UXT potentiates angiogenesis by attenuating Notch signaling

The blood vessel network is composed of arteries, capillaries and veins. Vasculogenesis is a de novo assembly of the network, whereas angiogenesis is the coordinated growth of endothelial cells (ECs) from pre-existing vasculature (Risau, 1997; Risau and Flamme, 1995). Both processes are essential for tissue growth and organ function in development, physiology and disease (Phng and Gerhardt, 2009).

Angiogenic sprouting in zebrafish is one of the best-characterized examples of angiogenesis both morphologically and mechanically (Childs et al., 2002; Lawson and Weinstein, 2002; Siekmann and Lawson, 2007). This model centers on the interplay between tip cell and stalk cells (Eilken and Adams, 2010; Gerhardt et al., 2003; Siekmann and Lawson, 2007). Displaying a gene expression profile distinct from that of the stalk cell, the tip cell is the spearhead selected to lead the emerging sprout, and it is followed by the endothelial stalk cells (Phng and Gerhardt, 2009). Studies of mouse and zebrafish development have firmly established that vascular endothelial growth factor (VEGF) and Notch signaling pathways are fundamental for the specification of endothelial cells into tip and stalk cells (Potente et al., 2011).

Notch is a family of large single-pass type I transmembrane protein receptors, which includes Notch 1-4 in mammals (Artavanis-Tsakonas et al., 1999), and Notch1a, Notch1b, Notch2 and Notch3 in zebrafish (Lorent et al., 2004; Westin and Lardelli, 1997). Correspondingly, there are multiple ligands that selectively bind to different Notch receptors. Engaging a ligand to Notch causes receptor proteolysis (trans-interactions), which is catalyzed by a series of proteases, resulting in the release of the Notch intracellular domain (NICD) and its translocation into the nucleus. When Notch signaling is off, RBP-Jκ/CSL mediates transcriptional repression by recruiting transcriptional co-repressor proteins (Bray, 2006). After Notch binding to the ligand and receptor proteolysis, the nuclear NICD interacts with RBP-Jκ/CSL through a conserved WxP motif in the Notch Rbp-associated molecule (RAM) domain, thus triggering the ‘transcriptional switch’ to activate the gene expression (Tamura et al., 1995). Notch target genes include the helix-loop-helix (HLH) family transcription factors (such as hey1, her6), which are important for regulating angiogenesis (Phng and Gerhardt, 2009). In zebrafish, Notch signaling potently suppresses vessel sprouting and represents an elegant model for studying cell fate determination in angiogenesis (Artavanis-Tsakonas et al., 1999; Siekmann and Lawson, 2007). We are intrigued to explore how the strength and duration of Notch signaling are modulated in the context of angiogenesis; in particular, we are interested in the molecular regulation of NICD in the nucleus.

The ubiquitously expressed transcript UXT (∼18 kDa, also known as ART27) was first identified in human and mouse (Markus et al., 2002; Schroer et al., 1999). UXT expression is markedly elevated in some human tumors (Zhao et al., 2005). Its functional characterization is largely unknown. We recently reported that UXT modulates innate immunity and inflammation, using in vitro cell models (Huang et al., 2011a,,b; Sun et al., 2007). Interestingly, a global knockout of UXT in mouse is embryonic lethal (Y.Z., unpublished data), suggesting that UXT might play an important role during development. In this study, we characterize the UXT protein as a novel repressor of Notch signaling in angiogenesis, which functions by impairing the interaction between NICD and RBP-Jκ/CSL.

Our bioinformatics analysis reveals that UXT is highly conserved across vertebrate species (Fig. 1A). The zebrafish UXT displays 50.6% sequence homology and 87.8% sequence similarities to human sequences (Fig. 1A). Using the whole-mount in situ hybridization (WISH) technique, we observed a ubiquitous expression of uxt mRNA as early as 6 hours post fertilization (hpf) in zebrafish embryos, and later on we saw a relative enrichment in the head region and parts of the trunk (Fig. 1B). The real-time quantitative PCR (RT-qPCR) results further substantiated the expression of uxt mRNA during zebrafish embryo development (Fig. 1C). A mouse monoclonal antibody against the zebrafish UXT protein (4B4) was generated and its specificity was confirmed (ABmart, Shanghai, China). Consistently, the protein expression of UXT mirrors that of the mRNA (Fig. 1D), but with a relatively stable expression level (Fig. 1E; supplementary material Fig. S3). In addition, we screened out three anti-sense morpholino oligonucleotides (MO1, MO2 and MO3, listed in supplementary material Table S1) that effectively knocked down UXT expression (Fig. 1G). A scramble morpholino sequence was included as a negative control (Ctrl MO). We observed that MO2 was the most efficient, and therefore it was used in the following experiments (Fig. 1G; supplementary material Figs S1 and S2). MO1 and MO3 were also used in some experiments (supplementary material Figs S4, S5, S7 and S10).

Depletion of UXT by injection of morpholino oligonucleotides caused a marked developmental abnormality, which was evidenced by the shortened trunk, curly tail, decreased pigmentation and hindbrain atrophy (Fig. 1F; supplementary material Fig. S4). UXT is an α-class prefoldin family protein, and its structure was predicted to contain two α-helix domains (Vainberg et al., 1998). By mutating two conserved leucine residues to proline residues (L45P/I54P in zebrafish; L50/59P in human) (Fig. 1A, red outline), we generated a loss-of-function mutant of UXT, UXT-2M. Interestingly, the developmental abnormalities were rescued by re-expression of wild-type uxt mRNA, but not by the UXT-2M mRNA (Fig. 1F,H). Taken together, these data suggest that UXT might play some role during zebrafish development.

To address the potential function of UXT in zebrafish development, we carried out microarray assays and compared relative mRNA expression levels between UXT-deficient and control embryos (72 hpf). Notably, the expression of genes associated with both Notch signaling (such as dla, dlb, dld) and angiogenesis (such as birc5a, igfbp2, vegfaa) were affected by UXT knockdown (data not shown).

It has been well established that the angiogenesis of intersegmental vessels (ISVs) mainly takes place at the time window from 18 hpf to 30 hpf in zebrafish (Fouquet et al., 1997). This led us to speculate that UXT might play a role in ISV development. To explore this possibility, we investigated mRNA expression of the angiogenesis markers, including cox2, nestin1, angpt1, flt1 and kdrl (Albini et al., 1996; Amoh et al., 2005; Kim et al., 2006; Luttun et al., 2002; Stefater et al., 2011; Wu et al., 2006), and of the somite marker myod, at 18 hpf, 24 hpf and 30 hpf. When the expression of UXT was knocked down, these angiogenesis markers were expressed at slightly lower levels at 18 hpf and 24 hpf (Fig. 2A,B), but were strikingly decreased at 30 hpf (Fig. 2C; supplementary material Fig. S5).

Next, we examined the expression of the angiogenesis marker kdrl as well as additional lineage markers by using WISH. At 22 hpf, the expression of kdrl mRNA in UXT-deficient zebrafish embryos was markedly reduced in ISVs, whereas the pronephric marker pax2.1 and the erythroid marker gata1 were unaffected in UXT-deficient embryos (Fig. 2D). Taken together, these data demonstrate the importance of UXT on the process of zebrafish angiogenesis.

To further dissect the function of UXT in angiogenesis, we analyzed the ISV development of Tg(kdrl:EGFP)^s843 transgenic zebrafish using confocal microscopy. As expected, ISVs of control embryos extended to the dorsal-most aspect of the neural tube at 30 hpf, and T-shape branches formed by dorsal longitudinal anastomotic vessels (DLAV) were normal (Fig. 3A,B). Remarkably, ISVs of the embryos that were injected with 4 ng of UXT MO sprouted to the horizontal myoseptum and did not reach the dorsal roof to form DLAVs (Fig. 3C,D). Furthermore, the embryos that were injected with 6 ng of UXT MO displayed severe defects, such as a shorter trunk and a more downward curly tail (Fig. 3E). Notably, the ISVs of UXT-deficient embryos projected up at the midline of the trunk, and almost no ISVs were generated at posterior part of the trunk (Fig. 3E,F).

The length of ISVs in UXT morphants was significantly shorter, approximately half that of the control ones (51.3±5.2 µm versus 101.8±9.6 µm, P<0.005; Fig. 3K). The lumen diameter of the dorsal aorta remained roughly the same for both UXT morphants and control embryos. However, the lumen of the posterior cardinal vein (PCV) became narrower in UXT morphant embryos than in control embryos (27.2±2.8 µm versus 23.6±3.4 µm, P<0.05; Fig. 3K). The defects in ISVs in UXT morphants were rescued by injection of the morpholino-resistant form of the full-length UXT mRNA (Fig. 3G,H,K), but were not rescued by injection of the mutant UXT-2M mRNA (Fig. 3I-K). These observations indicate that UXT plays an essential function in promoting angiogenesis.

To explore whether UXT affected the arterial-venous identity, we examined the expression of arterial and venous markers in 30 hpf zebrafish embryos. efnb2a, notch1b, notch3, hey2 and tbx20 are arterial markers that are selectively expressed in the dorsal aorta area, whereas venous markers, such as aplnra (also known as msr), flt4 and ephb4 are expressed exclusively in the PCV region (Fischer et al., 2004; Lawson et al., 2001; Saint-Geniez et al., 2003; Szeto et al., 2002; Thompson et al., 1998; Wang et al., 1998). The arterial markers efnb2a, notch1b and hey2 were expressed at a much higher level in UXT-deficient embryos than in control embryos (Fig. 3L).

To explore the role of UXT in determining the ISV cell fate, we carried out cell transplantation experiments. We isolated the cells from Tg(kdrl:EGFP)^s843 zebrafish into which either control MO or UXT MOs were injected, and transplanted these cells into wild-type host embryos at the sphere stage. The donor cells from Tg(kdrl:EGFP)^s843 were also labeled with lineage tracer in order to mark the amount of donor-derived cells in host embryos. The contribution of donor cells to different trunk vascular cell types was examined. GFP-positive donor cells from Tg(kdrl:EGFP)^s843 control embryos appeared at 30 hpf in all types of cells located in the trunk blood vessels (Fig. 3M-O). We confirmed this observation in 20 successfully transplanted embryos (Fig. 3S). By contrast, donor cells from UXT-deficient Tg(kdrl:EGFP)^s843 embryos appeared predominantly in the dorsal aorta (16/17 embryos, Fig. 3P-R), but were barely seen in the PCV (1/17 embryos). In addition, these donor cells were also observed in basal cells (10/17 embryos) and connector cells (4/17 embryos). Strikingly, no UXT-deficient donor cells had populated the DLAV (0/17 embryos, Fig. 3T). These results indicate that UXT is required cell autonomously for ISV formation. We also noticed that the phenotypes of UXT-deficient embryos echoed those observed following the activation of Notch signaling as reported recently (Siekmann and Lawson, 2007).

To elucidate the cellular function of UXT, we further investigated ISV formation by in vivo time-lapse confocal microscopy, using Tg(fli1:nEGFP)^y7 transgenic fish. Tip cells appeared from the dorsal aorta at around 19 hpf, then they migrated to the horizontal myoseptum (Fig. 4A, 21:31 and 23:43). The majority of tip cells underwent cell divisions two or three times. Ultimately, tip cells migrated dorsally and formed the DLAV normally during ISV formation (Fig. 4A, 26:02 and 29:26). By contrast, tip cells in UXT-deficient Tg(fli1:nEGFP)^y7 embryos sprouted late, at ∼23 hpf (Fig. 4B, 20 hpf and 23:28), and then initiated cell divisions at ∼28 hpf (Fig. 4B, 27:34). Remarkably, these tip cells did not migrate further over the horizontal myoseptum (Fig. 4B, 29:38).

We noticed that most ISVs in control embryos consisted of three or four cells (Fig. 4C), whereas those in UXT-deficient embryos contained only one or two cells (Fig. 4C). The migration speed of UXT-deficient tip cells (0.079±0.012 µm/min) was much lower than that of control ones (0.188±0.036 µm/min, P<0.005; Fig. 4D).

Given that UXT-deficient embryos displayed severe defects in cell migration and division, we speculated that UXT might regulate the cell cycle of endothelial cells. To test this, the endogenous UXT in human umbilical vein endothelial cells (HUVECs) was knocked down by using short hairpin (sh)RNA, and the alteration of the cell cycle was measured (Fig. 4E). Consistently, UXT-deficient HUVECs displayed a marked increase in the number of G0-G1 phase cells, and a corresponding decrease in the number of cells in S phase (Fig. 4E,F), which strongly suggested that knockdown of UXT blocked cell cycle at G0-G1 phase. This cell cycle blockage was relieved by treatment with a Notch signaling inhibitor, DAPT (Fig. 4E,F). Taken together, these data establish the crucial function of UXT in endothelial cell migration and division.

In light of the microarray analysis and the impact of UXT deficiency on Notch signaling activation, we reasoned that UXT might modulate the Notch signaling pathway. To explore this possibility, Notch-TP-1-luciferase reporter plasmids were introduced into UXT-deficient cells, in the presence or absence of NICD. Interestingly, knockdown of UXT potentiated the activation of the Notch-TP-1-luciferase reporter, whereas ectopic expression of UXT attenuated the same reporter expression (Fig. 5A). Notably, ectopic expression of UXT-2M displayed only a marginal effect on the expression of the Notch-TP-1-luciferase reporter (Fig. 5A), which was consistent with its inability to rescue the developmental abnormalities induced by the UXT morpholino (Fig. 1F). BMP signaling and Wnt signaling have been reported to be crucial signaling pathways modulating angiogenesis (Phng et al., 2009; Rothhammer et al., 2007). However, UXT did not affect the activation of BMP signaling or Wnt signaling, according to the results from a BRE reporter assay or a TOPflash reporter assay, respectively (Fig. 5B,C).

Next, we checked whether UXT could modulate Notch signaling in vivo. her6 and hey1 in zebrafish are homologous to HES1 and HEY1 in human, respectively. We observed that all of these genes were expressed at significantly higher levels at ∼18 hpf and 24 hpf when endogenous UXT expression was knocked down (Fig. 5D,E). We also evaluated other Notch target genes, such as hey2 and heyl, and obtained similar results (supplementary material Figs S6 and S7). In addition, we observed no obvious alterations in the expression of the signaling proteins in the Notch pathway (including notch1a, notch1b, rbpja and rbpjb) in control embryos, but a slightly lower level of notch1b expression at 24 hpf in UXT-deficient embryos (Fig. 5F-I). Interestingly, the expression of dll4 increased in UXT-deficient embryos (Fig. 5J), and reciprocally, the expression level of uxt was higher in dll4-deficient embryos (Fig. 5K). Consistently, vegfaa expression was downregulated in UXT-deficient embryos from 18 hpf to 30 hpf (Fig. 5L). These data indicate that UXT regulates Notch signaling per se.

To further address the action of UXT in Notch signaling, we performed a whole-embryo chromatin immunoprecipitation (E-ChIP) assay in 24 hpf zebrafish embryos, and the conserved promoter sequences of the RBP-Jκ binding site (5′-GTGGGAA-3′) (Ling et al., 1993) were analyzed. Consistently, UXT bound to the promoter regions of hey1 and her6, but not to the promoter regions of the control gene cdx4 (Fig. 5M). Interestingly, in notch1b-deficient embryos, an increase in UXT binding to the promoter regions of Notch target genes was observed (Fig. 5M).

Next, we explored whether UXT interacted with any components of the transcriptional protein complex. The endogenous UXT was pulled down by NICD in HUVECs (Fig. 6A), indicating that UXT interacted with NICD. In addition, UXT directly regulated the endogenous expression of Notch target genes, such as hes1, hey1 (Fig. 6B), hey2 and heyl (supplementary material Fig. S8). Knockdown of UXT enhanced the expression of Notch-responsive genes significantly, whereas the expression level of these genes was remarkably reduced by ectopic expression of UXT in HUVECs. To prove that UXT directly regulates Notch-responsive genes, we crossed the transgenic Tg(TP1:mCherry) fish with Tg(fli1:EGFP)^y1 fish, and observed a robust activation of the TP1 reporter in UXT-deficient embryos (Fig. 6C; supplementary material Fig. S10).

Next, HA-UXT was co-introduced with either Flag-NICD or Flag-RBP-Jκ into HEK293T cells. The co-immunoprecipitation experiments revealed that UXT strongly interacted with NICD, but interacted only marginally with RBP-Jκ (Fig. 6D). Consistently, UXT-2M failed to interact with either NICD or RBP-Jκ, indicating the functional relevance of the interaction between UXT and NICD.

In addition, knocking down or ectopically expressing UXT did not affect the stability of either Notch1 or RBP-Jκ. Notably, ectopically expressed UXT impaired the interaction between NICD and RBP-Jκ, as demonstrated by the endogenous immunoprecipitation experiment (Fig. 6E). Furthermore, UXT did not affect the cleavage or stability of the Notch1 receptor (supplementary material Figs S11 and S12).

To determine the mechanism by which UXT modulates Notch signaling activation, we generated several Flag-NICD truncation mutants (Fig. 6F). Each individual truncation mutant was co-expressed with HA-UXT in HEK293T cells, and then immunoprecipitated by using an anti-Flag antibody. We observed that HA-UXT was not pulled down by a transactivation region (TAD) domain deletion mutant of NICD (ΔTAD-Flag, Fig. 6G), indicating that the TAD domain of NICD mediates its interaction with UXT. To corroborate, deletion of the RAM or TAD domain of NICD significantly impaired the NICD-induced TP-1 activation (Fig. 6H), because the RAM domain mediates the interaction between NICD and RBP-Jκ. As expected, co-transfection of UXT with ΔTAD-NICD did not result in any further repression of Notch signaling activation. These data demonstrate that UXT directly interacts with the TAD domain of NICD, thus serving as a negative regulator of the Notch signaling.

To substantiate whether UXT modulates angiogenesis by targeting Notch signaling, an in vitro angiogenesis assay was employed. HUVECs were seeded on Matrigel and they migrated to establish capillary-like structures with a lumen. As expected, control HUVECs formed normal capillary-like structures (Fig. 7A, shNT and phage-vec). By contrast, UXT-deficient HUVECs became rounded as isolated cells and failed to establish contacts with adjacent cells (Fig. 7A, sh-UXT). When HUVECs ectopically expressed UXT (Fig. 7A, phage-UXT), these cells formed more-robust capillaries and cord-like structures (Fig. 7A, phage-UXT). Quantitatively, ectopic expression of UXT induced a 1.5-fold increase in the mean mesh area (Fig. 7B) and branching length (Fig. 7C), as compared with the control HUVECs. Consistently, ectopic expression of UXT-2M caused a marginal dominant-negative effect on the angiogenesis (Fig. 7A, phage-UXT-2M; Fig. 7B,C).

We further employed a three-dimensional (3D) angiogenesis assay based on cytodex microcarriers. UXT was either knocked down or overexpressed in HUVECs, and cytodex beads were coated with these cells and then embedded into collagen gels. The outgrowth of capillary-like structures was assessed. Consistent with the reduced tube formation in the Matrigel assay, silencing of UXT expression caused the loss of sprouting activity. By contrast, ectopic expression of UXT markedly potentiated the sprouting from cytodex beads (Fig. 7D-F).

Furthermore, we treated HUVECs with DAPT or DMSO. DAPT is a γ-secretase inhibitor that prevents the S3 cleavage of Notch and mimics the inhibition of Notch signaling. Notably, DAPT treatment reversed the defects caused by knocking down endogenous UXT (Fig. 7Gii,iv), which was substantiated by quantifying the increase in the mesh area (Fig. 7H) and total branching length (Fig. 7I) during in vitro angiogenesis.

Finally, to substantiate the function of UXT in modulating Notch signaling in vivo, we treated Tg(kdrl:EGFP)^s843 embryos with 60 µM DAPT or DMSO at 12 hpf, and then observed ISV formation by confocal microscopy at 30 hpf. For the control embryos, DAPT treatment led to morphological defects of prolonged trunks and upward curly tails (Fig. 7N). Notably, ISVs displayed excessive growth (Fig. 7O). For UXT morphants, DAPT treatment partially rectified abnormal morphology, resulting in approximately normal trunks (Fig. 7P) and ISV growth (Fig. 7Q,V). In addition, DAPT treatment relieved the cells from the cell cycle blockage in UXT-deficient HUVECs (Fig. 4E,F). Consistently, injection of the RBP-Jκ morpholino caused the robust growth of ISVs, which phenocopied the results of DAPT treatment on embryos (Fig. 7R,S). Interestingly, co-injection with RBP-Jκ morpholino partially rescued the ISV growth defects in UXT-deficient embryos (Fig. 7T-V). Collectively, these data establish that UXT potentiates angiogenesis by attenuating Notch signaling per se.

Angiogenesis is the growth of new blood vessels from pre-existing ones. It is a major challenge to identify and characterize pro- and/or anti-angiogenesis signals, and to elucidate how these signals influence the corresponding cellular responses in blood vessel formation. In particular, it remains elusive how the relevant signaling is modulated temporally and spatially in developmental and tumor angiogenesis. In this study, we have characterized UXT (also known as ART27) as a novel modulator that potentiates angiogenesis in vivo and in vitro, by inhibiting Notch signaling.

The sprouting of ISVs in zebrafish is a well-established model for studying angiogenesis. Notch and VEGF signaling play major roles in the vascular cell fate determination (Phng and Gerhardt, 2009; Williams et al., 2006). Tip cells are characterized by the expression of higher levels of mRNA encoding DLL4 and VEGFR2 (also known as KDRL – ZFIN), whereas stalk cells are rich in Notch and VEGFR1 (also known as FLT1 – ZFIN). In this study, we observed that the mRNA expression level of both flt1 and kdrl were significantly lower in UXT-deficient embryos at 22 hpf, but the expression of vegfaa apparently did not change at that time (Fig. 5L). In addition, the expression of notch1b was attenuated in UXT-deficient embryos as early as 12 hpf (Fig. 5F), whereas the expression of hey1 and her6 was increased in UXT-deficient embryos from 18 hpf to 24 hpf. Interestingly, we did observe the alteration of vegfaa expression after 48 hpf in UXT-deficient embryos. It has been reported that VEGF induces DLL4 (Lobov et al., 2007), whereas Notch selectively inhibits VEGFR expression (Williams et al., 2006). Interestingly, we observed the reduced expression of kdrl in UXT-deficient embryos at 22 hpf. The expression of vegfaa but not vegfab (supplementary material Fig. S9) was decreased during the time window between 18 hpf to 30 hpf, whereas the expression of Notch target genes and dll4 was markedly enhanced. The cell-autonomous role of Notch and VEGF signaling needs to be further addressed in future studies. It remains to be explored how UXT modulates VEGF signaling indirectly.

UXT interacts with NICD endogenously. Ectopic expression of wild-type UXT markedly inhibited the expression of Notch target genes, whereas ectopic expression of UXT-2M (L50/59P) failed to do so. Notch1-NICD contains a RAM (RBP-Jκ-associated-molecule) domain, an Ankyrin (ANK) domain with seven repeats, a TAD domain and a C-terminal PEST sequence (Choi et al., 2011). We identified that UXT interacts mainly with the TAD domain of NICD. Consistently, when using this TAD truncation mutant as the stimulus, UXT failed to repress the Notch-dependent activation, indicating that UXT potentiates angiogenesis by impairing Notch signaling. This is consistent with the reports that Notch signaling attenuates ISV development.

RBP-Jκ represses Notch target genes by recruiting transcriptional co-repressors, which include SMRT, SHARP, SKIP and CIR in mammals (Hsieh et al., 1999; Kao et al., 1998; Zhou et al., 2000). SMRT and SHARP directly interact with RBP-Jκ, whereas SKIP binds both SMRT and NICD. Unexpectedly, our E-ChIP assays revealed that UXT could constitutively bind the promoter region of Notch target genes. Furthermore, UXT predominantly interacted with NICD, but interacted only marginally with RBP-Jκ. Because DAPT directly reduced the endogenous NICD level, this treatment could rescue the angiogenesis defects caused by deficiency of UXT in vitro or in vivo. Knockdown of RBP-Jκ could also achieve the similar effects. Thus, we propose that UXT is another basal co-repressor of Notch target genes. Once NICD enters the nucleus, UXT is recruited onto the promoter sequences of Notch target genes. UXT interacts with the TAD domain of NICD and competes NICD away from RBP-Jκ, thus potently inhibiting the induction of Notch target genes. Future biophysical studies will hopefully provide mechanistic details for this hypothesis.

Angiogenesis is an indispensable vehicle for tumor metastasis, allowing oxygen and nutrients to be replenished and cancer cells to be mobilized (Carmeliet and Jain, 2011). Notch mutations were found in ∼40-70% of T-cell acute lymphoblastic leukemias, and in 10-15% of chronic lymphocytic leukemias and mantle cell lymphomas (Carmeliet and Jain, 2011). Notch induced cell cycle arrest in a variety of cell types, which could be reversed by mutations of Notch signaling (Sriuranpong et al., 2001). Notably, UXT is highly expressed in various tumor tissues, and serves as a co-activator of androgen receptor in prostate cancer. We have shown, in the context of angiogenesis, that deficiency of UXT blocks cell cycle at the G0-G1 stage, whereas DAPT treatment relieves cells from the blockage, demonstrating that UXT modulates the migration and division of endothelial cells through Notch signaling. It is intriguing for future studies to explore whether UXT promotes tumor metastasis by potentiating angiogenesis.

UXT is an α-class prefoldin family protein (Vainberg et al., 1998). The crystal structure of this protein is unknown. Bioinformatics analysis predicts that UXT harbors a large stretch of coil-coiled domain. Previously, we characterized UXT as a transcriptional co-factor for NF-κB, which plays important roles in inflammation and innate immunity. However, no investigation has addressed the in vivo function of UXT. We had failed to obtain UXT knockout mice, due to embryonic lethality. We are trying to generate a UXT conditional knockout mouse, so as to address whether UXT could modulate angiogenesis in mammals and, if so, whether UXT could attenuate Notch signaling in mice. Taken together, the present study uncovers UXT as a novel repressor of Notch signaling during angiogenesis, highlighting the yet-to-know essential functions of UXT in vertebrate development.

Zebrafish maintenance, breeding and staging were performed as described previously (Kimmel et al., 1995). Wild-type (WT) AB, Tg(kdrl:EGFP)^s843 (Picker et al., 2002) and Tg(fli1: nEGFP)^y7 (Roman et al., 2002) zebrafish lines were obtained from Jiulin Du's (Institute of Neuroscience, Chinese Academy of Sciences, China) laboratory; Tg(TP1:mCherry) (Parsons et al., 2009) and Tg(fli1: EGFP)^y1 (Lawson and Weinstein, 2002) fish lines were obtained from the laboratory of F.L. The establishment and characterization of the transgenic lines has been described elsewhere (Chen et al., 2012; Yu et al., 2010).

Antisense riboprobes for UXT were synthesized using the DIG RNA Labeling kit (Roche) according to the manufacturer's instructions. Whole-mount in situ hybridization was carried out as described previously (Gan et al., 2008; Lawson et al., 2001), and staining was performed with an alkaline phosphatase substrate kit (Promega).

Embryos were deyolked as described previously (Link et al., 2006). Cell pellets were lysed with RIPA buffer (150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 1% sodium deoxycholate, 0.1% SDS, 100 mM Tris-HCl pH 7.4) supplemented with protease inhibitor cocktail (Roche), and sonicated. Protein samples were resolved by SDS-PAGE, probed with anti-zebrafish UXT (ABmart, 4B4; 1:500) and anti-mouse GAPDH (ABmart; 1:1000), and visualized by enhanced chemiluminescence.

Embryos were fixed in 4% paraformaldehyde (PFA), and whole-embryo antibody staining of UXT was performed as described previously (Liu et al., 2003).

In vivo confocal imaging was performed on the Tg(kdrl:EGFP)^s843 line at 28.5°C as described previously (Isogai et al., 2003). The length of ISVs and lumen diameters of dorsa aortas and posterior cardinal veins were calculated by using ImageJ analysis software (NIH). The calculation of the migration speed of tip cells was performed as described previously (Yu et al., 2010).

Transplantation was performed as described previously (Siekmann and Lawson, 2007; Yu et al., 2010). The donor Tg(kdrl:EGFP)^s843 was injected with 4 ng of control morpholino or 4 ng of UXT MO2; the Tg(kdrl:EGFP)^s843 was also injected with Rhodamine-561 dextran (Molecular Probes) as a lineage tracer at the one-cell stage.

Total RNAs were extracted from 50 zebrafish embryos using TRIzol reagent (Invitrogen) as described previously (Yue et al., 2009). All values were normalized to the level of GAPDH mRNA. The forward and reverse primers used are shown in supplementary material Table S2.

E-ChIP was performed as described previously (Wardle et al., 2006; Yue et al., 2009). The E-ChIP primer sequences are provided in supplementary material Table S2.

Cells were seeded in 24-well plates and transfected with reporter gene plasmids combined with siRNAs and other constructs, as described previously (Sun et al., 2007).

BrdU flow cytometry was prepared following the APC BrdU Flow Kit (BD) manual instructions. HUVECs were transfected with shRNA or phage-UXT packaged in lentivirus. The concentration of DAPT or DMSO was 1.5 μM. HUVECs were stained with 1 mM BrdU (BD) for 2 h, and cytometric analysis was carried out on a BD FACScalibur Flow Cytometer.

Co-immunoprecipitaion was performed as described previously (Rustighi et al., 2009). See supplementary material methods for details.

The Matrigel assay (Arnaoutova and Kleinman, 2010) and 3D angiogenesis (Nehls and Drenckhahn, 1995) assay were performed as described previously. For details, please see supplementary material methods.

DAPT (Sigma) was dissolved in DMSO (300 μM stock solution) and added to system water with a final concentration of 60 μM. System water containing DMSO alone was used as a control. The embryos were dechorionated and added to system water containing DAPT or DMSO at 12 hpf. In vivo confocal images were taken at 30 hpf.

Quantitative data are expressed as the mean±s.e.m. Statistical significance was determined by one-way analysis of variance. A P-value of less than 0.05 was considered statistically significant.

We thank Drs Jiulin Du (IOS, CAS, China), Congjian Xu (FuDan University, China) and Wilhelm Krek (ETH, Zürich, Switzerland) for transgenic zebrafish strains and reagents. We are grateful to Drs Yun Zhao, Xing Liu and Mi Li (SIBCB, CAS) for critical comments, and to Drs Peng Liu, Shanye Gu, Panpan Zhang and Yunbin Zhang (the Science and Technology Commission of Shanghai Municipality, 11140900100) for technical assistance. We thank Dr Maria Lopez-Ocasio (NIH) for kindly editing our manuscript.