Slit2/Robo Signaling Restores Diabetic Erectile Function via Neurovascular Remodeling

ABSTRACT

Background

Diabetes mellitus‐induced erectile dysfunction (DMED) is a common and debilitating vascular‐neurogenic complication of diabetes. The poor responsiveness of DMED patients to phosphodiesterase type‐5 inhibitors underscores the need for therapies capable of restoring both endothelial integrity and cavernous nerve function. The Slit2/Roundabout (Robo) pathway, a conserved regulator of axon guidance and vascular remodeling, is well positioned to coordinate neurovascular repair; however, its involvement in DMED remains largely undefined.

Methods

Single‐cell RNA sequencing from embryonic genital tubercle to postnatal stages was used to map fibroblast‐endothelial‐neuronal signaling and identify Slit2 as a key secreted ligand. Diabetic and control mouse corpus cavernosum tissues were analyzed for endothelial and neuronal Slit2‐Robo1/4 activity. In vitro, recombinant Slit2‐N was applied to cells under high glucose to assess proliferation, migration, tube/sprout formation, and neurite outgrowth; Robo1/4 knockdown tested receptor dependency. In vivo, AAV‐Slit2 was delivered intracavernously to diabetic mice, with erectile function assessed by sodium‐nitroprusside perfusion and electrostimulation, and structural remodeling evaluated by endothelial, neural, and fibrosis markers.

Results

Single‐cell profiling identified Slit2 as a fibroblast‐derived developmental cue for corpus cavernosum morphogenesis. Diabetic mice exhibited markedly reduced endothelial and neuronal Slit2‐Robo signaling. Slit2‐N restored angiogenic and neuroregenerative functions in vitro under high glucose, enhancing endothelial activity and neurite growth; all effects were lost after Robo1/4 silencing. In vivo, AAV‐Slit2 re‐established local Slit2 signaling, improved penile perfusion and erectile responses, enhanced endothelial proliferation and nerve regeneration, and reduced fibrosis, indicating substantial neurovascular recovery.

Conclusions

Slit2/Robo signaling is a key regulatory axis disrupted in DMED. Local restoration of Slit2 via AAV delivery drives synchronized neurovascular regeneration and effectively rescues erectile function in diabetic mice. These findings establish Slit2 as a promising therapeutic target for reversing the underlying pathophysiology of DMED rather than merely providing symptomatic relief.

1. Introduction

Erectile dysfunction (ED) is one of the most prevalent and debilitating chronic complications in men with diabetes [1, 2]. Recent epidemiological evidence shows that the global prevalence of ED ranges from 3% to 76.5% and continues to increase [3], while the pooled prevalence of erectile dysfunction among patients with diabetes worldwide is as high as 65.8% (95% CI: 58.3–73.3%) [4]. Notably, diabetes mellitus‐induced erectile dysfunction (DMED) develops earlier, progresses more rapidly, and often serves as an early clinical warning signal of systemic endothelial injury, neurodegeneration, and metabolic deterioration [5]. The pathophysiology of DMED is multifactorial, driven by chronic hyperglycemia‐induced vascular endothelial dysfunction, degenerative changes and fibrosis of corpus cavernosum smooth muscle, impaired neurovascular coupling, and endocrine/metabolic abnormalities [6, 7]. These converging pathological events render DMED a complex disorder involving vascular, neurogenic, smooth muscle‐derived, and metabolic etiologies.

Although phosphodiesterase type‐5 inhibitors (PDE5i) remain the first‐line treatment and can substantially improve erectile function in the general ED population, diabetic patients exhibit markedly reduced responsiveness to PDE5i therapy, particularly those with severe cavernous nerve injury or advanced vasculopathy [8]. Current clinical interventions predominantly focus on symptomatic relief rather than reversing the underlying pathological deterioration, and thus fail to restore endothelial integrity or promote regeneration of degenerated cavernous nerves. In recent years, a variety of regenerative and microenvironment‐modulating strategies have emerged [9, 10], including proangiogenic and neurotrophic agents, stem cell and exosome‐based therapies, gene therapy, and localized tissue engineering approaches, aiming to rebuild neurovascular function at the disease origin. Preclinical studies have demonstrated that several angiogenic and neurotrophic factors—such as Ang1 [11], VEGF [12], BDNF [13], Latrophilin‐2 [14], and MT‐100 [15]—can partially improve vascular and neural structure and function within the corpus cavernosum. However, most of these approaches remain limited by single‐target activity, insufficient local bioavailability, and inadequate therapeutic efficacy under the complex diabetic pathological microenvironment. Therefore, there is an urgent need to develop novel therapeutic paradigms based on neurovascular synchronized regeneration to simultaneously restore nerve repair and vascular reconstruction in DMED.

Slit2, a secreted glycoprotein of the Slit family and the canonical ligand of Roundabout (Robo) receptors (Robo1–4) [16, 17], plays a pivotal role in neural guidance [18, 19], angiogenesis and endothelial remodeling [20, 21], immune‐inflammatory modulation [22, 23], and tissue regeneration [24], serving as a crucial signaling hub linking the nervous and vascular systems. Structurally, full‐length Slit2 undergoes proteolytic cleavage to yield two functionally distinct fragments: the N‐terminal fragment Slit2‐N (∼140 kDa), primarily binds to Robo1 to regulate axonal repulsion and neural pathfinding [25, 26, 27], and also primarily interacts with Robo1/4 to modulate angiogenesis, vascular remodeling, and endothelial permeability [28, 29, 30]; and the C‐terminal fragment Slit2‐C (∼55‐60 kDa), recently recognized for its metabolic regulatory function by promoting adipose browning and thermogenesis, thereby improving insulin sensitivity and restoring systemic glucose homeostasis [31, 32].

Accumulating evidence has implicated Slit2 in multiple diabetes‐associated complications, including diabetic peripheral neuropathy [33], nephropathy [34], and retinopathy [35, 36], suggesting its critical regulatory role in diabetes‐related tissue injury. Interestingly, the therapeutic outcomes of Slit2 appear to be highly dependent on pathological microenvironmental cues and the dynamic interplay among Robo receptor isoforms. For instance, in diabetic nephropathy, hyperglycemia‐induced activation of Slit2/Robo1 signaling drives aberrant angiogenesis, exacerbating glomerular endothelial proliferation and fibrosis [34]. In contrast, in diabetic retinopathy, Slit2 binding to Robo4 stabilizes vascular integrity and attenuates vascular leakage [37], underscoring its dual, receptor‐dependent regulatory effects in vascular remodeling. This isoform‐specific bidirectional modulation renders Slit2 a unique regulatory factor in neurovascular remodeling.

Given that the core pathological hallmarks of DMED involve endothelial injury and neurodegeneration, Slit2, as a convergent molecular hub orchestrating neuroguidance and vascular homeostasis, is highly likely to play a critical role in DMED onset and progression. It is therefore essential to elucidate the expression patterns, functional roles, and neurovascular‐coupling mechanisms of the Slit2‐Robo signaling axis within the corpus cavernosum under diabetic conditions. In this study, we aimed to systematically investigate the expression patterns, cellular sources, and functional roles of the Slit2‐Robo signaling axis within the corpus cavernosum under both developmental and diabetic conditions. By integrating single‐cell transcriptomic analysis with in vitro and in vivo functional approaches, we sought to elucidate how Slit2‐mediated signaling contributes to neurovascular homeostasis and to determine whether targeting this pathway could provide a therapeutic strategy for restoring erectile function in DMED.

2. Results

2.1. Single‐cell Transcriptomic Profiling Reveals Cellular Heterogeneity and Slit2‐Robo Signaling Axis During Penile Tissue Development

To delineate the cellular architecture of developing penile tissue, we performed single‐cell transcriptomic profiling across multiple developmental stages, spanning from the embryonic genital tubercle to postnatal sexual maturation. Unsupervised UMAP clustering first resolved six major lineages—mesenchymal/fibroblast, epithelial, endothelial, myeloid immune, smooth muscle, and Schwann cells—indicating substantial intercellular diversity (Figure 1A). We then applied higher resolution subclustering, which further uncovered 12 transcriptionally distinct populations, including region‐specific mesenchymal domains of the glans, corpus cavernosum (cc) glandis, cc urethra (corpus spongiosum) and prepuce, together with endothelial and immune compartments, Schwann cells, smooth muscle cells, and a proliferative mesenchymal subset (Figure 1B). Building upon this cellular atlas, we next interrogated the signaling cues governing penile development and identified Slit2 as a mesenchymal/fibroblast‐enriched ligand exhibiting stage‐dynamic expression, progressively declining from birth to sexual maturation (Figure 1C,D).

FIGURE 1.

Single‐cell transcriptomic profiling reveals cellular heterogeneity and Slit2‐Robo signaling axis during penile tissue development. (A) Unsupervised broad clustering of all profiled cells. UMAP visualization showing the major cell lineages identified from pooled tissue samples across all developmental stages. Cell clusters are broadly annotated as Mesenchyme/fibroblast, Epithelial cell, Endothelial cell, Myeloid immune cell, Smooth muscle cell, and Schwann cell. (B) Subclustering reveals fine‐grained cellular identity within the penile tissue. UMAP projection of the subclustered cells, resolving 12 distinct cell populations including: Schwann cells, SMC, Myeloid cells, EC, CC Urethra, CC Glans, Ventral Glans, Epithelial cells, General Prepuce, Subdermal Prepuce, Dorsal Glans, and a Proliferating mesenchyme population. (C) UMAP‐based density visualization of Slit2 expression. (D) Average expression of Slit2 across developmental stages. (E) Sender‐receiver communication architecture inferred through CellPhoneDB. (F) Heatmap of intercellular communication strength mediated by SLIT and NTRK‐like protein family. (G and H) Contribution of individual L‐R pairs to the total SLIT and NTRK‐like protein signaling. Bar chart and bubble plot quantifying the percentage contribution of the top ligand‐receptor pairs. Note the predominant contribution of SLIT2‐ROBO pairs. I. UMAP visualization of Robo1 expression. (J) Circos plot of SLIT2_ROBO1 mediated directional communication. Note: Uppercase gene symbols (e.g., SLIT2, ROBO) in panels F‐J are presented as defined by the CellPhoneDB ligand–receptor interaction database.

Guided by this observation, we reconstructed intercellular communication using CellPhoneDB. The inferred network highlighted SLIT/NTRK‐like signaling as a major communication module, with SLIT2‐ROBO interactions contributing the dominant ligand‐receptor strength (Figure 1E–H). Consistently, Robo1 expression was predominantly localized to epithelial and endothelial compartments, and circos analysis further confirmed a directional SLIT2‐ROBO1 communication stream within the developing penile microenvironment, primarily targeting epithelial and endothelial cells (Figure 1I,J). Together, these data establish a detailed cell map and identify Slit2‐Robo as a central communication axis orchestrating penile tissue development.

2.2. DMED Disrupts Fibroblast‐derived Slit2 and Endothelial/Neuronal Robo Signaling at Single‐cell and Tissue Levels

To determine whether diabetic erectile dysfunction (DMED) perturbs the Slit2‐Robo signaling axis, we performed single‐cell transcriptomic profiling on adult mouse penile tissues under control and DMED conditions. UMAP‐based global clustering revealed the major cellular compartments of the penile microenvironment (Figure 2A), and refined subclustering further resolved 12 transcriptionally distinct populations, enabling higher‐resolution discrimination of fibroblast (FB) and endothelial cell (EC) subsets (Figure 2B). We next examined the expression landscape of Slit2 and its receptors across conditions. Spatial UMAP visualization revealed that Slit2 expression in adult corpus cavernosum was primarily distributed within Slc1a3⁺ fibroblasts, whereas Robo1 was broadly distributed across fibroblasts, endothelial cells, and smooth muscle cells. In contrast, Robo4 expression was largely confined to endothelial populations (Figure 2C–F). To quantitatively assess alterations in Slit2 signaling under DMED, we compared ligand–receptor expression across defined cell populations. In healthy tissue, Slc1a3⁺ fibroblasts serve as the primary source of Slit2, which was markedly reduced along with associated downstream components in DMED (Figure 2G). Concurrently, vascular endothelial cells (VECs) exhibited significant downregulation of Robo1 and Robo4, indicating impaired receptor‐mediated ligand reception (Figure 2H).

FIGURE 2.

DMED disrupts fibroblast‐derived Slit2 and endothelial/neuronal Robo signaling at single‐cell and tissue levels. (A) Broad clustering of all profiled cells. (B) Subclustering identifies fine‐grained cell types. UMAP projection of the subclustered cells, resolving 12 distinct cell populations, further refining the identities of FB and EC populations. (C–F) UMAP visualization of Slit2, Robo1 and Robo4 expression. (G) Expression of Slit2/Robo components in Slc1a3⁺ fibroblasts under CON and DMED conditions. (H) Expression of Slit2/Robo components in vascular EC (VEC) cells under CON and DMED conditions. (I to P) Representative immunofluorescence costaining images and corresponding semiquantitative analysis in different groups (n = 6): I, M. Pdgfrα (red) and Slit2 (green). Scale bar: 100 µm. (J and N) Lectin (red) and Robo1 (green). Scale bar: 100 µm. (K and O). Lectin (red) and Robo1 (green). Scale bar: 100 µm. (L and P) Lectin (red) and Robo4 (green). Scale bar: 100 µm. Significance: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Immunofluorescence staining of the corpus cavernosum further validated the molecular alterations of Slit2, Robo1, and Robo4 at the protein level. Under diabetic conditions, Slit2 signals derived from fibroblasts were markedly diminished, as evidenced by the substantial reduction in Slit2 colocalization within Pdgfrα⁺ fibroblasts (Figure 2I,M). Correspondingly, the expression of its major receptor Robo1 showed a noticeable decline in two anatomically and functionally crucial regions: the dorsal penile nerve within the neurovascular bundle (Figure 2J,N) and the endothelial lining of the cavernosal sinusoids (Figure 2K,O). Robo4 exhibited a similar pattern, with significantly reduced expression in sinusoidal endothelial cells of DMED mice (Figure 2L,P), although it remained scarcely detectable in the dorsal nerve (Figure S1A).

Together, these findings demonstrate a coordinated failure of both the “source” and the “receiving machinery” of neurovascular regenerative signaling within the diabetic penile microenvironment. Specifically, compromised fibroblast‐derived Slit2 secretion and downregulation of endothelial and neural Robo1/4 receptors collectively dismantle the Slit2‐Robo signaling axis. This collapse of neurovascular guidance cues likely contributes fundamentally to the progressive vascular injury and neural degeneration characteristic of DMED.

2.3. Slit2 Activates Robo1/Robo4 Signaling to Reverse High‐Glucose‐Induced Endothelial Dysfunction and Restore Angiogenic Competence

To elucidate the functional significance of the Slit2/Robo signaling axis in endothelial repair and angiogenic regeneration, we systematically evaluated the effects of recombinant Slit2 (rSlit2) on human cardiac microvascular endothelial cells (HCMECs) under high‐glucose (HG) conditions, and further examined the respective contributions of Robo1 and Robo4. Five experimental groups were included: control (Con), high glucose (HG), Slit2 treatment (HG+rSlit2), Robo1 knockdown (HG+rSlit2+shRobo1), and Robo4 knockdown (HG+rSlit2+shRobo4).

In the EdU incorporation assay (Figure 3A,E), HG markedly suppressed endothelial proliferation (10.55 ± 2.26%), reflecting a low‐proliferative state characteristic of the diabetic milieu. Slit2 treatment effectively restored proliferative capacity (25.76 ± 2.21%), approaching the normal baseline (33.37 ± 3.51%). This pro‐proliferative effect was partially diminished when Robo1 was silenced (21.43 ± 1.38%), whereas Robo4 knockdown almost completely abrogated Slit2‐mediated rescue (13.84 ± 2.91%), indicating that Robo4 plays a dominant role in regulating endothelial proliferation. Consistent findings were observed in scratch‐wound assays (Figure 3B,F). HG severely impaired endothelial migration (16.45 ± 3.83%), while Slit2 robustly restored migratory capacity (62.94 ± 6.44%), nearly reaching control levels (73.70 ± 7.00%). Silencing Robo1 attenuated this recovery (49.16 ± 4.65%), whereas Robo4 knockdown caused a profound reduction (29.18 ± 5.48%), further underscoring the indispensable role of Robo4 in endothelial repair.

FIGURE 3.

Slit2 Activates Robo1/Robo4 signaling to reverse high‐glucose‐induced endothelial dysfunction and restore angiogenic competence. (A and E) Representative fluorescence images (A) and corresponding semi‐quantitative analysis (E) of EdU incorporation assay in HCMECs (n = 5). Scale bar: 100 µm. (B and F) Representative images (B) of the scratch assay in HCMECs and corresponding semiquantitative analysis (F) of migration rate (n = 5). Scale bar: 200 µm. (C and G). Representative images (C) and corresponding semi‐quantitative analysis (G) of endothelial outgrowth from ex vivo aortic ring sprouting assay (n = 5). Scale bar: 500 µm. (D, H, I) Representative images (D) and statistical analysis of branches (H) and nodes (I) from the tube formation assay of HCMECs (n = 5). Scale bar: 200 µm. Significance: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

To assess the impact of Slit2 on angiogenic potential, we conducted an ex vivo aortic ring sprouting assay (Figure 3C,G). HG significantly reduced microvessel sprouting (2.01 ± 0.33), whereas Slit2 markedly enhanced sprout formation and outward extension (4.67 ± 0.54), comparable to control levels (5.45 ± 0.48). Sprouting was reduced following Robo1 knockdown (3.84 ± 0.42) and further diminished by Robo4 knockdown (2.90 ± 0.50), confirming that Slit2 promotes angiogenesis through coordinated engagement of both Robo1 and Robo4, with Robo4 serving a more critical function. Similarly, HCMEC tube‐formation assays (Figure 3D,H,I) demonstrated that HG severely disrupted the formation of vascular networks, whereas Slit2 significantly increased branching complexity, node counts, and overall network integrity. Knockdown of Robo1 or Robo4 impaired these proangiogenic effects, with Robo4 silencing nearly completely blocking Slit2‐induced tube formation.

Collectively, these findings demonstrate that Slit2 effectively restores endothelial proliferation, migration, and angiogenic capacity compromised by diabetic conditions, and that intact expression of Robo1 and Robo4 is essential for this regenerative response. Notably, Robo4 functions as a central regulator within the Slit2‐mediated angiogenic program. These results underscore the therapeutic potential of the Slit2/Robo axis in reactivating vascular regenerative processes under diabetic stress, providing a molecular foundation for vascular repair in DMED.

2.4. Slit2 Activates Robo1 Signaling to Reverse High‐Glucose‐Induced Neuronal Impairments and Restore Axonal Regenerative Capacity

Given the prominent expression of Robo1 within the neuronal lineage, we further investigated the functional relevance of Slit2/Robo signaling in neural repair and axonal regeneration. PC12 cells, which exhibit neuron‐like characteristics, were used to model key regenerative processes. Similar to endothelial cells, high‐glucose conditions markedly suppressed the biological activity of PC12 cells. In terms of proliferation (Figure 4A,B), high glucose reduced PC12 proliferative capacity to a low‐activity state (10.18 ± 2.60%), whereas Slit2 treatment robustly restored proliferation (24.42 ± 3.08%), approaching physiological levels (30.07 ± 3.68%). This pro‐proliferative effect was substantially diminished when Robo1 was silenced (17.10 ± 2.82%), indicating that Robo1 is an essential mediating receptor in this process. A similar pattern was observed in cell motility (Figure 4C,D). High glucose severely impaired PC12 migration, while Slit2 significantly rescued migratory capacity (49.70 ± 5.02%), nearly reaching control levels (63.13 ± 6.02%). Knockdown of Robo1 almost completely abolished the Slit2‐mediated enhancement of motility (26.91 ± 5.36%), reinforcing Robo1's central role in the regulation of neuronal movement and repair.

FIGURE 4.

Slit2 activates robo1 signaling to reverse high‐glucose‐induced neuronal impairments and restore axonal regenerative capacity. (A and B) Representative fluorescence images (A) and corresponding semiquantitative analysis (B) of EdU incorporation assay in PC12 cells (n = 5). Scale bar: 100 µm. (C and D) Representative images (C) of the scratch assay in PC12 cells and corresponding semiquantitative analysis (D) of migration rate (n = 5). Scale bar: 200 µm. (E and F) Representative immunofluorescence images (E) and corresponding semiquantitative analysis (F) of β3‐Tubulin‐labeled neurite outgrowth from ex vivo dorsal root ganglion (DRG) cultures (n = 5). Scale bar: 200 µm. Significance: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

To validate these findings in a more physiologically relevant neural regeneration model, we further assessed the effect of Slit2 on β3‐tubulin‐positive neurite outgrowth in ex vivo DRG cultures (Figure 4E,F). Slit2 not only markedly increased neurite length but also enhanced axonal branching, indicating that it promotes multiple critical steps in axonal regeneration.

Collectively, these results demonstrate that Slit2 can effectively restore neuronal proliferation and migration impaired by high‐glucose stress, while also enhancing axonal elongation and network reconstruction. These regenerative effects are highly dependent on intact Robo1 expression. Thus, the Slit2‐Robo1 signaling axis emerges as a key regulatory mechanism in diabetic cavernous nerve repair, providing a strong mechanistic basis for promoting neuroregeneration in DMED.

2.5. Reactivation of Slit2 Signaling Restores Cavernosal Perfusion and Structural Remodeling in DMED

To systematically evaluate the therapeutic potential of Slit2 within the pathological microenvironment of DMED, we first established an AAV‐mediated Slit2 overexpression model (AAV‐Slit2) via intracavernous injection. Fluorescent signals from AAV‐Blank demonstrated uniform viral distribution and robust infection efficiency throughout the corpus cavernosum (Figure 5A), providing a reliable foundation for subsequent functional assessments. Western blot analysis further confirmed a marked increase in Slit2 protein levels within the corpus cavernosum following AAV‐Slit2 administration (Figure 5B,C), elevating its expression to supraphysiological levels, thereby validating the successful construction of the overexpression model. After the treatment period, diabetic mice in all experimental groups exhibited comparable trajectories of body weight and blood glucose levels (Figure S2A, B), thereby excluding the potential confounding influence of altered glucose metabolism on the observed phenotypic outcomes.

FIGURE 5.

Reactivation of Slit2 signaling restores cavernosal perfusion and structural remodeling in DMED. (A) Representative fluorescence images of AAV‐Blank infection in the mouse corpus cavernosum. (B and C). Representative Western blot bands (B) and corresponding semiquantitative analysis (C) of Slit2 expression (n = 3). (D and E) Representative laser speckle contrast images (D) and statistical analysis of Δperfusion (E) depicting sodium nitroprusside (NO donor)‐induced penile blood flow (n = 6). The color scale (blue to red) indicates the relative perfusion levels across anatomical subregions. (F–H) Representative curves of electrostimulation‐induced intracavernosal pressure dynamics (F) and hemodynamic quantification of max ICP/MAP (G) and total ICP/MAP (H) (n = 6). (I and J) Quantification of NO (I) and cGMP (J) levels in mouse cavernosal tissue homogenates (n = 6), with all measurements normalized to the protein concentration. (K and L). Representative images (K) and corresponding semiquantitative analysis (L) of Masson's trichrome‐stained mouse corpus cavernosum sections (n = 6). Scale bar: 100 µm. (M and N) Representative immunofluorescence costaining images (M) and corresponding semiquantitative analysis (N) of α‐SMA (red) and CollagenI (green) (n = 6). Scale bar: 100 µm. Significance: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

For functional evaluation, we employed laser speckle contrast imaging to assess sodium nitroprusside (NO donor)‐induced penile perfusion. DMED mice exhibited prominent perfusion deficits (155.50 ± 23.94), significantly lower than healthy controls (367.52 ± 37.74), indicative of severe vascular hyporeactivity. Remarkably, AAV‐Slit2 treatment substantially restored NO‐mediated corpus cavernosum perfusion (274.85 ± 33.45), approaching physiological levels (Figure 5D,E). To determine whether this improvement translated into functional erectile recovery, we further performed electrical stimulation‐induced cavernous pressure recordings. AAV‐Slit2 significantly increased both maximal ICP/MAP and total ICP/MAP in DMED mice (Figure 5F–H), demonstrating that Slit2 effectively reestablished hemodynamic filling responses and ameliorated DMED‐associated erectile dysfunction. To elucidate the molecular basis underlying this functional recovery, we evaluated the NO‐cGMP signaling cascade within the corpus cavernosum. AAV‐Slit2 markedly elevated NO and cGMP levels (Figure 5I,J), suggesting enhanced endothelial nitric oxide signaling—an essential mechanism driving the restoration of erectile function.

At the histological level, DMED mice exhibited classical structural deterioration of the corpus cavernosum, including reduced smooth muscle content and excessive fibrosis. Masson's trichrome staining revealed that AAV‐Slit2 significantly decreased collagen deposition and preserved smooth muscle architecture (Figure 5K,L). Consistently, α‐SMA/Collagen I dual immunofluorescence showed that Slit2 overexpression restored smooth muscle contractile components and suppressed pathological collagen accumulation, resulting in a marked increase in the α‐SMA/Col1a2 ratio (Figure 5M,N).

Collectively, AAV‐Slit2 enhances endothelial NO‐cGMP signaling, improves cavernosal hemodynamics, and reverses DMED‐associated structural abnormalities, ultimately leading to a robust recovery of erectile function in DMED mice. These findings highlight Slit2 as a promising therapeutic strategy capable of targeting multiple pathological dimensions of DMED.

2.6. Therapeutic Slit2 Delivery Activates Dual Neurovascular Repair Pathways in DMED

To further delineate the pro‐regenerative actions of Slit2 within the pathological microenvironment of DMED, we comprehensively evaluated its impact on neurovascular structure and function in the corpus cavernosum. At the protein level, representative Western blot analyses revealed that AAV‐Slit2 markedly increased the expression of PCNA, VEGF, and GAP43 (Figure 6A—D), corresponding to key markers of cellular proliferation, angiogenesis, and axonal growth, respectively. These findings indicate that Slit2 concurrently activates multiple biological programs essential for tissue regeneration. Consistently, PCNA immunofluorescence confirmed a pronounced increase in proliferative activity throughout the cavernosal tissue (Figure 6E,F), suggesting that Slit2 reinitiates reparative cellular turnover that is otherwise suppressed in DMED.

FIGURE 6.

Therapeutic Slit2 Delivery Activates Dual Neurovascular Repair Pathways in DMED. A—D. Representative Western blot bands (A) and corresponding semi‐quantitative analysis (B—D) of PCNA, VEGF, and GAP43 expression (n = 3). E, F. Representative immunofluorescence images (E) and corresponding semi‐quantitative analysis (F) of PCNA expression (n = 6). Scale bar: 500 µm. G, H. Representative immunofluorescence co‐staining images (G) and corresponding semi‐quantitative analysis (H) of CD31 (red) and eNOS (green) (n = 6). Scale bar: 100 µm. I, J. Representative immunofluorescence co‐staining images (I) and corresponding semi‐quantitative analysis (J) of NF200 (red) and nNOS (green) (n = 6). Scale bar: 100 µm. K—M. Representative Western blot bands (K) and corresponding semi‐quantitative analysis (L, M) of eNOS and nNOS expression (n = 3). Significance: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

In terms of vascular repair, eNOS/CD31 double immunostaining demonstrated that AAV‐Slit2 restored a more continuous and well‐organized endothelial network, accompanied by a robust enhancement of eNOS signaling (Figure 6G,H), underscoring its capacity to promote both endothelial integrity and vascular functionality. Parallel improvements were observed in neural regeneration. nNOS/NF200 double staining showed that the typically sparse dorsal nerve terminals characteristic of DMED were markedly replenished following Slit2 treatment, with denser axonal projections and elevated nNOS expression (Figure 6I,J). These structural improvements were corroborated at the molecular level, as Western blotting confirmed significant increases in both eNOS and nNOS in the AAV‐Slit2 group (Figure 6K—M), indicating that Slit2 simultaneously enhances endothelial and neuronal NOS activity—two critical determinants of neurovascular function in the corpus cavernosum.

Collectively, these findings demonstrate that AAV‐Slit2 promotes a coordinated regenerative response by enhancing cellular proliferation, stimulating angiogenesis, and rebuilding neural networks, thereby reversing the neurovascular degeneration characteristic of DMED. Slit2 thus emerges as a pivotal regenerative cue capable of synchronizing vascular and neural repair, offering strong structural and functional support for its therapeutic potential in DMED.

3. Discussion

This study systematically reveals the pivotal regulatory role of the Slit2/Robo signaling axis in maintaining neurovascular homeostasis within the corpus cavernosum and, for the first time, establishes the Slit2‐Robo axis as a key regulator of cavernosal neurovascular integrity in diabetes mellitus‐induced erectile dysfunction (DMED). As a canonical pathway governing axon guidance and vascular remodeling, the Slit‐Robo network has long been recognized for its essential functions in neuronal navigation [38, 39] and in maintaining vascular stability and endothelial barrier integrity [28, 40]. By integrating single‐cell transcriptomic profiling across developmental and diabetic states, in vitro mechanistic studies, and in vivo gene therapy in a DMED model, we demonstrate that Slit2 acts as a potent bidirectional neurovascular regenerative cue capable of reactivating intrinsic repair programs and restoring precise intercellular interactions under conditions of metabolic and inflammatory stress. These findings provide a robust mechanistic foundation for targeting the Slit2‐Robo axis as a therapeutic strategy for DMED and highlight its broader potential as a master regulator of neurovascular regeneration.

From a developmental perspective, our study shows that fibroblast‐derived Slit2 [41, 42] serves as a key molecular determinant orchestrating both endothelial spatial organization and neuronal projection patterns during penile morphogenesis, consistent with its established roles in coordinating neurovascular co‐development in other organ systems [21, 43, 44, 45]. This bidirectional instructive capacity positions Slit2 at the core of neurovascular assembly. However, under chronic hyperglycemia and inflammatory stress, fibroblasts, endothelial cells, and peripheral neurons all exhibit marked disruptions in the Slit2‐Robo1/4 signaling axis, including ligand depletion, receptor downregulation, and decoupling of intercellular signal transmission. Similar suppression of Slit‐Robo signaling has been reported in other diabetic vascular complications [34, 36, 46], underscoring its vulnerability to metabolic stress. These observations suggest that DMED is not merely a consequence of local structural damage or metabolic toxicity, but reflects a fundamental breakdown of developmental neurovascular coupling programs in the disease state—offering a transformative biological perspective on the nature of DMED.

At the mechanistic level, our findings illustrate how Slit2 executes compartment‐specific regenerative functions through distinct receptors. Mechanistically, Slit2 appears to re‐engage a coordinated regenerative program that couples endothelial proliferation, angiogenic activation, and axonal growth. Rather than acting on a single cellular compartment, Slit2 synchronizes vascular and neuronal repair. Robo1/4, the predominant receptors in endothelial cells, have been shown to stabilize vascular barriers, suppress endothelial inflammation, and promote angiogenesis [40]. Consistently, we demonstrate that activation of the Slit2‐Robo1/4 axis restores endothelial proliferation, migration, and structural integrity under diabetic stress. Robo1, on the other hand, is a major mediator of neural repair, in line with its established roles in axon guidance and neuronal regeneration [47]. We show that Slit2 promotes neurite outgrowth, axonal branching, and growth cone dynamics through Robo1, thereby supporting robust neural regeneration. This receptor‐specific division of labor endows Slit2 with the unique capacity to simultaneously repair the two major systems disrupted in DMED—endothelium and peripheral nerves—setting it apart from traditional growth factors that target only one compartment (e.g., VEGF [48, 49], NGF [50], or BDNF [13, 51]).

Therapeutically, we demonstrate that local AAV‐mediated delivery of Slit2 drives integrated, cross‐system regeneration. While similar gene therapy approaches have shown encouraging results in erectile dysfunction models [52], most operate through single pathways; Slit2 provides a higher‐order, coordinated regulatory mechanism. Functionally, reactivation of Slit2 signaling appears to reinstate a neurovascular coupling program that is normally compromised under diabetic conditions. This re‐engagement of developmental signaling architecture enables synchronized coordination of vascular perfusion, neuronal maintenance, and tissue structural integrity, thereby providing a systems‐level basis for functional recovery of erectile physiology. Here, AAV‐Slit2 not only markedly improved erectile function in diabetic mice but also produced extensive structural and functional restoration of the corpus cavernosum, including robust endothelial and neural regeneration, enhanced neurovascular function, increased smooth muscle content, and significant attenuation of fibrosis.

Despite these promising findings, further work is needed to evaluate the long‐term safety of Slit2 gene therapy, particularly with regard to organ‐specific expression control. Notably, the intracellular signaling cascades downstream of Robo activation remain to be fully elucidated. In endothelial cells, the mechanisms by which Robo1/4 activation promotes proliferative and regenerative responses are not yet clearly defined; it is plausible that canonical pro‐angiogenic and pro‐survival pathways [53], such as PI3K‐Akt, MAPK, or Src signaling, may be involved, although their specific contributions in the context of Slit2‐mediated endothelial repair under diabetic conditions require further investigation. Similarly, Robo1 activation is known to regulate axonal guidance and growth cone dynamics, yet the downstream effectors mediating these processes in diabetic cavernous nerve regeneration remain unclear. Cytoskeletal remodeling pathways [54], including those involving Rho GTPases and actin‐regulatory proteins such as Ena/VASP or WASP family members, represent plausible mediators. In addition, although restoration of Slit2 signaling was associated with increased expression of eNOS and nNOS in vivo, the mechanistic linkage between Slit2‐Robo signaling and nitric oxide synthase activation remains undefined; it is conceivable that Slit2 may influence eNOS/nNOS activity through phosphorylation‐dependent mechanisms or indirect modulation of endothelial and neuronal homeostasis, but these hypotheses require further mechanistic studies. It will also be important to elucidate the upstream mechanisms responsible for suppression of Slit2/Robo signaling in the diabetic milieu, including potential contributions from inflammatory cytokines, AGEs, and metabolic oxidative stress [55, 56].

In addition, although HCMECs were selected as an experimentally stable and transfection‐compatible microvascular endothelial model for mechanistic studies, they do not fully recapitulate the specialized biological properties of cavernosal endothelial cells [57, 58, 59]. Compared with cavernosal sinusoidal endothelium, cardiac microvascular endothelial cells differ in multiple structural and functional aspects, including endothelial fenestration, constitutively elevated eNOS activity, hemodynamic mechanosensing characteristics, and caveolin‐1‐associated signaling organization [60, 61, 62]. These phenotypic distinctions may influence endothelial responses to diabetic stress and Slit2‐Robo signaling, thereby limiting the direct extrapolation of the in vitro findings to native cavernosal endothelial biology. Nevertheless, HCMECs retain key endothelial characteristics and endogenous Slit2‐Robo signaling activity, allowing investigation of the conserved endothelial regenerative mechanisms involved in this pathway. Importantly, the consistency between the in vitro findings and the in vivo phenotypes observed in diabetic mouse corpus cavernosum partially supports the biological relevance of the current model. Future studies employing primary cavernosal endothelial cells, multicellular neurovascular co‐culture systems, or cavernosal organoid platforms will be important to further validate the cell‐type specificity and translational applicability of the proposed mechanism.

From a translational perspective, while AAV‐mediated Slit2 delivery provides a robust proof‐of‐concept for restoring neurovascular function, alternative therapeutic modalities may offer more clinically feasible strategies. The structural properties of Slit2 offer opportunities to engineer more stable or potent functional fragments or derivatives as therapeutic candidates [25, 63]. In particular, recent studies have highlighted recombinant Slit2 protein as a promising pharmacological approach [64] with potentially improved controllability, dosing flexibility, and safety profiles compared to gene‐based interventions. Such protein‐based delivery systems may facilitate more precise temporal and spatial modulation of Slit2 signaling, thereby enhancing translational applicability in clinical settings. Moreover, integration with advanced delivery platforms—such as biomaterial‐based local release systems or minimally invasive administration strategies [65, 66]—may further improve tissue targeting and therapeutic efficacy.

Overall, this study provides compelling evidence that the Slit2/Robo signaling axis represents a promising multi‐target therapeutic pathway for DMED and, potentially, for broader categories of diabetic neurovascular injury. The findings carry significant implications for both basic biology and translational medicine.

4. Materials and Methods

4.1. Animal Models

All experimental procedures involving animals were approved by the Academic Administration Committee of Tongji Hospital, Tongji Medical College, Huazhong University of Science and Technology (Ethical Approval No. TJH‐202505010) and conducted in accordance with the National Institutes of Health Guidelines for the Care and Use of Laboratory Animals. Male C57BL/6 mice were purchased from Beijing HFK Bioscience Co., Ltd., housed under specific pathogen‐free (SPF) conditions, and provided food and water at a 12‐h light:dark cycle.

Mice were randomly assigned into four experimental groups (n = 10 per group): Con (non‐diabetic), DMED (diabetes mellitus‐induced erectile dysfunction), AAV‐Blank (diabetic mice receiving control AAV‐GFP), and AAV‐Slit2 (diabetic mice receiving AAV‐Slit2). Diabetes mellitus was induced by intraperitoneal injection of streptozotocin (STZ; Sigma‐Aldrich, USA) at a dose of 55 mg/kg body weight for five consecutive days. One week after the final injection, mice with fasting blood glucose levels exceeding 16.7 mmol/L on two consecutive days were considered diabetic.

One month after successful establishment of the STZ‐induced diabetic model, mice in the AAV‐Blank and AAV‐Slit2 groups received corresponding intracavernosal injections of AAV vectors. At the endpoint (2 months after AAV administration), six mice per group (n = 6) that met the inclusion criteria were randomly selected for functional and histological evaluations, including erectile function assessment (ICP/MAP), penile perfusion analysis, and histological examination. The sample size was determined based on previous studies and statistical power considerations, and also in accordance with animal welfare principles to minimize the number of animals subjected to invasive intracavernosal pressure measurements.

4.2. Cell Culture

The PC12 cell line, derived from a male rat pheochromocytoma, exhibits a pronounced neuronal phenotype. Highly differentiated PC12 cells (Procell system, Wuhan, CL‐0481) were obtained from Procell and cultured in a humidified incubator (37°C, 5% CO₂) using PC12‐specific medium (Pricella, CM‐0481). Human cardiac microvascular endothelial cells (HCMECs, Warner bio, Wuhan, WN‐10341) were purchased from Warner Bio and maintained in endothelial cell medium (ECM, ScienCell, Cat. #1001), which contains a basal glucose concentration of 5.5 mM. To model diabetic conditions in vitro, endothelial cells were cultured under either control or high‐glucose conditions. Specifically, control (Con) cells were maintained in ECM supplemented with 24.5 mM mannitol as an osmotic control, whereas high‐glucose (HG) conditions were established by supplementing ECM with 24.5 mM D‐glucose, resulting in a final glucose concentration of 30 mM.

4.3. Single Cell RNA‐sequencing Analysis

For normal postnatal development, mouse corpus cavernosum tissues were collected at P3, P9, P21‐28, and P42‐49, with four animals pooled per time point. Tissues were washed in ice‐cold PBS and dissociated using SeekMate Tissue Dissociation Kit A Pro or Kit C (SeekGene), with DNase I treatment applied as needed. After erythrocyte removal and viability assessment (AO/PI, Countstar Rigel S2 or SeekMate Tinitan), single‐cell suspensions were washed, resuspended at 1×10⁶ cells/ml in RPMI1640 with 2% FBS, and libraries were prepared using the SeekOne Digital Droplet Single Cell 3' kit (SeekGene): cells were encapsulated with barcoded hydrogel beads in droplets, reverse‐transcribed, and the resulting cDNA was amplified, adaptor‐ligated, and indexed for sequencing. Libraries were quality‐checked and sequenced on Illumina NovaSeq 6000. The embryonic genital tubercle and P6 penile single‐cell RNA‐sequencing data were obtained from publicly available datasets (GEO accession: GSE174712 [67], GSE175498 [68], GSE278744 [69], GSE290862 [70]). For diabetic modeling, corpus cavernosum tissues were collected at 0, 4, 8, 12, and 16 weeks, with three animals pooled per group. Single‐cell libraries were prepared using the BD Rhapsody Single‐Cell system following the manufacturer's standard protocol.

Data were processed and analyzed using the OmicVerse integration framework [71]. Briefly, datasets from different developmental stages or experimental groups were integrated using the scvi‐tools model to correct batch effects and generate a unified latent representation. Cells were clustered based on the scVI latent space, visualized using UMAP, and annotated using canonical marker genes combined with reference‐guided validation. Differentially expressed genes were identified across clusters or conditions using the built‐in statistical modules of OmicVerse. Cell‐cell communication analysis was performed with CellPhoneDB, applying its permutation‐based framework to infer ligand–receptor interactions and characterize intercellular signaling networks.

4.4. Measurement of Erectile Function

Erectile function was assessed in vivo by recording intracavernosal pressure (ICP) in response to cavernous nerve stimulation in anesthetized male mice. ICP was recorded using a pressure transducer (BIOPAC MP160, USA) inserted into the corpus cavernosum, while systemic blood pressure was monitored using a noninvasive tail‐cuff system (Visitech Systems, Apex, NC, USA). Electrical stimulation (15 Hz, 5.0 V, 60 s) was applied using a bipolar platinum electrode. The maximum ICP and the area under the curve (total ICP) were both normalized to mean blood pressure (MBP) and analyzed. After recordings, mice were euthanized, and corpus cavernosum tissues were collected for further study.

4.5. Masson's Trichrome Staining

Masson's trichrome staining was performed on paraffin‐embedded penile corpus cavernosum sections following established protocols. Smooth muscle fibers appeared red, whereas collagen deposits were stained blue. Quantitative analysis of the smooth muscle and collagen areas was performed with ImageJ software to calculate the smooth muscle‐to‐collagen ratio, which served as a measure of fibrotic remodeling within the corpus cavernosum.

4.6. Western Blotting

Tissues or cells were lysed in RIPA buffer supplemented with protease and phosphatase inhibitors. Protein concentrations were determined using a BCA assay, and equal amounts of protein were separated by SDS‐PAGE and transferred to PVDF membranes. Membranes were blocked with protein‐free rapid blocking buffer (Boster Biological Technology, AR0041) for 15 minutes at room temperature, followed by incubation with primary antibodies overnight at 4°C (Table S1). After washing, membranes were incubated with HRP‐conjugated secondary antibodies for 1 hour at room temperature. Protein bands were visualized using enhanced chemiluminescence (ECL) and quantified using ImageLab software. β‐Actin was used as a loading control.

4.7. Immunofluorescence

The penile tissues were carefully dissected and fixed in 4% paraformaldehyde overnight at 4°C, followed by overnight dehydration in 30% sucrose. Frozen sections (5 µm) were prepared and for antigen retrieval, sections were incubated in 0.01 M sodium citrate buffer in a microwave oven. Sections were permeabilized with 0.1% Triton X‐100 for 15 minutes and blocked with 4% bovine serum albumin (BSA) for 1 hour at room temperature. Primary antibodies targeting specific markers (Table S1) were applied overnight at 4°C. After washing, fluorescently labeled secondary antibodies were incubated for 1 hour at room temperature. Nuclei were counterstained with DAPI, and slides were mounted with antifade medium. Images were acquired using a fluorescence microscope (Zeiss Axio Imager M2), and fluorescence intensity was quantified using ImageJ software.

4.8. Lentivirus Construction, Package, and Transfection

Plasmids were generated either by restriction enzyme digestion followed by ligation or through Gibson assembly. Target genes were first amplified with specific primers and subsequently cloned into expression vectors such as pLenti‐EnCMV and pLKO.1 to obtain fusion constructs (detailed information in Appendix Table S2). The lentiviruses were specifically constructed to knock down Slit2 receptors Robo1 and Robo4.

For lentiviral production, HEK293T cells were co‐transfected with pLenti‐EnCMV‐based plasmids together with the packaging plasmids pMD2.G and psPAX2 using Lipofectamine 3000 (Invitrogen, L3000015). After 48 hours, the viral supernatant was harvested, filtered, and concentrated by ultracentrifugation.

Stable transduction of HCMECs was achieved by infecting the cells with lentivirus at a multiplicity of infection (MOI) of 10, followed by puromycin selection. For transient transfection, plasmids were delivered into cells using Lipofectamine 3000 according to the manufacturer's protocol.

4.9. Protein Purification

For eukaryotic protein expression, pcDNA3.1‐based plasmids encoding SP‐Slit2‐N‐Flag were transiently transfected into HEK293F cells using PEI reagent, following the manufacturer's instructions. After 72 hours of incubation in serum‐free medium, the supernatant was collected for purification. Flag‐tagged proteins were captured using Anti‐Flag Affinity Gel (Beyotime, P2271). The bound proteins were eluted with Flag peptides, concentrated using 10 kDa cutoff membranes, and stored at ‐80°C for future use.

4.10. AAV Construction, Package, and Delivery

The mouse sp‐Slit2‐N (1∼1113 aa) coding sequence was cloned into an AAV expression plasmid under the control of an enhanced cytomegalovirus (EnCMV) promoter to ensure robust and ubiquitous expression, while an EGFP‐only construct was used as a negative control. The construct was verified by Sanger sequencing prior to packaging.

AAV9 serotype was selected due to its high tropism for penile fibroblast. Recombinant viruses were produced by General Biol Co. Ltd (China). Briefly, high‐titer AAV9‐Slit2‐N and control AAV9‐GFP vectors were generated through triple transfection of HEK293T cells, followed by purification via iodixanol gradient ultracentrifugation and dialysis. Final viral titers were quantified using quantitative PCR.

For in vivo delivery, mice with successful DMED modeling were anesthetized, and AAVs (1×10¹³ vg/ml) were administered via intracavernous injection in a total volume of 2 µL. Control DMED animals received equal volumes of AAV9‐GFP.

4.11. Tube Formation Assay and Scratch Wound Assay

Tube formation assay was performed to evaluate the angiogenic potential of HCMECs. Briefly, 96‐well plates were pre‐coated with 50 µL of Matrigel (ABW, 082704) and allowed to polymerize at 37°C for 30 minutes. HCMECs were harvested and seeded onto the Matrigel‐coated wells at a density of 2 × 10⁴ cells/well in ECM supplemented with indicated experimental conditions. After 6 hours of incubation at 37°C in a humidified 5% CO₂ atmosphere, tubular structures were visualized and imaged using an inverted phase‐contrast microscope.

Scratch wound assays were performed to evaluate the migratory capacity of HCMEC and PC12 cells under different experimental conditions. HCMECs and PC12 cells were seeded separately into 6‐well plates at a density of 3 × 10⁵ cells per well and cultured in their respective complete growth media until they reached approximately 90–100% confluence. A sterile 200 µL pipette tip was used to generate a linear scratch across the cell monolayer. Detached cells were carefully removed by washing twice with phosphate‐buffered saline (PBS), and cells were then incubated in serum‐reduced medium (1% FBS). Where applicable, cells were treated with the indicated reagents or transfected accordingly prior to the scratch. Wound closure was monitored by capturing images at 0 h and 24 h using an inverted phase‐contrast microscope.

4.12. Aortic Ring Assay

Thoracic aortas were rapidly isolated from mice and immediately transferred into culture dishes containing ECM. Under a stereomicroscope, surrounding adipose tissue, connective tissue, and branching vessels were meticulously removed using fine microforceps and microscissors. The aortic lumens were gently flushed with ECM to remove any remaining blood. Cleaned vessels were then transversely sectioned into rings approximately 0.5 mm in thickness using a sterile scalpel. For each assay, 50 µL of chilled growth factor‐reduced Matrigel was dispensed into the center of each well of a 96‐well plate using pre‐cooled pipette tips. Aortic rings were carefully positioned upright in the center of the Matrigel droplets and incubated at 37°C with 5% CO₂ for 1 hour to ensure full polymerization. Subsequently, 150 µL of ECM supplemented with the designated treatments was gently added to each well without disturbing the gel. Cultures were maintained at 37°C with media replenished every 48 hours. On day 10, endothelial sprouting and microvessel outgrowths were assessed using an inverted phase‐contrast microscope. Quantitative analyses were performed on representative images using ImageJ software.

4.13. DRG Explant Culture and Sprouting Assay

The entire spinal column, spanning from the cervical to lumbar segments, was rapidly excised from mice. Under a stereomicroscope, bilateral dorsal root ganglia (DRGs) were carefully isolated segment by segment, with precise removal of meninges and peripheral nerve branches to preserve the structural integrity of each explant. For the neurite outgrowth assay, 200 µL of ice‐cold Matrigel was dispensed into 24‐well plates containing sterile glass coverslips (Biosharp, BS‐14‐RC) using pre‐cooled pipette tips. DRG explants were positioned upright in the center of the Matrigel and incubated at 37°C with 5% CO₂ for 1 hour to allow complete polymerization. Subsequently, 500 µL of PC12‐specific medium, with or without designated treatments, was gently added to each well to fully submerge the explants. Cultures were maintained at 37°C with media replenished every 48 hours. On day 7, neurite outgrowth was evaluated using an inverted phase‐contrast microscope. For further visualization, explants were fixed and subjected to immunofluorescence staining with anti‐β3 Tubulin antibody to label neurites. Fluorescence images were captured using a Zeiss fluorescence microscope, and quantitative analysis of neurite extension was performed using ImageJ.

4.14. Statistical Analysis

All results were expressed as mean ± standard deviation, and data analysis was performed using R v4.3.1 and GraphPad Prism v9.0. Normality was assessed via Shapiro‐Wilk test. Normally distributed data (mean ± SEM) were analyzed by unpaired two‐tailed Student's t‐test (two groups) or one‐way ANOVA with Tukey's post hoc (≥3 groups), while non‐normally distributed data (median [IQR]) were analyzed using Mann‐Whitney U or Kruskal‐Wallis/Dunn's tests. Significance: *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Author Contributions

Sen Fu, Peng Hu: conceptualization, methodology, investigation, data curation, writing – original draft. Zizhong Yang: investigation, formal analysis, visualization. Zhenghui Jin, Xiaoyu Zhu: validation, resources, software. Tao Wang, project administration. Jihong Liu and Kai Cui: conceptualization, supervision, funding acquisition, writing – review & editing. all authors discussed the results, contributed to the manuscript revision, and approved the final version for publication.

Conflicts of Interest

The authors declare no conflict of interest.

Supporting information

Supporting File 1: andr70288‐supp‐0001‐SuppMat.docx

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.