ABSTRACT
MicroRNAs are key regulators of angiogenesis, as illustrated by the vascular defects observed in miR-126-deficient animals. The miR-126 duplex gives rise to two mature microRNAs (miR-126-3p and -5p). The vascular defects in these mutant animals were attributed to the loss of miR-126-3p but the role of miR-126-5p during normal angiogenesis in vivo remains unknown. Here, we show that miR-126-5p is expressed in endothelial cells but also by retinal ganglion cells (RGCs) of the mouse postnatal retina and participates in protecting endothelial cells from apoptosis during the establishment of the retinal vasculature. miR-126-5p negatively controls class 3 semaphorin protein (Sema3A) in RGCs through the repression of SetD5, an uncharacterized member of the methyltransferase family of proteins. In vitro, SetD5 controls Sema3A expression independently of its SET domain and co-immunoprecipitates with BRD2, a bromodomain protein that recruits transcription regulators onto the chromatin. Both SetD5 and BRD2 bind to the transcription start site and to upstream promoter regions of the Sema3a locus and BRD2 is necessary for the regulation of Sema3A expression by SetD5. Thus, neuronally expressed miR-126-5p regulates angiogenesis by protecting endothelial cells of the developing retinal vasculature from apoptosis.
INTRODUCTION
MicroRNAs are key regulators of blood vessel development (Kane et al., 2014). The Mir126 locus is embedded within the endothelial gene Egfl7 and gives rise to two mature microRNAs: miR-126-3p and miR-126-5p. The physiological functions of the miR-126 microRNAs have been studied in mice and zebrafish using gene inactivation; miR-126-3p is a main actor of angiogenesis that enhances angiogenic factor signaling in endothelial cells by repressing Spred1 and phosphoinositol-3-kinase regulatory subunit 2 (Fish et al., 2008, Wang et al., 2008). Although both mature miR-126-3p and miR-126-5p were inactivated in these in vivo studies, the vascular defects were attributed exclusively to the loss of miR-126-3p and the potential contribution of miR-126-5p to vascular development was not addressed. We subsequently established that miR-126-5p was actually expressed by mouse embryonic endothelial cells, although at lower levels than miR-126-3p (Poissonnier et al., 2014b), and we and others demonstrated that this microRNA was functional in various biological processes such as atherosclerosis (Schober et al., 2014), erythropoiesis (Huang et al., 2011), choroidal neovascularization (Zhou et al., 2016), HIV infection (Huang et al., 2017), cancer (Wang et al., 2017; Wu et al., 2014; Zhang et al., 2013; Zhou et al., 2014) and endothelial cell activation (Cerutti et al., 2017; Poissonnier et al., 2014b). We have identified Setd5 as a target gene of miR-126-5p in endothelial cells using transcriptomic analysis and have validated in vitro the contribution of SetD5 in the control of leucocyte adhesion onto endothelial cells by miR-126-5p (Poissonnier et al., 2014b). However, to date, the relevance of the SetD5 regulation by miR-126-5p has not been investigated in vivo.
SetD5 is a previously uncharacterized member of the protein lysine methyltransferase family, which is defined by a conserved SET domain (Xiao et al., 2003). These enzymes catalyze mono-, di- or tri-methylation of lysine ε-amine groups of proteins, particularly histones, which drive chromatin rearrangement and regulate gene expression. However, SetD5 shares higher homologies with mammalian MLL5 (KMT2E), Drosophila UpSET and yeast Set3, three members of this family of proteins that do not show methyltransferase activity (Pijnappel et al., 2001; Rincon-Arano et al., 2012; Sebastian et al., 2009). Therefore, it was recently proposed that MLL5 and SetD5 should be separated from the active methyltransferase family (Mas-Y-Mas et al., 2016). SetD5 functions are not yet known but mutations of its gene are associated with human neurodevelopmental disorders (Grozeva et al., 2014; Kuechler et al., 2015; Parenti et al., 2017; Szczałuba et al., 2016) and prostate cancer (Sowalsky et al., 2015). In addition, SetD5 is upregulated in pancreas ductal adenocarcinoma and breast cancers (Liu et al., 2015; Mazur et al., 2014). SetD5 expression was analyzed during early development in mice and appeared to be ubiquitous (Osipovich et al., 2016). Recently, a constitutive knockout of the Setd5 gene in mice showed that Setd5−/− embryos were not viable and suffered from abnormal development of the neural tube and heart, probably because of disturbed post-translational histone modifications, as seen in SetD5-deficient embryonic stem cells (Osipovich et al., 2016). Postnatal functions of SetD5 were not addressed in this report.
In the present study, we characterized the specific functions of miR-126-5p during vascular development using the postnatal establishment of the retina vasculature as a model. The retina is vascularized by the inner, deeper and intermediate vascular plexuses, which sequentially appear and expand during the first two postnatal weeks in mice (Sapieha, 2012). The inner plexus initially emerges from the optic nerve head and spreads onto the retinal surface using the astrocytic template as a guide during the first postnatal week (Fruttiger, 2007). Vascular sprouting is directed by specialized endothelial cells, which either migrate to orientate the vascular sprout (tip cells) or proliferate to favor vascular expansion (stalk cells) (Gerhardt et al., 2003). Vascular endothelial growth factor (VEGF), the angiogenic factor that stimulates tip- versus stalk-endothelial cell determination in retina through VEGFR-Dll4-Notch signaling, consequently governs retinal vascular expansion (Hellström et al., 2007; Jakobsson et al., 2010; Lobov et al., 2007; Suchting et al., 2007). Astrocytes, retinal ganglion cells (RGCs) and cells located in the inner nuclear layer of the retina, all express Vegf transcripts during retina blood vessel development and can therefore govern vascular growth (Scott et al., 2010). Astrocyte-specific deletion of VEGF led to a subtle vascular phenotype in the retina (Scott et al., 2010), suggesting that other cells could govern vascular development in the retina. The strong vascular defects observed in mice deficient for RGCs (Edwards et al., 2012; Sapieha et al., 2008) suggested that these cells were also good candidates for involvement in the control of retinal vascular expansion, probably through their VEGF secretion.
Here, we show that miR-126-5p is expressed by both endothelial cells and RGCs of the early postnatal retina. Specific repression of miR-126-5p in mouse retina during the expansion of the inner vascular network does not affect vascular covering; however, it reduces the vascular plexus density and increases endothelial cell apoptosis. These vascular density defects are due to the upregulation of semaphorin 3A (Sema3A) in RGCs, in which miR-126-5p controls Sema3A expression through regulation of its target gene Setd5. We demonstrate that SetD5 is a nuclear protein that binds to the Sema3a promoter and requires the double bromodomain protein BRD2 to control Sema3A expression.
RESULTS
miR-126-5p is expressed both in endothelial cells and in RGCs in the postnatal retina
The expression pattern of miR-126-5p in the retina was analyzed using in situ hybridization (ISH). In the whole retina, miR-126-5p was detected in the endothelium (Fig. 1B, arrowheads), as expected from previous reports (Poissonnier et al., 2014b), and was comparable to miR-126-3p expression (Fig. 1A,B) (Poissonnier et al., 2014a). However, miR-126-5p was also detected in cells outside blood vessels in both the vascularized and the non-vascularized areas of the retina (Fig. 1B, arrows). At higher magnifications, miR-126-5p was observed in round-shaped cells of the inner layer of the retina (Fig. 1C), which appeared to be RGCs as shown by whole retina immunostaining for the RGC-specific factor POU domain class 4 transcription factor Brn3A (Pou4f1) (Geeraerts et al., 2016) (Fig. 1D). Combined miR-126-5p ISH with Brn3A and blood vessel collagen IV immunostainings in whole retina confirmed the expression of miR-126-5p both in endothelial cells and in RGCs (Fig. 1E). The restricted expression pattern of miR-126-5p to the inner vasculature and to RGCs was also observed in retina transversal sections (Fig. 1F,G). To corroborate these observations, miR-126-5p levels were measured in endothelial and neuronal cells purified from mouse retina. Retinal endothelial cells were isolated using a magnetically labeled anti-CD31 (Pecam1) antibody. Neurons were negatively purified from dissociated retina after depletion of non-neuronal cells using a cocktail of biotin-conjugated antibodies. Enrichment for each cell type was confirmed by qPCR (Fig. S1) and miR-126-5p was indeed detected in neurons, though at lower levels than in endothelial cells (Fig. 1H).
Endothelial and neuronal expression of miR-126-5p in postnatal retina. (A) Left: miR-126-3p detection by in situ hybridization (black) of whole P6 retina. Middle: immunostaining for type IV collagen (Coll IV) to visualize the vasculature. Right: merged images. miR-126-3p is expressed in endothelial cells (arrowheads). (B) Left: miR-126-5p detection by in situ hybridization (black) in whole P6 retina. Middle: immunostaining for type IV collagen (Coll IV). Right: merged images. miR-126-5p is expressed in endothelial cells (arrowheads) and in cells outside the vasculature in the non-vascularized (black arrows) or in the vascularized area (white arrows) of the retina. (C) Higher magnification of miR-126-5p in situ hybridization (black) and type IV collagen immunostaining (blue) of P9 whole-mount retina. (D) Immunolocalization of the RGC marker Brn3A (green) in whole-mount P6 retina immunostained for type IV collagen (Coll IV, blue). (E) Immunolocalization of the RGC marker Brn3A (green) and the blood vessel marker type IV collagen (Coll IV, blue) in whole-mount P9 retina previously treated for miR-126-5p detection by in situ hybridization (black). miR-126-5p is detected in endothelial cells (arrowhead) and in RGCs (arrow). (F) miR-126-5p detection in P9 retina transversal sections (purple). miR-126-5p is exclusively detected in the RGC layer of the retina. (G) Higher magnification of the RGC layer showing the endothelial cells (arrowheads) and RGCs (arrows). miR-126-5p is detected both in RGCs and in endothelial cells. (H) RT-qPCR analysis of miR-126-5p levels in the whole retina (Retina cells), in purified retinal endothelial cells (CD31+ cells), and in purified retinal neurons (Neurons). Data are representative of three independent experiments; n=2 technical replicates.
Endothelial and neuronal expression of miR-126-5p in postnatal retina. (A) Left: miR-126-3p detection by in situ hybridization (black) of whole P6 retina. Middle: immunostaining for type IV collagen (Coll IV) to visualize the vasculature. Right: merged images. miR-126-3p is expressed in endothelial cells (arrowheads). (B) Left: miR-126-5p detection by in situ hybridization (black) in whole P6 retina. Middle: immunostaining for type IV collagen (Coll IV). Right: merged images. miR-126-5p is expressed in endothelial cells (arrowheads) and in cells outside the vasculature in the non-vascularized (black arrows) or in the vascularized area (white arrows) of the retina. (C) Higher magnification of miR-126-5p in situ hybridization (black) and type IV collagen immunostaining (blue) of P9 whole-mount retina. (D) Immunolocalization of the RGC marker Brn3A (green) in whole-mount P6 retina immunostained for type IV collagen (Coll IV, blue). (E) Immunolocalization of the RGC marker Brn3A (green) and the blood vessel marker type IV collagen (Coll IV, blue) in whole-mount P9 retina previously treated for miR-126-5p detection by in situ hybridization (black). miR-126-5p is detected in endothelial cells (arrowhead) and in RGCs (arrow). (F) miR-126-5p detection in P9 retina transversal sections (purple). miR-126-5p is exclusively detected in the RGC layer of the retina. (G) Higher magnification of the RGC layer showing the endothelial cells (arrowheads) and RGCs (arrows). miR-126-5p is detected both in RGCs and in endothelial cells. (H) RT-qPCR analysis of miR-126-5p levels in the whole retina (Retina cells), in purified retinal endothelial cells (CD31+ cells), and in purified retinal neurons (Neurons). Data are representative of three independent experiments; n=2 technical replicates.
Repression of miR-126-5p reduces retina vascular density
To characterize the functions of miR-126-5p during retina postnatal development independently of those of miR-126-3p, miR-126-5p levels were specifically reduced in the retina of 3-day-old pups [postnatal day (P) 3] by injecting a cholesterol-conjugated miR-126-5p locked-nucleic acid inhibitor. This experimental approach provided a much more specific knockdown of miR-126-5p compared with miR-126 knockout mice (Wang et al., 2008) as the miR-126-5p inhibitor strongly reduced miR-126-5p expression (normalized with U6 snRNA) without affecting miR-126-3p levels 3 days after retinal injection (P6) (Fig. 2A). When RT-qPCR analysis was performed on endothelial and neuronal cells isolated from retina treated with anti-miR-126-5p, miR-126-5p was downregulated in both cell types (Fig. 2B,C).
Repression of miR-126-5p induces vascular defects in the retina. (A) RT-qPCR of miR-126-5p and miR-126-3p in retina of mouse eyes treated with the specific miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). Data are representative of three independent experiments, n=6 retinas/condition. (B,C) Quantification of miR-126-5p in endothelial cells (CD31+ cells; B) and in neurons (C) purified from P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). Two independent experiments were performed, n=2 technical replicates; results obtained with the purest enriched cell population are shown. (D) CD31 immunostaining (red) of the vascular network in P6 retina in which miR-126-5p was repressed (αmiR-126-5p) or not (αmiRCtrl). n=6 retinas/condition. (E) Percentage of retina total vascularized area measured after repressing miR-126-5p (αmiR-126-5p) or in the control condition (αmiRCtrl). Percentages were calculated by dividing the vascularized area by the total area of the retina. n=6 independent biological samples. Data are representative of two independent experiments. (F) Vascular network density measured by counting intercapillary spaces in miR-126-5p-repressed P6 retina (αmiR-126-5p) or in controls (αmiRCtrl). n=5 retinas per condition. Data are representative of two independent experiments. (G) Endothelial cell apoptosis analyzed by cleaved-caspase 3 (cleaved Casp3, green), and CD31 (red) co-immunostaining of flat-mount P6 retina repressed for miR-126-5p (αmiR-126-5p) or in the control condition (αmiRCtrl). Lower panels show higher magnifications. n=4 retinas/condition. (H) Apoptotic vessel length measured per field in P6 miR-126-5p-repressed retina (αmiR-126-5p) or in the control condition (αmiRCtrl). n=4 retinas/condition. (I) Percentages of apoptotic HUVECs analyzed by flow cytometry of annexin V-positive/propidium iodide-negative cells after transfection of microRNA inhibitors (αmiR-126-5p or αmiRCtrl) and cultured in control (left panel) or serum-depleted medium (right panel). n=3 independent biological samples corresponding to three different HUVEC lots. Data are representative of two independent experiments. *P<0.05; **P<0.01; ns, not significant.
Repression of miR-126-5p induces vascular defects in the retina. (A) RT-qPCR of miR-126-5p and miR-126-3p in retina of mouse eyes treated with the specific miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). Data are representative of three independent experiments, n=6 retinas/condition. (B,C) Quantification of miR-126-5p in endothelial cells (CD31+ cells; B) and in neurons (C) purified from P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). Two independent experiments were performed, n=2 technical replicates; results obtained with the purest enriched cell population are shown. (D) CD31 immunostaining (red) of the vascular network in P6 retina in which miR-126-5p was repressed (αmiR-126-5p) or not (αmiRCtrl). n=6 retinas/condition. (E) Percentage of retina total vascularized area measured after repressing miR-126-5p (αmiR-126-5p) or in the control condition (αmiRCtrl). Percentages were calculated by dividing the vascularized area by the total area of the retina. n=6 independent biological samples. Data are representative of two independent experiments. (F) Vascular network density measured by counting intercapillary spaces in miR-126-5p-repressed P6 retina (αmiR-126-5p) or in controls (αmiRCtrl). n=5 retinas per condition. Data are representative of two independent experiments. (G) Endothelial cell apoptosis analyzed by cleaved-caspase 3 (cleaved Casp3, green), and CD31 (red) co-immunostaining of flat-mount P6 retina repressed for miR-126-5p (αmiR-126-5p) or in the control condition (αmiRCtrl). Lower panels show higher magnifications. n=4 retinas/condition. (H) Apoptotic vessel length measured per field in P6 miR-126-5p-repressed retina (αmiR-126-5p) or in the control condition (αmiRCtrl). n=4 retinas/condition. (I) Percentages of apoptotic HUVECs analyzed by flow cytometry of annexin V-positive/propidium iodide-negative cells after transfection of microRNA inhibitors (αmiR-126-5p or αmiRCtrl) and cultured in control (left panel) or serum-depleted medium (right panel). n=3 independent biological samples corresponding to three different HUVEC lots. Data are representative of two independent experiments. *P<0.05; **P<0.01; ns, not significant.
Repressing miR-126-5p induced morphological changes in the global vasculature of the retina. Although the total vascular area covering the retina was not significantly affected following miR-126-5p inhibition (Fig. 2D,E), the density of the vascular network was strongly reduced, as quantified by counting intercapillary spaces (Fig. 2D-F). This defective vascular branching correlated with increased apoptosis of retinal capillary endothelial cells, as revealed by cleaved-caspase 3 staining of whole retina (Fig. 2G,H). Of note, when miR-126-3p was specifically repressed in the retina after injection of a cholesterol-conjugated miR-126-3p locked-nucleic acid inhibitor at P3, vascular development was delayed (Fig. S2) at P6, as observed previously (Wang et al., 2008), but endothelial cell apoptosis was not different from control (Fig. S2), strengthening the specific negative impact of miR-126-5p repression on endothelial cell survival. Unexpectedly, repression of miR-126-5p in primary endothelial human umbilical vein endothelial cells (HUVECs) in vitro did not induce apoptosis when the cells were cultured either in complete or in serum-depleted medium (Fig. 2I), suggesting that the endothelial apoptosis observed in the retina was probably not due to the repression of miR-126-5p in endothelial cells but rather to an indirect effect of miR-126-5p inhibition. Thus, the contribution of VEGF levels, astrocyte network and perivascular cell covering were assessed in the miR-126-5p-repressed retina, as they all contribute to endothelial cell survival in this organ (Fruttiger, 2007; Scott et al., 2010). None of these parameters appeared to be modified when miR-126-5p was repressed in mouse retina, excluding them from being strongly involved in the observed endothelial cell apoptosis (Fig. 3A,B, Fig. S3).
Repression of miR-126-5p upregulates IL1β and Sema3A in the retina. (A) Dosage of VEGF protein by ELISA in retina of P6 mouse eyes treated with miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). Data are representative of two independent experiments, n=6 retinas/condition. ns, not significant. (B) Upper six panels: immunostaining of the vascular plexus (CD31, green) and the astrocytic network (GFAP, red) of retina from P6 mouse eyes treated with miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). n=6 biological replicates. Lower six panels: co-immunostaining of endothelial cells (CD31, red) and pericytes (NG2, green) in retina from P6 mouse eyes treated with miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). n=6 biological samples. (C) Eyes of P3 OF1 wild-type mice (WT, top), lymphocyte-deficient SCID mice (SCID, middle), and microglia/macrophage-deficient Csf1op/op mice (Op/op, bottom) were injected with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl) and retina co-immunostained at P6 for CD45 (red) and blood vessel collagen IV (Coll IV, blue). WT: n=6 retinas; SCID: n=4 retinas; Op/op: n=3 retinas. (D) Left: RT-qPCR analyses of Il1b (top) and Sema3a (bottom) transcripts in P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). Data are representative of three independent experiments, n=6 retinas/condition. Right: dosage of IL1β (upper panel) and Sema3A (lower panel) proteins by ELISA in P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). Data are representative of three independent experiments, n=6 retinas/condition. (E) Immunolocalization of IL1β in microglia/macrophage cells by co-immunostaining of IL1β (red) and the microglial marker Iba1 (green) in transversal sections of P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). n=4 retinas/condition. (F) Immunolocalization of Sema3A in RGCs by co-immunostaining of Sema3A (red) and the RGC marker Brn3A (green) in transversal sections of P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). n=5 retinas/condition. *P<0.05; ns, not significant.
Repression of miR-126-5p upregulates IL1β and Sema3A in the retina. (A) Dosage of VEGF protein by ELISA in retina of P6 mouse eyes treated with miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). Data are representative of two independent experiments, n=6 retinas/condition. ns, not significant. (B) Upper six panels: immunostaining of the vascular plexus (CD31, green) and the astrocytic network (GFAP, red) of retina from P6 mouse eyes treated with miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). n=6 biological replicates. Lower six panels: co-immunostaining of endothelial cells (CD31, red) and pericytes (NG2, green) in retina from P6 mouse eyes treated with miR-126-5p inhibitor (αmiR-126-5p) or control (αmiRCtrl). n=6 biological samples. (C) Eyes of P3 OF1 wild-type mice (WT, top), lymphocyte-deficient SCID mice (SCID, middle), and microglia/macrophage-deficient Csf1op/op mice (Op/op, bottom) were injected with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl) and retina co-immunostained at P6 for CD45 (red) and blood vessel collagen IV (Coll IV, blue). WT: n=6 retinas; SCID: n=4 retinas; Op/op: n=3 retinas. (D) Left: RT-qPCR analyses of Il1b (top) and Sema3a (bottom) transcripts in P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). Data are representative of three independent experiments, n=6 retinas/condition. Right: dosage of IL1β (upper panel) and Sema3A (lower panel) proteins by ELISA in P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). Data are representative of three independent experiments, n=6 retinas/condition. (E) Immunolocalization of IL1β in microglia/macrophage cells by co-immunostaining of IL1β (red) and the microglial marker Iba1 (green) in transversal sections of P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). n=4 retinas/condition. (F) Immunolocalization of Sema3A in RGCs by co-immunostaining of Sema3A (red) and the RGC marker Brn3A (green) in transversal sections of P6 retina treated with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl). n=5 retinas/condition. *P<0.05; ns, not significant.
Repression of miR-126-5p induces interleukin 1β and Sema3A upregulation in retina
Because lymphocytes and macrophages induce endothelial cell apoptosis respectively in the retina (Ishida et al., 2003) and in the hyaloid vasculature (Lobov et al., 2005), we investigated whether immune cells were recruited after miR-126-5p inhibition. Retina from immunocompetent OF1 mice were therefore stained for the pan-leukocyte marker CD45 (Ptprc) after treatment with the anti-miR-126-5p. There was a strong increase in the number of CD45+ cells, which accumulated around the vascular network in miR-126-5p-repressed retina but not within the blood vessels themselves (Fig. 3C). To identify the recruited immune cells, the same experiments were performed in lymphocyte-deficient SCID mice and in microglia/macrophage-deficient Csf1op/op mice (Fig. S4). There was a similar accumulation of CD45+ cells after miR-126-5p repression in the retina of SCID mice as that observed in OF1 mice (Fig. 3C), thus excluding lymphocyte recruitment. However, CD45+ cells did not accumulate in Csf1op/op mice upon miR-126-5p repression (Fig. 3C), indicating that the immune cells recruited in wild-type mice injected with miR-126-5p inhibitors were indeed microglia/macrophages.
In the retina, myeloid cells play opposite roles in blood vessel development, depending on the biological context. During physiological development, microglia/macrophages favor vascular branching (Fantin et al., 2010; Kubota et al., 2009) in the inner vascular plexus whereas during the vaso-obliterative phase of retinopathy of prematurity, microglia/macrophages induce endothelial programmed cell death by releasing inflammatory cytokines such as IL1β (Rivera et al., 2013). Here, IL1β transcript and protein levels were actually upregulated in retina following miR-126-5p repression (Fig. 3D). Furthermore, after immunostaining of retina sections, IL1β was detected in microglia/macrophages, with higher levels in retina treated with anti-miR-126-5p than in control retina (Fig. 3E). During oxygen-induced retinopathy, microglial-secreted IL1β generates signaling events in RGCs, which consequently secrete Sema3A and induce endothelial cell apoptosis (Rivera et al., 2013). Interestingly, repression of miR-126-5p in our retina model also induced an upregulation of Sema3A transcripts and proteins (Fig. 3D), exclusively in RGCs, as illustrated by Sema3a transcript level analysis performed in neuronal and endothelial cells purified from miR-126-5p-repressed retina (Fig. S5) and in retina sections immunostained for Sema3A (Fig. 3F). Altogether, these results suggested a possible contribution of microglial IL1β and RGC-derived Sema3A in the retinal vascular phenotype observed following miR-126-5p repression.
Blocking Sema3A in miR-126-5p-repressed retina rescues the vascular development
As Sema3A is known to induce endothelial cell apoptosis in vitro (Guttmann-Raviv et al., 2007) and as RGC-secreted Sema3A promotes endothelial cell death in oxygen-treated retina (Rivera et al., 2013), the contribution of this factor to the vascular defects observed after miR-126-5p repression was assessed by injecting a Sema3A-blocking antibody together with miR-126-5p inhibitor in mouse retina. Remarkably, the vascular defects induced when repressing miR-126-5p were abolished in mice treated with the Sema3A-blocking antibody, and a normal vascular density was restored in these retinas (Fig. 4A,C). This demonstrated that the vascular phenotype observed following inhibition of miR-126-5p in retina was actually due to overexpression of Sema3A. Of note, blocking Sema3A signaling concomitantly with miR-126-5p inhibition prevented Sema3a transcript upregulation in retina (Fig. 4B).
The vascular defects in retina treated with anti-miR-126-5p are due to Sema3A upregulation. (A) P3 eyes were co-injected with anti-miR-126-5p and with a control antibody (αmiR-126-5p/IgG) or an anti-Sema3A antibody (αmiR-126-5p/αSema3A) or with control anti-miR and a control antibody (αmiRCtrl/IgG). Retinas were immunostained at P6 for blood vessel collagen IV (Coll IV, red). n=6 retinas/condition. (B) RT-qPCR of Il1b and Sema3a transcripts in retina from mice injected with miR-126-5p inhibitor and control antibody (αmiR-126-5p/IgG) or with a Sema3A-blocking antibody (αmiR-126-5p/αSema3A) in the vitreous and compared with eyes injected with control anti-miR and control IgG (αmiRCtrl/IgG). Data are representative of two independent experiments, n=6 retinas/condition. (C) Vascular network density assessed by counting intercapillary spaces in retina injected with miR-126-5p inhibitor and with a control antibody (αmiR-126-5p/IgG) or with a Sema3A-blocking antibody (αmiR-126-5p/αSema3A) in the vitreous and compared with eyes injected with control anti-miR and control IgG (αmiRCtrl/IgG). Data are representative of two independent experiments, n=6 animals per condition, one field was analyzed per retina. (D) Eyes were injected with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl) at P3 and mice were treated with PBS or with an antagonist of the IL-1 receptor (IL-1Ra). Three days later, retinas were immunostained for collagen IV (Coll IV, red). n=7 retinas/condition. (E) RT-qPCR of Il1b and Sema3a transcripts in retina treated with anti-miR-126-5p or with the control microRNA inhibitor from mice treated with PBS (αmiR-126-5p/PBS or αmiRCtrl/PBS) or with the IL1 receptor antagonist (αmiR-126-5p/IL-1Ra). n=8 animals per condition. Data are representative of two independent experiments. (F) Vascular network density assessed by counting intercapillary spaces in retina. Data are representative of two independent experiments. Two fields were analyzed/retina. n=7 animals per condition. *P<0.05; **P<0.01; ns, not significant.
The vascular defects in retina treated with anti-miR-126-5p are due to Sema3A upregulation. (A) P3 eyes were co-injected with anti-miR-126-5p and with a control antibody (αmiR-126-5p/IgG) or an anti-Sema3A antibody (αmiR-126-5p/αSema3A) or with control anti-miR and a control antibody (αmiRCtrl/IgG). Retinas were immunostained at P6 for blood vessel collagen IV (Coll IV, red). n=6 retinas/condition. (B) RT-qPCR of Il1b and Sema3a transcripts in retina from mice injected with miR-126-5p inhibitor and control antibody (αmiR-126-5p/IgG) or with a Sema3A-blocking antibody (αmiR-126-5p/αSema3A) in the vitreous and compared with eyes injected with control anti-miR and control IgG (αmiRCtrl/IgG). Data are representative of two independent experiments, n=6 retinas/condition. (C) Vascular network density assessed by counting intercapillary spaces in retina injected with miR-126-5p inhibitor and with a control antibody (αmiR-126-5p/IgG) or with a Sema3A-blocking antibody (αmiR-126-5p/αSema3A) in the vitreous and compared with eyes injected with control anti-miR and control IgG (αmiRCtrl/IgG). Data are representative of two independent experiments, n=6 animals per condition, one field was analyzed per retina. (D) Eyes were injected with anti-miR-126-5p (αmiR-126-5p) or control (αmiRCtrl) at P3 and mice were treated with PBS or with an antagonist of the IL-1 receptor (IL-1Ra). Three days later, retinas were immunostained for collagen IV (Coll IV, red). n=7 retinas/condition. (E) RT-qPCR of Il1b and Sema3a transcripts in retina treated with anti-miR-126-5p or with the control microRNA inhibitor from mice treated with PBS (αmiR-126-5p/PBS or αmiRCtrl/PBS) or with the IL1 receptor antagonist (αmiR-126-5p/IL-1Ra). n=8 animals per condition. Data are representative of two independent experiments. (F) Vascular network density assessed by counting intercapillary spaces in retina. Data are representative of two independent experiments. Two fields were analyzed/retina. n=7 animals per condition. *P<0.05; **P<0.01; ns, not significant.
As miR-126-5p is expressed in RGCs (Fig. 1B), overexpression of Sema3A in these cells (Fig. 3F, Fig. S5) after miR-126-5p repression could be due either to a direct function of miR-126-5p in RGCs or to increased IL1β signaling in these cells subsequent to upregulation of this factor in microglia/macrophages (Fig. 3D,E). To address the latter hypothesis, IL1β signaling was repressed in mouse pups by injecting an IL1 receptor antagonist (IL-1Ra, anakinra) subcutaneously for 2 days following miR-126-5p inhibition. The blockage of IL1β in these conditions was confirmed as the induced IL1β overexpression was no longer detected (Fig. 4E), as previously described (Rivera et al., 2013) and the upregulation of MCP-1 (Ccl2), a downstream gene of IL1β signaling (Bian et al., 2001; Cuff et al., 2000) was abolished (Fig. S6). However, blocking IL1β signaling did not prevent Sema3A upregulation in retina where miR-126-5p had been repressed (Fig. 4E), indicating that Sema3A overexpression upon miR-126-5p repression was not dependent on IL1β signaling. More importantly, the vascular defects induced by the repression of miR-126-5p were not relieved by injecting IL-1Ra either (Fig. 4D,F), demonstrating that the vascular defects induced when repressing miR-126-5p did not depend on the observed IL1β overexpression.
Setd5 protection from miR-126-5p phenocopies miR-126-5p inhibition in retina
Altogether, the results described above pointed toward a possible control of Sema3A expression by miR-126-5p in RGCs. However, Sema3A is not a predicted target of miR-126-5p in the TargetScan, Pictar and Diana databases (Krek et al., 2005; Lewis et al., 2005; Maragkakis et al., 2009). We therefore analyzed the expression of reportedly known miR-126-5p target genes such as Setd5 (Poissonnier et al., 2014b), Dlk1 (Schober et al., 2014), Cxcl12 (Zhang et al., 2013), Mmp13 (Wu et al., 2014), Pthrp (Pthlh) (Zhou et al., 2014) and Mmp7 (Felli et al., 2013) in retina treated with anti-miR-126-5p. Setd5 and Mmp13 were the most upregulated transcripts following miR-126-5p repression, whereas Dlk1, Cxcl12, and Pthrp were less strongly regulated, and Mmp7 expression was not detected in either condition (Fig. 5A).
SetD5 is upregulated in miR-126-5p-repressed retina and expressed in RGCs. (A) RT-qPCR analysis of the indicated transcripts in retina treated with control (αmiRCtrl) or anti-miR-126-5p (αmiR-126-5p). Data are representative of two independent experiments, n=6 biological replicates. (B) RT-qPCR analysis of Mmp13 and Setd5 transcripts in the whole retina (Retina cells), in purified endothelial cells (CD31+ cells), or in neurons purified from whole retina (Neurons). Data are representative of three independent experiments, n=2 technical replicates. (C) Upper panels: Setd5 (black) in situ hybridization analysis in whole flat-mount retina with higher magnification after immunostaining for type IV collagen (Coll IV, red) to visualize the vascular network (right). n=3 retinas/experiment. Lower panels: β-galactosidase (βGal) staining of whole flat-mount retina from SetD5iLacZ/iLacZ P6 reporter mice with higher magnification after immunostaining for type IV collagen (Coll IV, red) (right). n=3 retinas/experiment. (D) Co-immunostaining of wild-type retina transversal sections for SetD5 (red) and for the RGC-specific marker Brn3A (green). INL, inner nuclear layer. n=3 retinas/experiment. *P<0.05; ns, not significant.
SetD5 is upregulated in miR-126-5p-repressed retina and expressed in RGCs. (A) RT-qPCR analysis of the indicated transcripts in retina treated with control (αmiRCtrl) or anti-miR-126-5p (αmiR-126-5p). Data are representative of two independent experiments, n=6 biological replicates. (B) RT-qPCR analysis of Mmp13 and Setd5 transcripts in the whole retina (Retina cells), in purified endothelial cells (CD31+ cells), or in neurons purified from whole retina (Neurons). Data are representative of three independent experiments, n=2 technical replicates. (C) Upper panels: Setd5 (black) in situ hybridization analysis in whole flat-mount retina with higher magnification after immunostaining for type IV collagen (Coll IV, red) to visualize the vascular network (right). n=3 retinas/experiment. Lower panels: β-galactosidase (βGal) staining of whole flat-mount retina from SetD5iLacZ/iLacZ P6 reporter mice with higher magnification after immunostaining for type IV collagen (Coll IV, red) (right). n=3 retinas/experiment. (D) Co-immunostaining of wild-type retina transversal sections for SetD5 (red) and for the RGC-specific marker Brn3A (green). INL, inner nuclear layer. n=3 retinas/experiment. *P<0.05; ns, not significant.
Interestingly, Setd5 was strongly expressed in isolated retinal neurons whereas Mmp13 expression was much lower in these cells (Fig. 5B). In situ hybridization analysis of Setd5 in wild-type retina showed that Setd5 was expressed in the round-shaped cells of the retina superficial layer (Fig. 5C), which looked like RGCs. This result was confirmed by β-galactosidase staining of retina isolated from SetD5iLacZ/iLacZ mice in which β-galactosidase expression was placed under the Setd5 promoter through an intronic insertion of a splice acceptor-IRES-LacZ cassette into the Setd5 gene. Those round-shaped cells were positively identified as RGCs after co-immunostaining for SetD5 and Brn3A in retina sections (Fig. 5D). Of note, SetD5 was also observed in the inner nuclear layer of the retina (Fig. 5D) but was not detected in the blood vessels (Fig. 5C).
To determine whether the vascular defects observed after miR-126-5p repression in retinas were actually due to a direct control of SetD5 expression by miR-126-5p, specific target blockers (TBs) were designed to protect Setd5 transcripts from miR-126-5p in vivo (Fig. S7). As expected from this design, injection of a cholesterol-conjugated TB-SetD5 in mouse eyes induced a specific upregulation of Setd5 (Fig. 6A) without affecting the expression of the other miR-126-5p target genes. Remarkably, injection of TB-SetD5 induced vascular density defects in retinas similar to those observed when injecting the anti-miR-126-5p, i.e. no effects on the global vascularized area but a marked decrease in the number of intercapillary spaces (Fig. 6B,C). TB-SetD5 also increased endothelial cell apoptosis in vivo (Fig. 6D) and upregulated Sema3A transcripts and protein compared with control non-targeting blocker (Fig. 6E,F). Of note, the injection in the subretinal space of another target blocker designed to protect Cxcl12 from miR-126-5p repression (TB-CXCL12) induced an upregulation of Cxcl12 transcripts without significantly affecting SetD5 expression. In addition, although this TB-CXCL12 induced vascular defects in the retina, such as the accumulation of endothelial cells at the vascular front (data not shown), it did not affect the vascular density or endothelial cell apoptosis in the retina (Fig. S8). Thus, Setd5 protection from miR-126-5p specifically phenocopied the repression of miR-126-5p in the retina, indicating that the vascular defects observed when repressing miR-126-5p in the retina occurred mostly through the endogenous regulation of SetD5 expression.
Vascular defects induced by protecting Setd5 from miR-126-5p. (A) RT-qPCR analysis of the miR-126-5p target genes Setd5, Dlk1, Cxcl12, Mmp13 and Pthrp in retina injected with a target blocker that protects Setd5 from miR-126-5p repression (TB-SetD5) or with a control target blocker (TB-Ctrl). Data are representative of two independent experiments, n=5 retinas/condition. (B) Top: percentage of vascularized area calculated by dividing the vascularized area by the total area of the retina injected with the target blocker protecting Setd5 from miR-126-5p repression (TB-SetD5) or with the control target blocker (TB-Ctrl). Data are representative of two independent experiments, n=6 retinas/condition. Bottom: Quantification of vascular network density by counting intercapillary spaces per field in retina treated with TB-SetD5 or TB-Ctrl. Data are representative of two independent experiments, n=6 animals per condition. (C) CD31 staining (red) of whole retina treated with TB-SetD5 or with TB-Ctrl. n=5 retinas/condition. (D) Images show co-immunostaining of apoptotic cells (cleaved-caspase 3, Casp3, green) and endothelial cells (CD31, red) in retina treated with TB-SetD5 or with TB-Ctrl. n=6 retinas/condition. Graph shows apoptotic vessel length measured per field in P6 retina treated with TB-SetD5 or with TB-Ctrl. n=4 retinas/condition. (E) Top: RT-qPCR analysis of Sema3a transcripts in retina treated with TB-SetD5 or with TB-Ctrl. Data are representative of four independent experiments, n=5 animals per condition. Bottom: Dosage of Sema3A protein by ELISA in retina treated with TB-SetD5 or with TB-Ctrl. Data are representative of two independent experiments, n=6 animals per condition. (F) Immunolocalization of Sema3A (Sema3A, red) and the RGC marker Brn3A (green) in sections of retina injected with the target blocker protecting Setd5 from miR-126-5p repression (TB-SetD5) or with the control target blocker (TB-Ctrl), n=5 retinas/condition. *P<0.05; **P<0.01; ns, not significant.
Vascular defects induced by protecting Setd5 from miR-126-5p. (A) RT-qPCR analysis of the miR-126-5p target genes Setd5, Dlk1, Cxcl12, Mmp13 and Pthrp in retina injected with a target blocker that protects Setd5 from miR-126-5p repression (TB-SetD5) or with a control target blocker (TB-Ctrl). Data are representative of two independent experiments, n=5 retinas/condition. (B) Top: percentage of vascularized area calculated by dividing the vascularized area by the total area of the retina injected with the target blocker protecting Setd5 from miR-126-5p repression (TB-SetD5) or with the control target blocker (TB-Ctrl). Data are representative of two independent experiments, n=6 retinas/condition. Bottom: Quantification of vascular network density by counting intercapillary spaces per field in retina treated with TB-SetD5 or TB-Ctrl. Data are representative of two independent experiments, n=6 animals per condition. (C) CD31 staining (red) of whole retina treated with TB-SetD5 or with TB-Ctrl. n=5 retinas/condition. (D) Images show co-immunostaining of apoptotic cells (cleaved-caspase 3, Casp3, green) and endothelial cells (CD31, red) in retina treated with TB-SetD5 or with TB-Ctrl. n=6 retinas/condition. Graph shows apoptotic vessel length measured per field in P6 retina treated with TB-SetD5 or with TB-Ctrl. n=4 retinas/condition. (E) Top: RT-qPCR analysis of Sema3a transcripts in retina treated with TB-SetD5 or with TB-Ctrl. Data are representative of four independent experiments, n=5 animals per condition. Bottom: Dosage of Sema3A protein by ELISA in retina treated with TB-SetD5 or with TB-Ctrl. Data are representative of two independent experiments, n=6 animals per condition. (F) Immunolocalization of Sema3A (Sema3A, red) and the RGC marker Brn3A (green) in sections of retina injected with the target blocker protecting Setd5 from miR-126-5p repression (TB-SetD5) or with the control target blocker (TB-Ctrl), n=5 retinas/condition. *P<0.05; **P<0.01; ns, not significant.
SetD5 controls Sema3A expression independently of its SET domain and interacts with BRD2
To evaluate whether SetD5 directly regulated the expression of Sema3A, HeLa and HEK293T (HEK) cells which spontaneously express both SetD5 and Sema3A, were silenced for SetD5 using two distinct siRNA (Fig. 7B, Fig. S9) and Sema3A expression was then measured. In both cell lines, either siRNA targeting SetD5 concomitantly repressed SETD5 and SEMA3A (Fig. 7A). In HEK cells, at least two SETD5 protein variants (size ranging between 135 and 260 kDa) are detected (Fig. 7B) but the larger variant is predominantly expressed (Fig. 7B). Therefore, the SETD5 sequence corresponding to the longer predicted protein (Fig. S10) and its mutated version (with deletion of the SET domain) were cloned and validated for expression of the corresponding protein (Fig. 7B). In rescue experiments, SEMA3A expression was restored in SetD5-silenced HEK cells after re-expression of wild-type SetD5 (Fig. 7C), confirming the specific contribution of SetD5 in the control of Sema3A expression. Furthermore, expression of the mutated version of SetD5 deleted of its SET domain (ΔSET; Fig. 7C) also rescued SEMA3A expression, indicating that the putative lysine methyltransferase activity of SetD5 was not involved in the regulation of Sema3A expression. The endogenous SETD5 protein was detected in vitro in HeLa cell nuclei (Fig. 7D) and a similar nuclear localization was observed when overexpressing the wild-type or ΔSET SetD5 variants (data not shown), indicating that SetD5 controls Sema3A expression in the cell nuclear compartment.
SetD5 regulates Sema3A expression. (A) RT-qPCR analysis of SETD5 and SEMA3A transcripts in HeLa or HEK cells transfected with two distinct siRNA directed against SetD5. Data are representative of two independent experiments performed on three biological replicates. (B) Left: western blot analysis of HEK cells repressed for SetD5 using two distinct siRNA (#1 and #2). Different SETD5 isoforms are detected (molecular weight between 135 and 260 kDa). The most strongly expressed SETD5 isoform has the higher apparent molecular weight. All isoforms are strongly downregulated in SetD5-silenced HEK cells. ns, non-specific band. Top right: schematic of wild-type SetD5 (SetD5) and SET domain-deleted SetD5 (ΔSET) constructs. Bottom right: western blot analysis of HEK cells transfected with pcDNA-IRES-GFP control plasmid (pMock), with HA-tagged wild-type SetD5-IRES-GFP plasmid (pSetD5), or with HA-tagged ΔSET-SetD5-IRES-GFP (pΔSET). Anti-HA antibody was used to detect the overexpressed SetD5 proteins. Data are representative of two independent experiments. (C) Left: RT-qPCR analysis of SETD5 and SEMA3A transcripts in SetD5-silenced HEK cells (siSetD5) transfected with pMock or rescued with wild-type SetD5 expression plasmid pSetD5. Right: RT-qPCR analysis of SETD5 and SEMA3A transcripts in SetD5-silenced HEK (siSetD5) transfected with pMock, or rescued with the SET domain-deleted expression plasmid pΔSET. Data are representative of two independent experiments, n=3 biological replicates. ns, not significant. (D) Immunolocalization of SETD5 in HeLa cells transfected with control siRNA (siCtrl) or with SetD5 siRNA (siSetD5) and counterstained with DAPI. SETD5 is detected in the nucleus. (E) Co-immunoprecipitation (IP) of SETD5 with BRD2 in HEK cells transfected with pSetD5-HA and a pcDNA3 vector coding for FLAG-tagged BRD2 (pBRD2-Flag) and analyzed by western blot (αHA). Input: 2% of total protein extract. Data are representative of four independent experiments. (F) Co-immunoprecipitation (IP) of endogenous SETD5 with BRD2 in HEK cells and analyzed by western blot using anti-SetD5 and anti-BRD2 antibodies. Input: 2% of total protein extract. (G) Co-immunoprecipitation (IP) of endogenous SetD5 with BRD2 in P6 retina and analyzed by western blot using anti-SetD5 and anti-BRD2 antibodies. Input: 1% of total protein extract. (H) Retina cryosections stained for BRD2 (green), the RGC marker Brn3A (red), and DAPI (blue). Higher magnification of the RGC layer is shown on the right. RGCs stain positive for BRD2. *P<0.05, **P<0.01.
SetD5 regulates Sema3A expression. (A) RT-qPCR analysis of SETD5 and SEMA3A transcripts in HeLa or HEK cells transfected with two distinct siRNA directed against SetD5. Data are representative of two independent experiments performed on three biological replicates. (B) Left: western blot analysis of HEK cells repressed for SetD5 using two distinct siRNA (#1 and #2). Different SETD5 isoforms are detected (molecular weight between 135 and 260 kDa). The most strongly expressed SETD5 isoform has the higher apparent molecular weight. All isoforms are strongly downregulated in SetD5-silenced HEK cells. ns, non-specific band. Top right: schematic of wild-type SetD5 (SetD5) and SET domain-deleted SetD5 (ΔSET) constructs. Bottom right: western blot analysis of HEK cells transfected with pcDNA-IRES-GFP control plasmid (pMock), with HA-tagged wild-type SetD5-IRES-GFP plasmid (pSetD5), or with HA-tagged ΔSET-SetD5-IRES-GFP (pΔSET). Anti-HA antibody was used to detect the overexpressed SetD5 proteins. Data are representative of two independent experiments. (C) Left: RT-qPCR analysis of SETD5 and SEMA3A transcripts in SetD5-silenced HEK cells (siSetD5) transfected with pMock or rescued with wild-type SetD5 expression plasmid pSetD5. Right: RT-qPCR analysis of SETD5 and SEMA3A transcripts in SetD5-silenced HEK (siSetD5) transfected with pMock, or rescued with the SET domain-deleted expression plasmid pΔSET. Data are representative of two independent experiments, n=3 biological replicates. ns, not significant. (D) Immunolocalization of SETD5 in HeLa cells transfected with control siRNA (siCtrl) or with SetD5 siRNA (siSetD5) and counterstained with DAPI. SETD5 is detected in the nucleus. (E) Co-immunoprecipitation (IP) of SETD5 with BRD2 in HEK cells transfected with pSetD5-HA and a pcDNA3 vector coding for FLAG-tagged BRD2 (pBRD2-Flag) and analyzed by western blot (αHA). Input: 2% of total protein extract. Data are representative of four independent experiments. (F) Co-immunoprecipitation (IP) of endogenous SETD5 with BRD2 in HEK cells and analyzed by western blot using anti-SetD5 and anti-BRD2 antibodies. Input: 2% of total protein extract. (G) Co-immunoprecipitation (IP) of endogenous SetD5 with BRD2 in P6 retina and analyzed by western blot using anti-SetD5 and anti-BRD2 antibodies. Input: 1% of total protein extract. (H) Retina cryosections stained for BRD2 (green), the RGC marker Brn3A (red), and DAPI (blue). Higher magnification of the RGC layer is shown on the right. RGCs stain positive for BRD2. *P<0.05, **P<0.01.
To establish how SetD5 controls Sema3a gene expression, we searched for proteins interacting with SetD5 using a yeast two-hybrid screen of a retina cDNA library and SetD5 as bait. The nuclear double bromodomain protein BRD2 was identified as a partner for SetD5. BRD2 is a member of the BET family of proteins, which share homology in their double bromodomain and their extraterminal domain (Belkina and Denis, 2012). BRD2 is a transcriptional regulator involved in multiprotein transcription complexes that associates with histone acetylated lysines (Belkina and Denis, 2012), leading to its recruitment onto chromatin (LeRoy et al., 2008). The interaction of SetD5 with BRD2 was confirmed by co-immunoprecipitation using HEK cells transfected with HA-tagged SetD5 and Flag-tagged BRD2 constructs (Fig. 7E) and by co-immunoprecipitation of HEK endogenous BRD2 and SetD5 proteins (Fig. 7F). The interaction of SetD5 with BRD2 was also confirmed in vivo by co-immunoprecipitation of BRD2 and SetD5 proteins from retina extracts (Fig. 7G). The detection of BRD2 in RGCs (Fig. 7H) in vivo, where SetD5 was colocalized, further supports the fact that these two proteins actually interact in these cells. This prompted us to check whether BRD2 was involved in the regulation of Sema3A expression by SetD5.
Both SetD5 and BRD2 bind the Sema3a promoter and SetD5 requires BRD2 to control Sema3A expression
To evaluate the contribution of both SetD5 and BRD2 to Sema3A expression regulation, the presence of SetD5 and BRD2 on the Sema3a promoter was investigated using chromatin immunoprecipitation. As BRD2 preferentially binds to acetylated histone H4 (Belkina and Denis, 2012), we scanned three regions of the human SEMA3A promoter corresponding to the transcription start site, and −1.2 kb and −1.7 kb upstream regions for the presence of this modified histone. Acetylated histone H4 levels were high at the transcription start site and at the −1.2 kb region of the SEMA3A promoter (Fig. 8A). As expected, BRD2 was detected bound to the same regions and at levels proportionally similar to acetylated histone H4. SetD5 also bound to these regions of the human SEMA3A promoter (Fig. 8A). The specificity of binding was confirmed after SetD5 silencing in HEK cells (Fig. S11). In the retina, the presence of BRD2 and SetD5 on the mouse Sema3a promoter was confirmed by chromatin immunoprecipitation (Fig. 8B) and repression of miR-126-5p in the retina increased the recruitment of SetD5 on the Sema3a promoter compared with control (Fig. 8C). SetD5 and BRD2 therefore co-occupied the same regions of the Sema3a locus in vitro and in vivo, suggesting that both proteins could collaborate in the regulation of Sema3A expression. Interestingly, when HEK cells were silenced for BRD2 using RNA interference or when they were treated with the JQ1 inhibitor of interactions of bromodomain proteins with acetylated histones (Filippakopoulos et al., 2010), SEMA3A expression was downregulated (Fig. 8D,E), indicating that BRD2 contributes to the regulation of Sema3A expression, possibly with SetD5. To confirm this latter point, SEMA3A expression was analyzed in cells silenced for both SetD5 and BRD2 and transfected with the SetD5 expression construct. SEMA3A expression was reduced by the inactivation of SetD5 but was not restored by the sole re-expression of SetD5 in the absence of BRD2 (Fig. 8F). In contrast, when expression of both SetD5 and BRD2 was rescued in silenced cells, SEMA3A expression was then restored (Fig. 8G), thus indicating that BRD2 is indeed necessary for the regulation of Sema3A by SetD5.
SetD5 requires BRD2 to control Sema3A expression. (A) Top: Schematic of the human SEMA3A region corresponding to 2 kb upstream of the transcription start site (TSS). Primer sets used in ChIP experiments are depicted with arrows. Bottom: Levels of acetylated histone H4 (IP Ac-H4), BRD2 (IP BRD2) and SetD5 (IP SetD5) detected on the SEMA3A promoter regions in HEK cells. Regions (A, B or C) were amplified by qPCR and normalized to the control IgG sample. (B) Top: Schematic of the mouse Sema3a region corresponding to 1.5 kb upstream of the TSS. Primer sets used in ChIP experiments are depicted with arrows. Bottom: Levels of acetylated histone H4 (IP Ac-H4), BRD2 (IP BRD2) and SetD5 (IP SetD5) detected on the Sema3a promoter regions in P6 retina. Regions (A′or B′) were amplified by qPCR and normalized to the control IgG sample. (C) Levels of SetD5 (IP SetD5) detected on the Sema3a promoter regions in P6 retina treated with miR-126-5p inhibitor or control. Regions (A′or B′) were amplified by qPCR and normalized to the control IgG sample. (D) Quantification by RT-qPCR of BRD2 and SEMA3A expression in HEK cells treated with DMSO or with increasing concentrations of the inactive bromodomain protein inhibitor JQ1− or of the active bromodomain protein inhibitor JQ1+. Data are representative of two independent experiments, n=2 technical replicates. (E) Quantification by RT-qPCR of BRD2 and SEMA3A expression in HEK cells silenced for BRD2 (siBRD2) or transfected with siRNA control (siCtrl). Data are representative of two independent experiments, n=3 biological replicates. (F) RT-qPCR of SETD5 and SEMA3A transcripts in HEK cells silenced for SetD5 (siSetD5) and transfected with control plasmid (pMock) or with SetD5-expressing vector (pSetD5) and with or without BRD2 siRNA. Data are representative of two independent experiments, n=3 biological replicates. (G) RT-qPCR of SETD5, BRD2 and SEMA3A transcripts in HEK cells silenced for SetD5 (siSetD5), BRD2 (siBRD2) or both and transfected with SetD5-expressing vector (pSetD5) and with or without a BRD2-expressing vector. Data are representative of four independent experiments, n=3 biological replicates. (H) Model for miR-126-5p function in the retina during vascular development. In normal conditions (left), miR-126-5p, which is expressed in retinal ganglion cells (RGCs), represses the expression of SetD5 in these cells leading to low expression levels of Sema3A and protection of endothelial cells (ECs) of the inner vascular plexus. After miR-126-5p repression (right), SetD5 is overexpressed in RGCs leading to the upregulation of Sema3A. High Sema3A levels induce endothelial cell death leading to impaired retinal inner vasculature. *P<0.05; **P<0.01; ns, not significant.
SetD5 requires BRD2 to control Sema3A expression. (A) Top: Schematic of the human SEMA3A region corresponding to 2 kb upstream of the transcription start site (TSS). Primer sets used in ChIP experiments are depicted with arrows. Bottom: Levels of acetylated histone H4 (IP Ac-H4), BRD2 (IP BRD2) and SetD5 (IP SetD5) detected on the SEMA3A promoter regions in HEK cells. Regions (A, B or C) were amplified by qPCR and normalized to the control IgG sample. (B) Top: Schematic of the mouse Sema3a region corresponding to 1.5 kb upstream of the TSS. Primer sets used in ChIP experiments are depicted with arrows. Bottom: Levels of acetylated histone H4 (IP Ac-H4), BRD2 (IP BRD2) and SetD5 (IP SetD5) detected on the Sema3a promoter regions in P6 retina. Regions (A′or B′) were amplified by qPCR and normalized to the control IgG sample. (C) Levels of SetD5 (IP SetD5) detected on the Sema3a promoter regions in P6 retina treated with miR-126-5p inhibitor or control. Regions (A′or B′) were amplified by qPCR and normalized to the control IgG sample. (D) Quantification by RT-qPCR of BRD2 and SEMA3A expression in HEK cells treated with DMSO or with increasing concentrations of the inactive bromodomain protein inhibitor JQ1− or of the active bromodomain protein inhibitor JQ1+. Data are representative of two independent experiments, n=2 technical replicates. (E) Quantification by RT-qPCR of BRD2 and SEMA3A expression in HEK cells silenced for BRD2 (siBRD2) or transfected with siRNA control (siCtrl). Data are representative of two independent experiments, n=3 biological replicates. (F) RT-qPCR of SETD5 and SEMA3A transcripts in HEK cells silenced for SetD5 (siSetD5) and transfected with control plasmid (pMock) or with SetD5-expressing vector (pSetD5) and with or without BRD2 siRNA. Data are representative of two independent experiments, n=3 biological replicates. (G) RT-qPCR of SETD5, BRD2 and SEMA3A transcripts in HEK cells silenced for SetD5 (siSetD5), BRD2 (siBRD2) or both and transfected with SetD5-expressing vector (pSetD5) and with or without a BRD2-expressing vector. Data are representative of four independent experiments, n=3 biological replicates. (H) Model for miR-126-5p function in the retina during vascular development. In normal conditions (left), miR-126-5p, which is expressed in retinal ganglion cells (RGCs), represses the expression of SetD5 in these cells leading to low expression levels of Sema3A and protection of endothelial cells (ECs) of the inner vascular plexus. After miR-126-5p repression (right), SetD5 is overexpressed in RGCs leading to the upregulation of Sema3A. High Sema3A levels induce endothelial cell death leading to impaired retinal inner vasculature. *P<0.05; **P<0.01; ns, not significant.
DISCUSSION
This study reports, for the first time, the functions of miR-126-5p in neuronal cells and its role in endothelial cell survival during the establishment of the retinal vasculature through the regulation of SetD5 and Sema3A (Fig. 8H). Our results strengthen the key roles played by neurons during vascular development of the postnatal retina. It was previously known that neuronal cells promoted retinal vascularization through the angiogenic factor VEGF, and two independent mouse genetic models demonstrated that mice lacking RGCs had severe vascular defects in the retina. Indeed, Math5 (also known as Atoh7) null mice, which are almost completely devoid of RGCs (Edwards et al., 2012), and transgenic mice in which RGCs are induced to degenerate (Sapieha et al., 2008), have an avascular retina. Furthermore, although RGCs and astrocytes are the main source of VEGF in the postnatal retina, the astrocyte-specific deletion of VEGF surprisingly did not strongly affect retinal vascular development (Scott et al., 2010). Although one cannot exclude a VEGF compensation by RGCs in astrocytic-specific VEGF-deficient mice, this study suggested that VEGF produced by RGCs was probably crucial for retina vascularization.
Here, we demonstrated the protective function of RGCs expressing miR-126-5p toward endothelial cells during development of the retina vasculature. Recently, vascular branching defects were described in Mir126−/− retina in which both miR-126-3p and miR-126-5p had been knocked out (Zhou et al., 2016) but the molecular mechanisms leading to this phenotype were not fully elucidated. From our present work, it is probable that the observed reduced vascular density was due in part to miR-126-5p deficiency.
miR-126-5p protects endothelial cells from apoptosis through a tight control of Sema3A expression. Sema3A is a secreted protein that mediates its signaling through receptor complexes that include neuropilin 1 and plexin A or D1. Actually, a role for Sema3A as an inducer of cell death has been consistently described in cardiomyocytes (Zhao et al., 2016), tumor (Maione et al., 2009; Moretti et al., 2008), neuronal (Shirvan et al., 2002; Wehner et al., 2016) and endothelial (Guttmann-Raviv et al., 2007) cells, and also in vivo during kidney- (Reidy et al., 2009), tumor- (Maione et al., 2009) and oxygen-induced retinopathy (OIR) (Rivera et al., 2013) development. In the OIR model, microglial cells release IL1β, which induces Sema3A secretion by RGCs and leads to endothelial cell death (Rivera et al., 2013). Here, we did not observe a similar induction of Sema3A by IL1β, although this factor was upregulated following miR-126-5p inhibition. This suggests that RGCs do not respond to IL1β stimulation in physiological development as they do during OIR. However, miR-126-5p and SetD5 expression are modulated during OIR (data not shown), suggesting that their involvement in this pathological process might occur through additional molecular mechanisms.
The functions of the miR-126-5p target gene Setd5 were previously unknown. SetD5 is essential to development, as illustrated by the lethality of Setd5 knockout in mouse embryos (Osipovich et al., 2016) and by the intellectual disabilities frequently found in humans bearing mutations in the SETD5 gene (Grozeva et al., 2014). We showed here that SetD5 is a transcriptional regulator of Sema3A postnatally and, like other members of the Set family, SetD5 was not expected to bind directly to DNA but rather to associate with DNA-binding partners in order to regulate gene transcription. We identified BRD2 as one of the proteins interacting with SetD5 and showed that this interaction was necessary to regulate expression of Sema3A. The interaction between SetD5 and BRD2 suggests that SetD5 participates in the scaffold generated by BRD2 when binding to acetylated histones (LeRoy et al., 2008) on chromatin and which includes various partners such as transcription factors and proteins involved in chromatin remodeling (Belkina and Denis, 2012). The possible association of SetD5 with such protein complexes suggests that SetD5 might also be involved in the regulation of other gene networks in the neuronal compartment. Of note, we found that the gene regulation mediated by SetD5 was independent of its putative protein methyl transferase activity, which is consistent with results obtained in Setd5-deficient embryonic stem cells (Osipovich et al., 2016) and with its homologs, which do not appear to have methyltransferase activity either (Pijnappel et al., 2001; Rincon-Arano et al., 2012; Sebastian et al., 2009).
BET proteins such as BRD2 have been characterized as transcriptional activators. BRD2 controls cell cycle by regulating the expression of cyclin A2 and D1, notably in proliferating neuronal progenitors (Garcia-Gutierrez et al., 2012). BRD2 is also expressed in differentiating neurons (Crowley et al., 2004) but its functions in these cells are not clear. The control of an axon guidance molecule such as Sema3A by BRD2 in neurons was so far uncharacterized for this transcriptional regulator. In addition, whether BRD2 is strictly required for SetD5 binding to the Sema3a locus remains to be determined. Although we did immunoprecipitate SetD5 and BRD2 from the same Sema3a locus regions in vitro and in the retina, we cannot exclude the possibility that SetD5 binds to the Sema3a locus independently of BRD2. Indeed, SetD5 might be recruited onto chromatin by other partners such as components of the PAF1C and NCoR complexes, as it was recently shown to interact with these complexes (Osipovich et al., 2016), components of which are also expressed in the retina (Fig. S12).
This study revealed an unexpected role for miR-126-5p in neurons. Interestingly, the miR-126-5p host gene Egfl7 is also expressed in mammalian brain and controls Notch signaling in neural stem cells (Schmidt et al., 2009), supporting the idea that, like its host gene, the miR-126 duplex is functional in a neuronal context. In addition, miR-126-3p has been associated with gene de-regulation in Parkinson's neurodegenerative disease (Briggs et al., 2015) and its overexpression is neurotoxic in dopamine neurons (Kim et al., 2014), supporting a role for this microRNA in neurological disorders. Further analyses are now required to establish whether miR-126-5p contributes to neuron development and differentiation, in addition to mediating neuronal protection of endothelial cells.
MATERIALS AND METHODS
Animals
Wild-type outbred OF1, SCID, Csf1op/op and Setd5iLacZ/+ [C57BL/6NTac-Setd5tm1a(EUCOMM)Wtsi/Wtsi] mice were from Charles River, Institut Pasteur de Lille, Jackson laboratory, and Wellcome Trust Sanger Institute/EUCOMM Consortium, respectively, and were housed according to European legislation. Genotyping was performed as described by the provider. Csf1op/op genotyping was confirmed by Iba1 detection on whole-mount retina (Fig. S4). Subretinal injections were performed in 3-day-old (P3) anesthetized mice by injecting 0.5 µl of cholesterol-conjugated LNA inhibitor or target blocker (1 mg/ml) as previously described (Poissonnier et al., 2014b). When indicated, mice were subcutaneously injected with IL-1 receptor antagonist using Kineret (anakinra, 150 mg/ml, 15 mg/kg) 2 days after subretinal injection of inhibitor or co-injected in the subretinal space with anti-Sema3A antibody (R&D Systems, MAB1250, 2 µg/eye). Six-day-old mice were sacrificed according to CNRS recommendations and eyes were enucleated. Retinal cups were dissected from eye tissues under binoculars after brief fixation in 4% paraformaldehyde (PFA) in PBS for whole-mount immunostaining or in situ hybridization, or without fixation for RNA extraction or protein analysis. For cryosections, enucleated eyes were embedded in OCT. No statistical method was used to predetermine sample size. All pups derived from a litter were used in one experiment. Retinas with obvious immature vasculature were excluded from the analysis. V.M. and F.S. have the level I animal experimentation diploma. Until 2013, V.M. was authorized to perform experimentation on animals from Préfecture de la région Nord-Pas-de Calais (#59-35066). Since 2013, experimental procedures have been deposited to Ministère de l'enseignement supérieur et de la Recherche (French government research department), as per regulations.
In situ hybridization
Retinas were harvested from mice and fixed overnight in 4% PFA then in situ hybridization was performed as previously described (Poissonnier et al., 2014a).
microRNA inhibitors, target blockers and siRNA
Cholesterol-conjugated miR-126-5p inhibitor (cgcgtaccaaaagt/3cholTEG), cholesterol-conjugated control inhibitor (cgcgtcaacaaagt/3cholTEG), cholesterol-conjugated TB-SetD5 (ttacatctcttgtg/3cholTEG), cholesterol-conjugated control TB (tttcaagtcttctg/3cholTEG), cholesterol-conjugated miR-126-3p inhibitor (acattattacttttggtacgcg/3Chol/TEG), cholesterol-conjugated TB-CXCL12 (tattattattttcatcactga/3chol/TEG) and control inhibitor (taacacgtctatacgccca/3chol/TEG) were from Exiqon. Similar unconjugated microRNA inhibitors were synthesized for in vitro experiments. siRNA were from Dharmacon [On-TargetPlus non-targeting siRNA #1 (D-001810-01), On-TargetPlus SetD5 siRNA #1 (J-028069-07), On-TargetPlus SetD5 siRNA #2 (J-028069-08) and On-TargetPlus BRD2 siRNA (J-004935-06)].
Retina immunostaining
Flat-mount retinas were incubated overnight at 4°C with the following antibodies: rat anti-mouse CD31 (BD Pharmingen, MEC13.3 #553370, 0.5 mg/ml 1/100), rabbit anti-mouse cleaved-caspase 3 (Cell Signaling, #9664, 1/500), rabbit anti-mouse type IV collagen (Abcam, ab6586, 1 mg/ml, 1/500), rat anti-mouse CD45 (BD Pharmingen, #550539, 62.5 µg/ml, 1/100), mouse anti-rat GFAP-Cy3 (Sigma, C9205, 0.01 M, 1/250), mouse anti-human smooth muscle cell alpha actin (Sigma, F3777, 2.3 mg/ml, 1/100), rabbit anti-mouse NG2 (Millipore, #AB5320, 1 mg/ml 1/500). After washes, flat-mount retinas were incubated with fluorescent secondary antibodies: donkey anti-rat A594/A488, donkey anti-rabbit A488/A594 (Invitrogen, 2 mg/ml, 1/500). Retinas were finally washed and mounted in Mowiol. For Brn3A immunostaining after in toto miR-126-5p in situ hybridization, retina were blocked in 0.5% bovine serum albumin (BSA) in PBS overnight and incubated in the same buffer with primary antibodies (goat anti-Brn3A, 1/750, Santa Cruz, sc-31984; and rabbit anti-collagen IV, 1/500) overnight. After washes, retina were incubated with Alexa Fluor-conjugated secondary antibodies: donkey anti-goat A488 and donkey anti-rabbit A350 (Invitrogen, 2 mg/ml, 1/500) before washing and mounting in Mowiol. For all experiments, retinas were imaged with a Carl Zeiss AxioImager Z1-Apotome.
For cryosections, retina were fixed overnight with 4% PFA, rinsed in PBS then incubated with 30% sucrose before OCT embedding. Retina sections were blocked and permeabilized for 2 h in PBS 0.25% Triton X-100, 5% fetal bovine serum (FBS), 1% BSA and incubated overnight at 4°C with the following primary antibodies: goat anti-mouse Iba1 (Abcam, ab107159, 1 mg/ml, 1/250), rabbit anti-mouse IL1β (Abcam, ab9722, 1/250), goat anti-human Brn3a (Santa Cruz, sc-31984, 100 µg/ml, 1/200), rabbit anti-mouse Sema3A (Abcam, ab23393, 1 mg/ml, 1/200), rabbit anti-human SetD5 (Abcam, ab204363, 50 µg/ml, 1/50) or rabbit anti-BRD2 (Sigma, HPA042816, 1/100). After washes, sections were incubated with secondary antibody [donkey anti-goat A488 (for Iba1) or biotinylated goat anti-rabbit revealed with Cy-3 TSA amplification kit (Perkin Elmer) (for SetD5, Il1β and Sema3A)] for 2 h at room temperature. Sections were then washed and stained with DAPI before mounting in Mowiol. Sections were imaged with a Carl Zeiss AxioImager Z1-Apotome.
Reverse transcription and relative quantification
Total RNAs were extracted from retina using Trizol (Thermo Fisher) and from cells using NucleoSpin RNA plus kit (Macherey Nagel) according to the manufacturer's instructions. Total RNA was reverse transcribed using either High Capacity cDNA Reverse Transcription Kit (Applied Biosystems) for mRNA quantification or with TaqMan microRNA Reverse Transcription Kit and U6, miR-126-3p and miR-126-5p specific primers (Applied Biosystems) for miRNA quantification. U6 is a small nuclear RNA that is commonly used for miRNA normalization in RT-qPCR. When indicated, reverse transcription of miRNA was performed using the TaqMan Advanced miRNA cDNA Synthesis Kit (applied) and specific TaqMan advanced miR-126-5p probe (applied) according to manufacturer's recommendations. Relative quantification was performed using TaqMan gene expression master mix and specific TaqMan probes (Table S1, Applied Biosystems) using a StepOne system. mRNA expression levels were quantified using the 2−ΔΔCt method and normalized to U6 and β-actin or B2M levels, respectively. Error bars shown in graph are calculated using s.d. of the target (s.d.t) and s.d. of the reference (s.d.r) and using the formula: s.d.=(s.d.t2+s.d.r2)1/2.
Total RNA was extracted from HeLa and HEK cells using NucleoSpin RNA plus kit (Macherey Nagel) according to the manufacturer's instructions. RNA quantification was performed as described above.
Flow cytometry
Cells were trypsinized, and labeled using the annexin V-FITC Dead Cell Apoptosis Kit (Invitrogen) according to the manufacturer's recommendations. Samples were analyzed using a FACS Calibur cytometer (BD Biosciences). Annexin V-positive and propidium iodide-negative cells were considered apoptotic.
ELISA assay
Retinas were homogenized and VEGF (R&D Systems), IL1β (R&D Systems) and Sema3A (USCN) assays were performed according to the manufacturer's instructions.
Cell separation from retina
Retinas were dissected from P6 mice, washed in PBS and digested using a neuron isolation kit (Miltenyi) and incubated with CD31 microbeads (Miltenyi) for positive purification of endothelial cells and before negative purification of neurons following manufacturer's instructions. Briefly, depletion of non-neuronal cells was performed by positive purification of endothelial cells, microglial cells, oligodendrocytes, fibroblasts and astrocytes using a cocktail of biotin-conjugated antibodies (Miltenyi) and a magnetically labeled anti-biotin antibody. Six retinas were pooled per experiment to isolate enough cells for RT-qPCR analysis.
Cell culture and transfection
Primary human umbilical vein endothelial cells (HUVECs, three different lots, Lonza) were cultured in EGM-2 (Lonza) at 37°C, 5% CO2 following the manufacturer's instructions. HUVECs were plated at 25,000 cells/cm2 1 day before transfection and transfected with unconjugated LNA inhibitor using Primefect siRNA transfection reagent (Lonza) following the manufacturer's instructions. Cells were then cultured in normal or serum-depleted medium before staining for annexin V and analysis by flow cytometry. HeLa cells (ATCC) were cultured following manufacturer's instructions and were transfected with jetPrime. HEK cells were a gift from Dr D. Leprince (CNRS-UMR8161, Lille, France) and were cultured in DMEM supplemented with 10% FBS, non-essential amino acids and gentamycin. Cells were transfected with Lipofectamine RNAi Max (Invitrogen) for siRNA transfection or Lipofectamine 2000 (Invitrogen) for DNA transfection and following manufacturer's instructions.
When indicated, cells were treated with (+)-JQ1 or (−)-JQ1 (Sigma) bromodomain protein inhibitor or control, respectively, 24 h before RNA extraction.
All cells were routinely tested for mycoplasma contamination.
Plasmids
Setd5 cDNA and mutated Setd5 cDNA were amplified by PCR using primers described in Table S2 and cDNA synthetized from microglial cells as described previously (Poissonnier et al., 2014b). PCR products were cloned in pcDNA IRES GFP (Addgene plasmid #51406 deposited by Kathleen L. Collins) (Schaefer et al., 2008). HA-tagged sequence was inserted at the 3′ end of cDNA. The BRD2 vector was a generous gift of Dr LeRoy (LeRoy et al., 2008).
Yeast two-hybrid screening
Yeast two-hybrid screening was carried out by Hybrigenics using ULTImate Y2H technique. Screening was performed on a human retina cDNA library using human SETD5 cDNA as bait.
Co-immunoprecipitation
HEK cells were transfected with SetD5, BRD2 expression plasmid or both 48 h before protein extraction. Cells were lysed in 1 ml of IP buffer (50 mM Tris pH 8, 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 0.5% Igepal) before complexes were immunoprecipitated using anti-FLAG M2 affinity Gel (Sigma, 25 µl), and used for HA immunoblotting analysis. Flagged protein expression was measured with monoclonal anti-FlagM2 antibody (Sigma, 1/1000) in re-probing experiments. HA-tagged protein expression was assessed in whole-cell extracts (input) used for HA immunoblotting (mouse anti-HA, Eurogentec, #MMS-101P-100, 1/1000). For co-immunoprecipitation of endogenous SetD5 and BRD2 proteins, complexes were immunoprecipitated using anti-SetD5 antibody (Abcam, 1 µg) and used for BRD2 (Sigma, HPA042816, 1/100) immunoblotting analysis. For in vivo experiments, retinas were dissected at P6, dissociated in 500 µl retina IP buffer, lysed for 30 min and finally sonicated (2×10 s, 22%, Digital Sonifier, Branson). Complexes were immunoprecipitated in the same conditions as for in vitro experiments using 1 µg of SetD5 antibody or IgG control and BRD2 immunoblotting analysis were performed with Novus rabbit anti-BRD2 antibody (NBP1-30475).
Western blotting
Proteins were extracted in RIPA buffer complemented with protease inhibitor (Complete, Roche), analyzed by 7.5% SDS-PAGE, blotted onto Immobilon-P (Millipore), incubated with antibodies and finally revealed using ECL-Prime (GE HealthCare) with detection of chemiluminescence by film exposure.
Chromatin immunoprecipitation
For in vitro experiments, HEK cells were cultured until confluency. Cells were fixed for 10 min with 1% PFA in culture medium. Nuclear lysate was sonicated and chromatin immunoprecipitations were performed using Magna-ChIP (Millipore) following manufacturer's instructions using the following antibodies: anti-acetyl-histone H4 (Millipore, #06-866, 10 µl/IP), anti-BRD2 (Sigma, HPA042816, 1 µg/IP) and rabbit anti-human SetD5 (Abcam, ab204363, 1 µg/IP) or the control antibody mouse IgG Ctl (Mouse IgG2B Isotype, R&D Systems, MAB004, 1 µg/IP). qPCR for ChIP analysis were performed using SYBR green (Applied Biosystems) and the following primers: Region A, 5′-ccggataatgaggcacaact-3′/5′-tagagactgccaccggctat-3′; region B, 5′-gtagttggctgtggcctctc-3′/5′-ggggtagggcagaatcattt-3′; region C, 5′-cctcaagcctattgatcagccagt-3′/5′-taatccatggaagacagacaagcc-3′. For in vivo experiments, retinas were dissected at P6 and processed immediately or frozen before analysis. Samples were then dissociated and DNA-protein complexes were cross-linked according to ‘Magna ChIP G Tissue kit’ (Millipore) recommendations. Each retina was dissociated in 500 µl of Tissue Stabilizing Solution, lysed in 500 µl of cell lysis buffer and finally sonicated in 500 µl of dilution buffer for 90 min (90 s ON/OFF high pulses, Bioruptor, Diagenode). Immunoprecipitations were performed with the same experimental conditions as those used for in vitro experiments. qPCR for ChIP analysis were performed using SYBR Green (Applied Biosystems) and the following primers: Region A′, 5′-actacacttgttgtagagcc-3′/5′-tgtgtattctcgcatcagtg-3′; region B′, 5′-cctgaaagcctattgatcag-3′/5′-gtgctcaggataacaagggt-3′. Fold change relative to control IP was calculated using the 2−ΔΔCT(ChIP) formula: ΔΔCT(ChIP)=(CT IPtarget−CT Input)−(CT IP control−CT Input).
Image analysis
Retina images were analyzed using ImageJ or ImageJ angiogenesis analyzer (Carpentier G).
Statistics
For microscopy analysis, all immunostainings were performed at least twice. Sample size (n) included in the figure legends correspond to the experiment shown. For quantitative experiments, the number of times experiments were replicated and sample size are included in the figure legends. Center values show means; error bars represent s.d. Statistical significances between conditions were evaluated by applying either unpaired Student's t-test or Mann–Whitney test (when the assumptions of the t-test are not reached) onto two groups of raw data, and one-way ANOVA test onto three or more groups of data. Significance is noted as *P<0.05, **P<0.01 or ns (not significant).
Acknowledgements
We thank the BioImaging Center Lille-Nord (BICel), Dr LeRoy (Department of Molecular Biology, Princeton University, USA) for providing the Flag-BRD2 expression vector, Ingrid Loison for helpful discussions, and Marie-Christine Bouchez for assistance. F.S. is Director of Research of Institut National de la Santé et de la Recherche Médicale.
Footnotes
Author contributions
Conceptualization: G.V., S.C., V.M.; Methodology: G.V., L.P., B.N., G.B., J.C.R., S.C., V.M.; Validation: G.V., L.P., B.N., G.B., J.C.R., S.C., F.S., V.M.; Formal analysis: L.P., V.M.; Investigation: G.V., L.P., B.N., G.B., J.C.R., S.C., V.M.; Resources: G.V., L.P.; Writing - original draft: F.S., V.M.; Visualization: F.S., V.M.; Supervision: S.C., V.M.; Project administration: V.M.; Funding acquisition: S.C., F.S., V.M.
Funding
This work was supported by Institut National du Cancer, Fondation ARC pour la Recherche sur le Cancer, Ligue Contre le Cancer, and Cancéropôle Nord-Ouest.
References
Competing interests
The authors declare no competing or financial interests.