Tiki proteins are glycosylphosphatidylinositol-anchored proteases

Abstract

Wnt signaling pathways play pivotal roles in development, homeostasis, and human diseases, and are tightly regulated. We previously identified Tiki as a novel family of Wnt inhibitory proteases. Tiki proteins were predicted as type I transmembrane proteins and can act in both Wnt producing and Wnt responsive cells. Here, we characterize Tiki proteins as glycosylphosphatidylinositol (GPI)-anchored proteases. TIKI1/2 proteins are enriched on the detergent-resistant membrane microdomains, and can be released from the plasma membrane by GPI-specific glycerophosphodiesterases GDE3 and GDE6, but not by GDE2. The GPI anchor determines the cellular localization of Tiki proteins and their regulation by GDEs, but not their inhibitory activity on Wnt signaling. Our study uncovered novel characteristics and potential regulations of the Tiki family proteases.

Graphical Abstract

Tiki proteins are membrane-tethered Wnt inhibitory proteases. This study characterizes Tiki proteins as glycosylphosphatidylinositol (GPI)-anchored proteases. Tiki proteins are enriched on the detergent-resistant membrane microdomains on the plasma membrane, and can be released from the plasma membrane by GPI-specific glycerophosphodiesterases GDE3 and GDE6. This study uncovered novel characteristics and potential regulations of the Tiki family proteases.

Introduction

Wnt signaling controls embryogenesis, homeostasis, and regeneration. Aberrant Wnt signaling pathway is involved in a variety of human diseases including cancer, inflammation, and metabolic disease (1,2). The human genome contains 19 Wnt genes that encode a family of secreted glycosylated and lipidated proteins. The Wnt ligands activate several intracellular signaling pathways including the Wnt/β-catenin pathway, Wnt/PCP pathway and Wnt/Ca2+ pathway, via engaging different receptor complex at the plasma membrane. The so-called canonical Wnt ligands (such as Wnt3a) induce the Frizzled family receptors (FZDs) and the low-density lipoprotein receptor-related protein 5/6 (LRP5/6) to form a complex, in which LRP5/6 is phosphorylated at its intracellular PPPS(/T)P motifs. LRP5/6 phosphorylation is sufficient to inhibit the function of the β-catenin destruction complex, and stabilized β-catenin is translocated into the nucleus and binds to the TCF/LEF transcription factors to activate gene expression (3-5). Wnt signaling pathways are tightly regulated by multiple extracellular Wnt inhibitors, including DKK proteins and SOST that bind to LRP5/6, the Sfrp/Frzb family proteins that bind to Wnt and FZD, and the recently identified Wnt enzymatic inhibitors Tiki and Notum (6-9).

We previously reported Tiki as a family of Wnt- specific proteases that hydrolyzes the amino terminus of Wnt3a and causes Wnt3a to form oxidized oligomers and loosing receptor binding ability (6). Tiki proteins are evolutionarily conversed among the animal kingdom. Most vertebrates have two Tiki genes named Tiki1 and Tiki2, but rodents only have Tiki2 gene (10). Tiki seems to have specificity towards different Wnt proteins as TIKI2 only cleaved some of the Wnt proteins (11). Tiki proteins contain a signal peptide at their amino terminus and a predicted transmembrane domain (hydrophobic motif) on their carboxyl terminus and Tiki proteins can be localized at the plasma membrane (6). We have previously shown that Tiki functions in both Wnt producing and Wnt responsive cells in Xenopus embryos (6). How Tiki proteins are regulated and biochemical properties of Tiki proteins are still largely unknow. It was recently reported that human TIKI1 is highly expressed in resting CD4+ T cells and may regulate human immunodeficiency virus type 1 (HIV-1) production (12), suggesting a potential functional diversity of Tiki proteins beyond inhibition of Wnt signaling.

Glycosylphosphatidylinositol-anchored proteins (GPI-APs) are a class of proteins that are localized on the plasma membrane via the GPI moiety (13). More than 150 GPI-APs have been characterized in humans and play important roles in development, homeostasis and diseases by acting as cell surface antigens, adhesion molecules, receptors, or hydrolytic enzymes. For example, two drosophila GPI-APs, Dally and Dlp, and their mammalian homologs, Glypican-3 and Glypican-4, function in Wnt pathways by titrating/concentrating Wnt proteins at the cell surface (14-16). Precursors of GPI-APs contain an amino terminal signal peptide and a carboxyl localized GPI signal motif that can often be predicted as a transmembrane domain of the precursor protein (13). The signal peptide leads the precursor protein to enter the ER lumen where the signal peptide is removed and the GPI signal motif is replaced by a GPI moiety precursor. The GPI-APs will be further processed and eventually be translocated to the plasma membrane. GPI-anchor confers GPI-APs several special characteristics and regulatory potentials, such as localization to a membrane microdomain enriched with cholesterol and sphingolipids (the detergent-resistant membrane (DRM) fraction), and releasing from the cell surface by GPI-specific phospholipases.

In this study, we found that Tiki proteins are GPI-APs and enriched in DRM fractions and inhibit Wnt3a signaling in the Wnt-receiving cells. We further showed that human TIKI1/2 can be hydrolyzed and released from the cell surface by GDE3 and GDE6, but not by GDE2, suggesting a potential mechanism by which Tiki proteins can be regulated.

Results

Tiki proteins are GPI modified

Our previous study showed that Tiki proteins are predicted as type I transmembrane proteins and are localized at the plasma membrane (6). In order to detect the TIKI2 expression, we fused a FLAG tag at the carboxyl terminus of the full-length TIKI2 and TIKI2N (TIKI2 extracellular region) (Fig. 1A). We were surprised that an anti-FLAG antibody detected the expression of TIKI2N but not TIKI2, although both proteins could be detected by a homemade anti-TIKI2 antibody (Fig. 1B). These results suggest that the carboxyl terminus of TIKI2 may be post translationally cleaved. We thus suspected that TIKI proteins may be GPI-APs. Indeed the predicted transmembrane domain of TIKI1/2 is similar to the signature GPI signal sequence located at the carboxyl termini of other characterized precursors of GPI-APs, which typically has an unstructured linker region, a region of small residues for propeptide cleavage and GPI-attachment, a spacer region of moderately polar residues and a hydrophobic tail (17). In fact, Tiki proteins from protostomes to humans can all be predicted as GPI-APs (Fig. 1C). As GPI-APs can be released from the plasma membrane by phosphatidylinositol-specific phospholipase C (PI-PLC) (18), we utilized PI-PLC to treat HeLa or HEK293T cells expressing TIKI1 or TIKI2 with an amino-terminal HA tag (HA-TIKI1 or HA-TIKI2) (Fig. 1A). Both TIKI1 and TIKI2 exhibit two bands on western blots after SDS-PAGE (Fig. 1D, E), consistent with our previous report that showed the upper bands (~100 KD) are the mature form at the cell surface, while the lower bands (~75 KD) likely represent the immature form (6). In the presence of PI-PLC, the mature forms of both TIKI1 and TIKI2 was released into culture medium, establishing experimentally that TIKI1 and TIKI2 are GPI-APs (Fig. 1D, E).

Fig. 1.

Tiki proteins are GPI-APs. A, schemes of TIKI proteins with FLAG or HA tag. SP, signal peptide; TM, predicted transmembrane domain. B, full length TIKI2 and TIKI2 extracellular domain (TIKI2N), both with a C-terminal FLAG tag, were expressed in HEK293T cells, and whole-cell lysates (WCL) were analyzed by immunoblotting with anti-FLAG or anti-TIKI2 antibodies. C, sequence alignment of the carboxyl termini of Tiki proteins from different species. Predicted GPI ω-site is represented in bold. D and E, HA-TIKI1/2 proteins were released into culture medium (CM) in the presence of PI-PLC in HeLa (C) or HEK293T (D) cells.

TIKI1/2 are enriched at DRMs

DRMs are highly dynamic membrane microdomains enriched with cholesterol, glycosphingolipids, and GPI-APs, and may function as platforms for interactions of signaling molecules (19). It was reported that DRMs play critical roles in Wnt signaling pathways (20,21). To examine whether Tiki proteins are enriched on DRMs, we preformed sucrose-density gradient centrifugation to isolate buoyant DRMs and detergent-soluble membrane fractions (22). It has been reported that Caveolin1 is localized at DRMs whereas transferrin receptor (TFR) is excluded from DRMs (23). These two proteins have been commonly used as indicators for DRMs and non-DRM membrane fractions, respectively. As shown in Fig. 2, mature TIKI1 and TIKI2 proteins (the upper band) are enriched in the DRM fractions in HeLa, HEK293T, and MEF cells. However, we noted that there was significant amount of mature TIKI2 protein also in the non-DRM fractions in HeLa (Fig. 2B) and MEF (Fig. 2D) cells, but not in HEK293T cells (Fig. 2C), implying differential regulation of TIKI1 and TIKI2 in DRM distribution in different cells.

Fig. 2.

TIKI1/2 proteins are enriched on the detergent-resistant membranes (DRMs). TIKI1 or TIKI2 expressing cells were lysed and fractionated by sucrose density gradient centrifugation, and aliquots of the lysates were probed with anti-HA (to detect TIKI1/2), anti-Caveolin1 or anti-TFR antibody. Caveolin1 or TFR represent DRM or non-DRM fractions respectively. This experiment was performed in HeLa (A, B), 293T (C) and MEF (D) cells. Note that HEK293T cells do not express Caveolin1.

TIKI1/2 can be regulated by GPI-specific lipases

Another characteristic of GPI-APs is that they can be released from the cell surface by GPI-specific lipases. In addition to PI-PLC, animal cells also express a family of GPI-specific glycerophosphodiesterases (GEDs or GDPDs) composed of GDE2, GDE3 and GDE6, which are membrane proteins with six transmembrane domains and a large extracellular catalytic domain (24,25). GDEs play important roles in development and diseases by specifically shedding certain GPI-APs (24,26-29). To examine whether TIKI can be regulated by GDEs, TIKI1 or TIKI2 was expressed alone or together with each of the three GDEs in HEK293T cells. Culture media (CM) and whole cell lysates (WCL) were collected for detection of TIKI proteins. The result showed that in the presence of GDE3 or GDE6, but not GDE2, TIKI1/2 protein amounts in CM were increased while their cell surface levels (the upper bands in WCLs) were reduced, suggesting that GDE3 and GDE6 can release TIKI1/2 from the cell surface (Fig. 3A, B). We further showed that only the wild type GDE3 but not its enzymatic dead mutant (GDE3/R231A) (30) could release TIKI2 from the cell surface (Fig. 3C), indicating that the GDE3 activity on TIKI relies on its enzymatic activity.

Fig. 3.

TIKI1/2 can be regulated by GPI-specific lipases. HA tagged TIKI1 (A) and TIKI2 (B) were expressed alone or together with FLAG tagged GDE2, GDE3 or GDE6 in HEK293T cells. WCL and CM were analyzed by immunoblotting with anti-HA (TIKI1/2) or anti-FLAG (GDEs) antibody. C, HA tagged TIKI2 was expressed alone or together with GDE3 or GDE3/R231A (inactivated form of GDE3) in HEK293T cells.

The GPI anchor is required for Tiki localization at DRMs

To evaluate the GPI anchor on the cellular localization and function of Tiki, we generated two transmembrane types of TIKI2 by replacing the GPI signal motif of TIKI2 with either of the transmembrane domain of KREMEN2 (31) or EGFR, referred to as TIKI2/KRM2-TM and TIKI2/EGFR-TM, respectively. Expression in HeLa cells of either of the TIKI2 variants yielded similar patterns of the WT TIKI2 on SDS-PAGE and western blots, with a mature form (upper band) and an immature form (lower band), but the mature form of neither TIKI2 variants could be released from the cell surface by PI-PLC, in contrast to the WT TIKI2 (Fig. 4A). Furthermore, TIKI2/KRM2-TM and TIKI2/EGFR-TM are both exclusively enriched in non-DRM fractions in sucrose-density gradient centrifugation (Fig. 4B), suggesting that GPI anchor determines the cellular localization of TIKI2 in DRMs.

Fig. 4.

GPI anchor determines the DRM localization of TIKI2. A, wild type TIKI2 but not transmembrane types of TIKI2 (TIKI2/KRM2-TM or TIKI2/EGFR-TM) could be released into CM by PI-PLC. B, transmembrane types of TIKI2 were exclusively localized on the non-DRM fractions. C, live cells expressing HA-TIKI2 or HA-TIKI2/KRM2-TM were labeled with anti-HA antibody at 4 degree and transferred to normal culture condition for 1 hour. The cells were fixed and labeled with a fluorescent secondary antibody. Scale bar, 20 μm.

We further performed immunofluorescent staining to visualize endocytic trafficking of TIKI proteins. Live cells expressing HA-TIKI2 or HA-TIKI2/KRM2-TM were labeled with anti-HA antibody, and then were cultured for one hour to visualize endocytosis. The results showed that both TIKI2 and TIKI2/KRM2-TM could be detected on the cell surface, while TIKI2/KRM2-TM, but not TIKI2, was readily internalized into the intracellular vesicles (Fig. 4C), suggesting that GPI anchor plays a role in controlling endocytosis and turnover of TIKI proteins.

GPI anchor is not required for Tiki activity in Wnt inhibition

We previously reported that Tiki functions in both Wnt-producing and Wnt-receiving cells in Xenopus embryos (6). In mammalian cell culture, Tiki inactivates Wnt proteins in Wnt-producing cells (6,11). Whether Tiki also functions in Wnt-receiving cells in mammalian cell culture has not been examined. As most GPI-APs function at the cell surface, we hypothesize that Tiki may also act at the cell surface to antagonize Wnt signaling and thus functions in Wnt-receiving cells. We found that expression of either TIKI1 or TIKI2 in MEF cells inhibited Wnt3a induced LRP6 phosphorylation, which is a prototypical indicator of Wnt activation of the β-catenin pathway (32) (Fig. 5A), suggesting that Tiki functions at the cell surface of Wnt-receiving cells, similar to its action during Xenopus embryogenesis (6).

Fig. 5.

GPI anchor is not required for Tiki activity in Wnt inhibition. A, TIKI1 and TIKI2 inhibited LPR6 phosphorylation in Wnt-receiving cells. TIKI1 or TIKI2 expressing MEF cells were treated with increasing doses of Wnt3a conditioned medium (CM). WCL were analyzed by immunoblotting with indicated antibodies. B, Both TIKI2 and its transmembrane types inhibited TOPFLASH reporter expression induced by Wnt3a CM. Error bars represent means ± S.D. from 3 independent biological samples. ****, P<0.0001. The experiment was performed 3 times with similar results. C, Both TIKI2 and its transmembrane types inhibited LPR6 phosphorylation in Wnt-receiving cells.

GPI anchor is often of significance to the function of the protein itself and/or confers regulation at multiple levels including protein localization on the plasma membrane (33). We next evaluated the significance of GPI anchor to Tiki function in Wnt-receiving cells. As shown in Fig. 5B and C, either TIKI2/KRM2-TM or TIKI2/EGFR-TM inhibited Wnt3a-induced TOPFLASH reporter expression and LRP6 phosphorylation as efficiently as the WT TIKI2 did. These results suggest that GPI anchor, although determines the cellular localization of Tiki to DRMs, is not required for Tiki activity in the Wnt-receiving cells.

Discussion

Tiki proteins represent a family of membrane tethered proteases and inhibit Wnt signaling by hydrolyzing the amino terminus of certain Wnt proteins (6,11). Tiki proteins were initially predicted as type I transmembrane proteins with a carboxyl terminal transmembrane domain. Here, we characterized Tiki proteins as GPI-APs that are enriched in DRMs at the plasma membrane. Indeed TIKI1/2 appear to function at the cell surface of Wnt-receiving cells to inhibit Wnt/β-catenin signaling, consistent with our previous observation during Xenopus early embryogenesis (6). While the GPI anchor results in Tiki1/2 enrichment in DRMs, we found that neither the GPI anchor nor DRM localization is essential for Tiki activity and that the GPI anchor can be replaced by a transmembrane domain without compromising Tiki1/2 function in Wnt inhibition. Interestingly however, the GPI anchor allows Tiki1/2 to be cleaved and released from the cell surface by GDE3 or GDE6 but not GDE2, suggesting a potential mechanism by which Tiki proteins can be regulated by specific GDEs.

GPI-APs can be released from the cell surface by proteolytic cleavage, lipolytic cleavage or extracellular membrane vesicles (34). Release of GPI-APs can occur spontaneously or in response to endogenous or environmental stress signals (34). GDEs are GPI-specific glycerophosphodiesterases that shed certain GPI-APs and exert diverse biological functions (24). For instance, GDE2 has been reported to cleave RECK and regulate the RECK-ADAM10-NOTCH pathway in neural development (26). GDE2 also cleaves Glypican-6 and plays a role in neuroblastoma (35). GDE3 was reported to cleave the urokinase receptor (uPAR) in breast cancer cells (27). Here, we found that TIKI1/2 can be cleaved by GDE3 or GDE6 but not GDE2 in transfected HEK293T cells, suggesting a potential mechanism of regulation of Tiki function. Whether this regulation occurs in vivo during development and/or homeostasis deserves future investigation. As GPI-APs can be enriched in extracellular vesicles (i.e. exosomes) (36), whether Tiki proteins can be regulated in this way is an interesting question for future studies.

DRMs are often referred to as lipid rafts, which represent dynamic nanoscale membrane microdomains rich in cholesterol and sphingolipids, and are considered platforms for cell signal transduction (37). Since the existence and properties of lipid rafts are still controversial (38,39), we used DRMs instead of lipid rafts in this study. It has been reported that DRMs play critical roles in Wnt signaling initiation and regulation (14,21,40). TIKI1/2 proteins are enriched on the DRMs when expressed in multiple cells. We also noted that significant portions of mature TIKI2 protein are localized on the non-DRM membrane fractions in HeLa and MEF cells, but not in HEK293T cells, suggesting that Tiki protein localization may be differentially regulated in cells. Similarly, GPC4 was reported to be localized in both DRM and non-DRM domains, which determines GPC4 specificity in β-catenin dependent and independent pathways, respectively (14). Whether the GPI anchors of Tiki proteins also regulate Tiki functional specificity remains to be further elucidated.

We note that the in vitro cell culture system and overexpression study have limitations. For instance, although GPI anchor does not determine the activity of Tiki proteins in cultured cells in vitro, it may regulate the cellular localization and thus determine the function of Tiki proteins in vivo. Our current study has uncovered novel characteristics of Tiki proteins and shed lights on potential regulations of the Tiki family proteases.

Experimental procedures

Expressing constructs

TIKI2N-FLAG-His, HA-TIKI1 and HA-TIKI2 were described previously (11). TIKI2-FLAG-His was constructed by fusing the full length TIKI2 with a FLAG-6xHis tag on the pCS2+ and pBABE (a gift from Hartmut Land & Jay Morgenstern & Bob Weinberg (Addgene plasmid #1764)) (41) vectors. FLAG tagged GDE2, 3 and 6 were kindly provided by Dr. Sungjin Park (26). GDE3/R231A was generated via PCR-based mutagenesis approach using FLAG-GDE3 as the template. TIKI2/KRM2-TM and TIKI2/EGFR-TM were constructed by replacing the GPI signal sequence of TIKI2 with the transmembrane domain of KREMEN2 (IGARVFSTVTAFSVLLLLLLSLL) and EGFR (IATGMVGALLLLLVVALGIGLFM) respectively.

Cell Culture and Transfection

HEK293T (ATCC, CRL-11268), HeLa (ATCC, CL-0101), U2OS (ATCC, HTB-96) and immortalized MEF cells were maintained in DMEM supplemented with 10% FBS and 1% penicillin-streptomycin-glutamine. For retrovirus packaging, pCL-Eco (a gift from Inder Verma (Addgene plasmid #12371)) (42) and a pBABE construct (expressing HA-TIKI1, HA-TIKI2, HA-TIKI2/KRM2-TM or HA-TIKI2/EGFR-TM) were co-transfected into HEK293T cells, and the conditioned medium containing the retrovirus particles was used to infect the MEF cells to generate stable cell lines. Lipofectamine 2000 (Invitrogen) was used for all transfections.

Immunoblotting and Antibodies

Immunoblotting was performed as described previously (11). Antibodies that were used in this article are: FLAG (#8146, 1:2000), HA (#3724, 1:1000), pLRP6 (#2568, 1:1000), LRP6 (#2560, 1:1000), TFR/CD71 (#13113, 1:1000) from Cell Signaling Technology; Caveolin1 (610493, 1:1000) from BD Transduction Laboratories. Anti-TIKI2 mouse monoclonal antibody was generated with the purified TIKI2N protein from CM of transfected HEK293T cells and this antibody can only detect the overexpressed TIKI2.

Isolation of DRMs by Sucrose-Density Centrifugation

Cell fractionation experiments were performed as previously described (21,22). Briefly, HeLa, HEK293T or MEF cells (in 150-mm plates) were lysed in 1ml TNE (20mM Tris-HCl pH7.5, 150mM NaCl, 0.4% Triton X-100, protease inhibitor cocktail (Roche)) on ice. The cell lysates were homogenized by Dounce homogenizer with 20 strokes. 1ml of homogenized sample was mixed with 1ml 90% (w/v) sucrose/TNE. 35% sucrose in TNE was added on top of the sample/90% sucrose/TNE layer, followed by 5% sucrose in TNE. The gradients were centrifuged at 16 hours at 260,000g at 4°C. Centrifuged samples were collected from the top down in 0.5ml volumes and yield 10 fractions. Aliquots were probed with the indicated antibodies.

Immunofluorescence

U2OS cells stably expressing HA-TIKI2 or HA-TIKI2/KRM2-TM were seeded on the coverslips in a 24-well plate for 24 hours. The cells were washed once with ice cold PBS buffer and incubated with ice cold DMEM containing anti-HA antibody for 1 hour. The cells were washed once with ice cold PBS and fixed with 4% paraformaldehyde or were transferred to the normal culture medium and cultured for 1 hour before fixation. The cells were permeabilized and stained with a fluorescent secondary antibody.

Dual-Luciferase Assay

TOPFLASH reporter assay was performed as previously described (11). Briefly, HEK293T cells were plated into 24-well plate and transfected the following day with a total of 250ng of DNA per well (100ng of TOPFLASH, 5ng of thymidine kinase promoter Renilla, and indicated plasmids). The lysates were collected 48 h post-transfection and used with the dual luciferase reporter system (Promega). The experiment was performed three times with similar results.

Abbreviations

CM

culture media

DRM

Detergent-resistant membrane

FZD

Frizzled

GDE

Glycerophosphodiesterase

GPI-AP

Glycosylphosphatidylinositol-anchored protein

HIV-1

human immunodeficiency virus type 1

LRP5/6

Low-density lipoprotein receptor-related protein 5/6

DVL

dishevelled segment polarity protein

DKK

dickkopf WNT signaling pathway inhibitor

SOST

sclerostin

TFR

Transferrin receptor

uPAR

urokinase receptor

WCL

Whole cell lysate

Data availability

All data are contained within the manuscript.