An opsin 5-dopamine pathway mediates light-dependent vascular development in the eye.

During mouse postnatal eye development, the embryonic hyaloid vascular network regresses from the vitreous as an adaption for high-acuity vision. This process occurs with precisely controlled timing. Here, we show that opsin 5 (OPN5; also known as neuropsin)-dependent retinal light responses regulate vascular development in the postnatal eye. In Opn5-null mice, hyaloid vessels regress precociously. We demonstrate that 380-nm light stimulation via OPN5 and VGAT (the vesicular GABA/glycine transporter) in retinal ganglion cells enhances the activity of inner retinal DAT (also known as SLC6A3; a dopamine reuptake transporter) and thus suppresses vitreal dopamine. In turn, dopamine acts directly on hyaloid vascular endothelial cells to suppress the activity of vascular endothelial growth factor receptor 2 (VEGFR2) and promote hyaloid vessel regression. With OPN5 loss of function, the vitreous dopamine level is elevated and results in premature hyaloid regression. These investigations identify violet light as a developmental timing cue that, via an OPN5-dopamine pathway, regulates optic axis clearance in preparation for visual function.

Photons from the sun reach our planet at high flux. In response, organisms have evolved detection systems that decode light information for adaptive advantage. Examples from mammals include the visual system[1], where photons bouncing off an object are detected to decode object identity, and the circadian system, where the 24 hour light cycle entrains time-of-day dependent physiology[2,3]. Most light detectors in metazoans are opsins[4,5], a class of G-protein coupled receptors that convert the energy of a photon into a cellular signalling response. Rhodopsin, the opsin of mammalian rod photoreceptors, is a well-characterized example of a visual opsin[6,7] while melanopsin (Opsin 4, OPN4) has a central role in circadian clock photoentrainment[8-10]. Neuropsin (Opsin 5, OPN5) is another opsin family member. Relatively little is known about OPN5 except that it responds to violet light wavelengths (λmax of 380 nm)[11-15], regulates seasonal breeding behaviour in birds[16] and the activity cycle in mice[17] but also mediates photoentrainment of the retinal circadian clock[18]. Here we have investigated OPN5 function in development of the mouse eye and shown, like the above examples, that it is required for normal biological timing. In this case, OPN5 is required for a light response that regulates vascular regression timing.

Results

Opn5 is expressed in a retinal ganglion cell subset.

Opn5 is expressed in retinal ganglion cells (RGCs) in adult mice[18]. To further assess the features of Opn5-expressing cells, we combined an Opn5 allele (Supplementary Fig. 1) with Ai14, a tdTomato expressing cre reporter. According to labelling with multiple markers (Supplementary Figure 2) the overall architecture of the Opn5 null retina is unchanged. In P5 retinal flat mounts, Opn5; Ai14 cells were at relatively low density throughout the inner retina (Fig. 1a). In P12 calretinin-labelled cryosections , Opn5 expressing cell bodies were in the ganglion cell layer (Fig. 1c, d, GCL). At P5, Ai14 expressing processes were immature (Fig. 1a) but at P12 were prominent and observed as bundles within the nerve fibre layer (NFL) and within several laminations of the inner plexiform layer (Fig. 1c, d, IPL, S1-S5). These morphological features are consistent with the characteristics of RGCs.

Figure 1.

Opn5 is expressed in a distinct subset of retinal ganglion cells.

a, b, Flat mount retina from P5, Opn5; Ai14 mice showing the tdTomato cre reporter (a, b, red), nuclear labelling with Hoechst 33258 (b, blue), and counter labelling for melanopsin (b, green). c, d, Retinal cryosections from P12, Opn5; Ai14 mice showing the tdTomato cre reporter (c, d, red), nuclei with Hoechst 33258 (c, d, blue), and labeling for calretinin (c, green). Retinal laminae are indicated by the abbreviation between the panels: NFL; nerve fibre layer, GCL: ganglion cell layer; S5-S1; sublaminae of the inner plexiform layer, INL: inner nuclear layer, OPL: outer plexiform layer. e-h, As in (a, b) except at P12. g, h, show magnified regions of (e) as indicated by white corner marks. i, j, Flat mount retinae showing labelling of cell bodies (asterisks), dendritic fields and axons (arrows) for retinal ganglion cells labelled by the Brainbow3.2 reporter in P24 Opn5 mice. Scale bars are 20 μm. Panels a-j are representative of at least 3 separate experiments. Additional examples of these images are available on Figshare.

Melanopsin antibody labelling in Opn5; Ai14 retinae indicated that OPN4 and Opn5 are expressed in distinct RGC subsets. At P5, the density of Opn5; Ai14 and OPN4-labelled cells was similar (Fig. 1b). At P12 (Fig. 1e-h) co-labelling again showed that largely, melanopsin and Opn5; Ai14 cells were two distinct subsets. Prominent bundles of axons from Opn5 and OPN4 RGCs are cofasciculated (Fig. 1e-g). Rare co-labelled cells were identified (Fig. 1e, h, about 50 cells per retina) but could result from cre lineage marking over-sampling[19]. At P24, Opn5; Brainbow[20] labelled retinal cells (Fig. 1i, j) have the appearance of mature RGCs with extensive dendritic arbours and axons. Brain cryosections from Opn5; Ai14 mice showed axons in the optic tracts, lateral geniculate nucleus and superior colliculus as might be expected for RGCs. Labelling of Opn5; Ai6 retinae at P8 with the RGC marker RBPMS[21] and the RGC/amacrine cell marker Calretinin[22] provided evidence that Opn5 is expressed exclusively in RGCs (Supplementary Figure 3).

Normal hyaloid vessel regression timing requires Opn5 and violet light.

We have previously shown that light stimulation of melanopsin (Opn4) regulates hyaloid vessel regression and retinal angiogenesis[23]. Prompted by this, we assessed hyaloid regression in the Opn5 null mouse. At P1, Opn5 null mice showed normal hyaloid vessel numbers (Fig. 2a, b, g, h) and normal vessel cellularity (Fig. 2c, d, control, 13.3±1.2, Opn5 null, 13.2±1.6 nuclei/100 μm length, p=0.92). At P8, Opn5 null mice had fewer hyaloid vessels (Fig. 2e-h) indicating precocious regression. This phenotype is unique as all hyaloid phenotypes described so far[24], including that of the Opn4 null mouse[23], show hyaloid persistence. Precocious hyaloid regression in the Opn5 null mouse is best illustrated when vessel numbers are quantified over a P1-P8 time-course (Fig. 2g, h, blue line) and compared with the control (Fig. 2g, h, grey line) and the Opn4 null (Fig. 2h, green line). When Opn5 is conditionally deleted in the retina using Chx10-cre or Rx-cre precocious hyaloid regression is observed (Fig. 2i-k). Thus, OPN5 is required locally within retinal neurons to regulate hyaloid regression.

Figure 2.

Precocious hyaloid vessel regression in the Opn5 null and absence of 380 nm photons.

a, b, Hoechst 33258 (blue) labelled hyaloid vessel preparations from P1 Opn5 and Opn5 mice. c, d, Higher magnification images of two examples each of P1 hyaloid vessel segments from Opn5 (c) and Opn5 (d) mice labelled with Hoechst 33258 (presented in red) and isolectin (green). e, f, Hoechst 33258 labelled hyaloid vessel preparations from P8 Opn5 and Opn5 mice. g, Quantification of hyaloid vessel number in Opn5 (grey), Opn5 (light blue) and Opn5 (dark blue) mice over a P1-P8 time-course. h, as in (e) but relative hyaloid vessel numbers for control (grey), Opn5 (blue), and Opn4 (green) mice. i, j, Hoechst 33258 labelled hyaloid vessel preparations from P8 Opn5 control and Opn5; Rx-cre mice. k, Quantification of P8 hyaloid vessel number in Opn5 control (grey), Opn5; Chx10-cre (light blue), and Opn5; Rx-cre (dark blue) mice. l, m, Hoechst 33258 labelled hyaloid vessel preparations from P8 mice raised from birth in full spectrum (V, violet + B, blue, + G, green, + R, red) or "minus violet" (BGR) lighting. n, Quantification of P8 hyaloid vessel number from P8 mice raised from birth in full spectrum (grey bar) or “minus violet” (blue bar) lighting. All p-values by Student T-test. The number at the base of each chart is n and represents the number of animals assessed. Error bars are SEM. Scale bars 400 μm (a,b,e,f,I,j,l,m) 20 μm (c,d).

To mimic Opn5 loss-of-function using lighting conditions, we raised mice from birth in the absence of the 380 nm wavelengths that maximally stimulate OPN5[11-14]. Control mice were raised in a light-dark cycle of “VBGR” lighting (violet (380 nm), blue (480 nm), green (520 nm) and red (630 nm)) and showed a typical hyaloid vessel number at P8 (Fig. 2l, and n, grey bar). By contrast, mice raised in “BGR” lighting that omitted violet light showed precocious hyaloid regression (Fig. 2m, and n, blue bar). This phenocopies the Opn5 null and is consistent with a model in which 380 nm photons stimulate OPN5 to suppress hyaloid vessel regression.

A light-OPN5 pathway regulates dopamine levels in the eye.

In Opn4 null mice, elevated levels of VEGFA explain hyaloid vessel persistence[23]. Assessment of the levels of VEGFA and its inhibitor FLT1[25] in the vitreous of Opn5 null mice indicated no change (Supplementary Fig. 4d, e). This suggested that the OPN5-light response pathway regulated hyaloid regression by a distinct mechanism. In the Opn5 null retina, we noted an unusual labelling pattern for tyrosine hydroxylase (TH). In WT retina at P8, TH immunoreactivity was faint and largely restricted to the perinuclear region of a subset of amacrine cells (Fig. 3a, b). In Opn5 null mice, TH labelling was stronger and prominent in cell processes (Fig. 3c, d). Elevated intensity of TH labelling in the Opn5 null retina was also evident at P15 when dopaminergic amacrine cells are more fully developed (Fig. 3e, f). TH is the rate-limiting enzyme that mediates the first step in the biosynthesis of dopamine. Since TH levels are under feedback regulation[26], these data suggested that dopamine levels might be modulated in the Opn5 null eye. Interestingly, dopamine is known to have in vitro anti-vascular activity via suppression of VEGFR2 signalling[27,28]. This meant that Opn5-dependent regulation of dopamine could be an explanation for Opn5-dependent regulation of hyaloid regression. Thus, we hypothesized that OPN5-dependent release of dopamine from the retina promoted hyaloid vessel regression by direct signalling.

Figure 3.

OPN5- and light-dependent pathways regulate dopamine levels in the neonatal mouse eye.

a-f, Immunofluorescent labelling for tyrosine hydroxylase (TH, greyscale) in flat mount retinae from control Opn5 (a, b, e) and Opn5 (c, d, f) mice at P8 (a-d) and P15 (e, f). Scale bars 100 μm (a-d) 20 μm (e,f). g-k, Charts showing ELISA quantification of dopamine in retinal lysates (g, j) and vitreous fluid (h, i k). g, h, Dopamine levels in each tissue over a P2-P8 developmental time-course. i, Dopamine levels over a P2-P6 developmental time course comparing normal lighting (LD) with the consequences of dark rearing (DD). j, k, Dopamine levels in retinal lysate (j) and vitreous fluid (k) from P6 Opn5 control (j, k, grey bars), Opn5 heterozygote (j, k, light blue bars), and Opn5 homozygote (j, k, light blue bars). p-values by (g, h) Two Way ANOVA, (i) Student’s T-test, (j, k) One Way ANOVA. Error bars are SEM. The number at the base of each chart is n and represents the number of animals assessed. Panels a-f are representative of at least 3 separate experiments. Additional examples of (a-d) are available on Figshare.

Dopaminergic amacrine cells that express TH develop in the first few days after birth in the mouse[29]. This also means that normally, dopamine levels in the retina climb rapidly after birth[29]. Using ELISA quantification, we confirmed that retinal dopamine levels rise over a P2 to P8 time-course (Fig. 3g). Dopamine in solubilized retinal tissue is primarily from intracellular stores[30]. ELISA quantification showed that dopamine levels in the vitreous, where the hyaloid vessels reside, also rise over the postnatal period (Fig. 3h). In adult mice, dopamine levels in the eye are light-regulated[31,32,33,34]. To evaluate this possibility for neonatal mice, we quantified vitreous dopamine over a postnatal time-course of LD and DD lighting. At P4 and P6, vitreal dopamine levels were significantly elevated in DD (Fig. 3i) indicating that postnatally, light stimulation normally suppresses vitreal dopamine. To determine whether light-regulated dopamine might be a consequence of OPN5 activity, we harvested retina and vitreous from an Opn5 allelic series at P6 and quantified dopamine (Fig. 3j, k). This showed that Opn5 homozygote null mice had lower levels of intracellular dopamine in the retina (Fig. 3j), but elevated levels in the vitreous (Fig. 3k). These data indicate that dopamine in both compartments is regulated by OPN5.

A light-OPN5 pathway suppresses dopamine release to the vitreous by enhancing DAT activity.

The biological effects of dopamine are regulated by its release, signalling and reuptake[35,36]. A key regulator of uptake, and thus a good candidate for an OPN5-dependent activity, is the dopamine transporter DAT/SLC6A3[36 ,37]. Threonine 53 of DAT is phosphorylated and enhances the rate of dopamine uptake by DAT[37]. This activation marker can be detected with a phospho-specific antibody[37]. Basal phosphorylation stoichiometry of T53-DAT is typically 50% but is increased by stimuli that elevate dopamine uptake[37]. To determine whether phospho-T53-DAT levels were regulated by light and were OPN5-dependent, we labelled retinal cryosections from cohorts of P8 littermate Opn5 and Opn5 mice after they were dark-adapted followed by ±30 minutes of 380 nm light at 1×1012 photons/cm2/sec. Labelling for phospho-T53-DAT was found throughout the inner plexiform layer (IPL) and nerve fibre layer (NFL) of all samples (Fig. 4a-d). Phospho-T53-DAT signal was quantified by averaging pixel intensity across horizontal pixel rows of images aligned like those shown (Fig. 4a-d), generating intensity profiles (Fig. 4e, grey and blue profiles) and calculating area under the peaks (indicated by the grey vertical dashed lines in Fig. 4e and f). This revealed that for the IPL (Fig. 4g, grey and blue bars) but not for the NFL (Fig. 4h, grey and blue bars), phospho-T53-DAT signal was significantly increased in response to 380 nm light.

Figure 4.

Light-dependent activation of phospho-T53-DAT/SLC6A3 in the IPL requires OPN5.

a, b, Labelling of P8 Opn5 retina for nuclei (blue) and phospho-T53-DAT (green) for dark adapted mice half of which were exposed to 30 minutes of 380 nm light. NBL: neuroblastic layer. IPL: inner plexiform layer. GCL: ganglion cell layer. NFL: nerve fibre layer. Scale bars 50 μm. Indicated areas of each panel are exported to chart (e). e, Chart showing averaged (n=3) quantification profiles (dark adapted, grey trace, light exposed, blue trace) for phospho-T53-DAT in retinal cryosections (a, b). Grey shading emphasizes the elevated signal of light-induced sample. c, d, As in (a, b) but for Opn5 retina. f, Chart as in (e) except for Opn5 mice (dark-adapted, orange trace, or dark adapted and exposed to 380 nm light for 30 minutes, red trace). Traces from chart (e) are reproduced on chart (f) for comparison. g, h, Charts showing under-the-peak quantification (n=3) of phospho-T53-DAT labelling in inner plexiform layer (g) and nerve fibre layer (h) for P8 mice of the labelled genotype and light exposure. The vertical dashed lines in (e, f) indicate the areas that were quantified. i, Immunoblot detecting DAT/SLC6A3 (DAT), phospho-T53-DAT/SLC6A3 (pDAT) and β-tubulin (TUBB) in retinae harvested from Opn5 (left lane) and Opn5 (right lane) mice during the light phase. phospho-T53-DAT/SLC6A3 is lower than normal in the Opn5 null. Unprocessed blot image in Supplementary Figure 6. j, Quantification of P8 hyaloid vessels in WT mice injected with vehicle or with the DAT/SLC6A3 inhibitor raised from E17 either in normal (LD) lighting, or in constant darkness (DD). k, Quantification of P8 hyaloid vessels in Opn5, Opn5, and Opn5 mice injected from P1-P8 with the DAT/SLC6A3 inhibitor. p-values by Two Way ANOVA. Error bars are SEM. The number at the base of each chart is n and represents the number of animals assessed. Panels a-d are representative of 3 separate experiments.

To determine whether light induction of phospho-T53-DAT signal was OPN5-dependent, we performed the same analysis in Opn5 null mice (Fig. 4c, d, f-h). This revealed that in the Opn5 null retina, phospho-T53-DAT signal for the IPL did not elevate in response to light exposure (Fig. 4g, orange and red bars) indicating OPN5-dependence. Interestingly, in the NFL, Opn5 null mice have a lower level of phospho-T53-DAT independent of light exposure (Fig. 4h, compare grey bar and orange bar). However, this low level of phospho-T53-DAT can be rescued by light exposure (Fig. 4h, compare orange and red bars). This indicates that NFL phospho-T53-DAT levels are regulated both by OPN5 (negatively, light-independent) and by a distinct light response pathway (positively). Though we do not currently understand the pathway for positive regulation of phospho-T53-DAT in the NFL (other opsins are obvious candidates), it serves as a useful internal control to show that the Opn5 null retina can be light responsive. An immunoblot detecting phospho-T53-DAT in total solubilized retina from the light phase (Fig. 4i) revealed that overall, the level is lower in the Opn5 null, suggesting that NFL phospho-T53-DAT is the smaller proportion of the total. Collectively, analysis of T53-DAT indicates that OPN5 is required for a light-dependent upregulation within the IPL. Since T53 phosphorylated DAT sequesters dopamine with higher efficiency, this finding is consistent with a model in which loss of OPN5 function results in diminished dopamine uptake by DAT/SLC6A3, and thus elevated levels of vitreal dopamine.

Inhibition of DAT/SLC6A3 promotes hyaloid regression in an Opn5-dependent manner.

To determine whether DAT had a functional role in the regulation of hyaloid regression in vivo, we took advantage of the DAT inhibitor GBR12909[38]. Based on data showing that dopamine signalling can suppress VEGFR2 activation via the phosphatase SHP2[28] we would predict that suppressing DAT activity would elevate dopamine levels in the eye and thus should counter the consequences of dark rearing because the latter elevates levels of VEGFA[39]. To test this, we compared the effects of P1-P8 GBR12909 injection on hyaloid vessel regression in C57BL/6J mice raised either in normal (LD) lighting or in constant darkness (DD). This showed that GBR12909 had no significant effect on mice raised in normal lighting (Fig. 4j, LD) but could reverse the hyaloid vessel persistence resulting from dark rearing (Fig. 4j, DD). This shows that DAT activity is an important light-dependent regulator of hyaloid vessel regression.

To assess DAT activity in OPN5-dependent regulation of hyaloid regression, we injected GBR12909 daily from P1-P8 into Opn5, Opn5, and Opn5 mouse pups. Quantification of hyaloid vessels at P8 showed that wild type (WT) mice did not respond significantly (Fig. 4k) but that in heterozygote mice, GBR12909 produced a precocious hyaloid regression equivalent to the Opn5 homozygote phenotype (Fig. 4k). GBR12909 produced no change in homozygote mice (Fig. 4k). Inhibitor activity may be buffered by the intact feedback regulation of the WT mouse[40]. However, in Opn5 heterozygote mice, where dopamine levels are elevated (Fig. 3k) and feedback regulation may be compromised, the DAT inhibitor was able to convert the hyaloid phenotype from normal to precocious regression. It is likely that the inhibitor produces no effect in the Opn5 homozygote because endogenous vitreal dopamine is already high (Fig. 3k) and signalling activity may be close to maximal. These data indicate a finely balanced interaction between OPN5 and DAT that is consistent with OPN5-dependent regulation of DAT activity via phosphorylation.

VGAT in OPN5 RGCs is required for regulation of phospho-T53-DAT and hyaloid regression.

Glutamate, γ-aminoburyric acid (GABA) and glycine are neurotransmitters important for visual function. In the adult mouse, glutamate is used as an excitatory neurotransmitter by canonical photoreceptors[41,42] and OPN4 RGCs[43]. GABA and glycine are inhibitory neurotransmitters and their receptors are detected in a variety of retinal neurons including amacrine cells and RGCs[42]. Glutamate and GABA/glycine are loaded into presynaptic vesicles by the VGLUT (vesicular glutamate transporter 2) and VGAT (vesicular GABA transporter) family of transporters, respectively. During mouse retinal development, VGAT is expressed soon after birth and precedes the expression of VGLUT2[44]. Loss-of-function of these transporters eliminates neurotransmitter activity[45,46]. To determine whether OPN5 RGCs might use one of these neurotransmitters for the vascular response pathway, we quantified hyaloid vessels in Opn5 conditional deletion mutants of Vglut2 and Vgat. Though deletion of Vglut2 had no consequence (Fig. 5a) homozygous deletion of Vgat phenocopied the Opn5 germ line null precocious hyaloid regression (Fig. 5b-f). Furthermore, the Opn5 heterozygote showed no hyaloid phenotype, but Opn5; Vgat mice showed a significant precocious regression (Fig. 5d, f). This transheterozygote phenotype is genetic evidence that Opn5 and Vgat function in the same pathway and in the same cell type. If this is true, then we would predict that Vgat conditional deletion would, like Opn5 loss-of-function, result in diminished levels of phospho-T53-DAT under lighted conditions. This was confirmed using immunofluorescence labelling of retina from Opn5; Vgat (Fig. 5g, h) and Opn5; Vgat mice (Fig. 5i, j). These data suggest that OPN5 RGCs use VGAT to signal within the vascular regression pathway.

Figure 5.

OPN5 RGCs use Vgat in a hyaloid regression pathway: A model for OPN4-VEGFA and OPN5-dopamine pathway integration.

a, Quantification of hyaloid vessels in P8 Opn5; Vglut2 (light grey), Opn5; Vglut2 (dark grey), Opn5; Vglut2 (light blue), Opn5; Vglut2 (dark blue) mice. b-e, Hyaloids from P8 Opn5; Vgat (b), Opn5; Vgat (c), Opn5; Vgat (d), Opn5; Vgat (e) mice. f, Quantification of hyaloid vessels in P8 mice with genotypes listed in b-e. g-h, Retinal cryosections imaged for phospho-T53-DAT (red) and Ai6 cre (green) from P8 Ai6; Opn5; Vgat (g,h) or Ai6; Opn5; Vgat (i, j) mice. Additional examples of (g-j) available on Figshare. Sample size (n) for a,f is shown at the base of each bar and represents mice. p values by one-way ANOVA. Error bars are SEM. Images in b-e represent at least six, and in g-j at least three, separate experiments. Scale bars 200 μm (b-e) 20 μm (g-j). k, l, Schematic describing integration of the OPN4-VEGFA and OPN5-dopamine hyaloid regression pathways. The schematic identifies two phases of development, E16-E18 (k), and P3-P8 (l), when OPN4 and OPN5 are each required. In late gestation, blue light stimulation of OPN4 RGCs suppresses retinal cellularity. In dark-reared or in Opn4 null mice, elevated cellularity increases oxygen demand ([O2]) and, via the hypoxia response pathway, increases VEGFA expression in amacrine cells and RGCs. Elevated VEGFA causes promiscuous retinal angiogenesis and suppresses hyaloid vessel regression. According to the present analysis, violet light stimulation of OPN5 RGCs postnatally suppresses dopamine in the vitreous by upregulating T53 phosphorylation of the dopamine transporter (P-DAT/SLC6A3) in neurons in the inner plexiform layer. Normally, OPN5-dependent phosphorylation of DAT results in elevated dopamine uptake and a reduced flux of dopamine from dopaminergic amacrine cells to the vitreous. In the absence of OPN5, or the violet light that stimulates OPN5, vitreous dopamine is precociously elevated. This results in premature activation of dopamine receptor DRD2 in hyaloid VECs, suppression of VEGFR2 survival signalling and precocious regression. These data indicate that both 480 nm blue light via OPN4, and 380 nm violet light via OPN5, function as developmental timing cues.

Dopamine has a direct action on hyaloid vascular endothelial cells (VECs) to promote hyaloid regression.

We have hypothesized that dopamine release from the neonatal retina is light- and OPN5-dependent and that dopamine then signals directly to hyaloid VECs to promote regression. To assess whether a retinal source of dopamine regulated hyaloid regression, we conditionally deleted a Th allele. Chx10-cre[47], though effective for studies of TH function in adult mice[48] did not delete Th efficiently during the postnatal period. However, Rx-cre[49] was effective (Fig. 6a, b) and resulted in a hyaloid vessel persistence (Fig. 6c-e) indicating that the active dopamine is produced locally in the retina.

Figure 6.

Retinal dopamine promotes hyaloid vessel regression via DRD2-dependent suppression of VEGFR2 activity.

a, b, TH labelling (green) in four regions of P8 flat mount retinae from Th (a) and Rx-cre; Th (b) mice. c, d, Hyaloids from P8 control Th (c) and Rx-cre; Th (d) mice. e, P8 hyaloid vessel numbers in control (Th or Th, grey), Rx-cre; Th (light blue), and Rx-cre; Th (dark blue) mice. f, g, Hyaloid vessels from Drd2-GFP mice showing reporter expression (green) in vessels but not macrophages (circles). h, i, Immunolabelling for DRD2 in P8 hyaloids from tamoxifen treated Drd2 (h) and Drd2; Pdgfb-icreERT2 (i) mice. j, P8 hyaloid vessel numbers in Drd2 (grey) and Drd2; Pdgfb-icreERT2 (blue) mice. k, Immunoblots for VEGFR2, VEGFR2 pY1173, and β-tubulin in P6 hyaloid vessel lysates from Drd2 and Drd2; Pdgfb-icreERT2 mice. Hyaloid lysate is loaded in successively halved volumes to ensure linear range detection. l, Band intensity of n=3 mice, pY1173 VEGFR2, normalized to total VEGFR2 from (k) confirms that in the Drd2 mutant, active VEGFR2 is higher. m, n, Immunoblots for VEGFR2, pY1173-VEGFR2 (m), AKT and pS473-AKT (n) from P6 hyaloids of indicated genotypes. m, Quantification of pY1173-VEGFR2 relative to VEGFR2 in Opn5 (grey) and Opn5 (blue) hyaloid vessels. pY1173-VEGFR2 and pS473-AKT levels are lower in Opn5. n=3 mice. Unprocessed blot images in Supplementary Figure 6. o-q, P8 hyaloids from tamoxifen treated Opn5; Drd2 (o), Opn5; Drd2 (p), and Opn5; Drd2; Pdgfb-icreERT2 (q) mice. r, P8 hyaloid vessel numbers in Opn5; Drd2 (grey), Opn5; Drd2 (light blue), and Opn5; Drd2; Pdgfb-icreERT2 (dark blue) mice. s-u, P8 hyaloid preparations from Flt1 (s), Rx-cre; Flt1 (t), and SKF38393 injected, Rx-cre; Flt1 (u) mice. v, P8 hyaloid vessel numbers in control Flt1 (grey), Rx-cre; Flt1 (light blue), and SKF38393 injected, Rx-cre; Flt1 (dark blue) mice. Scale bars 20 μm (a,b,f-i) 200 μm (c,d,o-q, s-u). (e,j,r,v) The number at the base of each chart bar is n and represents number of mice. p-values by One Way ANOVA (e, r, v), Student’s T-test (j, l, m). Error bars are SEM. Images are representative of at least three separate experiments.

To assess the involvement of dopamine receptors in hyaloid regression we first took advantage of SK38393, a receptor agonist[50]. SK38393 was injected daily into Opn5, Opn5, and Opn5 mouse pups from P1-P8. Though SK38393 had no significant effect on WT mice (Supplementary Fig. 4f), it produced precocious hyaloid regression in heterozygous mice (Supplementary Fig. 4f). SK38393 did not produce a significant reduction in hyaloid vessel numbers in Opn5 null mice (Supplementary Fig. 4f). This pattern of response is very similar to that observed with the DAT inhibitor (Fig. 4k). Again, this pattern of modulation is probably explained by the resilience of an intact dopamine feedback pathway in WT animals, the sensitized background of the heterozygote, and the already saturated level of dopamine signalling in the homozygote. Injection of two different dopamine receptor antagonists from P1-P8 in WT mice produced elevated numbers of hyaloid vessels (Supplementary Fig. 4g). Thus, pharmacological manipulations indicate that hyaloid vessel regression can be regulated both positively and negatively by dopamine receptor modulators.

One prediction of the hypothesis that retinal dopamine regulates hyaloid regression was that dopamine receptors would be expressed within the hyaloid vessels. Since DRD2 was implicated in suppression of VEGFR2 signalling[28], we focused on this member of the family. Vascular cells, but not hyaloid-associated myeloid cells, showed Drd2-GFP reporter expression (Fig. 6f, g). Furthermore, labelling with an anti-DRD2 antibody detected cells within the hyaloid vessels (Fig. 6h) and this was eliminated in the Drd2; PDGFB-icreERT2 conditional deletion that targets VECs (Fig. 6i). These data show that DRD2 is expressed in hyaloid VECs.

In Drd2; PDGFB-icreERT2 mice, the hyaloid vessels are persistent (Fig. 6j) but there are no quantifiable consequences for development of the superficial retinal vasculature (Supplementary Figure 5). This identifies hyaloid VECs as a dopamine responsive cell and indicates that dopamine signalling promotes hyaloid regression. A further prediction of the hypothesis is that in hyaloid VECs, dopamine signalling would suppress the activation of VEGFR2[51]. To test this, we performed immunblotting for VEGFR2 and the activated, phosphotyrosine-1173 form of VEGFR2, from both Drd2 control and Drd2; PDGFB-icreERT2 hyaloid vessels at P5. Pooling dissected hyaloids from 6 animals of each genotype allowed threshold detection of the pY1173-VEGFR2 in the control (Fig. 6k, left lane). To assess the reliability of comparative immunoblotting, we performed a three step, two-fold loading dilution and quantified immunoblot band intensities. Presented graphically, band intensities for VEGFR2, pY1173-VEGFR2 and β-tubulin showed high Pearson coefficients, indicating a linear relationship between lysate quantity and band intensity (Supplementary Fig. 6). When pY1173-VEGFR2 values were normalized to VEGFR2 (Fig. 6l), Drd2; PDGFB-icreERT2 genotype values were much higher (Fig. 6l) consistent with the observed band intensities on the immunoblot (Fig. 6k). These data show that deletion of Drd2 in hyaloid VECs permits elevated activation of VEGFR2 and indicates that normally, dopamine signalling suppresses VEGFR2 activity. In an additional test of this model, we assessed pY1173-VEGFR2 and pS473-AKT levels in the Opn5 null. pY1173-VEGFR2 levels were lower in the hyaloid vessels of the Opn5 null (Fig. 6m). In addition, across an allelic series, pS473-AKT levels were lower only in the Opn5 homozygote, consistent with precocious hyaloid regression only in this genotype (Fig. 6n). Since dopamine levels are high in the Opn5 null, these data are consistent with a model in which dopamine promotes hyaloid vessel regression by suppressing VEGFR2 activity and the downstream survival signalling mediated by AKT.

As a genetic test of the relationship between OPN5 and vascular signalling, we determined whether deletion of Drd2 in VECs would reverse precocious hyaloid regression in the Opn5 null. We compared P8 hyaloid vessel numbers in mice of genotype Opn5, with Opn5; PDGFB-icreERT2; Drd2 and with Opn5; PDGFB-icreERT2; Drd2. This experiment confirmed the precocious regression of hyaloid vessels due to Opn5 loss-of-function (Fig. 6o, p, and r, light blue bar) but showed that deletion of Drd2 in VECs could switch the hyaloid phenotype to persistence (Fig. 6q, p, dark blue bar). This outcome establishes that Opn5 and Drd2 function in the same developmental pathway and have opposing influences on hyaloid regression.

One implication of immunoblotting data for pY1173-VEGFR2 (Fig. 6k) and the genetic analysis (Fig. 6o-r) is that a balance of VEGFA and dopamine signalling determines the fate of the hyaloid vessels. To determine whether we could demonstrate this balance at the level of receptor ligands, we designed a rescue experiment. We generated an elevated level of VEGFA activity, and thus hyaloid vessel persistence, by conditionally deleting the gene encoding the naturally occurring VEGFA inhibitor FLT1 in the retina. Chx10-cre deletion of Flt1 produces hyaloid persistence (Supplementary Figure 4a-c) but in this case, we used Rx-cre (Fig. 6s, t, v, light blue bar). To determine whether dopamine receptor signalling could reverse the hyaloid persistence, we injected (each day, from P1-P8) a littermate cohort of Rx-cre; Flt1 mice with the dopamine receptor agonist SKF38393. This resulted in a reversal of the hyaloid persistence (Fig. 6s-v), an outcome that illustrates the balance of VEGFA and dopamine signalling that regulates hyaloid regression.

Discussion

We have identified an unanticipated vascular development pathway in the eye. OPN5, an atypical opsin known to respond to near-UV photons[11-14], initiates the pathway response and functions postnatally (Fig. 5l). Dopamine, a broadly functional neurotransmitter and neuromodulator, is a signalling intermediate that is regulated by OPN5 and elicits a direct response in hyaloid VECs to limit VEGFR2 signalling via DRD2 (Fig. 5l). Based on OPN5- and light-dependent phosphorylation at threonine 53, and its pharmacological inhibition, our data also suggest that the dopamine transporter DAT/SLC6A3 is a key component of this pathway that normally suppresses levels of dopamine in the vitreous (Fig. 5l). In addition, the GABA transporter VGAT is implicated in OPN5 RGC signalling as its conditional deletion in OPN5 RGCs phenocopies the Opn5 null precocious hyaloid regression and low phospho-T53-DAT. The light-OPN5-VGAT-dopamine-DRD2-VEGFR2-hyaloid pathway is thus characterized by two suppressive steps: light-OPN5 suppresses levels of dopamine in the vitreous while dopamine suppresses VEGFR2 signalling in the hyaloid vessels (Fig. 5l).

Light-dependent regulation of vitreal dopamine occurs against a backdrop of generally rising levels of dopamine in the postnatal eye. This means that while the function of dopamine is to promote hyaloid vessel regression, the effect of 380 nm photons and OPN5 is to suppress regression of the hyaloid vessels. It is likely that this has evolved as a mechanism to optimize the timing of hyaloid vessel regression and ensure that they remain functional postnatally until the superficial retinal vascular plexus is complete. We have previously shown that melanopsin (Opsin 4, OPN4) also mediates light-dependent vascular development in the eye[39] and suppresses VEGFA levels because it keeps retinal cellularity in check, and thus limits the oxygen demand that can elevate levels of VEGFA (Fig. 5l). The crucial window for activation of the OPN4 response is in late gestation and requires a direct light stimulation of the mouse fetus[39]. Thus, the OPN4 and OPN5 response pathways use distinct mediators to regulate vascular development and, since they function at different stages of development, can be thought of as developmental timing cues (Fig. 5l). Notably, the spontaneous waves of neuronal activity that arise in the neonatal mouse retina[52,53] are partly dependent on OPN4 modulation of gap junctions that are, in turn, regulated by dopamine[54]. In the future, it will be interesting to assess the relationship of retinal wave activity to vascular development.

Neuropsin is highly conserved and we might anticipate that the pathway we describe (Fig. 5l) will be relevant to human biology. The latter steps in the pathway involving DRD2-dependent suppression of VEGFR2 activity may be an explanation for the observation that premature infants treated with dopamine (for hypotension) have a higher risk of retinopathy of prematurity (ROP)[55,56], a vascular overgrowth disease. We suggest that therapeutic dopamine promotes regression of the hyaloid vessels and thus exacerbates the hypoxia that leads to rebound vascular overgrowth. Furthermore, we have previously shown that the risk of ROP in premature infants is partly dependent on their season of gestation with short days and lower light exposure associated with higher risk[57]. It is possible that the OPN5-dopamine pathway is a component of this risk equation because insufficient light would be expected to result in elevated vitreal dopamine, precocious hyaloid regression and thus a more profound hypoxia in the premature eye. An understanding of the relationship between OPN4 and OPN5-dependent regulation of vascular development in the eye (Fig. 5l) raises the interesting possibility that premature infants at risk for ROP might be treated with a light therapy that differentially targets each pathway response. Finally, it has been proposed that both violet light in the 360–400 nm range[58] and dopamine[59] are key regulators of refractive development and that each can suppress progression to myopia. The current observations suggest that the OPN5-dopamine pathway is likely to be involved.

Materials and Methods

Mice

Animals were housed in a pathogen-free vivarium and all pharmacological treatments were in accordance with protocols approved by the Institutional Animal Care and Use Committee at Cincinnati Children’s Hospital Medical Center. This study is compliant with all relevant ethical regulations regarding animal research. Day of birth is defined as P1. Genetically modified mice used in this study were: Chx10cre (Tg(Chx10-EGFP/cre-ALPP)2Clc/J) (Jax source #00515), PdgfbicreER(T2)[1], Rxcre[2], Ai14 (Jax stock #007914), Brainbow (Jax stock #021227 Brainbow 3.2), Drd2 (Tg(Drd2-EGFP)S118Gsat)[3], Drd2 (Jax stock #020631), Flt1 (Jax stock #02809 Vegfr-1), TH (Ref[4]), Vglut2 (Jax stock #012898), Vgat (Jax stock #012897), Opn4 (Ref[5]) and Opn5 that were generated from C57BL/6N ES cells obtained from KOMP (ES clone ID:KOMP-(HTGRS6008_A_B12-Opn5-ampicillin). The ES cells harbour a genetic modification wherein a Lacz-Neomycin cassette is flanked by FRT sites, between exon 3 and exon 4 and a loxp site separates Lacz from the neomycin coding region (Supplementary Fig. 1). Loxp sites also flank exon 4 of Opn5 allowing multiple mouse lines that can serve as reporter nulls, conditional floxed and null mice. The Opn5 allele was created by crossing the Opn5 mice to FLPeR (Jax stock #003946) to remove the LacZ cassette. The Opn5 line was created by crossing the Opn5 mice to E2a-Cre (Jax stock # 003724). Littermate control animals were used for all experiments with the exception of C57BL/6J mice reared under different lighting conditions.

The genotyping primers and protocol for alleles except Opn5 are described in the cited publication or on the Jackson Labs website. Primer sequences for genotyping the Opn5 or Opn5 alleles are: F1: CACAGTATGTGTGACAACCT; R1: GTGGACAGATTAACTGAAGC; R2: GAACTGATGGCGAGCTCAGA. F1-R1 gives a 626 bp wild-type band and also gives a 700 bp band from the Opn5 allele. F1-R2 gives a 376 bp for Opn5 and a 1617 bp band from the Opn5 allele. Primer sequences for genotyping the Opn5 allele are: Opn5creF1: TGGAAAGAGATGCATTTGTGAG; Opn5creF2: CACTGCATTCTAGTTGTGGTTTGTCC; Opn5creR1: ACAGCCTATGAATTCTCTCAATGC. F1R1 gives a 300 bp band for wild-type and F2R1 gives a 209 bp band for Cre allele.

The Opn5 was generated in-house using CRISPR-Cas9 technology. Four gRNAs that target exon 1 of Opn5 were selected to knock in the Cre cassette. Plasmids containing the gRNA sequence were transfected into MK4 cells (an in-house mouse cell line representing induced metanephric mesenchyme undergoing epithelial conversion). The editing efficiency of gRNA was determined by T7E1 assay of PCR products of the target region amplified from genomic DNA of transfected MK4 cells. The sequence of the gRNA that was subsequently used for the transfection is TGGAGTCCTACTCGCGGACG. Sanger sequencing was performed to validate the knock-in sequence of founder mice.

Mice were placed on normal chow diet (NCD: 29% Protein, 13% Fat and 58% Carbohydrate kcal; LAB Diet #5010) ad libitum with free access to water. Littermate controls were used for genetic crosses and both male and female pups were included in the study.

Lighting conditions

Animals were housed in standard fluorescent lighting (photon flux 1.62×1015 photons/cm2/sec) on a 12L:12D cycle except where noted. For full spectrum lighting (VBGR), LEDs were used to yield a comparable total photon flux of 1.68×1015 photons/cm2/sec. Spectral and photon flux information for LED lighting: violet (λmax=380 nm, 4.23×1014 photons/cm2/sec in 370–400 nm range), blue (λmax=480 nm, 5.36×1015 photons/cm2/sec in 430–530 nm range), green (λmax=530 nm, 5.82×1015 photons/cm2/sec in 480–600 nm range) and red (λmax=630 nm, 1.93×1015 photons/cm2/sec in 590–660 nm range). For wavelength restricted hyaloid assessment, C57BL/6J animals were housed in 12L:12D cycle starting late gestation (embryonic day E18) either in full spectrum (VBGR) or without violet (BGR) lighting. For dark-reared (DD) experiments, pregnant dams were moved to dark at gestation age E16. For light induction experiments, on P7 at lights off, nursing females and pups were moved to DD for 24-hour dark-adaptation. Opn5 and Opn5 pups were subjected to ±30 minutes of 380 nm light at 1×1012 photons/cm2/sec (approximately 1% of clear sky summer day sunlight at this wavelength) at two hours after subjective lights off.

Immunohistochemistry and imaging

Animals were anaesthetized under isoflurane and sacrificed by cervical dislocation or decapitation for early post-natal pups. Preparation and immunofluorescence staining of retinae and hyaloid vessels were as described previously[6]. For phospho-T53-DAT (pDAT) quantification, retinae of each genotype and light condition were collected in dim red light (dark-adapted) or normal light from at least 3 different induction experiments and mounted in the same OCT blocks. Retinal sections were processed, stained, and imaged together to compensate for batch differences. Alexa-conjugated secondary antibodies were purchased from Jackson ImmunoResearch (JIR). Images were captured using Zeiss ApoTome AX10 or Zeiss LSM700 confocal microscopes and processed by ImageJ (NIH) and Adobe Photoshop (Adobe Systems). Primary antibodies and lectins used in this study are listed in Supplementary Table 1.

ELISA

Vitreous and retinae from pups were collected and rapidly frozen on dry ice. To detect dopamine levels, vitreal samples were pooled from 3–6 pups depending on age and 6 retinae for each “n”. Dopamine extraction and ELISA assay were performed according to manufacture protocol using BA E-5300 (Rocky Mountain Diagnostics). For DD experiments, vitreous and retinae were collected under dim red light. To detect VEGFA and FLT1 levels, samples from P5 pups were pooled from 6 eyes for each “n”. Mouse VEGFA kit, Quantikine (MMV00) and mouse VEGFR1 (FLT1), Quantikine (MVR100) from R&D systems were used. ELISA was read by using EnVision Multimode Plate Reader (Perkin Elmer).

Pharmacological reagents and treatments

All dopamine pharmacological modulators, except for antagonists 2-CMDO (2-Chloro-11-(4-methylpiperazino)dibenz(Z)[b,f]oxepin maleate), were injected at 1 mg/kg body weight[7] intraperitoneally (i.p.) into nursing dams on day of birth and P2, then directly into pups until P8. 2-CMDO was injected into pups P5-P8 at 2 mg/kg body weight. Injection was done in dim red light one hour before lights on. Dopamine agonist SKF38393 hydrobromide; high affinity D2 antagonist L-741626; dopamine transporter 1 inhibitor, GBR12909 dihydrochloride; 2-CMDO were all purchased from Tocris Biosciences. For experiments with Pdgfb-icreERT2 mouse lines, 2 mg Tamoxifen was injected into nursing dams on the day of birth and on P2 to activate tamoxifen-dependent cre.

Western Blotting

Western blots were performed using standard protocols. Immediately after dissection, twelve hyaloid vascular tissues or retinae of P6 (six pups) were pooled in 100 μl of 1x Laemmli sample buffer and sonicated. After centrifugation, 20 μl of supernatant was loaded on 4 to 20% gradient protein gel (ThermoFisher Scientific). Separated protein bands were transferred to PVDF membrane, and bands were visualized by chemiluminescence (ThermoFisher Scientific). Unprocessed immunoblots are available in Supplementary Fig. 6. Band intensity was measured by ImageJ (NIH). Following antibodies were used for Western blotting: VEGFR2 (Cell Signaling Technology, #9698), phospho-VEGFR2 (Cell Signaling Technology, #2478), β-tubulin (Abcam, ab6046), DAT (Novus, NB300–254), phospho-DAT (ThermoFisher Scientific, PA5–35414), AKT (Cell Signaling Technology, #4691) and phospho-AKT S473 (Cell Signaling Technology, #4060). All antibodies were used at 1:1000 dilution.

Statistics and Reproducibility

Samples for immunoblots were pooled from multiple animals (six pups for hyaloid vasculature and six retinae) and each experiment repeated at least twice with independent samples. Unprocessed western blots are available in Supplementary Fig. 6. ELISA assessment was performed at least twice with independent biological samples. Each n in this analysis is a separate animal with the exception that for immunoblots and ELISA, pooled samples from animals of the same genotype represent one n. Retinal images with immunofluorescence labelling represent n=3 independent biological samples from separate litters. Data gathered for hyaloid vessel quantification represents samples from multiple litters to reach n indicated on charts for each genotype and condition. Data are presented as mean±s.e.m in aligned dot plots overlaid with bar or line graph. Statistical analyses were performed using GraphPad Prism version 4.00 (GraphPad Software) and Microsoft Excel for Two Tailed Student’s T-test, One-way or Two-way ANOVA as indicated. Two‐tailed distribution, two‐sample unequal variance t‐test was used to determine the statistical significance between two independent groups except for Fig. 6m which is one-tailed. Sidak’s or Tukey’s multiple comparison test were performed post-hoc when significance differences were found in ANOVA.

Data availability

Source data for all figures have been provided as Supplementary Table 2.

Additional experimental repeats for key retinal labelling experiments have been deposited on Figshare (doi:10.6084/m9.figshare.7450961).

All other data supporting the findings of this study are available from the corresponding author on reasonable request.