Endosomal sorting of Notch receptors through COMMD9-dependent pathways modulates Notch signaling

Copper metabolism MURR1 domain containing (COMMD) proteins are a group of highly conserved factors defined by the presence of a unique C-terminal homology domain (Burstein et al., 2005). Ten family members can be identified from mammals to unicellular protozoa (Maine and Burstein, 2007), but little is known about their cellular functions and the underlying reason for their conservation and diversification. Most of our understanding is centered on COMMD1, the first identified member of this family that was initially noted to be the site of a recessive mutation that results in copper toxicosis in a particular dog breed, the Bedlington terrier (van de Sluis et al., 2002).

The mechanism for the accumulation of copper in these animals was initially unclear; however, an interaction between COMMD1 and the copper transporter ATP7B was reported early on (Tao et al., 2003). Recently, we demonstrated that COMMD1 regulates the endosomal sorting of the copper transporter ATP7A, such that in the absence of COMMD1, ATP7A is trapped in endosomal vesicles and lacks copper-dependent trafficking to the trans-Golgi network and plasma membrane (Phillips-Krawczak et al., 2015). In addition to its control of ATP7A/7B trafficking, COMMD1 has been linked to the regulation of other transporters, including epithelial sodium channel, cystic fibrosis transmembrane conductance regulator, and sodium-potassium-chloride cotransporter 1 (Biasio et al., 2004; Drévillon et al., 2011; Smith et al., 2013). However, whether these other transporters are similarly regulated at the level of endosomal sorting remains to be examined. Furthermore, COMMD1 has also been linked to other pathways that are seemingly not connected to the endolysosomal system, including nuclear factor κB regulation (Maine et al., 2007; Starokadomskyy et al., 2013) and hypoxia adaptation (van de Sluis et al., 2010).

The role of COMMD1 in endosomal sorting is linked to its incorporation into a larger complex containing the coiled-coil proteins CCDC22 and CCDC93 (Phillips-Krawczak et al., 2015). This COMMD–CCDC22–CCDC93 (CCC) complex localizes to endosomes through interactions between CCDC22 and CCDC93 with FAM21 (Harbour et al., 2012; Freeman et al., 2014; Phillips-Krawczak et al., 2015), a component of the Wiskott-Aldrich syndrome protein and scar homolog (WASH) complex (Derivery et al., 2009; Gomez and Billadeau, 2009). WASH is a member of the Wiskott-Aldrich syndrome protein family of actin nucleation-promoting factors (Rottner et al., 2010). Through its ability to recruit and activate the ubiquitously expressed Arp2/3 complex, WASH promotes endosomal actin deposition and is required for the maintenance of the architecture of the endolysomal system and to facilitate receptor trafficking (Derivery et al., 2009, 2012; Gomez and Billadeau, 2009; Jia et al., 2010; Gomez et al., 2012).

COMMD proteins are known to interact with each other through their COMM domains (Burstein et al., 2005); moreover, all COMMD family members have been shown to interact with the CCC complex component CCDC22 (Starokadomskyy et al., 2013). However, whether other COMMD proteins form distinct CCC entities and can mediate unique functions that distinguish them from COMMD1 has not been elucidated. At the present time, studies addressing functions of other COMMD proteins are scant and include roles in epithelial sodium channel regulation (Liu et al., 2013), regulation of cell proliferation and repair in kidney injury models (Devlin et al., 2003), effects on hepatocellular carcinoma (Zheng et al., 2012), and regulation of IκB protein degradation (Starokadomskyy et al., 2013).

In this study, we performed an extensive functional analysis of COMMD9, a COMMD family member that is frequently deleted in WAGR syndrome (MIM 194072), a congenital condition caused by a microdeletion and haploinsufficiency at 11p13 (Xu et al., 2008). We report that COMMD9 is incorporated into the CCC complex and plays a critical role in endosomal sorting of Notch family members. Upon ligand binding, Notch proteins undergo proteolytic cleavage to release the Notch intracellular domain, which in turn translocates to the nucleus and along with cofactors binds gene-promoter regions of chromatin to regulate gene expression (Kopan, 2012). Furthermore, extensive work in Drosophila and other models has uncovered that trafficking of Notch proteins through the endolysosomal system is critically important in regulating their activity (Wilkin et al., 2008; Fortini and Bilder, 2009; Kandachar and Roegiers, 2012; Troost et al., 2012). Here we describe that in the absence of COMMD9, Notch expression is reduced at the cell surface and Notch2 is missorted into cytosolic vesicles from where it can reach lysosomes, resulting in reduced Notch-dependent signaling. Furthermore, we show that Commd9 is critically required during mammalian development and that it functions as part of a unique CCC complex.

To begin to understand what unique function COMMD9 might mediate, we used tandem affinity purification to define its protein interactome. To that end, COMMD9 was doubly tagged with a tandem HA tag in its N terminus and a short biotinylation sequence in its C terminus. The protein was immunopurified with an HA antibody first, and on elution, it was purified again using streptavidin resin. The final material, when resolved by SDS-PAGE and stained with silver nitrate, demonstrated several unique bands in addition to the purified bait (Fig. 1 A). These bands were excised and subjected to trypsin digestion for proteomic identification by liquid chromatography tandem mass spectrometry (LC/MS-MS). This analysis identified 37 interacting proteins with high confidence (Table S1). CCDC22, CCDC93, and COMMD5, components of a recently identified regulator of endosomal protein sorting (Starokadomskyy et al., 2013; Phillips-Krawczak et al., 2015), were found in this purification (Fig. 1 B). The Rab proteins RAB5C, RAB7A, and RAB11B, which are present in different endosomal subcompartments, were also found. In addition, several transmembrane surface proteins were also identified (Fig. 1 B). Among these proteins, Notch2 was identified through four peptides mapping to both the N-terminal extracellular portion of the receptor and its C-terminal tail. This receptor was of interest to us because of its important developmental roles (High and Epstein, 2008).

The interaction between COMMD9 and Notch2 was readily recapitulated by coimmunoprecipitation, which also detected that COMMD9 interacts with members of the CCC complex, as expected (Fig. 1 C). Furthermore, COMMD9 also coimmunoprecipitated Notch1, another family member (Fig. 1 D); in comparison, Notch3 was not coprecipitated to a significant degree (Fig. 1 E).

The CCC complex is a regulator of endosomal sorting events mediated by retromer and WASH, including the endosome-to-surface recycling of the copper transporter ATP7A (Phillips-Krawczak et al., 2015). With this in mind, we reasoned that COMMD9 might similarly regulate the recycling of Notch2 to the plasma membrane. To assess this possibility, we screened a series of Notch2 antibodies and cell lines to identify suitable reagents that allow immunofluorescence staining of the endogenous protein. Among different laboratory variants of HeLa cells, we identified a subline with high expression of Notch2 that is suitable for immunofluorescence staining. Next, we derived COMMD9 knockout clones of this cell line using CRISPR/Cas9 gene editing technology using two separate guide RNA sequences targeting exon 1 of COMMD9 (Fig. S1 A). All three clones had no detectable protein expression (Fig. 2 A), and genomic analysis demonstrated short deletions in the targeted region (Fig. S1, B and C). In these cells, we identified that the normal surface staining pattern of Notch2 was largely lost, and instead, this was replaced by prominent intracellular vesicular staining for Notch2 in all three clones (Fig. 2 B). Furthermore, although the overall Notch2 protein expression was only modestly decreased in these cells, surface protein biotinylation demonstrated a substantial reduction of Notch2 in this fraction (Fig. 2, C and D), consistent with the reduced plasma membrane staining pattern noted by immunofluorescence staining. In addition, we also examined the nature of the intracellular Notch2 accumulation seen in COMMD9-deficient cells. The intracellular Notch2 puncta in COMMD9-deficient cells partly colocalized with EEA1, and to a lesser extent with VPS35 (Fig. S2, A and B). Consistent with the notion that these puncta represent early endosomes, Rab5 colocalized with Notch2 in COMMD9-deficient cells, whereas the colocalization with the recycling endosome marker Rab11 was poor (Fig. S2 C).

Next, expression of COMMD9 was restored in two of the targeted clones to ensure that these observations were indeed specific to COMMD9 targeting (Fig. 2 E). Surface biotinylation experiments demonstrated that the rescued lines had increased Notch2 levels at the cell surface (Fig. 2, F and G), indicating that this cellular phenotype is rescued by COMMD9 re-expression. Furthermore, silencing of CCDC93 or C16orf62 (two additional components of the CCC complex) also led to reduced plasma membrane staining for Notch2, a pattern also observed on silencing of VPS35 (Fig. S3, A and B), a subunit of the retromer complex which is important for WASH recruitment and function (Harbour et al., 2012; Jia et al., 2012). Altogether, these findings indicate that the absence of COMMD9 and associated factors results in the accumulation of Notch2 in the early endosomal compartment.

Next we assessed the effects of COMMD9 on Notch-dependent gene expression. COMMD9-deficient HeLa cell clones displayed decreased basal expression of several Notch-dependent genes, but this HeLa subline proved to be poorly responsive to stimulation with exogenous Jagged1 ligand (unpublished data). Therefore, we used alternative models to investigate this question. First, U2OS cells with stable silencing or stable overexpression of COMMD9 were used (Fig. 3 A). Using Jagged1 as ligand for Notch stimulation, we noted that COMMD9 silencing led to reduced induction of HEY1, a Notch target gene, whereas overexpression of COMMD9 had the opposite effect (Fig. 3 B). Furthermore, these cells were transfected with a Notch-responsive luciferase reporter containing a promoter with multimerized RBPj binding elements. Here we observed that the effects of COMMD9 expression on the Notch reporter were concordant with the observed effects on HEY1 expression (Fig. 3 C), consistent with the notion that endogenous HEY1 gene expression was indeed affected through a Notch-dependent event.

For further analysis, we used mouse embryo fibroblasts (MEFs) derived from a gene trap Commd9-deficient model, where a gene trap cassette was inserted between exons 5 and 6 of the mouse Commd9 gene (Fig. S4 A). This targeted allele is in effect a complete knockout because Commd9 protein expression in mutant MEFs is undetectable (Fig. S4 B). Commd9-deficient MEFs from this gene trap model demonstrated substantial reductions in Jagged1-mediated induction of Notch target genes, such as Hey1, Hey2, and Hes1, whereas TGF-β1 responses were unaffected (Fig. 3 D). Furthermore, re-expression of Commd9 in these MEFs (Fig. S4 C) restored Hey1 expression in this cell model (Fig. 3 E). Finally, we tested a second gene inactivation model that relies on the introduction of LoxP sites flanking exon 3 of the mouse Commd9 gene (Fig. S4 D). MEF lines were derived from two embryos, and isogenic pairs lacking Commd9 expression were generated in culture by transient expression of Cre recombinase (Fig. S4 E). In agreement with previous observations, Commd9 deficiency was associated with blunted activation of Notch target genes. This was true in response to plate-bound recombinant Jagged1, or when OP9 cells that stably express Jagged1 were used as a source of ligand (Fig. 3 F). Altogether, these findings indicate that COMMD9 is required for optimal activation of Notch-dependent gene expression, in agreement with the role of this protein in maintaining surface expression of Notch2.

Inactivation of retromer is frequently associated with missorting of endosomal protein cargoes to lysosomes, resulting in protein degradation (Steinberg et al., 2013). We noted a modest reduction of Notch2 in HeLa cells (Fig. 2 C) and assessed this phenomenon in other cell models. In this regard, we found that COMMD9 silencing in HEK 293 cells led to more notable reduction in Notch2 protein expression (Fig. 4 A). Unlike Notch2, silencing of COMMD9 had little impact on Jagged1 expression (Fig. 4 B). A similar phenomenon was noted for Notch1 and Notch2 in Commd9-deficient fibroblasts (Fig. 4 C), and this reduction was also noted in the cell surface fraction (Fig. 4 D). Importantly, Bafilomycin A1, an inhibitor of lysosomal acidification and protease activity, restored Notch2 expression in MEFs (Fig. 4 E). Altogether, these data are consistent with the notion that COMMD9 deficiency results in missorting of Notch proteins through the endolysosomal system, leading to their lysosomal degradation.

In view of the aforementioned findings, and the important roles of Notch during embryogenesis (High and Epstein, 2008), we suspected that Commd9 deficiency in mice could impair several developmental processes, including in the cardiovascular system. First, we observed that gene trap targeting of the Commd9 locus was not viable at birth, and timed pregnancies indicated that the expected Mendelian rate was last observed at embryonic day (E) 10.5 (Fig. 5 A). At this stage, Commd9-deficient embryos displayed neural edema (Fig. 5 B), which was followed a day later by hemorrhages, starting in the head of the embryo and soon after observed throughout the body (Fig. 5 B, and C). Several other morphologic changes were noted, most notably in the cardiovascular system. Hypoplasia of the heart was a universal finding at E10.5 and beyond (Fig. 5 D), with foci of focal wall necrosis and pericardial bleeding noted by E11.5 (Fig. S4 F). Furthermore, mutant embryos displayed several vascular changes, including focal narrowing of the dorsal aortas (Fig. 5 E). Whole-mount immunohistochemistry staining for the endothelial marker Pecam1 revealed additional alterations in cranial blood vessels, particularly a dramatic reduction in truncal vessels connected to the distal capillary networks (Fig. S4 G). Further analysis with markers of arterial (Cx40) and venous (Apj) vessels revealed normal arterial structures but substantial changes in venous vasculature formation, with ectatic dilations noted in the embryo’s head (Fig. S4 H). To assess if these changes were at least in part related to altered Notch signaling, we isolated heart mRNA from somite-matched embryos at E10.5. Indeed, a reduction in Hey1 and Hey2 mRNA expression in the heart was noted (Fig. 5 F), in line with the findings made in MEFs derived from this mouse strain (Fig. 3).

Interestingly, germline targeting in the Commd9 floxed mouse recapitulated the embryonic lethality noted in the gene trap model (Fig. S4 I). However, endothelial (Tie2-Cre), myocardial (Nkx2.5-Cre), and smooth muscle (Sm22-Cre) targeting were all viable and did not recapitulate the embryonic lethal phenotype observed with whole-body deficiency (Fig. S4 I). Therefore, we concluded that the embryonic lethality we observed resulted from alterations in more than one discrete cell lineage.

Commd1 deficiency in mice has been shown to result in embryonic lethality (van de Sluis et al., 2007), with a phenotype that is earlier to that observed in the Commd9-mutant embryos studied here. Therefore, despite the similarities between these proteins, they serve nonredundant developmental functions that cannot be compensated by other COMMD proteins. To try to understand the potential basis for the specific effects noted for COMMD9 deficiency, we examined whether COMMD9 exists in a complex that is distinguishable from COMMD1. A side-by-side comparison of endogenous COMMD proteins associating with COMMD9 and COMMD1 indicated that both proteins coimmunoprecipitated with each other and result in similar recovery of several COMMD proteins, such as COMMD2, COMMD4, COMMD6, and COMMD8 (Fig. S5 A). However, COMMD5 and COMMD10 bound preferentially to COMMD9 over COMMD1. This suggested that COMMD9 and COMMD1 might coexist in overlapping but not identical complexes, and in agreement with this, different ratios of COMMD9 and COMMD1 recovery were noted in each immunoprecipitation (Fig. S5 A).

In an additional analysis, five COMMD proteins were immunoprecipitated, demonstrating once again that COMMD9 was best immunoprecipitated by COMMD5 and COMMD10 (Fig. S5 B). Furthermore, COMMD5 was also more abundant in COMMD9 and COMMD10 immunoprecipitations, and the equivalent observation was made regarding COMMD10 (Fig. S5 B). In contrast, COMMD4 and COMMD6 only coimmunoprecipitated minimal amounts of COMMD5, COMMD9, and COMMD10. Corresponding to these interactions, COMMD9 and COMMD5 were found to colocalize in cytosolic puncta (Fig. 6 A). Altogether, these observations suggested that COMMD5, COMMD9, and COMMD10 form preferential complexes.

Finally, we examined whether key COMMD complexes observed here form through direct protein-protein interactions or are mediated indirectly through other factors present in cells. To that end, COMMD proteins were expressed recombinantly in Escherichia coli, and their ability to form complexes was assessed by coprecipitation. This analysis indicated that COMMD9 can bind directly to COMMD5, in agreement with the coimmunoprecipitation from cell lysates (Fig. 6 B). Interestingly, in this in vitro system, COMMD1 was able to bind to both COMMD5 and COMMD6, suggesting that additional cellular mechanisms might regulate the preferential in vivo binding of COMMD5 to COMMD9.

Further examination indicated that loss of COMMD9 expression did not affect COMMD5 or COMMD10 expression in HeLa cells (Fig. 7 A) but abrogated their interactions with the CCC complex subunit CCDC22 (Fig. 7 B). In contrast, binding of CCDC22 with COMMD4, COMMD6, and COMMD8 was unaffected in these cells. Therefore, COMMD5 and COMMD10 are incorporated into the CCC complex in a COMMD9-dependent manner.

Next we examined the possibility that COMMD9-containing complexes are specifically involved in Notch2 regulation. To examine this, all 10 COMMD genes were individually silenced, and their effects on Notch2 expression were examined by immunoblotting. Only COMMD9 and COMMD5 silencing led to reduced Notch2 levels (Fig. 7 C). In agreement with this result, recombinant COMMD5-containing complexes were able to bind in vitro with an immunopurified C-terminal Notch2 polypeptide (Fig. 7 D). In contrast, COMMD1–COMMD6 complexes were unable to bind to Notch2. Collectively, these data suggest that COMMD9 forms unique COMMD heterocomplexes in cells that are required for the trafficking of Notch receptors.

This study sheds light into a previously unappreciated role for COMMD9 in vesicular sorting and the delivery of Notch proteins to the cell surface. Absent COMMD9, Notch is mislocalized in intracellular early endosomes, and the response to Notch ligands is substantially attenuated. Furthermore, our studies indicate that Notch is not only mislocalized but also undergoes lysosomal degradation. Interestingly, we found that similar phenotypes are observed in cells deficient in CCC components CCDC93 and C16orf62 and the retromer subunit VPS35. This latter complex, which is critically important for the endosomal recycling of a large number of receptors (Seaman et al., 2013), has not been previously linked to Notch trafficking (Steinberg et al., 2013).

The observation that COMMD9-dependent effects on Notch trafficking can dramatically affect signaling is in line with a large body of literature that demonstrates that trafficking through the endolysosomal system is an evolutionarily conserved mechanism to regulate Notch activity (Baron, 2012; Kandachar and Roegiers, 2012). Notch is constantly endocytosed (McGill et al., 2009), and the signaling outcome arising from this internalized pool of Notch is carefully controlled by several factors. On the one hand, the protein Numb is a conserved inhibitor of the pathway that acts at this level (Cotton et al., 2013; Couturier et al., 2013, 2014; Reichardt and Knoblich, 2013). This factor interacts with components of the AP-2 complex and prevents endosome-to-plasma membrane recycling of Notch; in addition, Numb also promotes the lysosomal degradation of the receptor (McGill et al., 2009). In contrast, Deltex, an E3 ligase, has been shown to promote Notch signaling in a ligand-independent manner (Sakata et al., 2004; Wilkin et al., 2008; Hori et al., 2011, 2012; Troost et al., 2012). Acting in concert with the homotypic fusion and protein sorting complex and Rab7, Deltex promotes the trafficking of Notch to late endosomes. Interestingly, in this pathway, the localization of Notch is restricted to the limiting membrane of late endosomes, preventing the entry of the receptors into the multivesicular body. From this location, a ligand-independent cleavage event is thought to occur, resulting in pathway activation. Countering Deltex are HECT domain–containing ligases, such as Itch in mammals and Suppressor of Deltex in Drosophila (Wilkin et al., 2004; Chastagner et al., 2006), which promote the entry of Notch into the multivesicular body, resulting in its degradation.

Another intriguing observation made here is that the CCC complex, previously shown to regulate retromer/WASH-dependent endosomal trafficking, can have distinct configurations dictated by the incorporation of specific COMMD protein complexes. We show that COMMD5 and COMMD10 interact preferentially with COMMD9 and depend on this protein for their interaction with the CCC complex. The data further suggest that COMMD9–COMMD5 complexes are particularly important for Notch trafficking and can bind with the cytoplasmic tail of Notch2 in vitro. This raises the possibility that distinct CCC subcomplexes are generated to regulate specific cargoes and that perhaps COMMD proteins assist in cargo selection in this pathway.

In addition, it is reasonable to speculate that COMMD9 is likely to regulate many more receptors besides Notch. The identification of additional COMMD9-regulated cargoes will require proteome-wide analyses, such as those recently used to uncover retromer-regulated cargoes (Steinberg et al., 2013). This information would likely provide key insights that will be important to understand the developmental phenotypes resulting from Commd9 deficiency. Our analysis identified several vascular and cardiac alterations, but the phenotype could not be ascribed entirely to the deficiency of a singular Notch receptor. For example, Notch1 deficiency has more prominent effects on arterial fate decisions (Lawson et al., 2001; Fischer et al., 2004; Gridley, 2010) rather than the phenotype in the venous vasculature observed here. Isolated Notch2 deficiency is embryonically lethal at about the same stage as Commd9 deficiency and is also associated with cardiac hypoplasia (Hamada et al., 1999; McCright et al., 2006), but only a small percentage of these mutant embryos exhibit neural and abdominal hemorrhages (Wang et al., 2012), as we routinely saw in Commd9-deficient embryos. Although Notch3 deficiency is viable, combined Notch2/Notch3 recapitulated more aspects of the phenotype of Commd9-deficient embryos. For example, concurrent deficiency of these factors results in frequent obstructions of the dorsal aortas and also leads to frequent cerebral and abdominal hemorrhages, as we saw here (Wang et al., 2012). However, this phenotype in Notch2/Notch3-deficient embryos is largely driven by smooth muscle–dependent signals; in contrast, smooth muscle deletion of Commd9 did not recapitulate the embryonic lethality. Altogether, our interpretation is that the developmental phenotypes in Commd9-deficient embryos are likely the result of alterations in additional receptors. In this regard, the fact that COMMD9 is frequently included in the genomic interval deleted in WAGR syndrome highlights the biomedical importance of understanding these events. Patients with WAGR are affected by a high risk for Wilms’ tumors and have several congenital anomalies, including aniridia, genitourinary abnormalities, intellectual disability, and frequent cardiac anomalies (Fischbach et al., 2005). We examined our heterozygote mice for any of these phenotypes and found no overt pathology, but cooperation between COMMD9 and other genes may be relevant to the human condition.

In summary, this study identifies that COMMD9, acting in concert with the CCC and retromer complexes, regulates the endosomal sorting of Notch and its delivery to the cell surface. This constitutes a regulatory step in the Notch pathway that has not been previously appreciated. Future studies are likely to uncover that members of the COMMD protein family play similarly critical roles in the endosomal sorting of other important cell-surface receptors.

Mice were housed in conventional facilities and fed a standard diet. All animal procedures were approved by the institutional animal care and use committee. Gene-trapped Commd9 mice were derived by the Transgenic Animal Model Core facility at The University of Michigan through blastocyst injection using an embryonic stem cell line (AQ0362) obtained from the Sanger Institute Gene Trap Resource (stock number 017754-UCD). The nature of the genomic targeting was confirmed by Southern blotting. Animals were genotyped by PCR using the following primers: 5′-TGCATGCCATAGAGCCCACTTGT-3′, 5′-TAGCAGCAGCCCCTCCATCACACT-3′, and 5′-GGGCTACCGGCTAAAACTTGAGAC-3′. This amplification yields a 530-bp product for the wild-type allele and a 771-bp product for the targeted allele. Mice with the conditional Commd9 allele (Commd9^F/F), which contains LoxP sites flanking exon 3 of the Commd9 gene, were generated by the Transgenic and Knockout Core at Mayo Clinic through blastocyst injection using an embryonic stem cell line (EPD0136_6_D10) obtained from The Knockout Mouse Project. Animals were genotyped by PCR using the following primers: 5′-AAGGTGGAAACACATAGCCAG-3′ and 5′-TTACTAGGCAACCCTGCATTG-3′. This amplification yields a 297-bp product for the wild-type allele and a 489-bp product for the targeted allele. Sox2-Cre transgenic mice are available from the Jackson Laboratory. This mouse strain contains a transgenic construct consisting of the promoter of the Sox2 gene, which drives expression of Cre in the epiblast as early as E6.5 (Hayashi et al., 2002). The Nkx2.5-Cre mouse strain was provided by E. Olson (University of Texas Southwestern, Dallas, TX). This mouse contains a transgene consisting of a 2.5-kb fragment encompassing the promoter and cardiac enhancer of the Nkx2-5 gene, which drives expression of Cre in the myocardium (McFadden et al., 2005). The Tie2-Cre mouse was provided by M. Yanagisawa (University of Texas Southwestern, Dallas, TX). This mouse contains a transgene consisting of the endothelial-specific promoter/enhancer of the Tek (Tie2) gene, which drives expression of Cre in endothelial cells (Kisanuki et al., 2001). The Sm22-Cre mouse was provided by H. Yanagisawa (University of Texas Southwestern, Dallas, TX). This mouse contains a transgene consisting of a smooth muscle–specific promoter from the Tagln (Sm22) gene and the coding sequence of Cre recombinase (Holtwick et al., 2002).

HEK293T, U2OS, and HeLa cells were obtained from the American Type Culture Collection. HeLa shRNA stable cells for CCC and retromer components were generated by lentiviral infection of HeLa cells. Viruses were produced using the shRNA expression vector pLKO.TRC, including specific targeting sequences that have been previously reported (Phillips-Krawczak et al., 2015). Mouse Commd1-deficient fibroblasts were generated by harvesting Commd1^f/f embryos at E13.5 and deriving primary MEFs cultures from the embryonic soma. These cultures were immortalized using lentiviral vectors expressing E1A and Ras. Immortalized MEFs were then transiently infected with an adenoviral vector expressing Cre, resulting in Commd1 deletion, as described previously (Vonk et al., 2011). Dermal fibroblasts from patients with the CCDC22 p.T17A mutation (Voineagu et al., 2012) were obtained and immortalized through stable expression of telomerase; these cells have been reported previously (Starokadomskyy et al., 2013). MEFs from gene trap and floxed Commd9 mouse embryos were harvested and immortalized as described here for Commd1-deficient fibroblasts (Vonk et al., 2011). OP9-Jagged1 cells were generated by transduction of the murine bone marrow stromal cell line OP9 with a retroviral vector expressing human Jagged1, as previously reported (Dontje et al., 2006). All cell lines were maintained in high-glucose DMEM containing 10% FBS. Treatment with Bafilomycin A1 (Tocris) consisted of adding this reagent to the growth media (100 nM) for 24 h. Stimulation with plate-bound Jagged1 (R&D Systems) consisted of first applying the ligand to empty growth plates for 24 h (0.5 µg per well of a six-well plate). Thereafter, the plates were rinsed, and cells were seeded onto these plates; 48 h later, RNA was extracted from these cells for analysis. Stimulation with OP9 cells expressing Jagged1 consisted of first seeding the stimulating cells onto plates and allowing them to reach confluency. At that point, cells were rinsed and fixed with sterilized 4% PFA at room temperature for 15 min. After thoroughly rinsing the fixative, target cells were seeded onto these coated plates; 48 h later, RNA was extracted from these cells for analysis.

The pEBB vector is a derivative of pEF-BOS (Mizushima and Nagata, 1990) and contains an EF-1α promoter for mammalian protein expression. A series of derivatives of the pEBB vector (pEBB-HA, pEBB-2xHA-TB, pEBB-HA-COMMD9, and pEBB-2xHA-COMMD9-TB) were generated for these studies through conventional cloning techniques and have been previously described (Starokadomskyy et al., 2013). pEBB-AcGFP1-Rab5c and pEBB-AcGFP1-Rab11b are derivatives of pEBB containing the coding sequence for the fluorescent protein AcGFP1 and the ORFs for Rab5c and Rab11b, respectively. These were cloned into the BamHI and NotI sites of the pEBB-AcGFP1 vector (Phillips-Krawczak et al., 2015). A series of bacterial expression vectors were generated through conventional cloning techniques and include pET30a-His-COMMD1 (expressing His-tagged COMMD1), pGEX-4T1-GST-COMMD6 (expressing GST-tagged COMMD6), pRSF-His-COMMD9 (expressing His-tagged COMMD9), and p15Cool2-MBP-COMMD5 (expressing MBP-tagged COMMD5. The Notch reporter plasmid Notch-CBF1 contains multimerized RBPj binding sites and was shared by L. Lum with permission from the original developers of the construct (Baladrón et al., 2005). The pcDNA-human Notch2FL vector, which expresses full-length untagged Notch2, was provided by Artavanis-Tsakonas Lab (Zagouras et al., 1995). pEBB-2xHA-human Notch2 C terminus (intracellular domain) was PCR amplified from pcDNA-human Notch2FL and cloned into BamHI–NotI sites of pEBB-2xHA vector. For lentiviral expression of HA-tagged COMMD9, we used the lentiviral FG9 vector (Wright and Duckett, 2009). After cloning into the XbaI sites, we generated the FG9-PuroR-HA-COMMD9 vector. A “CRISPR-resistant” version of COMMD9 was introduced into a lentiviral vector (FG9-HygroR-2xHA-Crispr resistant COMMD9) after introducing silent mutations in COMMD9 (site-directed mutagenesis kit; Agilent Technologies) using the following primer: 5′-GCAAGATGGCTGCCCTAACGGCAGAGCATTTTGCAGCAC-3′ (mutated nucleotides are underlined). The siRNA duplexes used in this study are described in Table S2. Stable shRNA-mediated silencing was performed using the lentiviral vector pLKO.TRC. CRISPR-mediated inactivation of COMMD9 was performed through stable expression of a targeting guide RNA (gRNA) and Cas9, using the pLENTI-CRISPR vector (Mali et al., 2013; Shalem et al., 2014). Deleted clones were isolated through limiting dilution and screened for COMMD9 expression by immunoblot. Genomic DNA was isolated from individual clones, and genomic regions around the targeted area were amplified using the following primers: forward (5′-TCTCGGGGTTTAGTCATCCAGCC-3′) and reverse (5′-AGCAGGAGGTTGTATTGCGCCT-3′). These PCR products were cloned into PCR 2.1 TA vector (Life Technologies), and individual clones were sequenced. Nested PCR to obtain a shorter 283-bp product was performed using primers as forward (5′-ACCACAACCGAAACGCCTTCCC-3′) and reverse (5′-AGCAGCAGAGTTGTGGCTCCAA-3′). The resulting amplified products were resolved in 3% Nu-Sieve agarose gels. All targeting sequences used in the various approaches mentioned here are noted in Table S2.

The following transfection reagents were used: calcium phosphate (for plasmids and siRNA in HEK293T cells), Fugene (Roche, for plasmids in U2OS cells), and RNAiMAX (for siRNA in HeLa cells; Life Technologies). Lentiviral production was achieved through transient transfection of HEK293T cells with packaging and expression lentiviral vectors. Culture supernatants containing viral particles were harvested 48 h later, and after filtration through a 0.45-µm filter, this supernatant was applied to the intended cell target. The lentiviral expression vectors used here included puromycin or hygromycin resistance as selection markers, which were used to select stably transduced cells as described previously in more detail (Maine et al., 2007).

RNA extraction from cultured cells was performed using the Trizol method (Life Technologies) according to standard protocols. RNA extraction from isolated embryo hearts was performed using RNeasy columns (QIAGEN). Quantitative RT-PCR was performed using the Sybr green method, and relative quantitation was performed according to the 2^(−ΔΔCt) algorithm as previously reported (Maine et al., 2007; Starokadomskyy et al., 2013). The primers used for each of the gene targets examined here are noted in Table S3. All experiments involved biological and technical replicates (n = 3) and were performed at least three independent times.

Cell lysate preparation, immunoprecipitation, and immunoblotting were performed as previously described (Maine et al., 2007). Immunoblotting and immunoprecipitation experiments were performed using a Triton X-100 lysis buffer (25 mM Hepes, 100 mM NaCl, 10 mM DTT, 1 mM EDTA, 10% glycerol, and 1% Triton X-100). All Western blots presented in this paper are representative of at least three independent iterations for each experiment. Luciferase assays were performed by transfecting the luciferase reporter plasmid vector (Notch-CBF1) into U2OS cells. After 24 h, the cells were lysed, and 10 µl whole-cell lysate was loaded onto a 96-well luminometer plate. Luciferase substrate (Promega) was injected, and light emission was measured using a luminometer plate reader (EnVision; PerkinElmer), as previously reported (Burstein et al., 2005).

Bacterial expressing plasmids were individually transformed or cotransformed into to E. coli BL21 (DE3) cells (Invitrogen). All cells were grown in LB at 37°C with appropriate antibiotics until OD600 reached 0.6, and they were then shifted to 25°C for at least 30 min. Proteins were expressed at 25°C for 12–16 h by induction with 0.4 mM IPTG. Cells were resuspended in PBS buffer supplied with β-mercaptoethanol and protease inhibitors and lysed by sonication. The lysate was cleared by centrifugation at 12,000 rpm at 4°C for 30 min. For MBP-COMMD5, MBP-COMMD5/His-COMMD1, and MBP-COMMD5/His-COMMD9, clarified supernatant was purified using Amylose resin (NEB). For His-COMMD1/GST-COMMD6, supernatant was purified using glutathione agarose affinity resin (Molecular Probe). After incubation at 4°C for 30 min, the resin was washed four times with PBS buffer. Proteins bound to the resin were eluted with maltose and glutathione.

pEBB-2xHA-Notch2 C terminus was expressed in HEK293T cells, and Notch2-C was immunoprecipitated with HA-affinity beads (Roche). After extensive washing, the protein was eluted from the column with excess HA peptide (Anaspec). GST and His-COMMD1/GST-COMMD6 elution were rebound to glutathione resin. MBP, MBP-COMMD5, and MBP-COMMD5/His-COMMD9 elution were rebound to Amylose resin. The beads bound to these recombinant COMMD complexes were blocked with 5% BSA for 1 h at 4°C. Eluted Notch2-C was then mixed for 1 h at 4°C with COMMD complexes immobilized onto beads. After extensive washing of these beads, the material was resolved by SDS-PAGE and immunoblotted with HA antibody to detect 2xHA-Notch2 C terminus. The SDS-PAGE gels were stained with Coomassie blue.

All other antibodies used in this study are listed in Table S4. COMMD protein antibodies were generated by immunizing rats (COMMD1, COMMD2, and COMMD3), rabbits (COMMD4, COMMD5, COMMD9, and COMMD10), and guinea pigs (COMMD8) with recombinant GST fusion proteins encoding the N-terminal regions (proximal to the COMM domains) of the respective COMMD proteins (Cocalico Biologicals). The obtained sera were thereafter affinity purified using a His₆-tagged version of the respective immunogen (AminoLink Coupling Resin and Immobilization kit; Thermo Fisher Scientific). The antibodies were characterized by testing their ability to recognize FLAG-tagged protein expressed in HEK293 cells without cross reactivity to other COMMDs and to recognize endogenous protein (using siRNA as an additional control). Previously made custom antibodies to COMMD1, COMMD6, and COMMD9 have been reported (Starokadomskyy et al., 2013).

Cell staining was performed as previously reported (Phillips-Krawczak et al., 2015). In brief, cells were seeded in chambered slides and fixed in 4% PFA, permeabilized, and thereafter stained using the specific antibodies indicated in each experiment. The fluorochromes used in the secondary antibodies used for staining were Alexa Fluor 488 and Alexa Fluor 555 (Invitrogen). Images were obtained at room temperature using any of the following confocal microscopes: LSM 510 (63×/1.4 oil immersion objective lens; Carl Zeiss), LSM 710 (63×/1.4 oil immersion objective lens; Carl Zeiss), or A1R (60×/1.4 oil immersion objective lens; Nikon). Images were acquired using LSM 510 (Carl Zeiss), ZEN software (Carl Zeiss), and NIS-Elements AR (Nikon) software, respectively. Obtained images were analyzed using the LSM image browser, ZEN software, or Fuji ImageJ.

Cell-surface biotinylation was performed as previously described (Phillips-Krawczak et al., 2015). In brief, NHSSS-biotin (Pierce) dissolved in biotinylation buffer (10 mM triethanolamine, 150 mM NaCl, and 2 mM CaCl₂, pH 8.0) was used for surface labeling at 4°C for 30 min. The biotinylated antigens were precipitated using NeutrAvidin agarose resin (Pierce). Precipitated protein-containing beads were then resuspended in LDS sample buffer (Life Technologies) with 250 mM DTT. Proteins were then subjected to SDS-PAGE, and membranes were probed for Notch2 and N-cadherin.

This was performed by sequential purification through HA binding resin (Roche) and Streptavidin agarose (Pierce). In brief, 10 plates of seeded HEK 293 cells (15 cm each) were transfected with pEBB-2xHA-TB or pEBB-2xHA-COMMD9-TB. Two days later, a cell lysate was prepared in Triton X-100 lysis buffer as described in the Protein extraction section. The lysate was incubated with HA binding resin at 4°C for 2 h. At that point, the bait was eluted three times using excess HA peptide (1 mg/ml in 20 mM Tris-HCl, pH 7.4, 100 mM NaCl, and 0.1 mM EDTA). This pooled eluate was applied to a streptavidin agarose beads for 2 h at 4°C. Subsequently, the beads were washed extensively, and proteins were resolved by SDS-PAGE. After Coomassie stain, bands were excised and submitted to the Proteomics core at University of Texas Southwestern for trypsin digestion and peptide identification through LC/MS-MS analysis protein identification. A repeat iteration of this experiment was performed nearly identically except that most of the precipitated material bound to beads was submitted to the Taplin Mass Spectrometry Facility. A small aliquot of the precipitated material was resolved by SDS-PAGE and stained using silver nitrate (Invitrogen).

Timed pregnancies were setup to allow embryo harvesting at specified embryologic stages. Yolk sacs were used as a source of DNA for genotyping. For conventional histologic analysis (HE sections), the embryos were rinsed in PBS and then fixed in 4% PFA in 1× PBS overnight at 4°C with gentle rocking. Thereafter, embryos were dehydrated and stored in 75% ethanol at −20°C. For paraffin embedding, embryos were washed twice in 100% ethanol and xylene for 1 h each, followed by a series of rinses in 100% Paraplast Plus tissue embedding medium (McCormick) at 60°C. Subsequently, these embryos were paraffin embedded, sectioned, and stained according to standard techniques. For immunofluorescence, fixed embryos were rehydrated in stepwise fashion to 1× PBS and rinsed in 30% sucrose overnight at 4°C. The embryos were then embedded in Tissue-Tek O.C.T. compound and sectioned at 10 µm using a cryostat. Antigen retrieval was performed using Buffer A solution (Electron Microscopy Sciences). The slides were washed in 1× PBS and blocked for 1 h in Cas-Block (Invitrogen) before primary antibodies were added for overnight incubation at 4°C. Signals were detected the following day using fluorescently labeled secondary antibodies. The slides were mounted with ProLong Gold Antifade (Invitrogen). Whole-mount in situ hybridization was performed as previously reported (Xu et al., 2009). In brief, the fixed embryos were treated with proteinase K, fixed in a 0.25% glutaraldehyde and 4% PFA solution, and prehybridized at 60°C for 1 h. The samples were transferred into hybridization mix, containing a digoxigenin-labeled probe. Further in situ hybridization steps were performed with a series of washing buffer, and then development of color reaction was done using BM purple (Roche). RNA probes were generated from the following clones: Apj (BC039224; Open Biosystems) and Cx40 (BC053054; Open Biosystems). Whole-embryo images were obtained with a Stemi 2000-C stereomicroscope (Carl Zeiss) at magnification of 12.5, using a Canon PowerShot G10 camera and AxioVision Rel 4.8 software (Carl Zeiss). In situ staining images were obtained with a NeoLumar Stereomicroscope (Carl Zeiss) with 50×, 75×, or 80× magnification and a DP-70 camera (Olympus) with DP manager software (Olympus). Images of histologic sections were obtained with a DM2000 microscope (Leica) with 4×/0.70, 10×/0.70, 20×/0.70, or 40×/0.75 objective lenses and a Microfire CCD camera (Optronics) with PictureFrame 2.0 software (Optronics).

Fig. S1 shows additional information regarding the CRISPR-targeting strategy used to generated COMMD9-deleted HeLa cell clones (Cr-COMMD9). In addition, characterization of the resulting indels is shown. Fig. S2 shows additional details of the intracellular localization of Notch2 in COMMD9-deficient HeLa cells. Fig. S3 displays the effects of CCC or retromer deficiency on Notch2 intracellular localization. Fig. S4 contains details regarding the generation of the two knockout mouse models used here (gene trap and conditional deletion of Commd9). Additional morphologic data concerning the phenotype of Commd9-deficient embryos are also shown. Fig. S5 shows comparative coimmunoprecipitation of selected COMMD proteins and their associated complexes. Table S1 includes all the hits identified in the tandem affinity purification screen for COMMD9-associated proteins. Table S2 provides sequences for siRNA, shRNA, and CRISPR reagents used in this paper. Table S3 provides sequences for primers used for quantitative RT-PCR (qRT-PCR). Table S4 includes detailed information for all antibodies used here, including custom-made antibodies against all COMMD proteins which were generated for this study.

We are grateful to several investigators that provided critical reagents for our studies, including Eric Olson, Rhonda Bassel-Duby (Nkx2.5-Cre mice), Masashi Yanagisawa (Tie2-Cre mice), Hiromi Yanagisawa (Sm22-Cre mice), Lawrence Lum (reporter plasmids), and Spyros Artavanis-Tsakonas (Notch2 expression vector). We are also grateful to Eric Olson, Rhonda Basel-Duby, and Doug Engel for their generous help with some of the early embryologic studies. In addition, we are indebted to James Richardson, John Shelton, and the staff of the Molecular Pathology Core for their assistance with several aspects of this study. We are also grateful to the personnel and facilities of the Live Cell Imaging Core at University of Texas Southwestern.

The work of E. Burstein is supported by National Institutes of Health (R01DK073639 and R56AI113274) and CPRIT (RP130409). The work of D.D. Billadeau is supported by National Institutes of Health (grant AI065474). The work of B. van de Sluis is supported by the Netherlands Organization for Scientific Research (NWO-ALW grant 817.02.022).

The authors declare no competing financial interests.