Dopamine signaling from ganglion cells directs layer-specific angiogenesis in the retina

Pregnant CD1 dams were obtained from Charles River Laboratory, and litters were administered exogenous agents through an intraperitoneal route starting at P0.5. To elevate dopamine, the analogous precursor drug L-3,4-dihydroxyphenylalanine-methyl ester (LDOPA-ME) was administered once a day, in the morning, at a dose of 100 mg drug per kilogram weight of animal 87 . Drug-treated animals were injected with L-DOPA-ME dissolved in 2.5 mg/mL Ascorbic Acid/phosphate buffered saline (PBS) solution, while control animals were treated with Ascorbic Acid/PBS alone. The pAAV-DIO-TH(Human)-Myc construct was generated by Dr. Joshua Ortiz, director of the Baylor College of Medicine Optogenetics and Viral Design Core. The same Th sequence was excised (Origene MG227681 ) where Th overexpression is driven by the DIO promoter and targeted cells were marked with a Myc tag. The virus was injected at a final concentration of 2.26 E12 GC/mL,with 2–3 uL volume administered per eye. Intravitreal delivery of pAAV-DIO-TH(Human)-Myc into C57Bl6J;Chrnb cre+/− pup eyes were performed at P0.5, and eyes were harvested at P3 and P9. The DJ serotype allowed for early expression of Th in Cre -expressing retinal ganglion cells (driven by Chrnb3 cre+/ ). Cre negative littermates injected with AAV-Dio-TH-Myc were used as controls. Immunofluorescence antibody staining was done as previously described 58 , 90 . Briefly, eyes were harvested from P3 and P9 animals and subsequently fixed in 4% PFA (paraformaldehyde). For P3 timepoint, eye enucleation was followed by “light fixation” where eyes were placed in 4% PFA for 10-minutes. Eyes were then placed in PBS, and the hyaloid vessels were dissected and removed as these vessels impinge on the imaging of the superficial vascular layer. Eyes were then placed back in 4% PFA for an additional 45-minute fixation. All immunohistochemistry was performed through previously described cross-section and whole-mount preparation protocols 58 . To stain whole-mount dissected retinas, whole retinas were submerged in Eppendorf tubes containing 300μL blocking buffer (0.5% TritonX-100 and 10% Normal Donkey Serum in PBS solution) for 1-hour at room temperature. Subsequently, solutions were replaced with primary antibody diluted in blocking buffer. Retinas were incubated in primary antibody solution for 3-days at 4°C and then washed three times with 1-mL of PBS at room temperature before placing in secondary antibody solution (also diluted in blocking buffer). Retinas were incubated in secondary antibody solution for 1-day at 4°C, washed three times with 1-mL PBS, mounted on to glass slides, overlayed with ProLong™ Gold Antifade mounting media (Invitrogen # P36930 ), and then cover slipped. To prepare retinas for cross-section analysis, dissected retinas were cryopreserved in 30% Sucrose/PBS solution and then embedded in Tissue-Tek ® OCT compound (VWR) and gradually frozen in methyl butane. Frozen tissue blocks were sectioned at 20μm, and sections and mounted directly onto charged glass slides. To stain retinal sections, tissue was first washed with PBS and then blocked with blocking buffer (0.3% TritonX-100 and 3% Normal Donkey Serum in PBS solution). Primary antibodies were diluted in blocking buffer and then overlayed onto the tissue sections. Slides at were incubated overnight at 4°C and then washed three-times in glass chambers filled with PBS. Secondary staining was performed at room temperature for 1-hour, and slides were washed three-times with PBS in glass chambers. Sections were overlayed with ProLong™ Gold Antifade mounting media (Invitrogen # P36930 ) and then cover slipped All immunofluorescence images were acquired using Olympus Fluoview FV1200 confocal microscope and processed using FIJI. RNAScope ® single-molecule fluorescence RNA in situ hybridization (smRNA) was performed on frozen 20 μm sections mounted on glass slides as previously described 58 . Retina cross sections were prepared as described above. In brief, tissue was first pre-fixed with 4% PFA for 15-minutes prior to proceeding with the RNAScope ® manufacturer protocol with the following modifications. Tissue was first dehydrated with ethanol in a stepwise progression of 10%, 30%, 50%, 70%, and 100% for 3 minutes each. During antigen retrieval, slides were boiled for 5-minutes. The RNAScope ® -mm-TH (Cat # 317621) probe was applied and paired with the fluorophore Opal 520™ (Akoya Biosciences FP148700KT). After incubation, tissue slides were washed and incubated in blocking buffer (0.5% TritonX-100 and 10% Normal Donkey Serum in PBS) for 1-hour prior to primary antibody staining. RNAScope slides were co-stained with anti-RBPMS. Primary and secondary antibody following the staining protocols described above. Immunohistochemistry was performed with antibodies against the vascular marker CD31 (i.e. PECAM-1) on whole mount prepared retina for quantifications of angiogenic sprouting and area coverage at P3 and P9. Stitched images were acquired with 10X magnification of the entire whole mount retina (5000 ✕ 5000 microns), and these images were used to quantify angiogenic sprouting at P3 and vessel coverage at P9. To quantify angiogenic sprouting, the number of endothelial sprouts was counted across the entire retina in each P3 animal. The average (+/−SEM) of the number of angiogenic sprouts from at least four mutant and control animals, respectively, were tested for statistical significance. To quantify area coverage, the area occupied by emerging vessels in the of deep vascular layer were measured using the FIJI freehand tool to circle the total area in which vessels were present. These values were normalized by the entire area of the retina. The average (+/−SEM) of the deep vessel area from at least four mutant and control animals were tested for statistical significance. To assess branching of the deep layer, two to three 10X images were captured per retina, and branch points were counted per field of view (1271.81x1271.81 microns). The branch points were averaged per animal. To assess the density of the ESM1 signal, whole mount P3 retinas were stained with anti-CD31, and anti -ESM1 antibodies. Two to three 20x magnification images (635.90 ✕ 635.90 μm) were captured per animal. FIJI and the Polygonal selection tool was used to select and measured ESM1-positive areas to assess ESM1 signal coverage in superficial vessels. We summed the areas of ESM1+ signal and normalized these values to the total area of vessel coverage per field of view. The total vascular area was calculated using AngioTool 91 . To assess TH mRNA expression in RGCs, 20X images were captured (635.90 x 635.90 μm) from retina tissue cross sections that had undergone the RNAScope assay to label TH transcripts. The number of anti-RBPMS positive cells containing Th mRNA signal was quantified and normalized to the total number of RBPMS cells in order to determine the percentage of Th -mRNA+/RBPMS+ cells. The percentages of Th -mRNA+/RBPMS+ cells across three animals were averaged per time point (P3-P15). P3 control (5 animals) and TH RGC (6 animals) retinas were dissected and transferred into a RIPA buffer containing Xpert protease inhibitor cocktail solution (GenDEPOT, 1:100) and Xpert phosphatase inhibitor cocktail solution (GenDEPOT, 1:100). Samples were manually homogenized with a Kimble Kontes Pellet Pestle homogenizer (DWK Life Sciences). For each sample, 25ug of protein was loaded and separated by SDS-PAGE on 4–15% tris-glycine gels before transfer onto nitrocellulose membranes. Blots were blocked in blocking buffer (5% BSA, 0.05% Tween 20 in TBST) for 1 h and then probed with primary antibodies overnight at 4C in 5% BSA/TBST. Blots were subsequently washed and stained with secondary antibodies for 1 h at room temperature. Image Lab software was used to perform densitometry analysis of bands. Local background was subtracted from each individual band and adjusted band volumes were measured. Each sample was normalized to GAPDH and normalized volumes were used to measure the percent of VEGFR2 which was phosphorylated (+/−SEM) To measure the signal intensity of DLL4, we co-stained tissue with anti-DLL4, -CD31, and -ERG. Vascular tip cells were captured in images with 60X magnification (211.97 x 211.97 microns). Using FIJI and the Polygonal selection tool, areas were drawn around ERG+ cells within tip cell filopodia as defined by areas that are positive for both DLL4 and CD31 at the leading edge of the vessel network. The mean fluorescence intensity was measured using the FIJI Measuring tool for both DLL4 and CD31 fluorescence and normalized by the number of ERG+ cells present within each selected area. The average of the DLL4:CD31 fluorescence ratio was taken from three to four images per animal. T0 measure the fluorescence intensity of Jagged1, images were captured with 10X magnification (1271.81x1271.81 microns). FIJI and the Rectangle tool were then used to select a representative area of the fluorescent intensity across the vascular plexus,. The fluorescent intensity was normalized by the area of the region of interest (μm 2 ). Retinal dopamine pathway expression analyses from developing mouse and human single cell Developmental scRNA-Seq atlases from mouse ( GSE118614 ; Clark et al., 2019) and human ( GSE138002 ; Lu et al., 2020) were accessed with corresponding post-processing metadata from the Gene Expression Omnibus (GEO). As a pre-filtering step, genes with no expression or expressed in fewer than 3 cells were filtered out. Corresponding expression matrices and their associated metadata were converted into a SeuratObject (Seurat v3; Stuart et al., 2019) for normalization, dimensionality reduction, and gene expression assessment. Normalization was performed using the function ‘SCTransform’ whilst regressing out mitochondrial gene expression. Dimensionality reduction was performed via the ‘RunPCA’ function with default parameters, with downstream UMAP embedding using the first 30 PCAs (both species). For cluster and cell identity assignments, labels were lifted from the ‘new_CellType’ variable (mouse) or ‘Cluster’ (human) in the associated metadata file. For downstream gene expression analyses, only retinal cell types were included. To determine dopamine pathway gene expression, gene ontology lists corresponding to both “cellular response to dopamine” (GO:1903351) and “dopamine metabolic process” (GO:0042417) were imported and assessed across both datasets and are displayed as a z-score of the log-transformed expression data. A Single-cell RNA-seq database (Sanes et al., 2022) was used to compare the dopamine synthesis pathway expression in different RGC precursor types at P0, P5, and P56. RGC dubtype determination was derived was the database with certainly calls differing according to ganglion cell maturation state. For earlier timepoints (e.g P0), precursors with the highest fate association were grouped together (annotated as cell type_traj) in the dataset. Dopamine related biosynthesis and transporter genes were searched in the database and their expression values were taken from the annotated scatterplot dat. To construct a heatmap, 20 discrete RGC types were selected in order to diplay reprsenative values for dopamine pathway genes over RGC development and adulthood. All statistical analyses were performed using an unpaired Student’s t-test and assessed for statistical significance (p-value <0.05) in Prism GraphPad 9. No statistical analysis was conducted to predetermine sample sizes. Randomization and blinding were not employed. Statistical differences were evaluated using Graphpad Prism 7 software.

Tip cells guide outgrowing capillaries toward gradients of matrix-bound vascular endothelial growth factor A (VEGFA), which binds and activates VEGF receptor 2 (VEGFR2) 36 . Because dopamine can modulate VEGFR2 activity 37 , 38 , we first asked whether precise dopamine levels were required for normal tip cell function using pharmacological modulators. To elevate retinal dopamine, we took advantage of the dopamine precursor analog L-DOPA-methyl ester (L-DOPA-ME, Figure 1B ). L-DOPA-ME is readily converted to dopamine and reaches the eye in biologically relevant quantities 39 , 40 . We found that daily L-DOPA-ME administration significantly reduced CD31 labeled vascular sprouts by 38% at P3 (p= 0.0019 Figure 1C , E ). To confirm and extend these findings, we co-labeled with ESM1, a tip cell marker and VEGF-regulated gene 41 – 43 . The area of ESM1 signal per CD31+ vascular area was reduced by 67% in L-DOPA-ME treated animals (p<0.0001, Figure 1D , F ). Since tip cells are necessary for the emergence of the deep plexus 44 , we asked whether the changes in the superficial vascular plexus might impact deep layer emergence. We found that elevating dopamine resulted in a significant decrease in deep layer coverage at P9, with a 23% reduction (p=0.0250, Figure 1G - I ). Dopamine modulation did not alter intermediate layer vessel growth, with no changes in vascular branching, nor was branching of the superficial layer affected ( Figure S1A - B ). Thus, a dopamine precursor analog specifically inhibits tip cell sprouting and layer emergence in the superficial and deep retinal layers. Retinal neurons are thought to produce the most, if not all, retinal dopamine, and removing the rate limiting enzyme for catecholamine production, tyrosine hydroxylase (TH), from these cells markedly reduces dopamine production 45 . Because increasing dopamine reduced angiogenesis, we next sought to determine whether reducing retinal dopamine signaling would promote vessel growth. To achieve this, we crossed conditional TH F/F mice to the retina neuron-specific Cre line Six3 Cre . We term this line TH RET , and we confirmed Cre specificity for all retinal neurons and the absence of TH immunoreactivity ( Figure 2A , Figure S2A - B ). We assessed vessel growth and maturation at P3 and P9. At P3, the number of endothelial sprouts in TH RET mutants was increased by 21% (p=0.009, Figure 2B , D ). In addition, ESM1 -positive signal area was increased and extended further into the vessel plexus (71% increase, p=0.0021, Figure 2C , E ). These changes impacted deep vessel layer emergence, with a reduction in vessel area coverage at P9 ( Figure S1G ). As in the L-DOPA-ME treated animals, branching of the superficial or intermediate vascular plexus was not affected in these animals ( Figure S1C - D ). This analysis shows that Th loss-of-function results in elevated angiogenesis in the superficial and deep vascular layers. The only population of dopaminergic cells in the retina described to date is a subset of interneurons called dopaminergic amacrine cells (DACs, or dopaminergic ACs). We thus investigated whether dopamine-dependent alterations in angiogenesis was dependent on amacrine cell derived TH. TH Cre lines do not reliably target TH-expressing cells in the retina 46 . To examine this question, we thus selectively deleted TH from all amacrines early in development using the Ptf1a Cre line to generate TH F/F ;Ptf1a Cre . This transgenic line expresses Cre exclusively in amacrine and horizontal cells as these cells become fate committed 47 . We confirmed the specificity of this line by crossing it to the Cre reporter Ai6 48 and validated the loss of TH from amacrine cells ( Figure S2C - D ). We hereafter refer to this line as TH AC ( Figure 2F ). Notably, vessel pattering, growth, and maturation were largely unaffected in TH AC mice. Angiogenic sprouting was similar between control and mutants (p=0.7406, Figure 2G , I ), and ESM1 staining in the superficial layer was indistinguishable from that of controls at P3 suggesting tip cell behavior was unchanged ( Figure 2H , J ). In addition, comparable vessel coverage area was observed in the deep vessel layer at P9 ( Figure S1G ), and the branching of the superficial and intermediate layer was also normal ( Figure S1E - F ). Together, these results suggest that dopamine derived from canonical dopaminergic interneurons is not required to modulate vessel growth. What retinal neuron cell type might be responsible for the vascular defects in TH RET mice? To begin to answer this question, we performed single-molecule fluorescent in situ hybridization (smFISH) for Th mRNA. Th was detectible as early as P3, just as the vasculature begins to be patterned. Notably, however, Th was not restricted to the inner nuclear layer (INL) where DACs reside. Instead, Th was present in similar numbers of cells in the ganglion cell layer ( Figure 3A ). Th levels in the ganglion cell layer increased from P3 to P5 as the superficial layer forms, and then decreased by P9, coinciding with deep layer emergence ( Figure 3A - B ). At the conclusion of vessel development at P15, Th levels remain low in the ganglion cell layer and become enriched in the INL where dopaminergic amacrines reside ( Figure S3A ). To confirm the identify of Th -expressing ganglion cell layer neurons, we performed dual smFISH together with antibody labeling for the RGC-specific marker RBPMS ( Figure 3A ). We found that at early developmental ages a high percentage of Th -positive neurons were RGCs (52% at P3, 59% at P5, 26% at P9, and 14% at P15 Figure 3B ). Data mining of retina neuron transcript databases confirmed the time restricted expression of Th in retinal ganglion cells but also other enzymes involved in dopamine biosynthesis ( Th , Ddc, Dao, Figure S3B ). The dopamine vesicular packaging transporters Slc18a2 and Slc6a3 similarly expressed in RGCs ( Figure S3B ). Re-analysis of retinal ganglion cell single cell sequencing datasets 49 , 50 suggests dopamine is produced and transported by a diverse pool of RGC types at these early P3 and P5 developmental time points ( Figure 3C ). Dopamine biosynthesis and transporter genes ( Th, Ddc, and Slc18a2,) have greater enrichment across several RGCs types at P0 and P5, whereas expression levels of these genes decrease by adulthood at P56 ( Figure 3C and Figure S3C ). Finally, antibody co-labeling for TH and RBPMS in the first postnatal week validated RGC protein production of TH ( Figure S3D - F ). To independently confirm that RGCs can produce dopamine that inhibits vessel growth, we combined the retinal ganglion cell specific Cre line Chrnb3 Cre 51 with our TH F/F allele to generate a line in which dopamine production is suppressed in RGCs. We refer to this line hereafter as TH RGC ( Figure 3D ). We confirmed it is specific for RGC, deletes genes as early as P1, and removes TH from RGCs ( Figure S2E - F ). We assessed the relative contribution of RGC-derived dopamine to tip cell sprouts. We found that angiogenic sprouting was markedly increased in TH RGC mice. TH RGC mutants have a 42% increase in tip cell sprouts during superficial layer emergence at P3 compared with controls, an increase similar to that observed when TH is removed from all retina neurons (p=0.0233, Figure 3E , G ). ESM1 staining was also elevated and extended further into the vessel plexus (96%, p=0.0237, Figure 3F , H ). As in the TH RET mouse line, these changes impacted deep vessel layer emergence. We observed an increase in area coverage at P9 in TH RGC mutants compared to controls ( Figure S1G ). Stalk cells do not extend filopodia but exhibit proliferative behavior that elongates the vasculature 41 , 52 . To assess the specificity of dopamine responses for tip cells, we quantified the total number of endothelial cells and endothelial cell proliferation by labeling for the endothelial cell marker ERG and the cell cycle marker Ki67, respectively. We observed no significant difference in the levels of total endothelial cells or cell proliferation in TH RGC animals ( Figure S4A - D ). We further assayed whether changes to tip cell growth may induce downstream changes in vessel stability. To examine this, we stained for collagen IV and CD31 to label empty collagen IV positive sleeves left by regressing endothelial cells 53 in mutants and controls at P3. We found no apparent change in the presence of empty sleeves (ColIV-postive/CD31-negative) in TH RET mutants versus controls ( Figure S4E ). Together, these data suggest that tip cells are the primary target through which RGC-derived dopamine signals to the growing vasculature, and that these signals may occur locally, while other aspects of vessel maturation are unaffected. We then addressed whether increasing the amount of dopamine produced by retinal ganglion cells was sufficient to reduce angiogenesis in the neonatal retina. To achieve this, we transduced Chrnb3 Cre retinas with an AAVD/J containing a Cre-dependent Th expression cassette at P0 (AAV-DIO-TH-Myc, Figure 3I ). We confirmed that transduction was specific for RGCs and resulted in increased TH expression only in these GCL localized, axon-containing neurons ( Figure 3I , arrows). Littermates without the Cre protein but transduced with the same AAV-DIO-TH-Myc virus served as controls. Elevating TH in RGCs markedly reduced angiogenesis. The number of angiogenic sprouts decreased by 40% (p<0.0001) in P3 C re+ retinas relative to control retinas without Cre ( Figure 3J , L ). Similarly, ESM1 staining was also reduced by 77%. (p<0.0001, Figure 3K , M ). Vessel growth in the deep plexus was also impacted, with a reduction in area coverage and branching at P9 ( Figure S3G ). These defects were not persistent, as prolonging TH expression using AAV-DIO-TH-Myc did not result in altered vessel organization in adults ( Figure S3H ). RGC-derived dopamine is thus necessary and sufficient to modulate vessel growth specifically in the early stages of sprouting and vascular layer emergence. Our experimental results suggest that dopamine has an inhibitory effect on sprouting angiogenesis and impacts the levels of ESM1, a tip cell marker and VEGF-regulated gene 42 . This could be explained by dopamine-mediated alterations to Jagged/Dll4-Notch signaling, which is known to occur downstream of VEGF receptor (VEGFR) 2 activation 36 , 54 . In particular, dopamine binding to dopamine receptor-D2 (DRD2) on endothelial cells can decrease the ability of VEGF to phosphorylate and activate VEGFR2 37 , 38 . This can directly alter vessel sprouting and growth, as VEGFA activation of VEGFR2 can decrease the levels of Notch ligand Jagged1 and upregulate Delta-like 4 (Dll4) in filopodia containing tip cells 36 , 55 . In turn, these changes suppress tip cell phenotypes in trailing stalk cells 36 , 56 . This process drives tip to stalk cell differentiation and vessel radial growth 17 , 54 , 57 . We thus tested whether RGC-derived dopamine interfaced with Dll4-Jagged Notch signaling pathway via suppressing VEGFR2 activation. To begin, we performed immunoblotting for VEGFR2 and the activated, phosphotyrosine-1173 form of VEGFR2, from both control TH F/F and TH RGC retinas at P3. We have previously shown that the antibodies used in this experiment can detect a linear relationship between lysate quantity and band intensity, confirming the reliability of comparative immunoblotting 29 , 58 . We found that while the amount of VEGFR2 declined in TH RGC mice, the levels of pY1173-VEGFR2 increased when bands were normalized to VEGFR2 levels ( Figure 4A - B ). These results are consistent with the idea that dopamine modulates VEGF receptor activation 38 . These results are also consistent with the increase in ESM1 staining in TH RGC mice, which reports on VEGF-A activity. We further confirmed that DRD2 is present on superficial vasculature endothelial cells during the first postnatal week using DRD2-GFP reporter mice ( Figure 4C ). These data show that RGC-derived dopamine normally suppresses VEGF signaling. If dopamine reduces VEGF signaling activity, we predict TH inactivation may alter the levels of the Notch ligands Jagged1 and Dll4, both of which directly impact tip cell behavior. The former interacts with Notch to promote the expression of genes that drive tip cell function 36 , 54 , 55 . The latter is elevated in tip cells relative to stalk cells and interacts with Notch to ultimately reduce tip cell proliferation and promote stalk cell maturation 17 , 54 . Immunohistochemistry for Jagged1 at P3 in control and TH RGC mice revealed a significant increase in Jagged1 levels throughout the retina vasculature (43% increase, P = 0.0005, Figure 4D , G ). These results are consistent with the increase in angiogenic sprouting observed in these animals and with the elevation in VEGFR2 phosphorylation. Since DLL4 is known to counteract Jagged1 signaling, we questioned whether the increase in Jagged1 may be associated with a compensatory increase in DLL4. DLL4 is normally higher at the angiogenic front of tip cells with low intensity expression in stalk cells 36 and more mature vessels. In TH RGC mutants, we observed less differential staining of Dll4 across the vascular plexus and a modest but significant increase in total Dll4 levels at the angiogenic front (p=0.0433, Figure 4E , H ). Consistent with these results, we also observed a decrease in the transcriptional repressor Hey1 54 , a downstream target that is positively regulated by Notch ( Figure 4F ). Together, these results suggest that RGC-derived dopamine is necessary and sufficient to regulate VEGF activity and coordinate Notch signaling via Jagged1 and Dlll4 to promote tip-to-stalk conversion and vessel growth ( Figure 4I ).

The lead contact is Melanie Samuel ( [email protected] ). All reagents used in this study are available from the lead contact. Animal lines, plasmids and viruses used in this study can be requested from the lead contact. Microscopy data reported in this paper will be shared by the lead contact upon request. No original code was used to generate data in this paper. Any additional information required to reanalyze the data reported in this paper is available from the lead contact.

To ensure proper vascularization outcomes, the nervous system limits vessel growth to developmental periods when neural circuits are established and refined, but the precise types and location of neural activity that influences these events is unclear. In the retina, distinct types of neural activity predominate at different developmental phases, allowing a direct investigation of how, when, and where different neuromodulators impact vessel development. Using this system, our studies reveal three surprising features of activity-regulated vessel growth. First, dopamine is required early in development to direct angiogenesis through negatively regulating tip cell sprouting and maturation. These data indicate that precise types of neuromodulators can signal to and regulate the vasculature. In the case of dopamine, this signaling may happen directly given the presence of DRD2 on endothelial cells themselves. Second, distinct types of neuromodulators can regionally influence vessel development despite overlaps in the timing of their production and release. For example, early cholinergic activity in the retina directs angiogenesis in the deep retinal layer but does not impact superficial layer growth 22 . Even though cholinergic and dopaminergic activity overlap in the first two postnatal weeks, dopamine appears to specifically impact superficial layer, and in turn, deep layer vessel growth. Whether this reflects spatial differences in the location of cholinergic amacrines in the INL and GCL relative to dopaminergic retinal ganglion cells, which are restricted to the GCL, is unclear. Alternatively, these regional effects may reflect differences in tip cell biology in the superficial layer relative to the inner vascular layers 16 , 44 . Third, our results indicate that vascular growth can be influenced by temporally restricted production of neuromodulators. In particular, our results indicate that the timing of RGC dopamine production in first postnatal week is critical for vascular outcomes. Too little RGC-derived dopamine, and vascular growth and tip cell sprouting is increased; too much, and vascular growth is restricted and deep layer emergence is decreased. The time restricted nature of RGC dopamine production may in part reflect the fact. RGCs are amongst the first retinal cells to differentiate. RGCs are born between E12 and P0, and dendritic and axonal outgrowth and refinement can be observed from E16-E17 until P14 when vasculature RGCs show spontaneous retinal ‘wave’ activity as early as E16, which persists until P14. At this time, bipolar cells become fully integrated into the circuit allowing light-driven photoreceptor activity. Together with our data, these results suggest that dynamic regulation of neuromodulator production in developing CNS neurons may be a key mechanism that controls when and where vessels grow as circuits mature. Indeed, there is increasing evidence for this type of so called ‘neurotransmitter switching’. Unlike classic synaptic plasticity, which involves changes in synapse strength, number, and excitability 59 , neurotransmitter switching involves the gain or loss of transmitter in the same neuron 60 – 63 . To our knowledge, our data are the first to demonstrate temporally restricted retinal ganglion cell dopamine production and thereby evidence for neurotransmitter switching, in the retina. What enables RGCs to temporally produce and release dopamine? While definitive answers to this question will require additional study, a few mechanisms may contribute. Dopaminergic RGCs do not appear to be restricted to particular subtypes based on scRNA sequencing analysis. If confirmed, dopamine production is unlikely to be RGC-subtype specific and thus regulated downstream of genetic programs that influence RGC identity. Because rod and cone photoreceptor synapses do not mature until after vessel development is complete 20 , RGC dopamine production is also unlikely to be driven by canonical rod and cone driven circuit activity. Previous studies in which dopamine production is induced or restricted in other neuron types may provide some alternatives. In general, neurotransmitter switching can be regulated by activity, cytokines and growth factors 60 , 63 – 65 , transcription factors 66 , microRNAs 62 , and epigenetic alterations 60 , 67 . In particular, exposure to a long-day photoperiod induces switching from dopamine to somatostatin in a subset of paraventricular hypothalamic nucleus (PaVN) neurons in the brain 62 . Activity changes in dopaminergic PaVN neurons appear causal to this switch 68 . Specific kinds of neural activity may also help drive dopamine production in the retina. For example, waves of retinal activity are propagated before eye opening downstream cholinergic signaling 69 . In turn, these waves can induce retinal dopamine release in the first two postnatal weeks 70 . Might there be co-regulation between acetylcholine and dopamine either at the circuit level or on the vasculature itself? While we cannot rule out co-regulation, it is likely that additional factors are at play given that cholinergic activity 22 and dopaminergic activity alter the growth of distinct vessel layers. None-the-less, it will be interesting to determine if retinal waves play a role in regulating ganglion cell-derived dopamine production and release. Angiogenesis requires a tightly coordinated balance between endothelial cell sprouting and tip to stalk cell conversion to drive vessel growth. In the superficial retinal vascular plexus, this process has been shown to be modulated by the precise signaling of the VEGF-DLL4-Jagged-Notch pathway. VEGFR-2 activation by VEGF-A will in turn upregulate DLL4 expression in tip cells, where the high levels of DLL4 will bind and activate Notch in trailing endothelial stalk cells cells. Previous work has shown that this equilibrium can be controlled by local Dll4 expression in endothelial tip cells, which suppresses sprouting in adjacent ECs, driving tip to stalk cell differentiation and vessel growth. An additional Notch ligand, Jagged1, also counteracts Dll4-Notch signaling interactions between stalk endothelial cells, which helps to sustain elevated VEGF receptor expression in the immature vascular plexus at the angiogenic front to increase tip cell sprouting and filopodia 36 . How might dopamine interface with and impact this pathway? Our data are consistent with a direct role for dopamine binding to DRD2 on endothelial cells. Dopamine binding to DRD2 decreases the ability of VEGFR2 to be phosphorylated and activated due to increased activity of the phosphatase SHP2 38 . In addition, DRD2 agonists can induce VEGFR-2 internalization 71 . Accordingly, we observed that reducing dopamine signaling broadly in retinal neurons, or in RGC specifically, increased the level of VEGF2R phosphorylation and thus activation, and also increased Jagged1 and ESM1, which reports on VEGF-A activity 42 . In addition, we observed ESM1 staining that extended beyond the angiogenic front in TH RGC animals, indicative of reduced or delayed maturation of tip cells to stalk cells. These changes also negatively impacted the transcriptional repressor Hey1, a downstream target positively regulated by Notch. While these and other data point to a role for dopamine in modulating canonical VEGF-Dll4/Jagged-Notch signaling, we cannot rule out potential contributions of non-canonical notch signaling, such as the regulation of Wnt/β-catenin signaling 72 . Several studies have found that modulating dopamine can directly impact endothelial cell function in vitro and in vivo beyond the CNS. In culture, dopamine inhibited VEGF-induced proliferation and migration of human umbilical vein endothelial cells (HUVEC), while DRD2 agonists had a similar result 38 . These effects extend to endothelial progenitor cells, as dopamine can reduce VEGF-induced endothelial cell progenitor migration 37 , 73 . In vivo, dopamine has also been used successfully to reduce vascularization in different disease models. These include would healing 74 and endometriosis 75 , 76 . In one human clinical trial, administration of the DRD2 agonist quinagolide was successful in reducing endometrial tissue lesion size by ~70% in a human clinical trial 77 . Dopamine modulation has also been used to reduce tumor growth in a large number of in cancer types ranging from lung to ovarian cancer models 78 – 82 . Our data suggest that vascular regulation by dopamine extends to nervous system. In line with idea, rodent models of Parkinson’s disease show increased levels of VEGF-A in the substantia nigra 83 and enhanced endothelial cell proliferation 84 . These findings seems to be recapitulated in human patients 85 . Finally, the value of the dopamine response in treating retina circuit dysfunction in diabetic humans with has been hinted at with a study showing that Sinemet (levodopa and carbidopa) can return the electroretinogram (ERG) abnormalities of diabetic patients to normal 86 . Given these results, understanding the regional, neuron-subtype, and vascular layer-specific consequences of dopamine signaling may provide new therapeutic opportunities for precisely intervening in vascular diseases.

All experiments were performed in both male and female mice. Animals were housed under 10 hr light:14 hr dark cycles. The control mice were C57BL/6J strain (JAX 000664), unless indicated otherwise. The following mouse lines were used in this study: TH F/F , Six3 cre , Ptf1a cre , Chrnb3 cre , CD1. TH F/F mice were given by Dr. Martin Darvas at University of Washington and originally generated by the lab of Dr. Richard Palmiter 88 . To achieve pan-neuronal deletion of Th from all retinal neurons, TH F/F was crossed to Six3 cre animals 89 , referred here as TH Ret . To generate cell-specific deletion of Th from dopaminergic ACs, Th F/F was crossed to Ptf1a cre animals, referred here as TH AC . Lastly, to generate cell-specific deletion of Th from retinal ganglion cells, TH F/F mice were crossed with Chrnb3 cre animals, referred here as TH RGC . For each of these lines, littermate TH F/F mice were used as controls. All transgenic animals were maintained in the C57Bl/6J background. Pharmacological administrations and plasmid electroporations were performed on CD1 wildtype animals obtained from Charles River Laboratories. All mice were used at the ages specified in the experimental procedures outlined below, with a mixture of male and female mice. Experiments were carried out in accordance with the recommendations in the Guide for the Care and Use of Laboratory Animals of the NIH under protocols approved by the BCM Institutional Animal Care and Use Committee.

Blood vessels in the central nervous system (CNS) support and are required for neural activity that underlies all CNS function 1 . In turn, the growth, patterning, and function of CNS vessels is regulated by reciprocal relationships between cells that comprise the vascular unit (endothelial cells, pericytes, and vascular smooth muscle cells) and resident neurons and glial in the CNS 2 – 4 . These interactions are thought to determine the features and timing of vascular development, leading to coordinated vessel growth, vessel maturation, and blood-brain-barrier (BBB) formation 4 . In particular, microglia 5 , 6 and astrocytes 7 – 9 contribute to the developmental patterning of vessels as they grow, which in turn determines where and how nutrients are available to local neuron populations. In addition, external sensory simulation of neural activity by whisker deflection can have moderate effects on vessel patterning in the retina and cortex 10 . However, whether and how temporally regulated induction of specific neural signals by defined neuron types can modulate vessel outcomes is unclear. Also unclear are the mechanisms by which these signals influence distinct phases of vessel development. Deciphering these relationships is important given the large number of eye and brain diseases associated with vessel dysfunction 11 – 15 . In mouse, like in human, the development of neuron subtypes and their vascular counterparts in the retina are temporally and spatially distinct, enabling precise investigation of the relationship between neural activity, neuron identity, vascular growth, and vascular maturation. At birth, the mouse retina is avascular. Vascular genesis is initiated by a single, superficial layer of blood vessels that grows progressively from the center toward the periphery from postnatal day (P) 1 until P7 ( Figure 1A ). A week later, vessels grow outward to form the deepest vascular plexus, followed by inward growth to generate the intermediate layer. Vascular tip cells are thought to help guide the formation of these networks. These specialized endothelial cells are morphologically and positionally distinct in the vascular plexus, with long filopodial protrusions that extend only at the leading edge of a radial spreading vasculature network 16 . The leading vascular edge is then converted into developing arteries and veins to establish the vascular plexus via Jagged1, Dll4, and Notch signaling 17 – 19 . Whether neurons help instruct these spatial endothelial cell transitions is unknown. To address the role of neurons in this process, we leveraged the precisely regulated developmental activity of retinal neurons. Outer and inner retina synaptic activity driven by rods and cones is not functional until after the superficial and deep vessel layers form, so canonical, visually-evoked neural activity is unlikely to drive angiogenesis in this circuit 20 , 21 . In contrast, spontaneous retinal neural activity overlaps with superficial and deep retina vessel layer development (P0-P10, Figure 1A ). Consistent with this, acetylcholine-regulated spontaneous activity has some impact on the deep vessel layer, but other retinal layers are unaffected 22 , 23 . What other developmentally regulated neurotransmitters may contribute to vessel growth? Dopamine is a good candidate because it is produced at the right place and time to affect retina vessel growth 24 , 25 . Within the CNS, dopamine functions as a synaptic neurotransmitter but also as a mass-action neuromodulator that signals through a sub-family of five G-protein coupled receptors (GPCRs) 19 , 26 . In the cerebral cortex 27 and retina 28 it can modulate microcirculation. We further showed that dopamine promotes hyaloid vessel regression in the developing eye 29 . Outside the CNS, cardiovascular smooth muscle cells 30 , 31 , renal vessels 32 , and tumor vessels 33 , 34 also show dopamine-mediated vascular responses. Whether dopamine can regulate vessel growth within the CNS and the mechanisms by which it may do so are uncertain. Here, we utilized the retina to ask whether dopaminergic signaling controlled by specific neuron subtypes contributes to the spatio-temporal regulation of angiogenesis. We found that dopamine is a potent neuron-derived signal that directly impacts where and how vessels grow. Excess dopamine restricts vessel growth, and reduced dopamine increases it by elevating vessel sprouting and angiogenesis. Surprisingly, we find that vessel-active dopamine is not produced by canonical dopaminergic amacrine interneurons. Rather, it is derived from retinal ganglion cells (RGCs) located in the nerve fiber layer and is expressed only during the early postnatal period when vessels develop. Removing dopamine only from this small neuron subset (<3–5%) 35 recapitulates the vascular effects of removing dopamine from all retinal neurons, while increasing RGC dopamine production prevents deep vessel layer formation. We show that RGC-derived dopamine modulates vessel growth by impacting VEGFR activation and Notch signaling via the ligand Jagged1, which in turn promotes tip-cell behavior. These data uncover how temporally regulated production of specific neurotransmitters in sparse neuron subtypes can direct vessel development. They further identify dopamine as a potential target for modulating vascular abnormalities in disease.