Atherosclerosis is fundamentally an endothelial disease driven by hemodynamic forces. Decades of vascular biology research have established that flow patterns critically regulate endothelial phenotype: laminar shear stress promotes quiescence and vascular homeostasis, whereas disturbed flow induces oxidative stress, inflammation, EndMT, and lesion-prone endothelial states that initiate and accelerate plaque formation [51, 52]. Mechanical forces not only modulate classical mechanosensors but also orchestrate complex transcriptional networks that determine vascular health and disease [53, 54]. Despite extensive knowledge regarding flow-mediated signaling pathways, the contribution of endothelial cannabinoid signaling to flow-responsive vascular dysfunction has remained largely unexplored. Our study identifies mechanosensitive CB1 as a previously unrecognized regulator linking disturbed flow to endothelial dysfunction and atherosclerosis. We demonstrate that disturbed flow activates endothelial CB1 and further enhances its transcription through a coordinated mechanism involving ZNF610 and Spi1 binding to the CNR1 promoter and dissociation of the atheroprotective transcription factor KLF4. These findings expand the current mechanobiology framework by positioning CB1 not only as a mediator of metabolic and inflammatory signaling but also as a flow-responsive molecular switch that amplifies atherogenic endothelial phenotypes. Building on this mechanistic insight and the causal role of CB1 activation, we developed water-soluble, orally bioavailable isoflavone monophosphate prodrugs (G7P and D7P) that function as endothelial-targeted CB1 antagonists. These compounds robustly attenuated disturbed-flow-induced inflammation, oxidative stress, and monocyte adhesion in human endothelial cells, and they reduced atherosclerotic lesion formation in both acute and chronic Apoe⁻/⁻ mouse models. Together, our findings position endothelial CB1 within the broader landscape of atherosclerosis research and demonstrate that isoflavone phosphates ameliorate disturbed flow-induced endothelial dysfunction in vitro and attenuate atherosclerosis in vivo, indicating their potential translational value.

Chemical synthesis of monophosphate derivatives of flavonoids often requires protection-deprotection steps and harsh reaction conditions, and is associated with poor regioselectivity, frequently generating side products such as oligophosphate esters [55]. Although G7P and D7P can be chemically synthesized, we intentionally selected a biotransformation-based production strategy for both scientific and translational reasons. Bacillus subtilis-mediated biotransformation converts soybean-derived substrates into phosphorylated isoflavones through an enzymatic, nature-mimicking pathway that avoids harsh chemical reagents. This enzymatic process offers high substrate specificity and generates fewer unwanted by-products compared with multi-step chemical synthesis. Moreover, when coupled with downstream purification and analytical validation, this strategy yields highly pure compounds suitable for biological investigation. Importantly, the biotransformation approach aligns with our long-term goal of developing safe, fermentation-derived CB1-modulating agents for preventing endothelial dysfunction and enables robust evaluation of the vascular protective effects of G7P and D7P.

Many studies have reported beneficial effects of genistein and daidzein on human health [56, 57], motivating us to study these soy-derived isoflavones. However, according to the BCS, both compounds belong to class IV, meaning they have low solubility, poor gastrointestinal permeability, and low bioavailability. To overcome this, various chemical, enzymatic, and microbial strategies have been used to improve their solubility and usability [58–60]. Microbial phosphorylation of isoflavones is rarely studied, with only two reports in the past decade [61, 62]. Kanakubo et al. produced water-soluble isoflavone analogs using Bacillus subtilis NCI-21011, initially thought to be sulfate esters but later found to be phosphates [63]. More recently, Hsu et al. demonstrated that B. subtilis var. natto BCRC 80517 can bioconvert genistein and daidzein into their 7-O-phosphate forms, which are subsequently dephosphorylated by intestinal ALP on enterocytes [31]. This was demonstrated by comparing the phosphate removal efficiency of G7P with that of fosphenytoin, an FDA-approved phosphate prodrug [64]. To evaluate G7P and D7P as vascular prodrugs, we assessed ALP activity in various endothelial cells and confirmed their conversion to genistein and daidzein. To determine whether isoflavone phosphates exert protective effects independent of cannabis exposure, HUVECs were subjected to disturbed flow without Δ⁹-THC treatment. Both G7P and D7P significantly reduced inflammation, oxidative stress, and EndMT under these conditions (Figure S10). These results indicate that the vascular protective actions of isoflavone phosphates are not limited to Δ⁹-THC-related endothelial dysfunction and may also be relevant for individuals who do not use cannabis. Further in vivo studies will be needed to confirm Δ⁹-THC-independent protective effects in atherosclerosis. In addition, these isoflavone phosphates were evaluated in healthy volunteers through oral administration at National Taiwan University Hospital, where they were found to significantly enhance the oral bioavailability of the parent isoflavones.

Cannabinoid-induced behaviors such as hypolocomotion, analgesia, hypothermia, and catalepsy are known as the Billy Martin tetrad [65, 66]. These effects are mainly mediated by central CB1 activation, and some, like sedation and analgesia, may have therapeutic value [67]. In our previous study, we assessed the neurobehavioral effects of isoflavones in C57BL/6 J mice using standard assays: activity chamber for locomotion, hot plate test for analgesia, rectal thermometer for hypothermia, and bar test for catalepsy [65]. Δ⁹-THC significantly reduced locomotion and induced analgesia, hypothermia, and catalepsy. Genistein alone had no effect on these behaviors and did not reduce the effects of Δ⁹-THC when co-administered [32]. Biodistribution analysis showed that genistein accumulated in abdominal organs, thoracic aorta, heart, and lungs, but was barely detectable in the brain, consistent with previous reports of poor BBB penetration [68]. These findings suggest that soy isoflavones may block the harmful peripheral cardiovascular effects of cannabinoid-induced endothelial CB1 activation while preserving beneficial central effects such as sedation and analgesia.

Rimonabant was the first CB1 antagonist approved by the European Medicines Agency for the treatment of obesity, but it was later withdrawn from the market due to severe psychiatric side effects, including depression, anxiety, and suicidal ideation [69]. To mitigate these CNS adverse effects, pharmaceutical companies have developed peripherally restricted CB1 antagonists by modifying the chemical structure of rimonabant or conjugating bulky groups to reduce BBB penetration [70]. For example, INV-202 (MRI-1891), a peripherally restricted CB1 antagonist developed by Novo Nordisk, is currently in phase 2 clinical trials [71]. It shows a reduced brain-to-plasma ratio (7%) compared to rimonabant (100%) and ibipinabant (22%) [72]. Although INV-202 and other rimonabant-derived compounds have reduced brain penetration, their partial brain exposure may still lead to CNS-related side effects. In contrast to rimonabant-derived compounds, soy isoflavones such as genistein offer a different approach. Several studies have shown that genistein has poor BBB permeability [68]. Consistent with these findings, our previous study demonstrated that genistein does not cross the BBB in mice or interfere with cannabinoid-mediated central effects. These properties make genistein a promising lead compound for the development of novel peripherally restricted CB1 antagonists. Using ligand-based virtual screening, we further identified daidzein, another natural isoflavone enriched in soy, which also exhibited anti-CB1 activity and vascular protective effects. However, compared with rimonabant (active in the nanomolar range), daidzein shows only micromolar activity and needs further optimization.

To investigate whether the vascular protective effects of G7P and D7P are CB1 dependent, CB1-deficient endothelial cells were generated using the CRISPR/Cas9 system (Figure S11A and S11B). Knockout of CB1 in HUVECs reversed Δ⁹-THC-induced endothelial inflammation and oxidative stress (Figure S11C). Likewise, disturbed flow-induced inflammation, oxidative stress, and EndMT were markedly abolished in CB1-deficient cells, further supporting the therapeutic potential of targeting endothelial CB1 (Figure S11D). Since CB1 deficient cells exhibited low levels of inflammation, oxidative stress, and EndMT, comparable to those of normal endothelial cells under laminar flow, it was difficult to detect further reductions in these parameters upon treatment with isoflavone phosphates. Nevertheless, SPR assays and radioligand binding assays confirmed the direct binding of isoflavones to CB1 receptors and their antagonistic effects (Fig. 7).

Multitarget effects are a common feature of small-molecule compounds, including flavonoid derivatives [73]. Studies have reported that flavonoids can bind to estrogen receptors due to their structural similarity to estrogen [74]. They also inhibit the kinase activity of several signal proteins, such as epidermal growth factor receptor (EGFR), ERK1/2, RSK2, MKK4, and Cot in an ATP-competitive manner [75]. To evaluate whether daidzein and genistein would bind to other targets, ligand-based virtual screening for target prediction was performed using the SwissTargetPrediction database [76]. The results showed that genistein may also target proteins such as thromboxane-A synthase, monoamine oxidase A, EGFR, and estrogen receptor (Table S4). Daidzein was predicted to interact with targets including aldehyde dehydrogenase, carbonic anhydrase, monoamine oxidase A, and estrogen receptor (Table S5). While isoflavones exert anti-atherosclerotic effects in part through CB1 inhibition, we cannot exclude the possibility that they also act via other additional mechanisms independent of the CB1-mediated pathway. Epidemiological studies have reported a lower incidence of cardiovascular and coronary heart diseases, including atherosclerosis, in populations with higher soy intake [77, 78]. Genistein and daidzein are the main isoflavones contained in soy-based foods that are regularly consumed in Asian countries [79]. Soy isoflavones are generally recognized as safe dietary supplements [80]. Although they exhibit multitarget effects, their safety profile supports further investigation as potential lead compounds for therapeutic optimization.

This study has several limitations. First, although isoflavone phosphates reduced disturbed-flow-induced endothelial dysfunction in vitro and attenuated atherosclerosis in vivo mouse models, we did not perform functional vascular assays such as wire-myograph studies in Apoe⁻/⁻ mice. These will be essential in future work to validate the translational relevance of these compounds. Second, although G7P and D7P exhibited comparable protective effects in vitro and in vivo, our radioligand binding assays revealed that genistein (IC₅₀ = 0.15 μM; Ki = 0.14 μM) exhibits approximately ten-fold stronger CB1 antagonistic activity than daidzein (IC₅₀ = 3.78 μM; Ki = 3.40 μM). Additionally, ligand-based virtual screening suggested that genistein and daidzein may interact with distinct protein targets. These findings indicate that CB1 antagonism alone does not fully account for the comparable vascular protective effects of G7P and D7P, and that additional mechanisms may contribute to their actions. Third, only male Apoe⁻/⁻ mice were used in this study. Epidemiological studies indicate that males develop atherosclerosis earlier, whereas the risk in females increases after menopause [81]. Because endothelial CB1 is expressed in both sexes and the protective effects of isoflavone phosphates act through blocking endothelial CB1 activation, we selected male mice to minimize biological variability in vivo. However, sex-dependent differences in vascular biology and atherosclerosis are well-recognized, and studies in female mice will be needed to determine whether similar protective effects occur across sexes. Finally, in both the acute and chronic Apoe⁻/⁻ models, treatment began at the onset of disease induction. Therefore, our experiments primarily evaluate preventive rather than therapeutic effects. Future studies are needed to determine whether isoflavone phosphates can reverse established atherosclerosis.