Deubiquitinase USP10 regulates Notch signaling in the endothelium

See allHide authors and affiliations

Science  12 Apr 2019:
Vol. 364, Issue 6436, pp. 188-193
DOI: 10.1126/science.aat0778

Deubiquitinase fine-tunes Notch signaling

The branching of blood vessels is controlled in part by signaling through Notch receptor proteins. When NOTCH1 binds its ligand DLL4, the intracellular domain (NICD1) is cleaved from the receptor and works with other proteins to regulate gene transcription. In a screen for proteins that interacted with NICD1 in human cells in culture, Lim et al. identified the deubiquitinase USP10. NICD1 is rapidly ubiquitinated and degraded in cells, but interaction of NICD1 with USP10 counteracts ubiquitination of NICD1 and stimulates Notch signaling. Genetic experiments in mice support a role for USP10 in fine-tuning Notch signaling during vascular morphogenesis.

Science, this issue p. 188


Notch signaling is a core patterning module for vascular morphogenesis that codetermines the sprouting behavior of endothelial cells (ECs). Tight quantitative and temporal control of Notch activity is essential for vascular development, yet the details of Notch regulation in ECs are incompletely understood. We found that ubiquitin-specific peptidase 10 (USP10) interacted with the NOTCH1 intracellular domain (NICD1) to slow the ubiquitin-dependent turnover of this short-lived form of the activated NOTCH1 receptor. Accordingly, inactivation of USP10 reduced NICD1 abundance and stability and diminished Notch-induced target gene expression in ECs. In mice, the loss of endothelial Usp10 increased vessel sprouting and partially restored the patterning defects caused by ectopic expression of NICD1. Thus, USP10 functions as an NICD1 deubiquitinase that fine-tunes endothelial Notch responses during angiogenic sprouting.

To add new vessel sprouts to an existing vascular network, endothelial cells (ECs) specialize into leading tip cells and following stalk cells (1), which dynamically interchange positions (2). Signaling by the Delta-like 4 (DLL4) ligand and NOTCH1 receptor helps to coordinate these processes. Inhibition of signaling causes excessive sprouting and vessel overgrowth, whereas forced activation leads to vascular rarefaction (310). DLL4 induces the cleavage of NOTCH1 by the γ-secretase protease, which releases the NOTCH1 intracellular domain (NICD1). This fragment forms a ternary complex with the DNA binding protein recombining binding protein suppressor of hairless (RBPJ) and the co-activators mastermind-like proteins 1 to 3 (MAML1 to -3) to drive the transcription of Notch target genes (1113). In many cellular contexts, the half-life of NICD1 is short because NICD1 undergoes rapid proteasomal degradation mediated by ubiquitin ligases, such as SCFFBXW7 [the SKP1–CUL1–F-box protein complex (SCF) containing the substrate recognition component F-box– and Trp-Asp (WD) repeat domain–containing protein 7 (FBXW7)] (1416). Inactivation of FBXW7 increases the abundance of NICD1 levels and causes severe defects in vascular sprouting (1719). Having the right amount of NICD1 at the right time is thus critical for the cooperative behavior of ECs.

To identify regulators of NICD1, we examined the NICD1 interactome in human embryonic kidney (HEK) 293T cells by using stable isotope labeling with amino acids in cell culture (SILAC)–based quantitative proteomics (fig. S1A). Liquid chromatography–tandem mass spectrometry (MS/MS) analysis revealed that NICD1 interacted with core components of the Notch transcription complex, including RBPJ and MAML1 (Fig. 1A). Ubiquitin-specific peptidase 10 (USP10), a deubiquitinase that removes ubiquitin modifications from its protein targets, was also among the NICD1-immunoprecipitated proteins (Fig. 1, A and B). To confirm this, we used primary ECs, in which USP10 is an abundant nucleocytoplasmic protein (Fig. 1, C and D, and fig. S1, B to D). We immunoprecipitated NICD1 from cultured human umbilical vein ECs (HUVECs) that expressed a V5 epitope–tagged NICD1 (NICD1V5). Immunoblotting for USP10 demonstrated that endogenous USP10 associated with NICD1V5, as did the ternary complex factors RBPJ and MAML1 (Fig. 1E). This interaction also occurred when endogenous NICD1 was immunoprecipitated from HUVECs, as demonstrated by a DLL4-inducible NICD1-USP10 protein complex that was enriched in the nuclear fraction (Fig. 1F and fig. S1E). Together, these findings suggest that USP10 interacts with NICD1 in ECs.

Fig. 1 USP10 interacts with NICD1 in ECs.

(A) Two-dimensional (2D) plot with log2 heavy (H)/light (L) SILAC ratios of the quantified proteins, correlating to an enrichment in NICD1 immunoprecipitation (NICD1-IP) versus control-IP on the x axis and log10 signal intensity in arbitrary units on the y axis. (B) MS/MS spectra of light-labeled (gray) and heavy-labeled (red) peptides, showing doubly charged EPLGEDSVGLKPLK and EGLVPVSEDPVAIK peptides from NICD1 and USP10, respectively. (Single-letter abbreviations for the amino acid residues are as follows: A, Ala; D, Asp; E, Glu; G, Gly; I, Ile; K, Lys; L, Leu; P, Pro; S, Ser; V, Val.) m/z, mass/charge ratio. (C) Immunoblot analysis of USP10 protein expression in several EC types. HAEC, human aortic ECs; HMVEC, human microvascular ECs; HDLEC, human dermal lymphatic ECs; VEGFR2, vascular endothelial growth factor receptor 2; Mr, relative molecular weight; K, thousands. (D) USP10 protein expression in cytoplasmic (cyto) and nuclear (nuc) fractions isolated from control or DLL4-stimulated HUVECs. Histone H3 and tubulin were used as nuclear and cytoplasmic markers, respectively. (E) Coimmunoprecipitation of adenovirus-delivered NICD1V5 (AdNICD1V5) and endogenous USP10, MAML1, and RBPJ from HUVECs. AdCtrl, control adenovirus; IB, immunoblotting. (F) Coimmunoprecipitation of NICD1 and USP10 from HUVECs treated with or without DLL4 for 24 hours. Endogenous NICD1 was immunoprecipitated by anti-NOTCH1 antibody. IgG, immunoglobulin G.

To investigate the role of USP10 in endothelial Notch signaling, we performed knockdown experiments in HUVECs by using small interfering RNAs (siRNAs) and assayed Notch signaling with various Notch-responsive luciferase (luc) reporters. Treatment of scrambled control siRNA (siCtrl)–transfected HUVECs with DLL4 induced a γ-secretase inhibitor [dibenzazepine (DBZ)]–sensitive and RBPJ-dependent increase in Notch reporter gene activity (Fig. 2, A and B, and fig. S2B). Depletion of USP10 with USP10 siRNA (siUSP10) did not affect the baseline activity of the reporters but reduced the DLL4-induced increase in Notch activity (Fig. 2, A and B, and fig. S2, A and B), suggesting that USP10 regulates canonical Notch responses in ECs. These findings were corroborated by a genome-wide transcriptomic analysis of siCtrl- and siUSP10-transfected HUVECs stimulated with DLL4 or unstimulated for 24 hours, which revealed that depletion of USP10 reduced the expression of canonical Notch target genes, such as HEY1 [encoding Hes-related family basic helix-loop-helix (BHLH) transcription factor with YRPW (Tyr-Arg-Pro-Trp) motif 1], HEY2, HES1 (encoding Hes family BHLH transcription factor 1), and NRARP (encoding Notch-regulated ankyrin repeat protein), as well as other genes that are regulated by DLL4-Notch (20) (Fig. 2C and fig. S2C). Quantitative real-time polymerase chain reaction (qRT-PCR) analysis confirmed this finding (fig. S2, D and E).

Fig. 2 USP10 deficiency reduces endothelial Notch signaling.

(A) Activation of Notch-responsive elements from the Epstein-Barr virus terminal protein 1 promoter (TP1)–luc reporter in siCtrl- or siUSP10-transfected HUVECs after treatment with vehicle or DLL4, in the presence or absence of a γ-secretase inhibitor (DBZ) for 24 hours. (B) Activation of a multimerized RBPJ binding site (4×RBPJ)– or 4×RBPJ mutant (RBPJmut)–luc reporter in siCtrl- or siUSP10-transfected HUVECs after treatment with vehicle or DLL4 for 24 hours. (C) Heat map of differentially expressed genes in siCtrl- or siUSP10-transfected HUVECs treated with vehicle (Ctrl) or DLL4 for 24 hours. Data were obtained by microarray analysis performed in triplicate for each condition. Differentially regulated genes with a cutoff P value of <0.05 and a fold change of >2.0 (for siCtrl versus siCtrl plus DLL4) are listed. (D) Immunoblot analysis of NICD1 expression in siCtrl- or siUSP10-transfected HUVECs stimulated with DLL4 or vehicle for 6 and 24 hours. (E) qRT-PCR analysis of NOTCH1 mRNA expression in control and USP10 knockdown HUVECs. Values are represented relative to the expression of β-actin. (F) NICD1 levels in control and USP10 knockout ECs generated by lentiviral transduction of HUVECs with control (gCtrl) or USP10-targeting (gUSP10) guide RNAs (gRNAs) as well as a FLAG-tagged CAS9 nuclease. HUVECs were stimulated with vehicle or DLL4 for 24 hours. (G) NICD1 immunoblot analysis of HUVECs transduced with a control lentivirus (LentiCtrl) or a GFP-tagged USP10–encoding lentivirus (LentiUSP10GFP). NICD1 abundance was determined after 24 hours of DLL4 stimulation. Error bars, mean ± SEM for n > 3 samples. ***P < 0.001; ****P < 0.0001; NS, not significant.

Loss of USP10 did not lead to compensatory expression changes of other deubiquitinases (fig. S2A), nor did we detect expression changes in core Notch signaling components in unstimulated HUVECs that might explain the reduction in pathway activity (fig. S2F). We surmised that USP10 regulates Notch signaling posttranscriptionally and focused on the short-lived NICD1 protein. Immunoblot analysis showed that the NICD1 protein concentration was low under basal conditions but that the protein became abundant in response to short (6-hour) or prolonged (24-hour) stimulation with DLL4 (Fig. 2D and fig. S3A). Inactivation of USP10 by siRNA or CRISPR-Cas9 gene editing reduced the abundance of NICD1 in response to DLL4 stimulation without affecting basal NOTCH1 mRNA expression (Fig. 2, D to F, and fig. S3, A to D). Similar results were obtained with the Notch ligands Delta-like 1 (DLL1) and Jagged 1 (JAG1), which trigger distinct Notch responses and signaling dynamics (8, 21). However, DLL1 and JAG1 were weaker NOTCH1 agonists than DLL4 in HUVECs, and higher concentrations were needed for receptor activation (fig. S3, E to G). Depletion of USP7, USP11, or USP15, three USP family members that associate with NICD1 in cancer cells (22), did not change NICD1 protein expression (fig. S4A). Thus, USP10 regulates endothelial Notch signaling by controlling NICD1 abundance, though we cannot exclude the possibility that alternative mechanisms may contribute to the observed effects on the Notch pathway.

This regulation appears to occur at the level of the NICD1 (downstream of γ-secretase cleavage) because the amounts of ectopically expressed NICD1V5 were also reduced in USP10-deficient ECs (fig. S4B). Overexpression of USP10, on the other hand, enhanced the abundance of NICD1 in HUVECs (Fig. 2G and fig. S4C). Because USP10 is a deubiquitinase, we tested whether USP10 interferes with NICD1 protein degradation. Treatment of ECs with the proteasome inhibitor MG132 or depletion of FBXW7 increased the abundance of NICD1 protein in USP10-deficient HUVECs (fig. S5, A and B). Moreover, measurement of the rate of NICD1 decay in cycloheximide-treated cells showed that the half-life of endogenous NICD1 protein was decreased upon depletion of USP10 (Fig. 3A). Depletion of endothelial USP10 did not alter the abundance of cellular tumor antigen p53 or sirtuin 6 (SIRT6) (fig. S5C), two substrates of USP10 in tumor cells (23, 24), suggesting that the signaling context in primary ECs differs from that in cancer cell lines.

Fig. 3 USP10 regulates NICD1 stability by altering its ubiquitin-dependent turnover.

(A) NICD1 abundance in HUVECs transfected with siCtrl or siUSP10. After 24 hours, cells were stimulated with DLL4 after pretreatment with cycloheximide. ECs were collected at 1-hour intervals, and NICD1 protein levels were analyzed by immunoblotting. The fold change in NICD1 levels is shown on the right. (B) NICD1 ubiquitination pattern in control (gCtrl) or USP10 knockout (gUSP10) HUVECs stimulated with vehicle or DLL4 for 24 hours. Cell lysates were incubated with TUBEHalo to enrich ubiquitinated proteins. Ubiquitinated NICD1 was detected by immunoblotting using an NICD1-specific antibody. (C) Endogenous NICD1 ubiquitination in cytoplasmic (Cyto) and nuclear (Nuc) fractions derived from control (gCtrl) or USP10 knockout (gUSP10) HUVECs stimulated with vehicle or DLL4 for 24 hours. (D) Ubiquitinated NICDV5 was purified by anti-V5 immunoaffinity chromatography from HEK293T cells transfected with NICD1V5, FBXW7FLAG, and UbHA and then incubated in vitro with vehicle, recombinant USP10 (rUSP10), or the recombinant catalytic core of USP2 (rUSP2 cc). Levels of NICD1 ubiquitination were determined by immunoblotting for HA. Error bars, mean ± SEM for n > 3 samples. ***P < 0.001; ****P < 0.0001.

To test whether USP10 modulates NICD1 ubiquitination in ECs, we captured ubiquitinated proteins from HUVECs by using Halo-tagged tandem ubiquitin binding entities (TUBEHalo) (25) followed by immunoblotting with an NICD1 antibody. This analysis revealed a high-molecular-size ubiquitin smear in DLL4-stimulated ECs that was enhanced by MG132 treatment (fig. S5D). Coexpression of USP10 decreased this ubiquitin signal (fig. S5E), suggesting that USP10 can deubiquitinate NICD1. Consistent with these data, the depletion of USP10 in DLL4-stimulated HUVECs enhanced the endogenous ubiquitination of NICD1 (Fig. 3B). This increase in NICD1 ubiquitination was observed in the nuclear fraction of HUVECs (Fig. 3C), where NICD1 is targeted by ubiquitin ligases such as SCFFBXW7 (1416) and where NICD1 and USP10 interact. To finally assess whether USP10 deubiquitinates NICD1, we purified ubiquitinated NICD1 from HEK293T cells that were cotransfected with NICD1V5, hemagglutinin (HA)-tagged ubiquitin (UbHA), and FLAG-tagged FBXW7 (FBXW7FLAG) (fig. S5F). Incubation of ubiquitinated NICD1 with recombinant USP10 under conditions allowing NICD1 deubiquitination by the promiscuous catalytic core of USP2 led to decreased polyubiquitination of NICD1 (Fig. 3D), suggesting that USP10 functions as an NICD1 deubiquitinase.

To understand the physiological role of USP10 in the regulation of Notch signaling, we generated mice in which exons 7 and 8 of Usp10 were flanked by loxP sites (Usp10fl/fl mice) (fig. S6A). Germline deletion of Usp10 (yielding the genotype Usp10−/−) caused perinatal lethality (fig. S6B) (26). To bypass lethality, we crossed Usp10fl/fl mice with mice expressing a tamoxifen-inducible cre deleter (creERT2) that is driven by the EC-selective platelet-derived growth factor subunit B (Pdgfb) promoter (Pdgfb-creERT2 deleter mice) and assessed vascular phenotypes in the retinas of the resulting mutants (Usp10iEC-KO). The mouse retina is a well-established model of Notch-dependent angiogenesis. We injected 4-hydroxytamoxifen (4-OHT) from postnatal day 1 (P1) to P4 and analyzed retinal angiogenesis at P7 (fig. S6C). Endothelially restricted deletion of Usp10 increased EC density and sprouting at the angiogenic front but had little effect or even opposite effects in the central parts of the retinal vasculature (Fig. 4, A and B). At the front, Usp10iEC-KO retinas displayed increased numbers of endothelial sprouts and filopodia (Fig. 4, C, D, and F) and showed enhanced immunostaining for EC-specific molecule 1 (ESM1) (Fig. 4E and fig. S6D), a tip cell marker that is suppressed by DLL4-Notch signaling (fig. S2C) (9, 27). In Usp10 mutant mice, ECs accumulated in veins and perivenous capillaries and veins were enlarged (fig. S6, D and E). Together, these phenotypes resemble aspects of reduced DLL4-NOTCH1 signaling in the retinal endothelium (3, 5, 6, 9) and indicate that USP10 deficiency reduces Notch responses in vivo.

Fig. 4 Deletion of Usp10 in ECs increases angiogenic sprouting and partially restores vascular defects caused by forced NICD1 expression.

(A) Immunofluorescence staining for platelet EC adhesion molecule (PECAM) in P7 mouse retinas of 4-OHT–injected control (Usp10fl/fl) and Usp10iECKO (Pdgfb-creERT2;Usp10fl/fl) mice. A, artery; V, vein. (B) Images of PECAM (red)– and erythroblast transformation–specific–related gene (ERG) (cyan)–stained P7 retinas of control and Usp10iEC-KO mutants showing clustering of ECs at the angiogenic front. (C) Higher-magnification confocal images of PECAM-labeled P7 retinas isolated from control and Usp10iEC-KO mice. Endothelial filopodia are denoted by the small red dots. (D) PECAM (blue), intracellular adhesion molecule 2 (ICAM) (green), and collagen IV (COL) (red) labeling of control and Usp10iEC-KO retinas at P7. The vasculature at the angiogenic front is shown. (E) Confocal images showing PECAM (gray), ERG (red), and ESM1 (yellow) immunostaining of the retinal vasculature in control and Usp10iEC-KO mutants at P7. ESM1 immunostaining behind the angiogenic front was enhanced in Usp10iEC-KO mice. (F) Quantifications of vascular parameters at the angiogenic front in P7 control and mutant retinas as indicated. (G) Confocal images of PECAM-labeled (gray) P6 retinas in control and NICD1iEC-OE mice. (H) Immunofluorescence staining for PECAM (blue), ICAM2 (green), and collagen IV (COL) (red) in P6 retinas from 4-OHT–injected control and NICD1iEC-OE mice. (I and J) Confocal images (I) of PECAM (red)– and ERG (cyan)–labeled retinas in control mice, NICD1iEC-OE mice, and NICD1iEC-OE; Usp10iEC-KO double mutants. The quantification (J) of the endothelial coverage at the angiogenic front in the respective genotypes is shown. Error bars, mean ± SEM for n > 3 samples. **P < 0.01; ****P < 0.0001.

We tested whether USP10 inactivation would restore endothelial sprouting behavior when Notch signaling was aberrantly activated. We used a Cre-inducible Notch gain-of-function mouse model in which NICD1 is expressed from the ubiquitously active cytomegalovirus early enhancer–chicken β-actin (CAG) promoter (fig. S7A). The expression of NICD1 [as well as those of membrane-targeted Tomato and nucleus-targeted green fluorescent protein (GFP)] is controlled by a floxed stop cassette, which can be removed by Cre-mediated excision (yielding Rosa26NICD1) (fig. S7, A and B). Pdgfb-creERT2–mediated recombination in Rosa26NICD1 (NICD1iEC-OE) mice led to severe retinal vascular defects, which correlated with the expression of the fluorescent reporters (Fig. 4, G and H, and fig. S7B). The vascular network in NICD1iEC-OE mutant mice was sparse and had fewer vessel sprouts (Fig. 4, G and H, and fig. S7, C and D), a phenotype that is consistent with increased endothelial Notch activity (9, 19). We also combined Rosa26NICD1, Usp10flox, and Pdgfb-creERT2 alleles (NICD1iEC-OE;Usp10iEC-KO) to generate compound mutants. Deletion of Usp10 reduced the sprouting defects caused by forced endothelial expression of NICD1 (Fig. 4, I and J). Compared with the NICD1iEC-OE mice, NICD1iEC-OE;Usp10iEC-KO double mutants had an increased number of vessel sprouts and branches, giving rise to a more dense vascular network at the angiogenic front (Fig. 4, I and J). Thus, USP10 deficiency can, at least in part, restore vessel density under conditions of enhanced NICD1 signaling, suggesting that USP10 is a regulator of Notch-dependent vascular morphogenesis in vivo.

In summary, we identify USP10-mediated deubiquitination of NICD1 as a mechanism to delay the rate of NICD1 degradation, which prolongs cellular Notch responses. Such regulation appears to be particularly relevant for angiogenic sprouting, during which dynamic changes in Notch activity determine the specification and position of tip and stalk cells (28). USP10 is itself a regulated protein, and recent studies reported that adenosine monophosphate (AMP)–activated kinase (AMPK)–mediated phosphorylation activates the enzyme (29). Because AMPK senses cellular energy levels and is activated by hypoxia and nutrient limitation, USP10 may be particularly active in ECs that face oxygen and nutrient deprivation, as they do when invading avascular tissues. We observed increased USP10 activity in ECs that were cultured in hypoxic or nutrient-starved conditions (fig. S8, A and B). These considerations could explain why the loss of USP10 in mice causes Notch loss-of-function phenotypes only at the sprouting front and not in the central plexus, where ECs are exposed to physiological oxygen and nutrient levels. Further studies on the regulation of USP10 may reveal in which other cell types and (patho)physiological contexts the USP10-dependent modulation of Notch signaling is at play (30).

Supplementary Materials

Materials and Methods

Figs. S1 to S8

References (3143)

References and Notes

Acknowledgments: We thank J. Yuan, Z. Lou, and N. Popov for expression plasmids and M. Gyrd-Hansen for advice on the TUBE assays. Funding: This work is supported by the Max Planck Society, European Research Council (ERC) starting grant ANGIOMET (311546), ERC consolidator grant EMERGE (773047), the Deutsche Forschungsgemeinschaft (SFB 834), the Excellence Cluster Cardiopulmonary System (EXC 147/1), LOEWE grant Ub-Net, the DZHK (German Center for Cardiovascular Research), the Stiftung Charité, the Cardio-Pulmonary Institute (EXC 2026project ID 390649896), and the European Molecular Biology Organization (EMBO) Young Investigator Programme. Author contributions: R.L., T.S., H.N., J.A., B.Z., C.S., A.D., Y.T.O., K.W., J.W.D.F., A.E., M.Ka., K.H., T.Bo., and S.G. performed experiments. R.L., T.S., H.N., J.A., B.Z., C.S., A.D., Y.T.O., K.W., J.W.D.F., A.E., M.Ka., K.H., T.Bo., S.G., and M.P. analyzed data. A.E., M.Ka., K.H., T.Br., M.Kr., R.B., and I.D. provided essential reagents and protocols. M.P. supervised the study. R.L., T.S., I.D., and M.P. wrote the paper. All authors discussed the results and commented on the manuscript. Competing interests: The authors declare no competing financial interests. Data and materials availability: The Rosa26NICD1 mice are available from R.B. under a material agreement with the CNIC. The proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository under accession number PXD013192. The microarray data have been deposited in the ArrayExpress repository under accession number E-MTAB-7774. Data from the RNA sequencing analysis have been deposited in NCBI Gene Expression Omnibus under accession numberGSE128179.
View Abstract

Stay Connected to Science

Navigate This Article