Hyaloid vessel pruning involves cell relocation and extravascular proliferation without increased apoptosis

While apoptosis has been implicated in pruning redundant vasculature, its temporal and functional relevance to hyaloid vessel involution remains incompletely defined [23, 24]. In mice, the postnatal regression of hyaloid vessels occurs concomitantly with the growth of the retinal vasculature (Fig. 1A). To investigate the cellular basis of hyaloid regression, we delineated the developmental trajectory and early regression of the hyaloid vasculature in C57BL/6J mice, spanning late embryogenesis (E18.5) to early postnatal stages (P8). Isolectin B4 (IB4) labeling of flatmounted hyaloids revealed a progressive decrease in hyaloid vascular density over time (Fig. 1B). Quantification revealed a progressive decline, with ~ 15% reduction by P0, ~ 25% by P4, and ~ 64% by P8 compared to E18.5. (Fig. 1C). In parallel, we observed a marked narrowing of vessels diameter (Fig. 1D) and a significant decrease in branching complexity relative to E18.5 as reflected by reduced vascular branch points (Fig. 1E). Although apoptosis has been reported in embryonic hyaloid vessels at mid-gestation, before the onset of postnatal regression, its contribution to postnatal involution remains unclear [9, 11]. We therefore quantified cleaved caspase-3-positive (C-Casp3+) cells across the regression timeline (Fig. 1F). 3D reconstructions at P0, P4 and P8 validated the presence of C-Casp3+ cells both along and outside the hyaloid vasculature throughout this period (Fig. 1G-I). Quantitative analyses showed that neither total apoptotic cells nor extravascular apoptotic cells increased between P0 and P8 (Fig. 1J, K). Given the lack of a measurable increase in apoptosis, we next asked whether proliferation accompanies postnatal hyaloid remodelling. We therefore quantified Ki67+ cells across the regression timeline (Fig. 1L). 3D reconstructions at P0, P4 and P8 revealed Ki67+ cells both along and outside the hyaloid vasculature throughout this period (Fig. 1L–O). Quantitative analysis showed that the number of proliferating cells per 100 μm of hyaloid vessel declined significantly from P0 to P4 and P8 (Fig. 1P). Interestingly, the proportion of extravascular proliferating cells increased over time and reached significance at P8 (Fig. 1Q).

Fig. 1

Cell relocation, apoptosis, and proliferation during postnatal hyaloid vessel pruning. A Schematic diagram of mouse hyaloid vessels (HV) regression and retinal vascular outgrowth at embryonic day 18.5 (E18.5) and postnatal days 0 (P0), 4 (P4), and 8 (P8). B Representative Isolectin B4 (IB4) and DAPI immunofluorescence staining of E18.5, P0, P4 and P8 HV flatmounts. C–E Quantification of HV density, vessel diameter and vascular branch points (n = 5–6). F Representative confocal micrographs of P0, P4 and P8 HV flatmounts labeled with cleaved caspase 3 (C-Casp3), IB4 and DAPI. G–I 3D reconstructions of HV at P0, P4, and P8 show the top and side views. Quantification of apoptotic cells (C-Casp3+) per 100 μm of HV length (J), and extravascular (extrav.) apoptotic cells among total extravascular cells K (n = 3). L Representative images of Ki67, IB4 and DAPI immunofluorescence staining of P0, P4 and P8 HV flatmounts show proliferating cells within (white arrowheads) and outside (yellow arrowheads) the vascular compartment. M–O 3D reconstructions show proliferating cells in vascular and extravascular regions of HV at P0, P4 and P8. Quantification of proliferating cells (Ki67+) per 100 μm of HV length (P), and extravascular proliferating cells among total extravascular cells (Q) (n = 3). Results are expressed as fold change (FC) relative to E18.5 (C–E) and as percentages (%) normalize to P0 (J, K, P, Q). Data are means ± SEM. Represented p values are *≤ 0.05, **≤ 0.01, ***≤ 0.001, ****≤0.0001 from ordinary one-way ANOVA test with Dunnett’s multiple comparison. Non-significant (ns). Scale bars, 500 μm (B, F, L) and 50 μm [for higher magnification images in (F–I, L–O)]

These data indicate that postnatal hyaloid vessel regression is not accompanied by an increase in apoptosis. To assess the specificity of Ki67 and C-Casp3 antibodies in the immunofluorescence staining, P0, P8 control samples processed in parallel with omission of the primary antibody were included. The absence of detectable signal in these negative controls supported staining specificity (Fig. S1A).

Taken together, these findings show that postnatal hyaloid vessel regression in mice occurs without a measurable increase in apoptosis. Instead, regression is characterized by reduced vascular cell proliferation and a spatial shift in proliferative activity to the extravascular compartment. These observations support a model in which hyaloid vessel involution is coupled to dynamic extravascular remodelling, rather than being driven primarily by apoptotic clearance as previously proposed.

Cell-type-specific extravascular proliferation accompanies hyaloid vessel regression

Because postnatal hyaloid vessel involution occurred without increased apoptosis, but with a shift in proliferative activity from the intravascular to the extravascular compartment (Fig. 1L–Q), we next analysed the dynamics of ECs, MCs and HCs to determine how each population contributes to vascular regression.

To assess whether extravascular cells arise from the endothelial lineage during postnatal hyaloid regression, we performed tamoxifen-inducible endothelial lineage tracing in Rosa26-mT/mG; Cdh5-CreERT2 mice. In the absence of tamoxifen, hyaloid vessels remained tdTomato+ with virtually no GFP signal, indicating negligible background recombination. Following tamoxifen administration, GFP+ cells were detected not only within the vessel wall but also in the extravascular compartment (Fig. 2A).

Fig. 2

Proliferative endothelial and mural cells delamination accompanies hyaloid vessels regression. A Representative confocal micrographs of P8 HV flatmounts from Rosa-mT/mG; Cdh5-CreERT2 mice in the absence (-Tam) or following tamoxifen induction (+ Tam), showing tdTomato (tdT), GFP, and DAPI labeling. B 3D reconstructions of HV illustrating tdT⁺, GFP⁺, and tdT⁺/GFP⁺ and their extravascular localization. Arrowheads indicate tdT⁺ cells (red), GFP⁺ cells (green), and tdT⁺/GFP⁺ cells (yellow). C Quantification of extravascular tdT⁺, GFP⁺, and tdT⁺/GFP⁺ cells, expressed as a percentage of total extravascular cells in -Tam and + Tam conditions (n = 3). D Representative confocal images of P0 and P8 HV flatmounts showing EdU incorporation in ECs (ERG1/2/3⁺). Nuclei are labeled with DAPI. E 3D reconstructions showing proliferating ECs (white circle) and non-proliferating ECs (green circle) in vascular and extravascular regions of HV at P0 and P8. F Quantification of EdU⁺ ECs relative to total number of extravascular ERG1/2/3⁺ cells in HV (n = 3). G Representative confocal images of P0 and P8 HV flatmounts showing EdU incorporation in mural cells (MCs; Desmin⁺). Nuclei are labeled with DAPI. H 3D reconstruction of HV at P0 and P8 showing proliferating MCs (white circle) and non-proliferating MCs (green circle). I Quantification of extravascular proliferating MCs among total extravascular MCs (n = 3–4). Results are expressed as percentages (%) from total cells (C) or as % normalized to P0 stage (F, I). Data are shown as means ± SEM. Represented p-values are ***≤0.001 from two-tailed parametric unpaired t-test. Scale bars, 50 μm

Extravascular cells included GFP+ cells, tdTomato+ cells and tdTomato+GFP+ cells. The tdTomato+GFP+ population is consistent with ongoing reporter conversion in endothelial-lineage cells, although we cannot exclude that some cells represent non-recombined hyalocytes/macrophages that have engulfed GFP+ endothelial material [9, 10]. 3D reconstruction confirmed the extravascular localization of GFP+ and tdTomato+GFP+ cells (Fig. 2B). Quantification showed that extravascular cells in control samples were almost exclusively tdTomato+, whereas tamoxifen-treated samples contained a marked increase in GFP+ and tdTomato+GFP+ extravascular cells, with only a small residual tdTomato-only population (Fig. 2C). Together, these results identify endothelial-lineage-labelled cells within the extravascular population that accumulates during hyaloid vessel involution, while also suggesting that this compartment remains heterogeneous and may include non-recombined cells and phagocytosed vascular material.

To independently validate the identity and localization of these extravascular endothelial-derived cells, we next combined endothelial and basement membrane markers. CD31 and Collagen IV co-staining identified Collagen IV+CD31− empty vascular sleeves, consistent with regressing vessels, and further revealed expansion of the extravascular compartment at P8 relative to P0 (Fig. S1B). Notably, CD31+ ECs were frequently detected outside these sleeves at P8, supporting their displacement from the regressing vessel wall. Quantification further showed that the number of empty sleeves increased significantly at P8 relative to P0 (Fig. S1C).

Having established that endothelial-lineage-labelled cells accumulate outside the regressing vessel wall, we next asked whether proliferating ECs are redistributed to the extravascular compartment during postnatal hyaloid involution. EdU incorporation combined with ERG immunostaining identified proliferating ECs both within and outside the vasculature and revealed a marked increase in extravascular EdU+ERG+ cells at P8 relative to P0, when EdU+ERG+ cells were largely confined to the vessel wall (Fig. 2D–F). We independently confirmed this result using a second set of complementary EC and proliferation markers (Fig. S1D, E). ERG marks EC nuclei, whereas CD31 labels the EC membrane, allowing more precise positional assignment of ECs relative to the vessel wall (Fig. S1D). In parallel, EdU labels S-phase cells, whereas Ki67 identifies cycling cells more broadly (Fig. S1E). Although these markers are not expected to fully colocalize, their complementary distributions provide independent support for the same conclusion and validate the increase in extravascular proliferating ECs at P8 compared to P0. Consistent with the results obtained using EdU and ERG (Fig. 2D–F), Ki-67 and CD31 co-immunostaining revealed increased displacement of proliferating ECs to the extravascular compartment at P8 compared to P0 (Fig. S2A–C). Together, these data indicate that hyaloid regression is accompanied not only by EC displacement, but also by proliferative EC redistribution beyond the vascular scaffold.

We next asked whether this remodelling program is restricted to ECs or also involves MCs. Desmin immunostaining combined with EdU incorporation identified proliferating MCs both along the vascular scaffold and outside the vessel wall (Fig. 2G). 3D reconstruction validated the presence of extravascular EdU+Desmin+ cells at P8 (Fig. 2H). Quantification revealed a significant increase in extravascular proliferating MCs at P8 relative to P0 (Fig. 2I). To exclude marker-dependent bias, we repeated the analysis using Ki67 co-immunostaining with Desmin and a second MCs marker, CD13. Desmin and CD13 yielded the same overall pattern, identifying proliferating MCs both along the vessel wall and within the extravascular compartment (Fig. S2D–I). 3D and quantification likewise showed a significant increase in extravascular Ki67+Desmin+ (Fig. S2E, F) and Ki67+CD13+ cells at P8 compared with P0 (Fig. S2H, I). These findings indicate that MCs, like ECs, participate in the extravascular proliferative remodelling that accompanies vessel involution.

We next examined whether HCs also contribute to the expanding extravascular proliferative compartment. F4/80 and Ki67 co-immunostaining identified proliferating HCs outside the vascular wall at both P0 and P8, and 3D reconstruction validated the extravascular localization of Ki67+F4/80+ cells (Fig. S2J, K). However, in contrast to ECs and MCs, the number of proliferating HCs remained unchanged over this period (Fig. S2L). Thus, unlike ECs and MCs, HCs are already extravascular at early postnatal stages and do not show evidence of progressive redistribution during hyaloid regression.

Together, these data define a cell-type-specific proliferative remodelling program during postnatal hyaloid vessel involution. ECs and MCs undergo extravascular proliferative shift but HCs remain largely unchanged. These findings demonstrate that ECs and MCs contribute to the expanding extravascular proliferative compartment during hyaloid vessel regression.

Hyaloid vessel regression proceeds through multicellular remodelling rather than increased apoptosis

To determine whether hyaloid vessel involution is driven by apoptosis of specific vascular cell types, we profiled dissociated hyaloid vessels by flow cytometry using CD31, Annexin V and 7-AAD (Fig. 3A). The total number of recovered 7-AAD−CD31+ ECs did not significantly change from P0 to P8, and CD31 median fluorescence intensity within this population was also unchanged (Fig. 3B, C). Among recovered 7-AAD−CD31+ ECs, the proportion of Annexin V+ apoptotic cells remained stable over this period (Fig. 3D), indicating that hyaloid regression is not accompanied by a detectable increase in apoptosis within this population. Similarly, apoptotic cells within the 7-AAD−CD31− compartment, which includes MCs and HCs, did not vary detectably across this period (Fig. 3A, D).

Fig. 3

Phagocytic uptake of ECs and MCs by hyalocytes and steady state apoptosis during hyaloid regression. A Flow cytometry (FACS) analysis of ECs and non-ECs apoptosis from HV at P0 and P8. Representative FACS plots show the gating strategy followed. Numerals shown in the plots indicate the percentages of the indicated cell population within each plot. B Quantification of total ECs (7-AAD−PE-CD31⁺) and non-ECs (7-AAD− PE-CD31⁻), C CD31 median fluorescence intensity (MFI) and D apoptotic (FITC-AnnexinV+) cells within each population at P0 and P8 (n = 3). E Representative confocal micrographs and F 3D reconstructions of ERG1/2/3, F4/80, and DAPI immunofluorescence in HV at P0 and P8 reveal two EC populations: ERG1/2/3⁺ cells positive for the hyalocyte (HC) marker F4/80 (white circles; ERG1/2/3⁺F4/80⁺) and ERG1/2/3+cells negative for F4/80 (green circles; ERG1/2/3⁺F4/80⁻). G Quantification of extravascular ERG1/2/3+ cells among total extravascular cells, and H extravascular ERG1/2/3+F4/80+ cells, expressed relative to the total number of extravascular ERG1/2/3+ cells (n = 3). I Representative confocal micrographs and J 3D reconstructions of P0 and P8 HV flatmounts labeled for CD13, IBA1, and DAPI reveal two MCs (CD13⁺) populations: MCs co-expressing the HC marker IBA1 (white circles; CD13⁺IBA1⁺) and MCs lacking IBA1 expression (green circles; CD13⁺IBA1⁻). K Quantification of extravascular CD13⁺ cells among total extravascular cells, and L extravascular CD13⁺IBA1⁺ cells, expressed relative to the total number of extravascular CD13⁺ cells (n = 3). Results are expressed as percentages (%) (B, D) and normalized to P0 stage (G, H, K, L) or as arbitrary units (a.u.) (C). Data are shown as means ± SEM. Represented p-values are **≤0.01 from two-tailed parametric unpaired t-test. Non-significant (ns). Scale bars, 50 μm

To directly assess apoptosis in defined vascular cell types in situ, we performed hyaloid flatmount co-immunostaining for CD31 and C-Casp3 to identify apoptotic ECs (Fig. S3A). Three-dimensional reconstruction and quantification showed a decrease in C-Casp3+ ECs in the extravascular compartment (Fig. S3B, C). Consistent with the flow cytometry analysis, whole-mount CD31/C-Casp3 co-immunostaining showed that the total number of C-Casp3+ ECs did not change significantly between P0 and P8 (Fig. S3D). In contrast, immunostaining of the hyaloid flatmount revealed a significant decrease in total CD31+ cells over the same period (Fig. S3E), in agreement with western blot analysis of whole isolated hyaloid vessel lysates showing reduced CD31 protein abundance from P0 to P8 (Fig. S3F). Together, these findings indicate that hyaloid regression is accompanied by a tissue-level reduction in CD31+ cells and CD31 protein abundance, without a detectable decrease in CD31 expression or increase in apoptosis within the recovered CD31+ EC population by flow cytometry.

We next performed co-immunostaining for CD13 and C-Casp3 to identify apoptotic MCs (Fig. S3G). Three-dimensional reconstructions and quantification likewise confirmed the presence of MCs in the extravascular compartment (Fig. S3H, I) and the frequency of C-Casp3+ MCs remained unchanged between P0 and P8 (Fig. S3J). Thus, during hyaloid regression, ECs and MCs accumulate in the extravascular compartment without a detectable increase of apoptosis; overall apoptotic extravascular ECs and MCs remain low at the analysed stages.

We next asked whether these displaced vascular cells are subsequently cleared by HCs. To distinguish extravascular ECs and MCs from HCs and quantify their spatial association, we performed whole-mount immunostaining using ERG or CD13 together with the HC markers F4/80 (Fig. 3E–H) or IBA1 (Fig. 3I–L) respectively. Consistent with our earlier results, the fraction of extravascular ERG+ cells increased significantly by P8 (Fig. 3G). However, the proportion of extravascular ERG+ cells associated with F4/80+ HCs did not change (Fig. 3H). Similarly, extravascular CD13+ cells accumulated during involution (Fig. 3K), whereas their association with IBA1+ HCs remained unchanged (Fig. 3L). Together, these findings indicate that postnatal hyaloid regression is not accompanied by a detectable increase in apoptosis or HC-associated clearance. Instead, it is characterized by the progressive displacement of ECs and MCs to the extravascular compartment during the regression process.

Notch1 promotes an EndoMT-like transcriptional program during hyaloid regression

The accumulation of proliferative ECs and MCs outside the regressing vessel wall at later stages of hyaloid involution suggested that vascular regression may involve not only cell displacement but also changes in endothelial cell state. We therefore asked whether this remodeling process is associated with an endothelial-to-mesenchymal transition (EndoMT)-like program.

This transition could facilitate EC displacement by coupling loss of endothelial identity to acquisition of mesenchymal features. qRT-PCR analysis of whole hyaloid lysates revealed marked downregulation of the endothelial markers Vegfr2, Pecam1, Vwf and Tie2 between P0 and P8 (Fig. 4A). This was accompanied by induction of the mesenchymal regulators Snail1, Slug and Acta2/α-Sma, whereas Twist1 remained unchanged (Fig. 4B).

Fig. 4

Notch1 drives endothelial identity loss and plasticity-associated gene expression during hyaloid regression. A Quantitative reverse transcription polymerase chain reaction (qRT-PCR) shows downregulation of endothelial cell (EC) identity markers Vegfr2, Pecam1, Vwf and Tie-2 during hyaloid regression. B qRT-PCR analysis shows induction of endothelial mesenchymal transition (EndoMT)-associated transcription factors: Snail1, Slug and Acta2 in hyaloid from P0 to P8. Twist expression remains unchanged. C Western blot analysis of hyaloid cell lysates at P0, P4 and P8 showing expression of Notch1, its active form (Notch1 intracellular domain, N1ICD), the ligands DLL4 and JAG-1 and the downstream target gene HES1 (n = 2–3). Densitometric quantification of each protein is shown relative to the loading control, β-ACTIN. D qRT-PCR analysis shows the levels of Notch1, E Dll4, F Jag1 and G Hes1 transcripts in HV at P0, P4, and P8. H Schematic diagram of tamoxifen-induced Notch1 deletion in ECs in Notch1fl/fl; Cdh5- CreERT2 C57BL/6 transgenic mice. I Validation of Notch1 deletion by qRT-PCR in Notch1cKO versus Notch1cWT P8 HV. J Western blot analysis of P8 hyaloid cell lysates from Notch1cWT and Notch1cKO shows the protein expression of Notch1 and VEGFR2 (n = 2–3). Densitometric quantification of each protein is shown relative to the loading control, β-ACTIN. K qRT-PCR shows upregulation of EC identity markers Vegfr2, Pecam1, Vwf and Tie2 in Notch1cKO versus Notch1cWT P8 HV. L qRT-PCR analysis shows reduction of EndoMT-associated transcription factors Snail1, Slug, Acta2 and Twist in Notch1cKO P8 HV versus Notch1cWT. β-actin was used as a reference gene. Results are presented as fold change (FC) normalize to P0 stage (A, B, D–G) or P8 Notch1cWT (I, K, L). Data are shown as means ± SEM. Represented p values are *≤ 0.05, **≤ 0.01, ***≤ 0.001, ****≤0.0001 from ordinary one-way ANOVA test with Dunnett’s multiple comparison (A, B, D, E, G), Kruskal-Wallis test with Dunn’s multiple comparisons (F) or two-tailed parametric unpaired t-test (I, K, L). n = 3–5 for Vegfr2 and Tie2; n = 3 for Pecam1 and Twist; n = 3–4 for Vwf, Snail1 and Slug; n = 4 for Acta2 (A, B); n = 3–4 for Notch1 D and Hes1 (G); n = 3–6 for Dll4 (E); n = 4–6 Jag1 (F); n = 3 for Notch1 (I); n = 3 for Vegfr2 and Pecam1; n = 4–6 for Vwf; n = 4–5 for Tie2; n = 3–4 for Snail1; n = 3 for Slug and Twist1; n = 4–6 for Acta2 (K, L). Non-significant (ns)

Although these measurements were obtained from whole hyaloid lysates and therefore do not assign these transcriptional changes to a specific cell type, the overall expression profile is consistent with selective engagement of an EndoMT-like remodeling program during hyaloid regression. The induction of Snail1 and Slug in the absence of Twist1 upregulation further supports a plastic, early transition state rather than a fully stabilized mesenchymal program [25,26,27,28].

Prompted by these observations, we next investigated upstream pathways that might regulate endothelial remodeling during hyaloid regression. Notch1 was of particular interest given its established roles in vascular stabilization and EndoMT [29,30,31] However, whether Notch1 contributes to hyaloid vessel regression has not, to our knowledge, been previously addressed. We therefore examined the temporal expression of Notch1, its ligands and canonical downstream targets across the regression window from P0 to P8. Western blot analysis of isolated hyaloid vessels revealed a progressive increase in total NOTCH1 protein from P0 to P8, with a peak at P4. The cleaved active form, N1ICD, was likewise maximal at P4 (Fig. 4C). DLL4 and JAG1 displayed similar temporal dynamics at the protein level, with transient upregulation at P4 before returning towards baseline by P8. HES1 was also elevated at P4, indicating a transient increase in Notch signalling activity at this stage (Fig. 4C). Consistent with these protein changes, qRT-PCR analysis showed a transient increase in Notch1 mRNA at P4 that declined to baseline by P8 (Fig. 4D). Dll4 transcripts also decreased by P8, whereas Jag1 expression was significantly increased at this stage, indicating divergent ligand regulation during hyaloid regression (Fig. 4E, F). Transcriptional profiling of canonical Notch1 targets further revealed upregulation of Hes1 at P4 (Fig. 4G), consistent with ongoing Notch1 pathway activity during involution.

To study the role of Notch1 in EndoMT and regression we used a tamoxifen-inducible, EC-specific Notch1 knockout mouse model (Notch1fl/fl; Cdh5-CreERT2) identified as Notch1cKO. This mouse model allows for temporally controlled deletion of Notch1 specifically in vascular ECs upon administration of tamoxifen. To target the period of active hyaloid vessels remodeling, tamoxifen was administered via intragastric injection to neonatal pups from P1 to P3, and hyaloid vessels were collected at P8 as demonstrated in the model (Fig. 4H). Efficient deletion was confirmed by the near-complete loss of Notch1 mRNA in Notch1cKO hyaloids relative to tamoxifen-treated Cre-negative littermate controls (Notch1cWT) (Fig. 4I), accompanied by NOTCH1 protein levels at P8 (Fig. 4J). To assess whether the loss of endothelial identity was similarly evident at the protein level, we next analysed VEGFR2. Although Vegfr2 mRNA was significantly increased in Notch1cKO hyaloids, VEGFR2 protein was not detectably upregulated at P8 (Fig. 4J, K), indicating that the effect of endothelial Notch1 deletion on endothelial marker VEGFR2 recovery is more evident at the transcriptional than the protein level at this stage. This model provides a robust platform to dissect the endothelial-specific function of Notch1 in postnatal vascular regression without confounding developmental lethality.

Consistent with this, mRNA profiling of P8 hyaloid lysates showed that endothelial Notch1 deletion increased the expression of the endothelial markers Vegfr2, Pecam1, Vwf and Tie2 in Notch1cKO hyaloids relative to Notch1cWT controls (Fig. 4K). In parallel, the mesenchymal regulators Snail1, Slug, Acta2 and Twist1 were significantly reduced (Fig. 4L). Together, these data are consistent with a role for endothelial Notch1 in promoting the EndoMT-like transcriptional changes associated with physiological hyaloid regression, including the attenuation of endothelial identity marker expression during involution.

Endothelial Notch1 is required for hyaloid regression and postnatal retinal vascular development

Having established that endothelial Notch1 promotes an EndoMT-like transcriptional program during hyaloid regression (Fig. 4), we next asked whether Notch1 is functionally required for hyaloid vessel involution in vivo. In striking contrast to the regressed hyaloid network observed in P8 Notch1cWT controls, IB4 staining of flatmounted hyaloids revealed marked vascular persistence in Notch1cKO littermates (Fig. 5A). Quantitative morphometric analysis confirmed a significant increase in vessel density, branching complexity and vessel diameter in Notch1cKO hyaloids relative to controls (Fig. 5B–D). These findings identify endothelial Notch1 as a key driver of physiological hyaloid vessel regression.

Fig. 5

Endothelial-specific Notch1 deletion in vivo disrupts hyaloid vascular regression and retards retinal vascular superficial layer formation A Representative confocal micrographs of P8 HV from Notch1cKO and Notch1cWT mice immunostained with IB4 and DAPI. B Quantification of hyaloid vascular density, C vessels diameter, and D hyaloid vascular branch points of confocal microscopy data in (A) (n = 6). E Representative IB4 staining and F quantification of retinal vascular area, and G vascular branch points in P8 flatmounted retinal vasculature in Notch1cKO versus Notch1cWT (n = 5). H, I 3D reconstructions of P14 and P21 retinal vasculature labeled with IB4 show three vascular layers [first superficial layer (yellow), second deep layer (blue) and third intermediate layer (green)] in Notch1cKO versus Notch1cWT (n = 3). Results are presented as fold change (FC) relative to Notch1cWT control HV. Data are shown as means ± SEM. Represented p-values are **≤ 0.01, ****≤0.0001 from two-tailed parametric unpaired t-test (B, C, F, G) or Mann-Whitney test (D). Scale bar, 500 μm (A, E) and 200 μm [for higher magnification images in (A)] and 100 μm [for higher magnification images in (E)], and 50 μm (H, I),

Because endothelial Notch1 is already known to control postnatal retinal vascular development, we analyzed the retinal vasculature in parallel with the hyaloid phenotype [32,33,34]. IB4 staining of flatmounted P8 retinas revealed a marked delay in formation of the superficial vascular layer in Notch1cKO mice relative to Notch1cWT controls (Fig. 5E). Quantitative morphometric analysis confirmed a marked reduction in retinal vascular outgrowth in Notch1cKO retinas, with an approximately 70% decrease in vascular area and a 50% decrease in branching relative to controls (Fig. 5F, G). Consistent with this early defect, 3D reconstruction of IB4-stained retinas showed that Notch1cKO mice failed to fully establish the deep vascular plexus at P14 and P21, in contrast to Notch1cWT littermates (Fig. 5H, I). Together, these data validate the requirement for endothelial Notch1 in postnatal retinal vascular development and identify a previously unrecognized role for Notch1 in fetal hyaloid clearance, with endothelial Notch1 loss leading to persistent fetal vasculature.

Loss of endothelial Notch1 impairs endothelial and mural cell redistribution during hyaloid regression

Having established a requirement for endothelial Notch1 in hyaloid vessel regression, we next sought the cellular basis of the persistence phenotype in Notch1cKO mice. To determine the identity and localization of proliferating cells, we analysed ERG and EdU whole mounts labeling by 3D reconstruction. Notch1cWT hyaloids displayed a clear population of extravascular ERG+ ECs, whereas in Notch1cKO hyaloids ERG+ cells remained largely restricted to the vascular scaffold (Fig. 6A, B). Quantification confirmed a near-complete loss of extravascular ERG+ cells in Notch1cKO hyaloids (Fig. 6C). In parallel, the proportion of total proliferating ECs (EdU+ERG+) was significantly increased in Notch1cKO hyaloids relative to controls (Fig. 6D). Together, these findings indicate that endothelial Notch1 deletion prevents EC displacement from the regressing vessel wall while retaining proliferating ECs within the persistent vasculature.

Fig. 6

Loss of endothelial Notch1 alters hyaloid endothelial and mural cells positioning and proliferation. A Representative confocal micrographs of P8 HV flatmounts from Notch1cWT and Notch1cKO mice showing EdU incorporation in ERG1/2/3⁺ cells. Nuclei are labeled with DAPI. B 3D reconstructions show ERG1/2/3+EdU⁺ cells (white circle) and ERG1/2/3+EdU− cells (green circle) in P8 HV from Notch1cWT and Notch1cKO mice (n = 3). C Quantification of extravascular ERG1/2/3+ among total extravascular cells and D ERG1/2/3+EdU+cells, expressed relative to total ERG1/2/3+ in P8 HV in Notch1cKO versus Notch1cWT (n = 3). E Representative confocal micrographs and F 3D reconstructions of Desmin, Ki67 and DAPI immunofluorescence in P8 from Notch1cWT and Notch1cKO mice, showing distinction between proliferating MCs (Desmin+Ki67+) (white circle) and non-proliferating MCs (Desmin+Ki67−) (green circle). G Quantification of extravascular Desmin+ among total extravascular cells and H Desmin+Ki67+, expressed relative to total Desmin+ of confocal microscopy data in (E) (n = 3). I Representative confocal micrographs and J 3D reconstructions of F4/80, Ki67 and DAPI immunofluorescence in P8 from Notch1cWT and Notch1cKO mice, showing distinction between proliferating HCs (F4/80+Ki67+) (white circle) and non-proliferating HCs (F4/80+Ki67−) (green circle). K Quantification of extravascular F4/80+Ki67+ of confocal microscopy data in (I) (n = 3). Results are expressed as percentages normalized to Notch1cWT control HV. Data are shown as means ± SEM. Represented p-values are *≤0.05, **≤0.01 and ***≤ 0.001 from two-tailed parametric unpaired T-test. Scale bars, 50 μm

We then examined whether this defect extends to MCs. Desmin and Ki67 co-staining showed that, in Notch1cWT hyaloids, proliferating MCs were distributed both along the vessel wall and within the extravascular compartment. By contrast, in Notch1cKO hyaloids, most Ki67+Desmin+ MCs remained closely associated with the vascular scaffold (Fig. 6E). 3D reconstruction and quantification revealed a significant increase in total MC number together with a reduction in extravascular MC redistribution in the absence of Notch1 (Fig. 6G, H). These data indicate that endothelial Notch1 deletion disrupts not only EC displacement, but also the normal redistribution of MCs during hyaloid regression.

Co-immunostaining for F4/80 and Ki67 revealed increased HC proliferation in Notch1cKO hyaloids compared with Notch1cWT controls at P8, as confirmed by 3D reconstruction and quantification of F4/80+Ki67+ cells (Fig. 6I–K). Notably, these HCs remained extravascular but closely apposed to the persistent vessels, resembling the early postnatal configuration observed before the onset of regression. Together, these data suggest that, in the absence of endothelial Notch1, the hyaloid vasculature fails to progress into the regression program and instead retains immature pre-regression features.

We further asked whether the persistent hyaloid vasculature in Notch1cKO mice results from altered apoptosis. Cleaved caspase-3 staining revealed comparable numbers of apoptotic cells in Notch1cKO and Notch1cWT hyaloids at P8 (Fig. S4A, B), excluding defective apoptotic cell loss as the primary cause of persistence. Instead, Collagen IV and CD31 co-staining showed that, whereas control hyaloids contained Collagen IV+CD31− empty sleeves and extravascular CD31+ ECs, Notch1cKO hyaloids retained CD31+ ECs within Collagen IV+ vascular tracks and formed fewer empty sleeves (Fig. S4C, D), consistent with impaired vascular disengagement.

Together, these findings show that endothelial Notch1 is required for the remodelling events that drive hyaloid regression. In its absence, ECs and MCs fail to disengage from the vessel wall, collagen sleeve formation is reduced without a change in apoptosis, and the hyaloid network persists in an immature, pre-regression configuration.

Notch1 and Wnt pathway mutants converge on defective endothelial and mural cell delamination during hyaloid regression

Given the established role of Wnt signaling in hyaloid vessel regression [13, 14, 35, 36], we next asked whether endothelial Notch1 loss alters the expression of core components of this pathway. As essential co-receptors for canonical Wnt signaling, LRP5 and LRP6 are key determinants of cellular responsiveness to Wnt ligands. qRT-PCR analysis revealed a ma