Extraocular delivery of bioswitchable tri-miR-22-loaded tetrahedral DNA nanostructures for intraocular neovascular and neurodegenerative repair

Pathological angiogenesis and progressive neurodegeneration are two interlinked hallmarks of a broad spectrum of ocular diseases that threaten vision, including diabetic retinopathy (DR), wet age-related macular degeneration (wAMD), retinal vein occlusion (RVO) and retinopathy of prematurity (ROP),^1,2,3 characterized by aberrant blood vessel growth, vascular leakage, macular edema, retinal ischemia, and irreversible neural damage.^4,5 The initiating factors of these diseases are diverse, such as hyperglycemia, inflammation, vascular occlusion, oxygen exposure during the neonatal period, and genetic factors. However, all these factors can induce local retinal ischemia and hypoxia. Retinal cells respond to hypoxia by inducing the upregulation of hypoxia-inducible factor-1α (HIF-1α), driving vascular endothelial growth factor (VEGF) secretion, and thereby promoting the proliferation and migration of endothelial cells, which ultimately leads to the formation of pathological angiogenesis. The ischemia and hypoxia also induce retinal neuronal dysfunction and neurodegeneration prior to the appearance of overt vascular changes on ophthalmoscopic examination, such as loss of retinal ganglion cells (RGCs) and photoreceptor cells. Both factors collectively contribute to the impairment of the retinal neurovascular unit (NVU).⁶ Current first-line therapies targeting VEGF have achieved remarkable progress in controlling neovascular leakage; multiple studies have demonstrated that patients’ visual acuity can be significantly improved after standardized anti-VEGF therapy. ^7,8 Even so, 30–50% of patients remain non-responsive, which indicates that in the process of pathological angiogenesis, besides VEGF, more pathological signaling pathways such as Ang2/Tie and Wnt/β-catenin are involved.⁹ However, by their mechanism, they primarily target vascular abnormalities without restoring vascular architecture and providing neuroprotection.^10,11 Moreover, frequent invasive intravitreal injections carry procedural risks such as endophthalmitis, require specialized clinical resources, and place a burden on patients and caregivers, often resulting in poor long-term compliance.^12,13 These limitations highlight the urgent unmet need for novel therapeutics that address both vascular dysregulation and neurodegeneration through safe, minimally invasive administration.

MicroRNAs (miRNAs) are a class of endogenous non-coding single-stranded RNAs approximately 22 nucleotides in length. By specifically binding to the 3’-UTR of target mRNAs, they mediate the degradation or translational repression of mRNAs, thereby negatively regulating gene expression at the post-transcriptional level. Previous studies have demonstrated that miRNAs have emerged as attractive regulators of biological pathways involved in complex neurovascular disorders.^14,15 Among them, microRNA-22-3p (miR-22) has drawn particular interest for its dual ability to suppress pathological neovascularization and enhance neuronal survival.¹⁶ Previous studies have indicated that miR-22 has neuroprotective and reversal effects on a 6-hydroxydopamine-induced cell model of Parkinson’s disease.¹⁷ Gu. et al. revealed that endothelial miR-22 functions as a potent angiogenesis inhibitor that inhibits all of the key angiogenic activities of endothelial cells (ECs) and consequently non-small cell lung cancer growth.¹⁸ Despite this potential, the clinical application of miR-22 has been hampered by its intrinsic limitations, including poor stability, rapid degradation, and low efficiency of cellular uptake—challenges that severely hamper the exploration of miR-22 in retinal neovascular disease and restrict its bioavailability when delivered via conventional methods.

Tetrahedral framework nucleic acids (tFNAs), a class of self-assembling DNA nanostructures with precise dimensions and programmability, were first proposed by Goodman, Turberfield et al.¹⁹ Owing to their unique spatial tetrahedral configuration, tFNAs exhibit exceptional mechanical stability, biocompatibility, and the capability for efficient cellular internalization without the need for transfection agents. In our previous investigations, we have successfully established tFNAs as versatile vehicles to deliver diverse therapeutic agents, including small interfering RNAs and small molecules such as quercetin, demonstrating precisely defined nanostructure, excellent biocompatibility, exceptional tissue permeability and significant therapeutic efficacy in ocular diseases.^20,21 In recent years, a modified design based on the structural characteristics of traditional tFNAs has been developed. By encapsulating three identical microRNA inhibitors around the nucleic acid core, this design retains the structural properties of tFNAs through integrating miR inhibitors into its interior while maintaining a size consistent with that of conventional tFNAs. To facilitate tracking the localization of miR inhibitors upon their entry into cells, a DNA-RNA hybrid switch at the tail of each inhibitor was designed, enabling this nanostructure to be triggered upon stimulation with RNase H.²²

To this end, we developed a tetrahedral framework DNA-based bioswitchable Tri-miRNA-22 mimic delivery system (BiRDS), which is precisely loaded with three copies of miR-22 mimics, engineered to achieve stable miR-22 transport, facilitate efficient cell penetration, and enable enzyme-triggered cargo release.^22,23 Here, we present BiRDS as a next-generation, minimally invasive nanoplatform that simultaneously targets pathological angiogenesis and neuronal degeneration, with the potential to penetrate ocular tissues via extraocular administration—thereby circumventing the need for repetitive intraocular injections. This study systematically evaluated the synthesis, ocular delivery pathway, and therapeutic efficacy of BiRDS in preclinical models of retinal and choroidal neurovascular diseases, and it also explored the potential mechanisms of action associated with BiRDS. The findings set the stage for a safer, microinvasive strategy that simultaneously suppresses pathological angiogenesis and preserves neuronal integrity.

Synthesis and characterization of BiRDS

We developed a unique tFNA-based bioswitchable Tri-miR-22 mimic delivery system, BiRDS, in which three identical miR-22 mimics are integrated into the tFNA interior without altering its overall size or structural integrity. To enable intracellular identification and release, each mimic was equipped with a DNA–RNA hybrid “switch” at its terminus, allowing RNase H-triggered activation. Upon cellular entry, BiRDS undergoes structural collapse from a 3D tetrahedron to a 2D configuration, facilitating the release of miRNA mimics. (Fig. 1a, Supplementary Fig. 1).

The formation of BiRDS was validated by native polyacrylamide gel electrophoresis (PAGE). As revealed by gel electrophoresis, BiRDS showed lower mobility than other permutations of the ssDNA assembly did, indicating that these RNA‒DNA molecules self-assembled into a single nanostructure (Fig. 1b, c). Transmission electron microscopy (TEM) and atomic force microscopy (AFM) were employed to investigate the synthesis of BiRDS. The images revealed distinct triangular particles alongside some polymers, with clear triangular structures confirming successful synthesis (Fig. 1b, d, e). To further confirm successful assembly, dynamic light scattering (DLS) was used to measure the zeta potential and particle size. Both tFNAs (−11.3 ± 5.95 mV) and BiRDS (−16.6 ± 8.04 mV) presented negatively charged surfaces (Fig. 1b, f). Furthermore, the sizes of the tFNAs and BiRDS were 11.58 ± 3.045 nm and 15.19 ± 1.845 nm, respectively (Fig. 1b, g). These results collectively suggest that the BiRDS were successfully self-assembled.

To ensure the fidelity of the synthesized BiRDS nanocomplex, a series of structural analyses was performed. High-resolution tandem mass spectrometry (HR-MS/MS) confirmed the accurate sequencing of all four single-stranded oligonucleotides (S1, S2, S3, and miR-22), with coverage rates greater than 92.31%. The isotopic distribution of miR-22 is shown in Fig. 1h and Supplementary Fig. 2 and Fig. 3. These results demonstrate the high sequence fidelity necessary for the precise self-assembly of the tetrahedral nanostructure.

An efficient targeted delivery system is crucial for miRNA loading, especially considering the poor stability of small RNAs and their low cellular uptake. A comprehensive denaturation analysis was conducted on the basis of the unique structure of BiRDS and the lack of reference standards for the samples loaded with miR-22 (Fig. 1h, Supplementary Fig. 4). Under denaturing conditions, the single-chain components mixed at the predetermined drug substance (DS) ratio were consistent with the DS spectrum and showed no interference (Supplementary Fig. 5). The theoretical loading efficiency was quantified via the calibration curve generated from the self-mixing standard (y = 2.2858x – 0.0459, R² = 0.9945). This calibration curve plotted the molar ratio of the self-mixed samples as the y-axis for the loading rate, with the UV response of miR-22/S2 as the x-axis (Fig. 1h, Supplementary Fig. 6). The calculated values represent the linear loading rate, which was ultimately adjusted for the actual loading rate through size-exclusion high-performance liquid chromatography (SEC-UPLC) purity assessment.

To evaluate their ability to penetrate cells, human umbilical vein endothelial cells (HUVECs) were incubated with Cy5-loaded BiRDS. Progressive Cy5 fluorescence accumulation was observed within the cytoplasm after 4, 16, and 24 h of incubation, corresponding to low, moderate, and strong signal intensities, respectively (Fig. 1i). To further confirm the effective delivery of BiRDS to the retina, we performed fluorescence in situ hybridization (FISH) in rabbit eyes. The results showed that intravitreal injection of BiRDS enabled efficient delivery of miR-22 mimics to the retinal surface, followed by gradual penetration into the deeper retinal layers (Fig. 1j).

CCK8 assays were used to evaluate the minimum effective concentration of BiRDS. The results demonstrated that BiRDS at a concentration of 100 nmol/L effectively inhibited the activity of HUVECs (Fig. 1k). The anterior segment was examined to evaluate the safety of BiRDS, and the results revealed that BiRDS did not cause any detectable damage to the conjunctiva, cornea, or other anterior segment structures (Supplementary Figs. 7, 8).

BiRDS inhibits hypoxia-induced HUVEC proliferation, migration, and pathological angiogenesis in vitro

To demonstrate the antiangiogenic effects of BiRDS, in vitro cellular experiments were conducted using a hypoxic HUVEC model. Aflibercept (AFL), a first-line anti-VEGF drug currently used in clinical practice, was used as the positive control. Cell proliferation was evaluated via the incorporation of EdU (Fig. 2a). The results showed that, compared with VHCL, BiRDS significantly inhibited the proliferation of HUVECs. In contrast, both the miR-22 and tFNAs resulted in only a modest reduction in proliferation (Fig. 2d, g). These findings suggest that BiRDS markedly reduces hypoxia-induced cell proliferation. A tube formation assay was used to evaluate the ability of HUVECs to create capillary-like networks (Fig. 2b). Compared with the VHCL group, the BiRDS group significantly reduced both the total length of the networks and the number of nodes formed. This inhibitory effect was comparable to that of the tFNAs and AFL groups. (Fig. 2e, h, i). Additionally, the migration of HUVECs was observed (Fig. 2c). Compared with VHCL and miR-22, BiRDS significantly reduced cell migration. The effect of BiRDS was similar to that of tFNAs and AFL (Fig. 2f, j). These data suggest that BiRDS has a prominent ability to inhibit the migration of hypoxic HUVECs.

Invasive intravitreal BiRDS effectively inhibits laser-induced CNV

We conducted dose-optimization studies of BiRDS in an oxygen-induced retinopathy (OIR) model and reported that a concentration of 5 μM effectively inhibited pathological neovascularization (Supplementary Fig. 9). To further verify the antiangiogenic potential of BiRDS, a laser-induced choroidal neovascularization (CNV) model was established in rats. Vascular leakage was monitored by fluorescein angiography (FFA) at days 7, 14, and 21 (Fig. 3a). Seven days post-laser irradiation (D7), there was no difference in the area of CNV lesions among the VHCL, BiRDS, and AFL groups before treatment (Fig. 3b, c). Following treatment, the BiRDS group presented a reduction in lesion area on both day 14 and day 21 (Fig. 3b, d, e), with an inhibitory effect comparable to that observed in the AFL group (Fig. 3f, g). These results indicate that intravitreal BiRDS effectively suppresses pathological CNV growth and leakage, achieving results comparable to those of standard anti-VEGF therapy.

Ocular penetration and retinal targeting of subconjunctivally administered BiRDS: from intraocular to extraocular delivery

After confirming the robust therapeutic efficacy of BiRDS via intravitreal injection, we further explored its potential as a minimally invasive delivery strategy to overcome the clinical limitations associated with repeated intraocular injections, such as the risk of severe complications (e.g., endophthalmitis), patient discomfort, and caregiver burden. The subconjunctival route was chosen for its feasibility in outpatient clinical settings, the convenience of office-based administration, and its reported ability to sustain intraocular drug exposure. To verify the effectiveness of this extraocular approach, we subsequently evaluated the ability of BiRDS to traverse ocular barriers and reach the neural retina following subconjunctival administration in vivo. Cy5-labeled BiRDS or miR-22 was injected into the subconjunctival cavity near the sclera in mice via a 34-gauge needle. After 24 h, the eyes were harvested, sectioned, and stained with DAPI (blue) for nuclei and Cy5 (red) for miR-22 and BiRDS (Fig. 4a). In contrast to the absence of a signal for miR-22, confocal microscopy revealed strong Cy5 fluorescence throughout the ocular tissues in the BiRDS-treated group, which demonstrated that BiRDS has excellent penetration (Fig. 4b). The hyperfluorescence detected in the neural retina also confirmed its ability to pass through multiple layers of the posterior eye (Fig. 4b).

To further map the intraocular distribution, fluorescence imaging was performed at various time points. As shown in Fig. 4c–e, Cy5 fluorescence beneath the conjunctiva appeared immediately after administration and gradually weakened over time. In the conjunctivae, iris, and corneal epithelium, fluorescence peaked within 1 h and decreased over time to 12 h. The negligible signal observed in the corneal stroma and endothelium indicated that BiRDS did not enter the anterior chamber via the iris or aqueous humor and thus did not distribute to the corneal endothelium or stroma. On the scleral surface, fluorescence peaked within 1 h and was sustained for up to 18 h, suggesting that the sclera may act as a reservoir contributing to sustained and prolonged release. Notably, strong Cy5 fluorescence was predominantly localized in the choroid beginning at 1 h postinjection and remained stable for 18 h. In the retina, fluorescence became evident at 6 h and increased progressively over time, indicating that BiRDS successfully penetrated the outer blood–retinal barrier–retinal pigmented epithelium (RPE).

Importantly, no morphological changes were observed in the retinal structure, suggesting that BiRDS delivery did not induce substantial structural damage. Collectively, these results showed that subconjunctivally administered BiRDS can effectively reach the innermost retina by traversing the sclera-choroid-retina pathways, providing a minimally invasive and promising strategy for treating CNV or retinal neovascularization (RNV).

Choroid targeting of extraocularly administered BiRDS: inhibition of CNV

Given the significant intraocular penetration and antiangiogenic effects of BiRDS, we further evaluated its potential as a minimally invasive, extraocular therapy for CNV in mice (Fig. 5a). The mice received treatment 4 days after laser irradiation (D4) and were observed at 7 days after laser irradiation (D7). The mean leakage area in the BiRDS group, as assessed through FFA, was significantly reduced. This reduction was comparable to that observed in the AFL group and was greater than that in the VHCL, miR-22, and tFNA groups (Fig. 5b, d). Moreover, OCT was conducted to measure CNV lesion size (Fig. 5c). The results of the BiRDS group were consistent with those of the AFL group (Fig. 5e, f). These results demonstrate that subconjunctival BiRDS delivered to the choroid significantly reduces both the size of CNV lesions and vascular leakage in a laser-induced mouse CNV model, achieving therapeutic effects comparable to those of intravitreal anti-VEGF therapy.

Choroid targeting of extraocularly administered BiRDS: inhibition of RNV

To further assess the effect of BiRDS on RNV, OIR pups received subconjunctival injections (SCI) of VHCL, miR-22, tFNAs, or BiRDS or intravitreal injections of AFL at P12. Eyes were harvested on P17 (Fig. 6a), and whole-mount retinal staining with isolectin B4 (IB4) was performed (Fig. 6b, c). The substantial pathological neovascular areas and avascular areas observed in the VHCL OIR mice confirmed successful model induction. Compared with VHCL, BiRDS-SCI can effectively reduce the area of neovascularization and avascular regions. The effectiveness of BiRDS was comparable to that of AFL in reducing pathological neovascularization (p = 0.883) (Fig. 6c, h).

Persistent avascular areas significantly contribute to poor responses to anti-VEGF agents, and current treatments do not adequately address this problem. We found that, compared with the VHCL treatment, the AFL treatment did not significantly reduce avascular areas, which indicated a limited effect of the AFL on retinal ischemia. In contrast, tFNAs and BiRDS reduced the avascular area. Notably, BiRDS significantly outperformed the other treatments in restoring retinal perfusion (p < 0.0001 vs. VHCL, p < 0.01 vs. miR-22, p < 0.0001 vs. AFL) (Fig. 6d, i). These findings highlight that BiRDS not only inhibits pathological neovascularization but also effectively reduces retinal nonperfusion, demonstrating its potential to alleviate ischemia and promote retinal vascular recovery—an advantage not observed with standard anti-VEGF therapy with AFL.

Extraocularly administered BiRDS suppresses pathological vessel sprouting and promotes healthy angiogenesis

To further assess the effect of BiRDS on pathological neovascularization, we quantified abnormal vessel sprouting from retinal veins. Compared with the VHCL, miR-22, and tFNA groups, the BiRDS treatment significantly reduced the sprouting area, and the sprouting area was comparable to that of the AFL group (Fig. 6e, j). The significant decrease in abnormal sprouting, along with reduced neovascular tufts and avascular areas, showed that BiRDS effectively antagonizes pathological neovascular growth.

Endothelial tip cells are key drivers of healthy vessel sprouting. Compared with the VHCL, miR-22, tFNAs, and AFL groups, the BiRDS group presented a significantly greater number of tip cells (Fig. 6f, k). Additionally, high-magnification confocal imaging revealed that, compared with the VHCL, miR-22, tFNAs, and AFL groups, the BiRDS group presented significantly greater filopodia extension at the leading edge of growing vessels in the avascular retinal zones (Fig. 6f, l).

In addition, RNV, characterized by abnormal vessels breaching the inner limiting membrane (ILM), was visualized via retinal cryosections and IB4 staining. Normal retinal capillaries, including the superficial capillary plexus and the deep capillary plexus, were accessed (Fig. 6g). Treatment with BiRDS significantly reduced the number of RNV endothelial cells in OIR mice, with effects similar to those in the AFL group (Fig. 6m). Additionally, the BiRDS group presented more intact deep capillaries than the other groups did (Fig. 6n). Together, these findings indicate that BiRDS effectively enhances physiological retinal revascularization while suppressing pathological retinal neovascularization, underscoring its potential as a minimally invasive therapeutic strategy for treating retinal neovascular diseases.

Extraocularly administered BiRDS protects retinal neurons in OIR

Hypoxia–ischemia leads to an increase in the apoptosis of retinal ganglion cells (RGCs) and promotes gliosis, as indicated by the upregulation of glial fibrillary acidic protein (GFAP).^24,25 In addition, oxygen deprivation disrupts the structure of the outer segment of rod photoreceptors. To evaluate the neuroprotective effect of BiRDS, eyes from OIR mice were collected and analyzed via immunofluorescence staining of frozen sections (Fig. 7a).

GFAP-positive astrocytes were primarily found in the nerve fiber layer, whereas GFAP-positive Müller cells were mainly identified in the inner nuclear layer, extending throughout the entire thickness of the retina. Compared with the VHCL and miR-22 groups, the BiRDS group presented a significant reduction in GFAP expression, indicating a decrease in astrocyte activation and Müller cell gliosis (Fig. 7b, c). Semiquantitative analysis revealed no significant difference in GFAP levels between the BiRDS and AFL groups. These findings suggest that BiRDS effectively suppresses reactive changes in astrocytes and alleviates Müller cell gliosis, thereby contributing to neuroprotection under hypoxic–ischemic conditions.

In addition, Tuj1+ ganglion cells were observed after different treatments. Compared with all the other groups, the BiRDS-treated group presented significantly greater numbers of Tuj1-positive ganglion cells (Fig. 7b, d). Compared with those in the VHCL and AFL groups, the number of PKC-α-positive bipolar cells increased significantly after BiRDS treatment (Fig. 7b, e). Compared with the VHCL and AFL groups, the BiRDS group presented a dramatically greater number of rhodopsin-expressing rods. (Fig. 7b, f). Moreover, for Calbindin+ amacrine cells, the BiRDS group presented significantly higher levels than the VHCL, miR-22, and AFL groups did (Fig. 7b, g). Overall, these results suggest the considerable potential of BiRDS as a neuroprotective agent for ocular neovascular diseases.

To determine whether morphological neuroprotection translated into improved visual function, electroretinography (ERG) was performed to assess scotopic (rod) and photopic (cone) responses, as well as oscillatory potentials (OPs) (Fig. 8a). The dark-adapted ( + 1.0 cd·s/m²) ERG primarily reflects rod function, whereas the light-adapted ( + 10.0 cd·s/m²) response reflects cone function. In addition, OPs are indicative of inner retinal activity. The OIR model mice were treated at P12 and subjected to ERG measurements at P17. As shown in Fig. 8b, c mice treated with BiRDS presented a greater scotopic a-wave than did the VHCL, miR-22, and tFNAs groups. The b-wave amplitude measured 161.03 ± 15.14 μV, which was significantly greater than that of the other groups. This finding indicates enhanced activity in rod photoreceptors and bipolar cells.

Similarly, the photopic a-wave and b-wave amplitudes were significantly enhanced, reflecting preserved cone function (Fig. 8b, c). Additionally, the amplitude of OPs was 1.94 ± 0.47 μV in the BiRDS group (Fig. 8b, c), significantly exceeding those of the other groups, suggesting improved inner retinal activity and signal processing. Together, these results demonstrate that BiRDS not only mitigates morphological damage but also sustains retinal electrical function, highlighting its therapeutic potential for protecting vision in ischemic retinal diseases.

Transcriptome analysis reveals BiRDS-mediated neurovascular protection via Wnt pathway modulation

To elucidate the mechanistic basis of the dual antiangiogenic and neuroprotective effects of BiRDS, we performed RNA-seq analysis on three groups of HUVECs: a normoxic group, a hypoxic group treated with VHCL, and a hypoxic group treated with BiRDS (Fig. 9a, b). The DEGs are shown in Fig. 9c, 9d. DEG analysis (FDR < 0.05, |log2FC | >1) revealed significant downregulation of canonical Wnt pathway components. In the gene ontology (GO) analysis of the DEGs, the top categories of the enrichment results were associated primarily with the canonical Wnt pathway, such as FZD4, GSK3β, and CTNNB1, in HUVECs under hypoxic conditions compared with those under normoxic conditions or those treated with BiRDS. Our quantitative polymerase chain reaction (qPCR) results demonstrated a similar trend: BiRDS effectively suppressed the mRNA expression of FZD4 and CTNNB1 while increasing the mRNA expression of GSK3β (Fig. 9f).

Then, experimental validation was carried out on protein expression levels. The protein levels of FZD4, GSK3β, and β-catenin in the different groups were examined by western blot analysis (Fig. 10a, b) and immunofluorescence staining (Fig. 10d, e). There was evident upregulation of FZD4 and β-catenin protein expression in VHCL and a reduction in FZD4 and β-catenin protein expression after BiRDS treatment (Fig. 10c, d). Moreover, increased FZD4 and β-catenin immunofluorescence was detected in VHCL hypoxic HUVECs, but a significant decrease was detected in BiRDS-treated cells (Fig. 10b, e). Similarly, OIR model mice were used to examine protein changes following BiRDS administration (Fig. 10f, g). The western blot and immunofluorescence results revealed that FZD4 and β-catenin proteins were markedly downregulated, whereas GSK3β expression was significantly upregulated after BiRDS treatment (Fig. 10h, g). The coordinated downregulation of Wnt/β-catenin signaling explains the dual effects of BiRDS on suppressing pathological EC activation and preventing hypoxia-induced neuronal damage.

In this study, we successfully developed and validated BiRDS, a tetrahedral framework DNA-based bioswitchable miRNA-22 mimic delivery system. We demonstrated robust intraocular penetration via a minimally invasive extraocular subconjunctival route. BiRDS effectively and durably suppressed pathological neovascularization, promoted physiological vascular remodeling, preserved the integrity of the retinal neurovascular unit, restored retinal neuronal function, and maintained overall retinal structural and functional integrity. This work highlights a paradigm shift from conventional purely vascular-targeted treatments toward a more comprehensive strategy that simultaneously stabilizes the retinal microenvironment and supports neuronal function, which is crucial for preserving visual performance in progressive retinal neurovascular diseases.

Importantly, by achieving effective delivery and functional efficacy through subconjunctival administration, our findings provide compelling evidence that nonintravitreal nucleic acid therapeutics can overcome both biological and patient compliance barriers that have historically hindered the clinical translation of RNA-based ocular therapies.

This is the first application of our bioswitchable nanoplatform in a highly complex neurovascular organ system that targets retinal diseases and features tightly regulated blood‒retina barrier and dual pathology involving both vascular proliferation and neuronal degeneration. Unlike previous studies that focused on superficial tissue models (e.g., skin fibrosis or aging), this study addressed deep-tissue delivery, blood‒ocular barrier penetration, and neurovascular crosstalk, which require distinct design refinements and biological considerations.

Another significant innovation of this work is that BiRDS facilitates effective penetration into the posterior segment of the eye through the scleral–choroidal–retinal pathway after subconjunctival administration. The greatest advantage of this less invasive approach is the avoidance of many complications caused by intravitreal injection, such as intraocular inflammation and endophthalmitis, which are the most devastating complications associated with traditional intravitreal injection of anti-VEGF agents. Here, BiRDS can penetrate cell membranes within 24 h and rapidly accumulate in the cytoplasm in vitro by leveraging the small size and high tissue permeability of the tetrahedral nanostructure.²⁶ We also provide clear histological evidence that BiRDS reaches the posterior retina through the scleral-choroidal-retinal pathway. This design has far-reaching clinical implications, suggesting that repeated intraocular injections could be replaced by a less invasive, office-based, more patient-friendly delivery method without compromising drug bioavailability.

The enhanced miRNA-loading capacity and improved stability are other advantages of this architecture. miR-22, a potent modulator, is known to inhibit proangiogenic signaling and exert neuroprotective effects. The unique tetrahedral architecture of BiRDS can effectively deliver and stabilize miR-22, enhance its cellular uptake, and enable its enzyme-responsive release at target sites.^22,27 This overcomes the inherent drawbacks of naked miRNA, such as rapid degradation and poor intracellular delivery.¹⁴ Each tetrahedral structure was designed to carry three miR-22 molecules in this study, which enables BiRDS to deliver a greater payload of therapeutic agents. These favorable cargo capacities and penetration capabilities form the foundation for the use of BiRDS for the treatment of progressive retinal neurovascular diseases.

BiRDS has a remarkably significant antiangiogenic effect, surpassing or matching that of intravitreal aflibercept, the current clinical standard. In vitro assays revealed that BiRDS effectively inhibited the proliferation, migration, and tube formation of endothelial cells under hypoxia, which aligns with the ability of miR-22 to target multiple angiogenic pathways.²⁸ In vivo studies further demonstrated that BiRDS effectively inhibits pathological neovascularization in both laser-induced CNV and OIR models. Aflibercept intravitreal injection can be active for up to approximately 3 months. Owing to the self-limiting nature of CNV in rodent models, we are unable to observe longer-term effects. However, on the basis of the effects observed in rats on day 21, we speculate that the duration of therapeutic efficacy could be comparable to that of aflibercept.

Compared with aflibercept, the most significant difference in BiRDS is that it not only effectively inhibits pathological neovascularization but also, more importantly, exerts a remodeling effect on normal blood vessels and a protective effect on neural units. Restoring a functional and efficient vascular architecture is key to restoring proper perfusion and neural tissue function.^29,30 In this respect, devising strategies to accelerate the revascularization of ischemic nervous tissue with healthy vessels is beneficial for salvaging the function of ischemic tissue.³¹ Our results demonstrated that BiRDS can significantly increase the number of endothelial tip cells and filopodia extension at the leading edge of growing vessels. Beyond vascular regulation, BiRDS has substantial neuroprotective benefits, which addresses a critical unmet need in current treatment paradigms. Immunofluorescence analyses revealed preservation of key retinal neurons, including Tuj1⁺ ganglion cells, PKCα⁺ bipolar cells, rhodopsin-expressing rod photoreceptors, and Calbindin⁺ amacrine cells, indicating maintenance of the retinal circuitry. Correspondingly, ERG recordings revealed significant improvements in both scotopic and photopic responses, suggesting that the morphological neuroprotection conferred by BiRDS translates into functional visual preservation. Together, these findings indicate that BiRDS not only suppresses pathological neovascularization but also helps maintain synaptic integrity and neuronal viability, contributing to the preservation of visual function within the 7–14-day experimental period. Owing to these multiple effects, BiRDS has expanded the therapeutic spectrum of neurovascular retinal diseases. While we focused on laser-induced CNV and OIR as representative models of choroidal and retinal neovascularization, the dual vascular and neural benefits suggest that BiRDS could be adapted for diseases such as diabetic retinopathy, retinal vein occlusion, or even early-stage AMD, where neurodegeneration and microvascular instability coexist. Furthermore, the modularity of the tetrahedral scaffold means that BiRDS could be re-engineered to deliver other nucleic acid payloads or combined with complementary therapeutics, opening avenues for personalized and combinatorial treatment strategies.

The therapeutic payload (miR-22) is fundamentally different in function and mechanism. Unlike the miR-27a or miR-31 used in past studies, miR-22 modulates the Wnt/β-catenin axis and GSK3β signaling, enabling simultaneous antiangiogenic and neuroprotective effects—a combination not addressed in prior applications. In this study, transcriptome analysis revealed that the therapeutic effects of BiRDS are closely associated with the modulation of the canonical Wnt/β-catenin signaling pathway, a pivotal regulator of vascular development, endothelial proliferation, and neurovascular stability. This pathway is activated when Wnt ligands bind to Frizzled (FZD4) receptors and LRP5/6 coreceptors, leading to the stabilization and nuclear translocation of β-catenin, which in turn drives the transcription of angiogenic and inflammatory target genes.^32,33 Aberrant activation of Wnt signaling has been implicated in the pathogenesis of numerous retinal diseases, including diabetic retinopathy, familial exudative vitreoretinopathy (FEVR), and retinal neoplasms, where it promotes pathological neovascularization and disrupts neurovascular homeostasis.³³ Notably, several studies have demonstrated that targeting components of this pathway—such as blocking FZD4, LRP5, or β-catenin—can attenuate pathological angiogenesis and preserve retinal structure.³⁴ In our study, downstream validation via Western blotting and immunofluorescence confirmed significant suppression of β-catenin expression in retinal tissues following BiRDS treatment, supporting the hypothesis that BiRDS-mediated delivery of anti–miR-22 exerts its therapeutic effect, at least in part, through Wnt pathway inhibition.

Despite these promising results, our study has several limitations. The experiments were conducted in self-limited rodent models with relatively short follow-up periods, which may not fully capture the long-term safety, effectiveness, biodistribution, or therapeutic durability in larger eyes or human-like anatomy. Some ocular safety evaluation parameters, such as intraocular pressure, FFA and ERG, were not assessed over longer observation windows. Future studies should also evaluate repeated dosing safety and assess efficacy in larger preclinical models, such as nonhuman primates. Some experiments may have relatively low statistical power due to the limited number of animals used. In addition, we plan to include a scrambled (nontargeting) miRNA payload and an inactive bioswitch variant to validate the carrier effects and the RNase-H–dependent release mechanism. Furthermore, evaluations of assembly yield, SEC profiles, polydispersity, and stability in biological matrices are currently being performed as part of our efforts toward comprehensive preclinical development.

In summary, this work introduces BiRDS as a novel, rationally engineered, minimally invasive nanoplatform that delivers stable miR-22 to the retina, achieving simultaneous inhibition of pathological neovascularization, remodeling of normal vessels, and protection of neural function. By overcoming the key limitations of current therapies, BiRDS represents a promising next-generation strategy for the safe and effective treatment of retinal neurovascular dysfunction diseases.

Assembly and Characterization of the BiRDS

The BiRDS was constructed through rational design of complementary DNA/RNA hybrid strands based on Watson-Crick base pairing principles. All oligonucleotides were custom-synthesized and HPLC purified (Sangon Biotech, Shanghai, China). For assembly, the component strands were dissolved in TM buffer (10 mM Tris-HCl, 50 mM MgCl2, pH 8.0) at equimolar ratios (S1:S2:S3 = 1:1:1), with a 3-fold molar excess of miR-22a strands to ensure complete hybridization. The mixture underwent a thermal annealing process consisting of: (1) denaturation at 95 °C for 10 min to disrupt secondary structures, followed by (2) rapid cooling to 4 °C for 20 min to facilitate controlled self-assembly into tetrahedral nanostructures. The successful assembly of BiRDS was first confirmed by native polyacrylamide gel electrophoresis (PAGE). Subsequently, to verify the tetrahedral nanostructures of the materials, their morphologies were observed using TEM and AFM. For TEM (Hitachi Ltd., Tokyo, Japan), approximately 20 μL of BiRDS sample was dropped onto the TEM sample holder, and a copper grid was placed face-down on the droplet for 3 minutes, then removed and air-dried at room temperature. The grid was then placed onto 2% phosphotungstic acid staining solution for 1–2 minutes, removed, and the excess stain was wicked off with filter paper before imaging. For AFM (Bruker, Billerica, MA, USA), samples were deposited onto freshly cleaved mica surfaces, allowed to dry, and imaged in tapping mode under ambient conditions. Finally, sizes and zeta potentials were measured using a Zetasizer Nano ZS90 system (Malvern Panalytical Ltd., Malvern, United Kingdom).

Cellular localization

Human umbilical vein endothelial cells (HUVECs) were seeded on sterile glass coverslips placed in 6-well plates at a density of 2 × 10⁵ cells per well and allowed to adhere overnight. The cells were cultured in DMEM/F-12 containing 1% fetal bovine serum and a complete antibiotic solution at 37 °C in a humidified atmosphere with 5% CO₂. HUVECs were then incubated with 100 nM Cy5-labeled BiRDS for 4, 16, 24 h. After incubation, the cells on the coverslips were washed three times with phosphate-buffered saline (PBS), fixed with 4% paraformaldehyde (PFA) for 15 min at room temperature (RT), incubated with DAPI solution for 5 min to stain the cellular nucleus, then rinsed with PBS before fluorescence imaging.

Cell Culture and administration

HUVECs were acquired from American Type Culture Collection (ATCC) and cultured in Dulbecco’s modified Eagle’s medium nutrient Ham’s F-12 (1:1) (DMEM/F-12) (Gibco, Thermo Fisher Scientific, Waltham, MA, USA) containing 10% fetal bovine serum (FBS) and a complete antibiotic solution (100 U/mL penicillin and 100 μg/mL streptomycin) (Gibco, Thermo Fisher Scientific, Waltham, MA, USA). Cells were maintained under conditions of 37 °C in a humidified 5% CO₂ atmosphere. Cells between passages 3-6 were used for all experiments.

HUVECs were divided into five groups for the following experiments. (1) vehicle control (VHCL), which consisted of PBS, (2) miR-22 (miR-22-only, without tFNAs), (3) tFNA, (4) BiRDS, (5) AFL. The final concentration of miR-22, tFNAS and BiRDS was 100 nmol/L, and the concentration of AFL was 1 μg/μL. After incubation under hypoxic conditions (37 °C, 1% O₂, 5% CO₂) for 6-24 h, different experiments were carried out.

Cell counting Kit-8

HUVECs were prepared at a density of 1 × 10⁴ cells per 100 µL, mixed thoroughly, and 100 µL of each suspension was seeded into wells of a 96-well plate. Plates were incubated for 24 h before assigning wells to the respective experimental groups, with three replicates per group. At the end of the treatment period, 10 µL of CCK-8 reagent (Sigma-Aldrich, St. Louis, MO, USA) was added to each well, followed by incubation for 2 h at 37 °C. Absorbance was measured at 450 nm using a microplate reader (Tecan, Männedorf, Switzerland).

EdU Assay for Cell Proliferation

Cell proliferation was assessed using the 5-ethynyl-2′-deoxyuridine (EdU) Alexa Fluor 555 Click-iT EdU Imaging Kits (Thermo Fisher Scientific, Massachusetts, USA) following the manufacturer’s protocol. HUVECs were seeded on coverslips in 6-well plates at a density of 3 × 10⁵ cells/well and allowed to adhere overnight. Cells were treated with experimental conditions and incubated under hypoxic conditions (37 °C, 1% O₂, 5% CO₂) for 24 h. Prewarm the EdU solution, then add it to the medium to obtain a 1X EdU solution. After incubating for 2 h, cells were incubated with Alexa Fluor 555 reaction cocktail (containing Click-iT^® reaction buffer, CuSO₄, fluorescent azide, and buffer additive), and the nuclei were counterstained with Hoechst 33342 (5 μg/mL). Images were captured using Confocal laser microscopy (Carl Zeiss, Oberkochen, Germany), and EdU-positive cells (red) were quantified relative to total nuclei (blue) using ImageJ software. Cell proliferation was assessed via the rate of EdU-positive cell. The experiment was replicated four times.

Tube formation assay

The tube-forming ability of HUVECs was evaluated using growth factor-reduced Matrigel (#356231; Corning Incorporated, Corning, NY, USA). Matrigel was thawed overnight at 4 °C and pipetted into pre-chilled 96-well plates (50 μL/well). The plate was incubated at 37 °C for 30 min to allow polymerization. HUVECs (2 × 10⁴ cells/well) were resuspended in 50 μL DMEM/F12 (with or without treatments) and seeded onto the Matrigel-coated wells. The plate was divided into the group VHCL, miR-22, tFNAs, BiRDS, and AFL. After 6 h of incubation under either hypoxic conditions (37 °C, 1% O₂, 5% CO₂) or normoxic conditions (37 °C, 5% CO₂), tube formation was observed using an Inverted Biologic Microscope (IX2-SL; Olympus Corporation, Tokyo, Japan). Images were analyzed using Angiogenesis Analyzer (ImageJ) to quantify total branching length and branching points per field. The experiment was replicated a minimum of four times.

Transwell migration assay

Cell migration was assessed using 8 μm-pore, 24-well Transwell inserts (#3422, Corning Incorporated, Corning, NY, USA). The lower chamber was filled with 500 μL of DMEM/F12 containing 10% FBS. HUVECs (5 × 10⁴ cells/well) were dispersed in serum-free DMEM/F12 along with different concentrations of treatment agents and seeded into the upper chamber. After incubation for 18 h at either normoxia (37 °C, 5% CO₂) or hypoxia (37 °C, 1% O₂, 5% CO₂) conditions. Non-migrated cells on the upper membrane surface were removed with a cotton swab. Migrated cells on the lower surface were fixed with 4% PFA for 15 min and stained with 0.1% crystal violet for 20 min. Six random fields per insert were imaged (100× magnification) using an Inverted Biologic Microscope (IX2-SL; Olympus Corporation, Tokyo, Japan), and cells were counted using ImageJ. The experiment was.

Animals and ethical approval

C57BL/6 J mice (6–8 weeks old, 20-25 g) were purchased from GemPharmatech Co., Ltd (Nanjing, China). All animals were kept in a SPF environment for a week for follow-up experiments and were housed under a 12 h light/dark cycle with ad libitum access to food/water. The mice were anesthetized with 1% pentobarbital sodium (30 ~ 60 mg/kg body weight) via intraperitoneal injection before the procedure. Mice undergoing intravitreal injection, subconjunctival injection, laser photocoagulation or electroretinography (ERG) testing also received topical anesthesia with 0.5% proparacaine hydrochloride. All experimental procedures were performed unilaterally and adhered to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research and were approved by the Institutional Animal Care and Use Committee (IACUC) of Zhongshan Ophthalmic Center, Sun Yat-sen University and Sichuan Provincial People’s Hospital (Z2023073, LS-2025-428).

Ocular Distribution of BiRDS

A total of 5 μL of Cy5-labeled BiRDS was injected under the conjunctiva of mice at the age of 6–8 weeks. The mice were sacrificed at 10 min, 1 h, 3 h, 6 h and 12 h, and 18 h post-injection via intraperitoneal injection of an overdose of 1% pentobarbital sodium (150 mg/kg) followed by cervical dislocation. Subsequently, the eyeballs were carefully dissected. Then the eyeballs with removal of the surrounding periorbital tissue embedded in optimal cutting temperature (O.C.T) compound (Sakura Finetek, Torrance, CA, USA) and frozen on −80 °C. Eyeballs were cryosectioned at 6–8 µm thickness using a Leica CM1950 cryostat (−20 °C chamber temperature). Then the prepared sections were fixed with 4% PFA for 15 min at RT and rinsed with PBS three times. After incubated with Hoechst 33342 for 5 min at RT, the stained sections mounted with anti-fluorescence quencher. CLSM images were captured using a CLSM (LSM980; Carl Zeiss, Oberkochen, Germany). Respective fluorescence channels (dapi, 340/488 nm; Cy5, 650/670 nm) and all parameters were kept consistent throughout the experiment. Mean fluorescence intensity (MFI) of the labeled drug in the cornea, conjunctiva, iris, sclera, choroid and retina were quantified using ImageJ software.

Laser-induced CNV mouse /rat model

Laser photocoagulation was performed bilaterally in 6–8 week-old mice /rat on day 0 to induce CNV according to the standard protocol. To prepare for the procedure, pupils were dilated with 1% tropicamide eye drops and the mice were anesthetized using 1% pentobarbital sodium (50 mg/kg). A 532 nm wavelength laser (Phoenix Micron IV; Phoenix Research Laboratories, Pleasanton, CA, USA.) was used to generate 4–6 burns per eye (50 μm spot size, 50 ms duration, 350 mW intensity) at a distance of 2–3 papilla diameters around the optic disc (avoiding retinal vessels). The formation of a small bubble confirmed rupture of Bruch’s membrane, indicating successful CNV induction. For mice, on day 4, CNV formation was examined and treatments were administered via subconjunctival or intravitreal injection, as indicated. Follow-up examinations and sample collection were conducted on day 7 post-treatment to evaluate CNV progression and treatment effects. For rat, on day 7, CNV formation was examined and treatments were administered via intravitreal injection, as indicated. Follow-up examinations and sample collection were conducted on day 14 and day 21 post-treatment to evaluate CNV progression and treatment effects.

Drug Administration

Adult mice were anesthetized with 1% pentobarbital by intraperitoneal injection, while mouse pups with 0.5% pentobarbital. Then Proparacaine Hydrochloride (Alcaine; Alcon, TX, USA) eye drops were added to the mouses’ conjunctival sac for topical anesthesia before injection. Three (mouse pups) or 5 μL (adult mice) of drugs was injected into the subconjunctival space using a 34-gauge metal needle and a 100 μL syringe under a surgical microscope. The injection site was 1 mm posterior to the limbus to avoid leakage. The filling of the conjunctival sac regarded as a sign of successful injection. Tobramycin ointment was applied to prevent infection. The mice were divided into follow groups randomly: (i) VHCL (vehicle 5 μL; subconjunctival injection), (ii) miR-22 (miR-22-only, without tFNAs) (5 μL 5 μM; subconjunctival injection), (iii) tFNAS(5 μL 5 μM; subconjunctival injection), (iv) BiRDS (5 μL 5 μM; subconjunctival injection), and (v) anti-VEGFA (Aflibercept, AFL (Eylea, Bayer, Leverkusen, Germany), intravitreal injection: 1 μL 40 μg/μL). All procedures were conducted by the same surgeon. Anterior segment was examined once a day with a hand-held slit lamp following published methods.³⁵ A diffuse beam was used to assess adnexa, conjunctiva, cornea, and iris. Corneal staining was evaluated with fluorescein under cobalt blue illumination. Lesions of the clear media-including corneal opacity or neovascularization, anterior chamber flare and cells, lens opacity, and anterior vitreous cells-were scored accordingly.

Optical Coherence Tomography (OCT)

Optical coherence tomography (OCT) images were captured by using an image-guided tomographer (Micron IV-OCT2; Phoenix Research Laboratories, Pleasanton, CA, USA) on days 4 (D4) and days 7 (D7) after laser. CNV lesions were identified as hyperreflective spindle-shaped structures oriented parallel to the retinal pigment epithelium (RPE). For consistent measurements, we analyzed only the central scan of each lesion where the maximum dimensions could be visualized. Both the thickness (vertical dimension) and length (horizontal dimension) of CNV lesions were measured using ImageJ software.

Fundus Fluorescein Angiography (FFA)

Fundus Fluorescein Angiography (FFA) was used to examine the leakage area of CNV lesion on day 4 and day 7 post-laser induction. Following intraperitoneal administration of 2% fluorescein sodium (0.1 mL/10 g body weight), retinal images were acquired using a Micron IV fundus camera (Phoenix Research Laboratories) during two distinct phases: early (1-2 min post-injection) and late (4-5 min post-injection). The leakage area of CNV was quantified by measuring the fluorescein intensity across all laser spots relative to the optic disc area using ImageJ software.

Oxygen-Induced Retinopathy (OIR) Model

On postnatal day 7 (P7), C57BL/6 J mouse pups with nursing dams were transferred to a precisely regulated hyperoxic environment (75 ± 2% O₂) maintained within a sealed chamber (PRO-OX 110 oxygen controller; Biospherix Ltd., Redfield, NY, USA) for a standardized 5-day exposure period (P7-P12). Following hyperoxic induction, pups were returned to normoxic conditions (21% O₂) until P17, corresponding to the established peak of pathological retinal vaso-obliteration and neovascularization in this model. Therapeutic interventions (e.g., subconjunctival injections, intravitreal injections) were administered at P12. Terminal endpoints were assessed at P17, with subsequent enucleation of ocular tissues for further experiments.

Intravitreal injection

Mice or rats were anesthetized with an intraperitoneal injection of 1% pentobarbital sodium (25 mg/kg for neonatal animals and 50 mg/kg for adult animals) and received topical anesthesia with 0.5% proparacaine hydrochloride. Pupils were dilated using 0.5% tropicamide. Under a surgical microscope, a 33-gauge Hamilton syringe was inserted approximately 1 mm posterior to the limbus, the indicated volume of solution was slowly injected into the vitreous cavity as follows: 1 µL of AFL for OIR mice, 2 µL of AFL for adult CNV mice, and 4 µL of BiRDS or AFL for rats. Care was taken to avoid contact with or injury to the lens. After the injection, antibiotic ointment was applied to the ocular surface to prevent infection. Only one eye of each animal was injected. All animals were monitored postoperatively until they had fully recovered from anesthesia.

Electroretinogram (ERG)

ERG was conducted to examine the retinal function of the P17 pup mice using the Celeris system (Diagnosys LLC, Lowell, MA, USA).³⁵ Mice were dark-adapted overnight ( > 12 h) prior to ERG. The mice were anesthetized with intraperitoneal 1% pentobarbital sodium (25 mg/kg) and the pupils were dilated using 1% tropicamide under dim red light. Then, celeris proprietary contact lens electrode was placed on the cornea with 2% methylcellulose to maintain hydration. The amplitude and latency of a-, b-waves and oscillatory potentials (OPs) were measured according to the International Society for Clinical Electrophysiology of Vision Standard.³⁰ Mice with the OIR model underwent ERG measurement five days after the treatment.

Retinal flat-mounts and immunofluorescence

Pups were sacrificed at P17. Eyes were immediately enucleated using microdissection scissors and fixed in 4% PFA for 45 min to preserve tissue architecture. The cornea and lens were removed, then the retina was carefully separated from the RPE. Four radial cuts were made to flatten the retina as a “cloverleaf” configuration. Retina tissues were incubated in blocking buffer (1% bovine serum albumin (BSA) + 0.5% Triton X-100 in PBS) for 1 h at RT to prevent nonspecific binding. Retinas were incubated with Isolectin B4-AlexaFluor 568 (1:250; I21412, Thermo Fisher Scientific, Massachusetts, USA) overnight at 4 °C with agitation to label blood vessels. Retinas were washed with PBS and mounted on slides. Images of retina flat-mount were performed using a Fluorescence Inversion Microscope (TS2R-FL, Nikon, Tokyo, Japan). Vaso-Obliteration and neovascularization were quantified using ImageJ software.

Retinal sections and immunofluorescence

Immediately following enucleation, a precise corneal incision was created using micro-dissection scissors to permit intracameral perfusion of fixative, thereby optimizing tissue preservation while reducing requisite fixation duration. Then, eyes of mice were fixed in 4% PFA for 45 min at RT. Following fixation, specimen were subjected to graded sucrose dehydration (10%, 20%, and 30% w/v in PBS; Biofroxx, Einhausen, Germany) at 4 °C until complete tissue saturation was achieved. Subsequently, anterior segment structures (cornea, iris, and lens complex) were carefully excised to yield intact retinal eyecups. The processed eyes were embedded in O.C.T. compound followed by freezing in liquid nitrogen-cooled isopentane (-80 °C). Serial sections (6-8 μm thickness) were obtained using a cryostat (Leica CM3050S) maintained at −20 °C chamber temperature and mounted on microscope slides.

Tissue sections were air-dried at RT for 15 min prior to rehydrating in PBS, and incubated in blocking solution containing 1% BSA and 0.5% Triton X-100 in PBS for 60 min at RT to minimize nonspecific binding and enhance antibody penetration. Then, tissues incubated with the primary antibodies at 4 °C overnight in a humidified chamber. After thorough PBS washing, appropriate species-matched secondary antibodies were applied for 120 min at RT under light-protected conditions. All sections were subsequently stained with Hoechst 33342 (Solarbio, Beijing, China; 1 μg/mL in PBS) for 10 min to visualize nuclear architecture. Processed sections were imaged using a confocal laser scanning microscope (LSM 980; Carl Zeiss, Oberkochen, Germany). The following antibodies were used for immunofluorescence: Isolectin GS-IB4, Alexa Fluor™ 568(I21412, Thermo Fisher Scientific, Massachusetts, USA), CD31(SC-376764, Santa Cruz Biotechnology, Dallas, TX, USA), VEGFR (#AF644; R&D Systems, Minneapolis, MN, USA), Tuj1 (#4466; Cell Signaling Technology, Danvers, MA, USA), GFAP (#12389; Cell Signaling Technology, Danvers, MA, USA), PKC-α (#sc-8393; Santa Cruz Biotechnology, Dallas, TX, USA), Rhodopsin (#ab221664; Abcam, Cambridge, UK), Calbindin (#ab82812; Abcam, Cambridge, UK), Alexa Fluor 488-labeled goat anti-rabbit IgG (#4412S; Cell Signaling Technology, Danvers, MA, USA), and Alexa Fluor 488-labeled goat anti-mouse IgG (#4408S; Cell Signaling Technology, Danvers, MA, USA), Alexa Fluor 555-conjugated goat anti-rabbit IgG (#25363S, Cell Signaling Technology, Danvers, MA, USA), Alexa Fluor 555-conjugated goat anti-mouse IgG (#37459S, Cell Signaling Technology, Danvers, MA, USA).

RNA sequencing (RNA-seq)

To elucidate the molecular mechanisms mediating BiRDS’s therapeutic effects, we conducted comprehensive transcriptomic analyses using an in vitro hypoxia model. HUVECs were subjected to hypoxic conditions (1% O₂, 5% CO₂, 37 °C) for 24 h in the presence of either VHCL or 100 nmol/L BiRDS. Total RNA was isolated from both normoxic controls and treated cells using the RNeasy Plus Kit (#74106; Qiagen, Hilden, Germany), with all samples demonstrating excellent RNA integrity (RIN > 7.0). High-throughput RNA sequencing was performed following rigorous quality control measures. RNA-Seq libraries were prepared according to the manufacturer’s protocol (TruSeq^® RNA Sample Preparation Kit, Illumina). The resulting libraries were purified and amplified by PCR, followed by stringent quality assessment using both the Qubit^® 2.0 Fluorometer (Life Technologies) for quantification and the Agilent 2100 bioanalyzer (Agilent Technologies) for size distribution analysis. Bioinformatic analysis revealed significant alterations in gene expression profiles, with differential expression defined by |log2(fold change)| > 1 and adjusted p-value < 0.05 (DESeq2 algorithm). Subsequent pathway analysis using clusterProfiler identified functionally enriched Gene Ontology (GO) terms and KEGG pathways, providing mechanistic insights into BiRDS’s mode of action. To confirm these transcriptomic findings, we performed orthogonal validation through qRT-PCR and Western blot analyses of selected differentially expressed genes and pathway components, ensuring robust reproducibility of our results. This integrated multi-omics approach provides a comprehensive understanding of BiRDS-mediated molecular changes under hypoxic stress.

RNA isolation and gene expression analysis

Total RNA was extracted using Trizol Reagent (#15596018; Thermo Fisher Scientific, Massachusetts, USA) and quantified with a spectrophotometer (NanoDrop ND- 1000; Thermo Fisher Scientific, Massachusetts, USA). cDNA synthesis was performed with 500 ng of total RNA using PrimeScript RT Master Mix (#KR116-02; TIANGEN, Beijing, China). Gene expression was then measured by quantitative PCR (qPCR) with SYBR Green Supermix (#1708880; Bio-Rad Laboratories, Hercules, CA, USA) nd calculated using the 2^ΔΔCt method. β-actin was used as a reference gene. The sequences of the primers (5′ − 3′) used were as follows:

Western blot analysis

Total proteins were extracted from the harvested cells using RIPA lysis buffer (#R0010-20; Solarbio, Beijing, China) containing protease and phosphatase inhibitors. Protein concentration was quantified via a BCA protein assay kit (#KGP903; KeyGEN, Nanjing, China) according to the manufacturer’s protocol. Equal amounts of protein (20-30 μg) were denatured in loading buffer by heating at 95 °C for 5 min, then separated electrophoretically on 10% SurePAGE™ Bis-Tris gels (#M00656; GenScript, Nanjing, China) and transferred to PVDF membranes (Millipore, Burlington, MA, USA). After blocking with 5% non-fat milk in TBST (#P0023B; Beyotime, Shanghai, China) for 1 h at RT, membranes were probed with the following primary antibodies overnight at 4 °C: FZD4 (#bs-13217R; Invitrogen, Carlsbad, CA, USA), GSK-3β (#ET1607-71; HuaBio, Hangzhou, China), and β-catenin (#71-2700; Invitrogen, Carlsbad, CA, USA). Following three washes with TBST, membranes were incubated with HRP-conjugated secondary antibodies (#7076S; Cell Signaling Technology, Danvers, MA, USA) for 1 h at RT. Protein bands were detected using enhanced chemiluminescence (ProteinSimple) and quantified densitometrically using ImageJ software (NIH). β-actin served as the loading control for normalization.

Statistical analysis

All data were analyzed using SPSS Statistics 25.0 (IBM, USA), GraphPad Prism 8.0 (GraphPad Software, USA), and R programming language. The data normality was assessed using the Shapiro-Wilk test. Appropriate statistical tests were applied according to data distribution: one-way ANOVA with Tukey’s post hoc test for normally distributed data and Kruskal-Wallis test for non-normal data. Continuous variables are presented as mean ± SD from at least three independent experiments, with each experiment including three biological replicates. Post-hoc power analysis was conducted. Data are presented as mean ± SD, and p < 0.05 was considered significant.