVOs for investigating targeted drug therapies for CADASIL.

Discussion
In this study, we established a robust CADASIL model using hBVOs derived from Notch3 mutant hiPSCs.
By employing CRISPR/Cas9 base-editing technology, we introduced the R153C or R182C mutation into hiPSCs, ensuring a controlled genetic background.These mutant hiPSCs were subsequently differentiated into hBVOs to investigate the vascular phenotype associated with CADASIL.Precise editing of the Notch3 gene was achieved using speci c sgRNAs and the base editor AncBE4max, resulting in the desired amino acid substitution.
Characterization of the Notch3 mutant hBVOs revealed several key features reminiscent of the vascular pathology of CADASIL.The mutant hBVOs exhibited a decreased diameter of endothelial vessels and a signi cant decrease in the population of PDGFRβ-positive mural cells compared to WT hBVOs.These ndings provide evidence that the Notch3 mutant hBVOs accurately recapitulate the vascular abnormalities seen in CADASIL.Furthermore, cytoskeletal alterations were detected in the Notch3 mutant hBVOs, suggesting additional cellular changes associated with CADASIL pathology.These observations highlight the multi-faceted nature of the disease and emphasize the value of using hBVOs as a disease model to study CADASIL.
To further investigate the characteristics of CADASIL, we focused on the accumulation of Notch3ECD and the presence of GOM.Immunolabeling revealed a signi cantly higher ratio of Notch3ECD expression per total mural cells in the Notch3 R153C homozygous mutant hBVOs.This accumulation of Notch3ECD occurred despite a decrease in Notch3 receptor-expressing mural cells, indicating an altered balance in Notch3 processing and shedding.Detecting GOM, a pathological hallmark of CADASIL, proved challenging in hBVOs due to their capillary-like nature and low density.Although GOM deposits are most prominent in arterioles, they were not detected in hBVOs.Therefore, future investigations should focus on arteriolar structures to assess the role of GOM in CADASIL.
To identify therapeutic targets for CADASIL, a range of inhibitors and activators targeting various signaling pathways were tested in the Notch3 mutant hBVOs.Among them, ROCK inhibitors signi cantly restored the interaction between mural cells and endothelial cells.Although the restoration did not reach the level in WT samples, the ndings indicate that ROCK inhibition has therapeutic potential for ameliorating the vascular phenotype in CADASIL.
In conclusion, we generated a robust CADASIL model using Notch3 mutant hiPSCs and hBVOs.Characterization of the mutant hBVOs revealed vascular abnormalities and cytoskeletal alterations and altered Notch3ECD expression, all reminiscent of the pathology of CADASIL.The identi cation of ROCK inhibition as a potential therapeutic strategy underscores the utility of the hBVO model for investigating pathophysiological mechanisms and identifying novel target drugs for CADASIL and other human vascular diseases.Continued research using the Notch3 mutant hBVO platform will provide insight into CADASIL pathogenesis and will promote the development of effective treatments for this debilitating condition.

Design of sgRNAs
To target the regions R153C and R182C of the human NOTCH3 gene, sgRNAs were designed.The Benchling web tool (https://www.benchling.com/) was used to select the optimal target sites for the CRISPR-Cas9 base editing system.Five sgRNAs were used in subsequent experiments.The sgRNAs were ligated to a backbone plasmid containing the U6 promoter, which was derived from the pSpCas9(BB)-2A-GFP plasmid (Addgene plasmid #48138).The U6-sgRNA construct was ampli ed by PCR for further use.

Construction of a base-editing vector
The base editor was created by assembling components from multiple plasmids.First, the AncBE4max gene was isolated from the pCMV-AncBE4max-P2A-GFP vector (Addgene plasmid #112100), and the nSpCas9-NG gene was obtained from the pSI-Target-AID-NG plasmid (Addgene plasmid #119861).The AncBE4max gene was fused to the 5'-end of the nSpCas9-NG gene using the In-Fusion® HD cloning kit (TaKaRa).To generate the nal construct, the EF1α-AncBE4max-P2A-GFP construct was fused to the U6-sgRNA construct.
Base-editor e ciency test HEK293 cells were maintain in DMEM containing 5% fetal bovine serum (FBS) medium at 37°C with 5% CO 2 .Cells were seeded onto 12-well plates at a density of 5 × 10 5 per well and allowed to attach and grow for 16-24 hours prior to transfection.
For transfection, the following conditions were employed: Lipofectamine 3000 (Thermo Fisher Scienti c) was used at 3 µL, and the base editor plasmid containing the sgRNA sequence was added at 1.2 µg.The transfection mixture was prepared by combining and diluting these components with Opti-MEM to a total of 106 µL, following the manufacturer's protocol.After 2-3 days, GFP-expressing cells were sorted using ow cytometry to isolate single cells.Each individual colony derived from the sorted cells was subjected to genomic DNA isolation.The DNA samples were subsequently sequenced at the target location to analyze the genetic alterations.

Off-target analysis
Base editing can generate nonspeci c and unintended genetic modi cations due to mismatch tolerance.To characterize potential off-target effects in the hiPSCs, we selected top-ranking off-target exonic sites in the human genome using the benching web tool (Extended Data Table 1).We performed targeted sequencing of the top 13 potential off-target sites from control, R153C monoallelic, and R153C biallelic hiPSCs (Extended Data Fig. 1).None of the off-target sequences had mutations, indicating that off-target cleavage was unlikely to have contributed to CADASIL-like pathology.

Targeted NGS/sample processing for deep sequencing
The primers used for PCR are listed in Supplementary Table 3.Genomic DNA of GFP + hiPSCs was ampli ed by two-step PCR.For the rst PCR (adapter PCR), speci c staggered primers were used to amplify the integrated fragment.For the second PCR (index PCR), Illumina barcoded sequences were added to distinguish the samples.

Generation of spin embryoid bodies
MEF-independent hiPSCs were washed twice with DPBS to remove debris and remaining medium.The hiPSCs were dissociated by adding 1 mL of Accutase (Stemcell Technologies, 07922) to each well and incubating for 3-5 minutes at 37°C.The detached hiPSCs were ltered using a 70 µm strainer and resuspended in 9 mL of KO-DMEM and transferred to a 15 mL tube.The tube was centrifuged at 1,200 rpm for 2 minutes, and the medium was aspirated.The hiPSCs were resuspended in 1 mL of E8 medium containing 5 µM Y-27632 (MCE, HY-10583).For manual cell counting, 10 µL of cell suspension were mixed with 10 µL of 0.1% Trypan blue solution in an extra-at 96-well plate.A hemocytometer (Merck, Z359629) was used to enumerate live cells under a microscope.The desired number of cells (2000 or 4000) was transferred to each well of an ultra-low attachment 96-well plate, and the volume was adjusted accordingly.To each well, 50 µL of E8 medium with 5 µM Y-27632 were added after transferring hiPSCs.
The plate was centrifuged at 1500 rpm for 5 minutes and incubated overnight at 37°C.
On day 7 or 8, ECs sprouted from the aggregates, and vascular networks were established.On day 10, individual BVOs were isolated from the gel and transferred to 96-well ultra-low attachment plates for further maintenance until sampling.

Whole-mount immunostaining of organoids
The organoids were subjected to immuno uorescence staining.First, they were rinsed twice with phosphate-buffered saline (PBS) and treated with cell recovery solution (Corning, 354253) at 4°C for 1 hour to eliminate non-speci c signals caused by Matrigel.After the PBS rinses, the organoids were xed in 4% paraformaldehyde at room temperature (RT) for 1 hour and washed with PBS for a minimum of 30 minutes.
The organoids were cleared using the CytoVista™ 3D Culture Clearing Kit (Invitrogen, MAN0017942), following the manufacturer's instructions.All procedures were carried out on a shaker.The organoids were subjected to a gradient series of methanol (50%, 80%, and 100%) at 4°C to achieve permeabilization.Subsequently, they were washed at RT using a series of methanol (80%, 50%, PBST, and PBS).The samples were next immersed in antibody penetration buffer at RT for 1 hour and blocked with blocking buffer at 37°C.Primary antibodies were diluted in antibody dilution buffer, and the samples were incubated overnight at 37°C in the antibody solution.
Next, the organoids were washed ve times for 10 minutes each in a washing buffer and incubated overnight at 37°C with a diluted solution of secondary antibody and DAPI.Subsequently, the samples were washed 10 times for 10 minutes in washing buffer.Samples were dehydrated using increasing concentrations of methanol (50%, 80%, and 100%) and incubated in CytoVista tissue clearing reagent overnight at 4°C.Z-stack imaging was performed using a confocal microscope.

Quantitative reverse-transcription PCR (qRT-PCR)
Samples were washed with PBS and immediately ash-frozen in liquid nitrogen.Total RNA was extracted using the RNeasy Mini Kit (Qiagen), and cDNA was synthesized from the RNA using random primers and reverse transcriptase (Toyobo).qRT-PCR was performed on a real-time PCR system (Thermo Fisher Scienti c), and relative mRNA quanti cation was determined using the 2∆∆CT method.

Mural cell differentiation
Mural cell differentiation was carried out using a neuroectodermal intermediate protocol as described in a previous study (Kelleher, Dickinson et al. 2019).hiPSCs were dissociated using ReleSR and plated onto Matrigel-coated six-well plates at a density of approximately 30,000 cells per well in E8 medium supplemented with 10 µM Y-27632.After 24 hours, the culture medium was switched to Y-27632-free E8 medium.The following day, the cells were cultured in E6 medium containing 10 µM SB-431542 and 20 ng/mL FGF2.The medium was refreshed daily until day 6 of differentiation when the supplements were replaced with 2 ng/mL TGF-β and 5 ng/mL PDGF-BB.Daily medium changes were performed until day 18 of differentiation.

Flow cytometry
To assess the proportions of cell types in hBVOs, ow cytometry was performed.hBVOs were dissociated using collagenase B for 30-40 minutes.The dissociated cells were pipetted and washed in PBS supplemented with 5% FBS.Subsequently, single cells were stained with uorescence-conjugated antibodies for 30 minutes at 4°C.For intracellular antigen staining, cells were xed using 4% paraformaldehyde and permeabilized with 0.1% saponin.Antibodies (CD31 and CD140b), diluted in PBS containing 5% FBS, were used to stain the cells.Flow cytometry data were acquired using the BD FACS Aria ow cytometer and analyzed using FlowJo software.

Preparation of for TEM
To GOM in Notch3 mutant hBVOs, we followed a previous protocol [17] for sample preparation and imaging using TEM.hBVO samples were xed in a solution of 2% glutaraldehyde and 2% paraformaldehyde in 0.1 M phosphate buffer.Subsequently, the samples were post xed in buffered osmium tetroxide.Dehydration of the samples was performed using a series of graded alcohols.Finally, the samples were embedded in an Epon-Araldite mixture.Semi-thin sections (2 µm) were obtained from the embedded samples using a microtome and stained with toluidine blue for examination.Thin sections were cut from the blocks using an ultramicrotome and stained with lead citrate for TEM imaging.The

Figure 1 Base
Figure 1

Figure 3 Generation
Figure 3