Apelin signaling drives vascular endothelial cells toward a pro-angiogenic state

To form new blood vessels (angiogenesis), endothelial cells (ECs) must be activated and acquire highly migratory and proliferative phenotypes. However, the molecular mechanisms that govern these processes are incompletely understood. Here, we show that Apelin signaling functions to drive ECs into such an angiogenic state. Zebrafish lacking Apelin signaling exhibit defects in endothelial tip cell morphology and sprouting. Using transplantation experiments, we find that in mosaic vessels, wild-type ECs leave the dorsal aorta (DA) and form new vessels while neighboring ECs defective in Apelin signaling remain in the DA. Mechanistically, Apelin signaling enhances glycolytic activity in ECs at least in part by increasing levels of the growth-promoting transcription factor c-Myc. Moreover, APELIN expression is regulated by Notch signaling in human ECs, and its function is required for the hypersprouting phenotype in Delta-like 4 (Dll4) knockdown zebrafish embryos. These data provide new insights into fundamental principles of blood vessel formation and Apelin signaling, enabling a better understanding of vascular growth in health and disease.


Introduction
Endothelial cell sprouting is a fundamental process of physiological and pathological blood vessel growth. Attracted by growth factors such as vascular endothelial growth factor-A (VEGF-A) secreted from hypoxic tissues, endothelial cells (ECs) break out of the quiescent vessel wall to form new vessel branches (Ferrara et al., 2003;Koch and Claesson-Welsh, 2012). ECs with higher levels of VEGF-A signaling become invasive tip cells that lead new vascular sprouts, while neighboring ECs with lower VEGF-A signaling become trailing stalk cells (Gerhardt et al., 2003). This process is coordinated by Delta-like 4 (DLL4)/Notch signaling. Activation of Notch receptors by their ligand DLL4, expressed by tip cells, represses tip cell behavior in stalk cells (Hellströ m et al., 2007;Leslie et al., 2007;Siekmann and Lawson, 2007;Suchting et al., 2007). Loss of Notch signaling, on the other hand, causes excessive tip cell formation and vascular overgrowth (Hellströ m et al., 2007;Leslie et al., 2007;Siekmann and Lawson, 2007;Suchting et al., 2007).

Apelin signaling is required for angiogenic sprouting
To examine the expression pattern of the apelin ligand and receptor genes during angiogenic sprouting in zebrafish embryos, we first performed whole-mount in situ hybridization during intersegmental vessel (ISV) formation. We detected clear alpn, but no apela, expression within the sprouting ISVs (Figure 1-figure supplement 1 arrowheads). For the receptor genes, we could only detect aplnrb expression in the ISVs (Figure 1-figure supplement 1 arrowheads).
To visualize apln and aplnrb expression at single cell resolution, we developed reporters using Bacterial artificial chromosome (BAC) recombineering ( Figure 1-figure supplement 2). To this end, we replaced the ATG of an apln containing BAC with an EGFP cassette. Similarly, we replaced the stop codon of an aplnrb containing BAC with a tandem fluorescent timer (TagRFP-sfGFP) cassette leading to a fusion protein. We injected both modified BACs into one-cell stage zebrafish embryos to generate stable transgenic lines, Tg(apln:EGFP) and Tg(aplnrb:aplnrb-TagRFP-sfGFP) (hereafter referred to Tg(aplnrb:aplnrb-sfGFP)) ( Figure 1-figure supplement 2). We first detected weak apln: EGFP expression in sprouting ISVs at 30 hpf ( Figure 1A). At 54 hpf, all ECs within the dorsal longitudinal anastomotic vessel (DLAV) -a vessel formed by tip cells -were labeled ( Figure 1A, arrowheads) while some stalk cells also exhibited weak apln:GFP expression ( Figure 1A, arrows). Of note, aplnrb:Aplrnb-sfGFP expression at 26 hpf was visible in the entire ISV sprout ( Figure 1B arrowheads), but it was absent from non-angiogenic ECs within the dorsal aorta (DA). At 54 hpf, aplnrb: Aplrnb-sfGFP expression was detected in all ECs that had sprouted out of the DA but also weakly in ECs within the DA ( Figure 1B). These results suggest that apln is expressed in tip cells while aplnrb is expressed in all sprouting ECs.
To examine the function of Apelin signaling during sprouting angiogenesis in zebrafish, we used mutants for aplnra (Helker et al., 2015), aplnrb (Helker et al., 2015), apln (Helker et al., 2015) and apela (Chng et al., 2013). Homozygous aplnra mutant embryos exhibited no obvious defects during ISV formation (Figure 1-figure supplement 3). However, homozygous aplnrb mutant embryos exhibited reduced ISV length and failed to form the DLAV (Figure 1-figure supplement 3). This phenotype was more severe in embryos lacking both aplnra and aplnrb ( Figure 1C, Figure 1-figure supplement 3), indicating partial compensation. We also analyzed apln and apela mutants. Homozygous apela mutant embryos displayed only a mild delay in ISV outgrowth (Figure 1-figure  supplement 3), while homozygous apln mutant embryos exhibited defects in ISV outgrowth and failure to form the DLAV (Figure 1-figure supplement 3). Loss of both ligands increased the severity of the phenotype leading to ISV stalling at the horizontal myoseptum ( Figure 1C, Figure 1-figure  supplement 3). Consistent with studies in the mouse retina (del Toro et al., 2010), our studies identify apln expression as a marker of endothelial tip cells in zebrafish and show that Apelin signaling is required for angiogenic sprouting.

Apelin signaling regulates tip cell morphology
To investigate when the sprouting defects in Apelin signaling-deficient embryos first appear, we analyzed developmental time points when tip cells start to sprout out of the DA. However, no differences in sprout initiation or tip cell specification were observed in double homozygous receptor or ligand mutants (Figure 2-figure supplement 1, Figure 2A). Instead, we found that sprout      Previously, we reported a role for Apelin signaling in establishing blood flow-induced EC polarity (Kwon et al., 2016). To determine whether the observed defects during sprouting were caused by defects in EC polarity, we analyzed the location of the Golgi apparatus during ISV formation in wildtype and mutant embryos. However, we could not detect obvious differences in EC polarity during angiogenic sprouting (Figure 2-figure supplement 2, arrowheads point to polarized ECs). Next, we asked whether Apelin signaling regulates the number of ECs, and so combined aplnr mutants with the Tg(fli1a:nEGFP) reporter line (Roman et al., 2002) to visualize EC nuclei. Compared to controls, aplnr mutants exhibited a reduction in ISV EC numbers of 1 cell at 30 hpf (4 instead of 5) and 2 cells at 52 hpf (5 instead of 7) ( Figure 2D,E). We next assessed whether apln overexpression leads to ectopic sprouting. To this end, we generated an inducible transgenic line to overexpress apln under the control of the hsp70l promoter. However, global overexpression of apln did not lead to ectopic sprouting of blood vessels but led to mispatterned lymphatic vessels (Figure 2-figure supplement 3, arrows). Altogether, these data indicate that the angiogenic defects in Apelin signalingdeficient embryos are caused by filopodia defects and impaired cell migration. Apelin signaling also regulates the number of ECs within the ISV sprouts.

Apelin signaling drives the sprouting behavior of ECs
We hypothesized that aplnrb expression ( Figure 1B) provides an advantage for ECs to sprout. To test this hypothesis, we generated chimeric embryos using wild-type and aplnr deficient embryos ( Figure 3A). Upon transplantation of wild-type donor cells into wild-type hosts, 34,5% of the donorderived ECs were present in the ISVs at 24 hpf ( Figure 3B,C). In contrast, upon transplantation of wild-type donor cells into aplnr-deficient hosts, 80% of the donor-derived ECs were present in the ISVs at 24 hpf ( Figure 3B,C). Together these data show that the apelin receptors function cell-autonomously in endothelial sprouting. The Apelin receptor has been shown to signal mainly through the G-protein Gai (Habata et al., 1999). Therefore, we blocked Gai function through the mosaic and vascular-specific overexpression of pertussis toxin (PTX). Our results show that ECs deficient for signaling though Gai behave similarly to aplnr mutant ECs indicating that the Apelin receptor mediates its angiogenic effect through Gai (Figure 3-figure supplement 1). Notably, wild-type donorderived ECs in aplnr deficient embryos populated the entire dorsal part of the vasculature which is usually missing in these mutants, further confirming the cell-autonomous function of the Apelin receptors during angiogenesis ( Figure 3-figure supplement 2). Together, these results indicate that apelin signaling primes ECs toward a sprouting state.   Apelin signaling functions downstream of Notch signaling It has been previously reported that Notch-deficient ECs outcompete wild-type ECs during ISV sprouting (Siekmann and Lawson, 2007), an observation consistent with data in mouse (Jakobsson et al., 2010;Pitulescu et al., 2017). Since wild-type ECs similarly outcompete aplnr mutant ECs, we wanted to investigate potential links between Apelin and Notch signaling. Hence, we first blocked Notch signaling in Tg BAC (apln:EGFP) embryos by injecting a dll4 MO. As previously reported (Leslie et al., 2007;Siekmann and Lawson, 2007), dll4 morphants exhibited a hypersprouting ISV phenotype ( Figure 4A). Notably, we also observed a clear increase in apln:EGFP expression in the ectopic sprouts ( Figure 4A). To test whether Apelin signaling is required as a downstream effector of Notch signaling during angiogenesis, we injected the dll4 MO into the offspring of apln and aplnrb heterozygous parents and compared the phenotype in homozygous mutant embryos versus their wild-type siblings. Strikingly, the hypersprouting phenotype of dll4 morphants was not present when Apln or Aplnrb function was lost ( Figure 4B   independent of Apelin function. Similar results were obtained when we analyzed ectopic venous sprouting in response to bmp2b overexpression (Wiley et al., 2011;Figure 4D,F). Together these data indicate that Apelin signaling is specifically required for Notch-modulated angiogenesis. To investigate whether apln expression is itself regulated by Notch signaling, we performed cell culture assays. We treated HUVECs with the Notch inhibitor DAPT and analyzed APLN expression by RT-qPCR. Consistent with the observations in zebrafish ( Figure 4A), we observed an increase in APLN expression upon Notch inhibition ( Figure 4G, Figure 4-source data 1). Next, we activated Notch signaling by stimulating HUVECs with the Notch ligand DLL4. Conversely to the Notch inhibition data, activating Notch signaling in HUVECs suppressed APLN expression ( Figure 4H, Figure 4source data 1). Together these data suggest that the increased sprouting in response to Notch inhibition is, in part, driven by the upregulation of apln.

Apelin signaling positively regulates EC metabolism
Because EC sprouting requires an increase in metabolic activity (Dobrina and Rossi, 1983;Krützfeldt et al., 1990;Mertens et al., 1990;De Bock et al., 2013a;Vandekeere et al., 2015) and Apelin signaling has been shown to control cell metabolism in other contexts (Dray et al., 2008;Sawane et al., 2013), we asked whether Apelin signaling promotes EC metabolism. Previous studies have demonstrated that ECs rely on glycolysis for sprouting (De Bock et al., 2013a;Vandekeere et al., 2015). Therefore, we measured the extracellular acidification rate (ECAR) as a surrogate parameter of glycolysis in Apelin signaling-deficient HUVECs ( Figure 5A,B). Notably, we observed a marked reduction in glycolysis after knockdown of Apelin signaling ( Figure 5A), whereas mitochondrial oxygen consumption appeared unchanged ( Figure 5B). To gain insight into the underlying mechanisms, we analyzed key regulators of metabolism and found a reduction in c-MYC protein levels after depletion of Apelin signaling ( Figure 5C). Furthermore, expression of PFKFB3, which encodes an enzyme that sustains high glycolytic rates, was also reduced in Apelin signalingdeficient HUVECs ( Figure 5D). In order to analyze whether a reduction in EC metabolic activity causes the vascular phenotype observed in aplnrb mutants, we performed mosaic rescue experiments and overexpressed pfkfb3 in ECs. In agreement with our in vitro data, we found that overexpression of pfkfb3 in endothelial tip cells leads to a partial rescue of the vascular phenotype in aplnrb mutants ( Figure 5E arrowheads, 5F, Figure 4-source data 1). Thus, Apelin signaling controls the expression of regulators of glucose metabolism as well as glycolytic activity in developing endothelial cells.

Discussion
During the formation of the first embryonic blood vessels, angioblasts migrate to the midline where they coalesce to form the future DA and cardinal vein. We have previously reported that vasculogenesis relies on the function of the ligand Apela (Helker et al., 2015). Here, we show that angiogenesis depends mostly on the function of the ligand Apln. However, Apela can partially compensate for the loss of Apln. This stage-specific ligand usage is in agreement with previous studies showing that apela expression is reduced by the end of vasculogenesis when apln starts to be expressed (Chng et al., 2013;Pauli et al., 2014).
During angiogenesis in embryos lacking Apelin signaling, we observed a severe sprouting defect with a reduction in EC numbers and filopodia. As ECs proliferate, extend filopodia, and migrate during ISV formation, it is challenging to assign the cause of the sprouting defect to the EC proliferation or filopodia formation defects. However, Phng et al., 2013 reported that the inhibition of filopodia formation by Latrunculin B treatment reduces ISV sprout length, suggesting that the ISV sprouting defects in apln mutants is caused by the filopodia defects. However, one cannot exclude the possibility that defects in EC numbers are also contributing to the ISV sprouting defects. While we observed a severe angiogenesis phenotype when Apelin signaling was impaired, global overexpression of apln did not lead to ectopic sprouting. However, these experiments were done in the presence of endogenous Apelin, and thus, it is possible that the endogenous Apelin gradient prevents ECs from ectopic sprouting. In addition, Apelin might need to be expressed from a discrete source, rather than globally, to elicit a sprouting response.
During sprouting angiogenesis, ECs within a sprout are highly heterogenous in their shape, gene expression and function, which led to the model of tip and stalk cells (Gerhardt et al., 2003). While differences in expression between tip and stalk cells have been reported for several genes (Tammela et al., 2008), (Hellströ m et al., 2007;Siekmann and Lawson, 2007;Leslie et al., 2007;Suchting et al., 2007;del Toro et al., 2010;Herbert et al., 2012), little is known about the molecular differences between sprouting and resting ECs (Schlereth et al., 2018). By analyzing novel reporter lines for apln and aplnrb expression, we observed high apln expression in tip cells while we could not observe any difference in aplnrb expression between tip and stalk cells ( Figure 5-figure supplement 1). Interestingly, aplnrb is highly expressed in sprouting ECs in ISVs while being absent from non-angiogenic ECs in the DA ( Figure 5-figure supplement 1). These observations are in line with a recent study showing that ECs during tumor angiogenesis can be labeled by a CreERT2 transgene in the Aplnr locus while quiescent blood vessels in the surrounding tissue are not labeled (Zhao et al., 2018).
At the molecular level, vascular sprouting and cell positioning within the sprout is tightly regulated by VEGF and Notch signaling (Hellströ m et al., 2007;Lobov et al., 2007;Siekmann and Lawson, 2007;Suchting et al., 2007;Jakobsson et al., 2010). In addition to these signaling pathways, we propose Apelin signaling as a molecular switch to drive ECs into a pro-angiogenic state. In line with the expression of aplnrb in sprouting but not quiescent ECs, we show that aplnrb function regulates the ability of ECs to sprout or stay quiescent. Similarly, Notch signaling regulates the behavior of ECs (Siekmann and Lawson, 2007): rbpj deficient ECs contribute to the ISVs while wild-type ECs stay within the DA (Siekmann and Lawson, 2007). Of note, we found that Notch signaling regulates the expression of apln in vitro as well as in vivo and that Apelin signaling is a key downstream effector of Notch signaling during angiogenesis ( Figure 5-figure supplement 1). However, it is very unlikely that apln is a direct Notch target gene since activation of Notch signaling leads to a downregulation of APLN expression. Thus far, two downstream effectors of Notch signaling have been reported to control angiogenesis namely PTEN (Serra et al., 2015) and CXCR4, another GPCR (Hasan et al., 2017;Pitulescu et al., 2017). While PTEN has been shown to be required for Notch induced arrest in EC proliferation (Serra et al., 2015), CXCR4 mediates Notch-controlled EC migration (Hasan et al., 2017;Pitulescu et al., 2017). PTEN and Apelin both regulate AKT phosphorylation (Davies et al., 1998;Masri et al., 2004). Thus, one might speculate that AKT function is a common effector of PTEN and Apelin signaling in EC proliferation. Furthermore, we found that Apelin was required for EC migration in the absence of Notch signaling. Similarly, CXCR4 is required for EC migration in the absence of Notch signaling (Hasan et al., 2017;Pitulescu et al., 2017). CXCR4 and APLNR both signal through the G-protein Gai (Moepps et al., 1997;Habata et al., 1999), and they might therefore have similar effects. Gpr124, another GPCR, has been reported to be required in tip cells during zebrafish angiogenesis (Vanhollebeke et al., 2015), similar to Aplnr. However, Gpr124 is required in tip cells only in the brain (Vanhollebeke et al., 2015), while Aplnr is required in tip cells in the ISVs, where it is most highly expressed.
Sprouting angiogenesis is controlled by genetically encoded signal transducers as well as by the metabolic state. However, how environmental signals modulate the metabolic activity of ECs is incompletely understood. Here, we show that Apelin signaling regulates the expression of PFKFB3 and c-MYC, two powerful drivers of EC metabolism (Wilhelm et al., 2016;De Bock et al., 2013b). Recently it has been shown that Apelin signaling promotes FOXO1 phosphorylation (Hwangbo et al., 2017), which negatively regulates its activity. Consistent with these findings, FOXO1 has been shown to suppress c-MYC expression (Wilhelm et al., 2016). Together these data raise the possibility that Apelin signals through FOXO1 to regulate c-MYC levels. Of note, genetic deletion of Pfkfb3 in mouse ECs leads to a reduction in their number as well as defects in filopodia formation and extension (De Bock et al., 2013b), phenocopying aplnr mutant embryos.
Taken together, our findings provide novel insights into a druggable pathway regulating angiogenesis and suggest that manipulating the angiogenic state of ECs by controlling Apln signaling might have therapeutic potential to control vascular growth in pathological settings.

Generation of the
To generate the Tg(kdrl:HsHRAS-EGFP) and Tg(hsp70:apln) lines, the gateway recombination system (Invitrogen) using entry vectors and the pTolDest destination vector (Villefranc et al., 2007) was used. The apln coding sequence was amplified from cDNA. 100 pg DNA of the plasmids and 50 pg of tol2 mRNA were injected into one-cell stage zebrafish embryos for stable germline transmission.

Morpholino injections
Morpholinos were obtained from Gene Tools, resuspended in distilled H 2 O and around 2 nl was injected into 1 cell stage embryos. The following morpholinos were used: aplnrb MO (Helker et al., 2015) at 0.5 ng/embryo, dll4 MO (Hogan et al., 2009) at 3 ng/embryo. An equal amount of the standard control MO: 5'-CCTCTTACCTCAGTTACAATTTATA-3' was used for each experiment.

Transplantation experiments
At the sphere stage, cells were removed from Tg(kdrl:HsHRAS-mCherry) donor embryos and transferred to Tg(fli1a:EGFP) aplnr mutant hosts using a glass capillary. Transplanted ECs were identified by transgenic mCherry expression.

Confocal microscopy
Zebrafish larvae were mounted in 1% low melt agarose. Egg water and agarose were supplemented with 19.2 mg/l Tricaine. All fluorescent images were acquired using an upright Zeiss LSM 780, 800 or 880 or a Leica SP5 or SP8 confocal microscope. Maximum projection images were analyzed and generated using Imaris (Bitplane).

Quantification of filopodia
Only filopodia with more than 10 mm in length were used for quantification.
ISVs were categorized as filopodia rich ISVs (more than six filopodia) or filopodia poor ISVs (less than six filopodia). A total of 49 ISVs were quantified for the Ctr MO and 103 ISVs for the aplnrb MO.
pfkfb3 rescue experiments pfkfb3 was cloned downstream of a bidirectional fli1a promoter driving EGFP in one direction and pfkfb3 in the other direction. aplnrb mutant embryos were injected with 20 pg EGFP:fli1a:pfkfb3 DNA and 30 pg Tol2 mRNA to generate mosaic blood vessels. EGFP positive tip cells were analyzed to quantify the percentage of connected DLAV segments. Neighboring EGFP negative tip cells were used as controls.

RNA interference
To silence APLN and APLNR gene expression, HUVECs were transfected with 50 nM APLN and APLNR ON-TARGET SMARTpool siRNA (Dharmacon). As a control, a non-targeting siRNA pool was used (Dharmacon). HUVECs were grown to 70% confluency and transfected with Lipofectamine RNAiMAX (Life Technologies) according to manufacturer's instructions.

RT-qPCR
Total RNA from HUVECs was extracted using a RNeasy Mini Kit (Qiagen). Reverse transcription polymerase chain reaction (RT-PCR) was performed using a SuperScript III First-Strand Synthesis System (Invitrogen) according to manufacturer's instructions. RT-qPCR was carried out to quantify gene expression levels on a CFX connect Realtime System (Bio-Rad) with the following Taqman probes: APLN Hs00175572_m1, APLNR Hs00270873_s1, PFKFB3 Hs00270873_s1. Each sample was normalized to the housekeeping probe ACTB Hs01060665_g1.

Metabolic assay
The metabolism of cells was assessed by the measurement of extracellular acidification (ECAR) and oxygen consumption rates (OCR) using a Seahorse XFe96 analyser (Agilent). Four hours before the measurement, 40.000 HUVECs per well were seeded in a fibronectin-coated XFe96 microplate. The measurement was done following manufacturer's protocol. To monitor glycolysis, the glycolysis stress test kit was used. The following substances were sequentially injected after a baseline measurement: Glucose (10 mM), Oligomycin (3 mM) and 2-Deoxyglucose (2-DG; 100 mM). The oxygen consumption rate was assessed using the Mito stress test kit. After a baseline measurement, the following substances were sequentially injected: Oligomycin (3 mM), the mitochondrial uncoupler carbonyl cyanide-4-(trifluoromethoxy)phenyl-hydrazone (FCCP; 1 mM) as well as a mixture of antimycin A (1.5 mM) and rotenone (3 mM).

Statistics
Standard error of the mean and P-values from a two-tailed t-test were calculated using Prism.