Angiogenesis describes the formation of new blood vessels from pre-existing capillaries governed by the balance of pro- and anti-angiogenic factors. While tightly regulated during physiological processes such as tissue repair and embryonic development, dysregulated angiogenesis contributes to pathological conditions including rheumatoid arthritis, diabetic retinopathy, and atherosclerosis. Moreover, aberrant angiogenesis plays a key role in the progression of malignant tumors. Despite the widespread use of first-line anti-angiogenic therapies targeting the VEGF pathway, their efficacy is limited by adverse effects and the emergence of drug resistance [1]. These limitations underscore the requirement for the identification and development of novel therapeutic agents.

In drug discovery, high-throughput molecular screening has been an effective approach for identifying inhibitors of specific proteins or simple biological processes. However, to enhance translational potential, there is a critical need for reliable, cost-effective angiogenesis assays with straightforward readouts. Current available assays, such as tube formation assay, fibrin bead assay, and aortic ring assay either lack the ability to recapitulate the 3D architecture of human blood vessels or are limited by complicated workflows. Additionally, the quantifications rely heavily on morphological assessments and image analysis, which further constrain their utility in high-throughput screening [2]. Recent advancements in stem cell-derived vascular organoids (VOs) have enabled the engineering of functional human blood vessels and modeling the progression of diabetic vasculopathy [3]. This highlights the potential usage of VOs in drug screening.

To develop a novel angiogenesis assay based on VOs for drug screening, we engineered endogenous PECAM1 and ACTA2 loci to fluorescently label endothelial cells (ECs) and smooth muscle cells (SMCs) within the VOs (Fig. 1A). This design also correlates the expression level of CD31 with the level of secreted nanoluciferase (secNluc) in the supernatant which could further reflect the angiogenic effect. Since the vascular network is induced by embedding VOs in hydrogel, using the supernatant for quantification simplifies the assay and makes it suitable for high-throughput screening. By applying our previously developed orthogonal selection strategy and the reported ACTA2-EGFP cell line construction method [4, 5], we efficiently established the PECAM1-mRuby3-secNluc; ACTA2-EGFP dual reporter cell line on human embryonic stem cell (hESC) line H9 via CRISPR/Cas9 technology (Fig. 1A). Junction PCR results reveal specific knock-in bands and the knock-in sequence was verified via Sanger sequencing on the PCR product (Supplementary Fig. 1A, B). The edited hESC line maintained normal expression of pluripotency markers OCT4 and SOX2 (Supplementary Fig. 1C). Further 2D differentiation demonstrated accurate fluorescent labeling of ECs and SMCs and the fidelity of labeling was validated via immunostaining and Fluorescence-activated cell sorting (FACS) analysis (Fig. 1B and Supplementary Fig. 1D-H). Overall, the above data demonstrated that we successfully edited the PECAM1 and ACTA2 loci for ECs and SMCs labeling.

Establishment and validation of the visualized and quantifiable in vitro angiogenesis model. A Schematic representation of the genome editing strategy targeting the PECAM1 and ACTA2 loci for dual fluorescent labeling. B Representative fluorescence images of induced endothelial cells (iECs) through 2D differentiation of the dual reporter line. C Fluorescence imaging of vascular networks formed by vascular organoids embedded in hydrogel after 5 days. D The coupling of ECs and SMCs in the vascular network. E Comparative analysis of vascular network morphology with or without SU5416 treatment. F Quantification on the diameter of the vascular network. G Quantification of the mean fluorescence intensity of mRuby3 (for ECs) and EGFP (for SMCs) in the vascular network H Bioluminescent intensity correlating with varying numbers of ECs seeded in the plate. I Workflow for high-throughput drug screening using secreted nanoluciferase (secNluc) as the final readout. J Bioluminescent intensity of the supernatant from day 3 and day 5 vascular network with or without SU5416 treatment. K Normalized scores reflecting pro- and anti-angiogenic effects of various treatments after 5 days. L Representative fluorescence images of vascular network with Repsox or PHA-665752 treatment. Data are presented as mean ± SEM. Statistical analysis was performed using two-way ANOVA for F, G and J. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001

Next, we developed a visualized and quantifiable in vitro angiogenesis model using VOs derived from the dual reporter cell line. The vascular networks formed within the hydrogel were easily visualized under confocal microscopy without the need for complex immunostaining, and the signals of ECs and SMCs were clearly separated (Fig. 1C). This allowed rapid morphological assessment of the vascular networks. Additionally, the coupling of ECs and SMCs within the same z-slice indicated that the model accurately recapitulated normal vessel structure (Fig. 1D). To determine whether the fluorescent signal accurately reflects the treatment, we evaluated the response of the vascular network to the VEGFR inhibitor SU5416. After five days of network formation, treated VOs showed significantly smaller vascular networks (Fig. 1E). And quantification on mRuby3 and EGFP fluorescent intensities indicated that SU5416 inhibited both EC and SMC network formation, as reflected by reduced network diameters and decreased fluorescence intensity (Fig. 1F, G). After evaluating the model’s potential for assessing vascular morphology, we sought to design a high-throughput screening workflow using secNluc as a readout. We first aimed to determine whether the level of secNluc in the supernatant could faithfully reflect the number of ECs, thereby reflecting the pro- and anti-angiogenic effects. Purified ECs were seeded in EGM2 and secNluc levels in the supernatant were quantified after 24 h. The bioluminescence intensity showed a positive correlation with cell number and the increase was nearly proportional (Fig. 1H). We also found that approximately 1,000 ECs could make a slight increase in bioluminescence intensity, although the change was not statistically significant. And the difference was distinct with 10,000 ECs (Supplementary Fig. 2A, B). Notably, the sensitivity of the assay depends on the amount of supernatant used for detection and the duration before detection as well. Using SU5416 as a positive control, we established a screening workflow and observed differences in bioluminescence intensity at day 5 (Fig. 1I, J). However, the structure of the vascular network was already disrupted by day 3 (Supplementary Fig. 2D, E). This may be due to the secNluc secreted by the pre-formed network. And SU5416 may inhibit the extension of the network without affecting the survival of ECs or the expression level of CD31. This highlights the importance of evaluating treatment effects on both day 3 and day 5 by using secNluc and morphology together to comprehensively capture the full response.

To assess the assay’s utility in identifying novel targets, we tested a panel of inhibitors targeting pathways including TGF-β, FGF, PDGF, Wnt, BMP, Notch, YAP, and c-Met. Bioluminescence intensity was normalized to day 1 to account for initial cell number variations, and the results revealed that inhibition of YAP, BMP, c-Met or activation of Wnt strongly suppressed angiogenesis (Fig. 1K and Supplementary Fig. 2C-E). Intriguingly, the selective TGFβR1 (ALK5) inhibitor, Repsox exhibited strong pro-angiogenic effects, while SB431542, an inhibitor of ALK4, ALK5, and ALK7, only mildly inhibited angiogenesis and suppressed the expansion of SMCs (Fig. 1K and Supplementary Fig. 2C-E). And the effect of Repsox and c-Met inhibitor, PHA-665752, can be directly observed (Fig. 1L). Repsox has been previously identified to promote SMC maturation and enhance endothelial barrier formation [6, 7]. We further investigated the effect of Repsox on purified 2D induced endothelial cells (iECs) to determine whether Repsox promotes angiogenesis via ECs or SMCs. We found that Repsox had no significant effect on EC proliferation and slightly downregulated the expression of CD31 (Supplementary Fig. 2F, G). Furthermore, the secNluc level in the supernatant decreased after treatment (Supplementary Fig. 2H). These findings suggest that Repsox potentially promotes angiogenesis via SMCs and the effect may depend on the structure of vascular network. Together, our results indicate that this angiogenesis assay could aid in the discovery of novel therapeutic compounds.

In conclusion, we successfully established a dual reporter cell line and developed a visualized and quantifiable in vitro angiogenesis model. Using this model, we validated the anti-angiogenic effects of VEGFR inhibitor and several regulators of angiogenesis, underscoring the potential of this assay for high-throughput drug discovery. Compared to traditional angiogenesis assays, this assay requires a two-step genome editing process on stem cells. The editing efficiency may vary between hESCs and human-induced pluripotent stem cells (hiPSCs), as well as between healthy donor-derived hiPSCs and patient-derived iPSCs. However, our donor construct enables the rapid enrichment of edited cells, which helps overcome the challenge of editing efficiency, making this assay feasible for multiple hPSC lines.