The Sec14-like phosphatidylinositol transfer proteins Sec14l3/SEC14L2 act as GTPase proteins to mediate Wnt/Ca2+ signaling

The non-canonical Wnt/Ca2+ signaling pathway plays important roles in embryonic development, tissue formation and diseases. However, it is unclear how the Wnt ligand-stimulated, G protein-coupled receptor Frizzled activates phospholipases for calcium release. Here, we report that the zebrafish/human phosphatidylinositol transfer protein Sec14l3/SEC14L2 act as GTPase proteins to transduce Wnt signals from Frizzled to phospholipase C (PLC). Depletion of sec14l3 attenuates Wnt/Ca2+ responsive activity and causes convergent and extension (CE) defects in zebrafish embryos. Biochemical analyses in mammalian cells indicate that Sec14l3-GDP forms complex with Frizzled and Dishevelled; Wnt ligand binding of Frizzled induces translocation of Sec14l3 to the plasma membrane; and then Sec14l3-GTP binds to and activates phospholipase Cδ4a (Plcδ4a); subsequently, Plcδ4a initiates phosphatidylinositol-4,5-bisphosphate (PIP2) signaling, ultimately stimulating calcium release. Furthermore, Plcδ4a can act as a GTPase-activating protein to accelerate the hydrolysis of Sec14l3-bound GTP to GDP. Our data provide a new insight into GTPase protein-coupled Wnt/Ca2+ signaling transduction. DOI: http://dx.doi.org/10.7554/eLife.26362.001


Introduction
Wnt ligands, a large family of secreted lipoglycoproteins, control a large number of developmental events in animals, including cell fate, migration and polarity, embryonic patterning, organogenesis and stem cell renewal (MacDonald et al., 2009;Clevers and Nusse, 2012). Mammals express 19 Wnt members that bind to a corresponding receptor among 10 Frizzled (Fz) receptors (Niehrs, 2012). These receptors have a seven transmembrane span motif characteristic of G proteincoupled receptors, and, following binding of a Wnt ligand, activate different downstream pathways (Semenov et al., 2007;Loh et al., 2016). In the canonical Wnt pathway, Wnt signaling stabilizes cytoplasmic b-catenin and thereby promotes its nucleus translocation and accumulation to activate downstream target genes transcription (MacDonald et al., 2009). Wnts also signal through at least two b-catenin-independent (non-canonical) branches, the Wnt/Planer Cell Polarity (PCP) and Wnt/ Ca 2+ pathways, during vertebrate development. They are both devoted to modulate cytoskeleton organization to coordinate or polarize cell convergent and extension movements (Veeman et al., 2003;Kühl et al., 2000b;Angers and Moon, 2009). In the Wnt/PCP pathway, the monomeric small GTPases such as RhoA and Rac1 are required for transducing Wnt-Fz signaling to c-Jun N-terminal Kinase (JNK) to direct cell polarity and cell movement (Huelsken and Birchmeier, 2001;Wodarz and Nusse, 1998).
The Wnt/Ca 2+ signaling pathway emerged with the observations that overexpression of Xenopus Wnt5A and rat Fz2 in zebrafish embryos stimulated intracellular calcium flux and calcium-activated intracellular pathway (Kohn and Moon, 2005;Slusarski et al., 1997aSlusarski et al., , 1997b. It is demonstrated that pertussis toxin-sensitive heterotrimeric G protein subunits, e.g., Gao and Gat, are required for transducing the specific Wnt-Fz-Dishevelled complex signals downstream to activate phospholipase C (PLC) (Halleskog et al., 2012;Halleskog and Schulte, 2013;Liu et al., 1999). The activated PLC cleaves PIP 2 , a membrane-bound inositol lipid, into diacylglycerol (DAG) and inositol 1,4,5-triphosphate (InsP 3 ). DAG stimulates protein kinase C (PKC) to induce ERK phosphorylation, while InsP 3 binds to its corresponding receptor on the ER membrane, opening calcium channels for Ca 2+ release and increasing cytoplasmic Ca 2+ ion levels (Kadamur and Ross, 2013). Besides regulating cytoskeleton organization, Wnt/Ca 2+ also can influence CE movements through modulating calcium-dependent cell adhesion or dynamics of calcium storage and release (Kühl et al., 2001;Slusarski and Pelegri, 2007;Tada and Heisenberg, 2012;Wallingford et al., 2002). However, until now the interplay between Fz and heterotrimeric G/GTPase proteins in the Wnt/Ca 2+ signaling pathway is very controversial (Oner et al., 2013;Schulte and Bryja, 2007). Therefore, it remains mysterious and debated about how to orchestrate the upstream Wnt/Fz stimulation with downstream PLC/Ca 2+ components.
In recent years, a number of Sec14-like proteins have been identified and characterized. It has been demonstrated that dysfunction of Sec14-like proteins would cause various human diseases, such as breast cancer, prostate cancer, ataxia, and retinal degeneration syndromes (Cockcroft, 2012). Sec14-like proteins belong to atypical class III phosphatidylinositol transfer proteins (PITPs) (Allen-Baume et al., 2002) and consist of the versatile Sec14 domain associated with a GTPase motif of uncertain biological function (Habermehl et al., 2005;Novoselov et al., 1996;Merkulova et al., 1999). PITPs can transfer phosphatidylinositol (PI) and phosphatidylcholine (PC) between membranes, exchanging PI for PC and vice versa, in order to maintain balanced membrane lipid levels upon consumption of phosphoinositides (Wiedemann and Cockcroft, 1998). Although it has been established that PITPs integrate the lipid metabolome with phosphoinositide signaling cascades intracellularly, only very few studies indicate that PITPs can respond to extracellular molecular cues to initiate intracellular signaling (Xie et al., 2005;Kauffmann-Zeh et al., 1995). So far, there is no evidence about the biological importance of the GTPase motif in the Sec14-like proteins, as well as the crosstalk between PITP family proteins and Wnt/Ca 2+ signaling.
In this study, we investigated the role of the zebrafish Sec14l3 in embryonic development and the molecular mechanism of its action. We demonstrate that genetic depletion of maternal sec14l3 results in defects in embryonic convergent and extension (CE) movements by impairing Wnt/Ca 2+ signaling. Biochemical and genetic data indicate that Sec14l3 can transduce, via its intrinsic GTPase activity, Wnt-Fz signaling to activate phospholipase C.

Depletion of maternal sec14l3 impairs the gastrulation CE movements in zebrafish
We are interested in roles of maternal genes in early development of zebrafish embryos. Our previous RNA-seq data suggested that, among three SEC14-like phosphatidylinositol transfer protein genes (sec14l1, sec14l2, and sec14l3), sec14l3 is highly expressed in zebrafish eggs. Whole mount in situ hybridization (WISH) confirmed the abundance of sec14l3 transcripts during cleavage and early blastula stages ( Figure 1A). Thereafter, sec14l3 mRNA decreases to undetectable levels by the shield stage and increases again after the bud stage with enrichment in the vasculature cells ( Figure 1A).
To substantiate the knockdown effect, we generated four sec14l3 mutant lines by targeting the sequence around the translation start codon using transcriptional activator-like effector nucleases (TALENs) technology ( Figure 1-figure supplement 1C). The first line we obtained was sec14l3 tsu-Figure 1. Sec14l3 depletion impairs CE movements and Wnt/Ca 2+ signaling in zebrafish. (A) Spatiotemporal expression pattern of sec14l3. Embryos were laterally viewed with animal pole to the top or with anterior to the left. Scale bars, 100 mm. (B) Morphological defects in sec14l3 morphants during gastrulation. Scale bars, 100 mm. (C) Morphological defects in Msec14l3, MZsec14l3 and Zsec14l3 mutants. Scale bars, 100 mm. (D) sec14l3 mRNA rescue assay. 150 pg sec14l3 mRNA was injected into Msec14l3 mutants for rescue, and then the morphology and dlx3b/ntl marker gene expression were examined. First two panels: lateral views; last panel: dorsal views. Blue and red two-way arrows indicate the width of neural plate and the length of notochord respectively. Scale bars, 100 mm. (E) Phosphorylation level of Jnk in Msec14l3 mutant embryos. p-Jnk (Thr183/Tyr185) and total Jnk were examined at the shield (Morphology comparable) and 75% epiboly stage (Time point comparable) by western blot. Quantification of relative protein levels is shown on the right, represented by mean ± SEM in three separate experiments (see also Figure 1-source data 1, ns, non-significant). (F) Differential induction of calcium transient activity in zebrafish embryos. Representative calcium release profiles of embryos at the sphere stage in wildtype and Msec14l3 mutant background with or without wnt5b mRNA overexpression. The color bar represents the number of transients: red represents high numbers, blue represents lower numbers, and the peaks represent more active regions. In all panels, the ratio in the right corner indicated the number of embryos with altered phenotypes/the total number of observed embryos. DOI: 10.7554/eLife.26362.002 The following source data and figure supplements are available for figure 1: Source data 1. Numerical data for Figure 1E Figure 1 continued on next page td10 , which carried a 10 bp deletion (À3 to +7). Some zygotic mutants (Zsec14l3) of this line survived to adulthood, allowing production of maternal (Msec14l3) and maternal-zygotic mutants (MZsec14l3). In two-cell stage Msec14l3 or MZsec14l3 embryos, sec14l3 transcripts were almost eliminated ( Figure 1-figure supplement 1D), which was likely due to unstable property of mutant mRNAs. Both Msec14l3 and MZsec14l3 mutant embryos showed a slower epibolic process, which mimicked sec14l3 morphants, whereas Zsec14l3 embryos appeared normal ( Figure 1C). Therefore, the contribution of sec14l3 to gastrulation cell movements is a strictly maternal-effect. Interestingly, this maternal effect lasted through larval stages as evidenced by a reduced body length in Msec14l3 mutants compared with control embryos (Figure 1-figure supplement 1E). It appeared that cell proliferation and cell cycle progression in Msec14l3 mutant embryos were unaffected, then we focused on the event of cell movements ( Figure 1-figure supplement 2A). At the bud stage (about 10 hpf), Msec14l3 mutant embryos had a broader and shorter embryonic axis, which was marked by the midline marker ntl and the neural plate border marker dlx3b ( Figure 1D), indicative of impaired CE movements. Moreover, the defective CE movements of Msec14l3 embryos were not caused by cell adhesion defects between envelop cell layer (EVL) and deep cells (Figure 1-figure supplement 2B) and could be rescued by sec14l3 overexpression ( Figure 1D). Maternal mutants of other sec14l3 lines (sec14l3 tsu-td4 , sec14l3 tsu-td8 and sec14l3 tsu-td9 ), which were obtained later on, also exhibited similar phenotypes. Taken together, these data strongly suggest that the maternal, but not the zygotic, contribution of sec14l3 is critical for normal epiboly and CE movements during gastrulation.

Sec14l3 is required for Wnt/Ca 2+ signaling transduction
Since both Wnt/PCP and Wnt/Ca 2+ can regulate cell movements during embryonic development (Cha et al., 2008;Niehrs, 2012;Lin et al., 2010;Webb and Miller, 2003), we then examined which pathway was affected in sec14l3 mutant embryos. Results showed that Wnt/PCP signaling readout, the phosphorylated Jnk, p-Jnk(Thr183/Tyr185), was almost unaffected in Msec14l3 mutant embryos, compared to wild-type embryos both at the same developmental stage and time point ( Figure 1E). Human SEC14L2, rather than SEC14L3, is expressed in HEK293T and Wnt5-responsive PC3 cells, allowing easier examination of the effect of SEC14-like proteins on related signaling pathways (Figure 1-figure supplement 3A) (Ye et al., 2004). Like zebrafish sec14l3, knockdown of SEC14L2 in PC3 cells had little effect on p-JNK expression levels (Figure 1-figure supplement 3B and C). Therefore, we speculate that Sec14l3/SEC14L2 may not be crucial for the Wnt/PCP signaling pathway.
Next, we used the calcium indicator protein GCaMP6 to visualize calcium transients by confocal microscopy in zebrafish embryos, according to the method reported by Slusarski et al. (Chen et al., 2013;Nakai et al., 2001;Slusarski et al., 1997a). We found that the frequency of calcium release was attenuated obviously in Msec14l3 mutant embryos either at the basal level or upon Wnt5b stimulation ( Figure 1F). We therefore conclude that maternal Sec14l3 plays a role in Wnt ligand-dependent calcium release during embryogenesis. Similarly, Wnt5a-induced calcium signal in PC3 cells was also decreased when SEC14L2 was knocked down by shRNA (Figure 1-figure supplement 3D). Thus, Sec14l3/SEC14L2 take part in Wnt/Ca 2+ signaling transduction by promoting intracellular calcium release.

Sec14l3 forms complexes with Fz, Dvl and PLC
Fz, Dishevelled (Dvl) proteins and PLC are all implicated in Wnt-induced calcium release (Kühl et al., 2000a;Komiya and Habas, 2008), but the underlying mechanisms remain elusive. We wondered  whether Sec14l3 could associate with these proteins and performed co-immunoprecipitation (Co-IP) in mammalian cells. We found that Sec14l3 associated with C-terminal of human Fz5 (hFz5-CT) and rat Fz2 (Rfz2-CT) in HEK293T cells ( Figure 2A, Figure 2-figure supplement 1A). The in vitro-synthesized hFz5-CT and Rfz2-CT could be pulled down by GST-Sec14l3 ( Figure 2B). These results support the idea that Sec14l3 directly interacts with Fz proteins. Our domain mapping analysis revealed that the N-terminal CARL-TRIO domain and the C-terminal GOLD domain of Sec14l3 were crucial for its interaction with RFz2-CT (Figure 2-figure supplement 1A-C).
Then, we tested physical interaction of Sec14l3 with different Dvl proteins. Co-IP results revealed that Sec14l3 physically interacted with human DVL1, DVL2 or DVL3 in HEK293T cells ( Figure 2C). Used as a representative, Myc-tagged mDvl2 was determined to directly interact with GST-Sec14l3 ( Figure 2D). And in MCF7 cells, endogenous SEC14L2 interacted with DVL2 ( Figure 2E). In addition, this interaction was further validated in zebrafish embryos by overexpressing xDvl2-Myc and Flag-Sec14l3 mRNAs (Figure 2-figure supplement 1D). Deletion analysis showed that the Sec14l3's Sec14 domain, but neither the CARL-TRIO nor GOLD domain, was essential for interaction with mDvl2 ( Figure 2-figure supplement 1A,B and E). On the other hand, the DEP form of xDvl2, consisting of the DEP domain and the C-terminal region of xDvl2, was sufficient for the interaction with Sec14l3 ( Figure 2-figure supplement 1A and F).
Next, we tested the physical interactions between zebrafish Sec14l3 and Plcd family members, including Plcd1a, Plcd3b and Plcd4a that are essential for PIP 2 hydrolysis into DAG and InsP 3 . As shown in Figure 2F, immunoprecipitation of Flag-tagged Sec14l3 in HEK293T cells retrieved HAtagged Plcd4a and Plcd3b, but not Plcd1a. Furthermore, Plcd4a-HA and GST-Sec14l3 fusion proteins were expressed in E.coli and purified for pull down assay, which showed a direct interaction between them ( Figure 2G). Additionally, we also found their interaction in zebrafish embryos ( Therefore, Sec14l3 utilizes different domains to interact with xDvl2 and Plcd4a. To test whether Sec14l3, Fz and Dvl form a complex, we performed two-step Co-IP experiment. Results showed that hFz5-CT-Myc was present in the GST-Sec14l3-Dvl2-Flag complex ( Figure 2H), suggesting the presence of the hFz5/Dvl2/Sec14l3 ternary complexes. Furthermore, we found that mDvl2 was present in the Plcd4a complexes, as well as in the Plcd3b complexes, but absent in the Plcd1a complexes ( Figure 2I), which were similar to Sec14l3-Plcd selective interactions ( Figure 2F). DVL2 was also proved to interact with endogenous PLCd4a in MCF7 cells ( Figure 2E). Moreover, we found that different Wnt ligands stimulation could result in distinct calcium responses. Among of them, Wnt5a had a strong capacity to promote calcium release in PC3 cells (Figure 2-figure supplement 2A, and [Thrasivoulou et al., 2013]). Upon Wnt5a stimulation, mDvl2-Plcd4a complex formation could be enhanced in PC3 cells, and knockdown of SEC14L2 led to a drastic reduction of mDvl2-Plcd4a complexes ( Figure 2J), indicating the presence of the mDvl2/Sec14l3/Plcd4a ternary complexes.
Interestingly, Sec14l3-GFP could be co-immunoprecipitated with Sec14l3-Flag through its CARL-TRIO domain and GOLD domain (Figure 2-figure supplement 2B and C), suggesting oligomerization of Sec14l3. It is possible that Sec14-like protein oligomers may facilitate the formation of complexes with Fz, Dvl and Plcd proteins.

Sec14l3 is required for PLC-catalyzed hydrolysis of PIP 2
Sec14-like proteins are members of PITP and assumed to transfer phosphoinositides (PIs) to the plasma membrane (PM) (Nile et al., 2010;Kearns et al., 1998;Wiedemann and Cockcroft, 1998;Wirtz, 1991). To test whether human SEC14L2 and zebrafish Sec14l3 have an effect on PI derivatives accumulation at the PM, we first measured levels of PIP 2 in HEK293T cells, the lipid substrate of PLC, using a PH probe, which consists of a GFP-tagged PH domain from PLCd1 that specifically binds to PIP 2 (Várnai and Balla, 1998;Idevall-Hagren and De Camilli, 2015). If the PI transfer activity of SEC14L2 is blocked, the PIP 2 level at the PM should be reduced. However, compared to control cells, transfection of PLCd1-PH-GFP DNA into HEK293T cells depleted of SEC14L2 resulted in more PLCd1-PH-GFP protein in the PM fraction ( Figure 3A), indicative of more PIP 2 at the PM. Confocal imaging also revealed that the PLCd1-PH-GFP fluorescence at the PM was about 2-fold brighter in SEC14L2 shRNA stable cells than in the control cells ( Figure 3B, top panel), which was These results indicate that SEC14L2 may be required for PIP 2 hydrolysis rather than for PIP 2 transfer.
Then, we switched to detect PIP 2 in zebrafish embryos by injecting PLCd1-PH-GFP mRNA. Compared to wild-type embryos, the Msec14l3 mutant and sec14l3 morphant embryos accumulated more PLCd1-PH-GFP/PIP 2 at the PM both at 4 and 6 hpf ( Figure 3C and F, Figure 3-figure supplement 1A and C). With regard to PIP 3 , the metabolic product of PIP 2 , we also observed a similar PM elevation, indicated by AKT1-PH-mCherry probe, in Msec14l3 mutants or morphants (Figure 3-figure supplement 1B and C). Therefore, consistent with SEC14L2 in mammalian cells, depletion of Sec14l3 leads to the PIP 2 accumulation at the PM likely due to inefficient activation of PLC but not deficiency of its PI transfer activity.
PLC-catalyzed hydrolysis of PIP 2 produces InsP 3 and DAG, two important secondary messengers in cell signaling transduction. InsP 3 signals release calcium from intracellular stores, while DAG induces PKC-mediated ERK phosphorylation (Seitz et al., 2014). In conjunction with measuring calcium levels ( Figure 1F), we also measured phosphorylation of ERK (p-ERK) in vitro and in vivo. As expected, SEC14L2 knockdown in HEK293T cells caused a reduction of p-ERK (Thr202/Tyr204), but had no effect on phosphorylated AKT (p-AKT Ser473) (Figure 3-figure supplement 1D). Similarly, both western blots and immunostaining of zebrafish embryos showed a significant decrease in p-Erk, but not p-Akt in Msec14l3 mutants (Figure 3-figure supplement 1E-G). Collectively, these data establish a role for Sec14l3 in regulation of PLC catalytic activity.

Wnt/Dvl-induced PLC activation is dependent on Sec14l3
To investigate whether Sec14l3 mediates Wnt5b/Dvl2-induced PLC transduction in vivo, PIP 2 probe mRNA was injected into blastomeres at the 1 cell stage to visualize PIP 2 , which was followed by injection of a cocktail of wnt5b and GFP mRNA into one cell at the 16-32 cell stage to produce mosaic expression of Wnt5b. While injection of GFP mRNA alone has no effect on the PIP 2 distribution (Figure 3-figure supplement 2A and E), PIP 2 at the PM in the region with ectopic Wnt5b was significantly reduced compared to in the region without ectopic Wnt5b in the same wild-type embryos ( Figure 3D, upper panels and Figure 3F), suggesting that Wnt5b stimulates PLC-catalyzed PIP 2 hydrolysis. However, Wnt5b-dependent PLC activation was obviously inhibited in Msec14l3 mutant embryos, as evidenced by a much minor reduction in the PIP 2 level ( Figure 3D, lower panels and Figure 3F). Like Sec14l3 depletion, wnt5b morphant embryos accumulated more PLCd1-PH-GFP/PIP 2 at the PM both at 4 and 6 hpf, compared to std-MO injected embryos ( . Taken together, these epistatic analyses indicate that Sec14l3 can transduce Wnt5 signal to activate PLC in embryos. Next, we used a truncated form of Xenopus Dishevelled, xDsh-DelN, to stimulate PLC-catalyzed PIP 2 . xDsh-DelN is a N-terminal DIX domain deletion form that is sufficient to activate the Wnt/PCP and Wnt/Ca 2+ but not Wnt/b-catenin pathways (Sheldahl et al., 2003). As expected, xDsh-DelN lysates. The protein complexes were sequentially pulled down using GST (1 st IP-a-GST) and Flag antibody (2nd IP-a-Flag). Finally the second round immunoprecipitated proteins were detected using the corresponding antibodies. (I) mDvl2 interacts with Plcd3b and Plcd4a in HEK293T cells. (J) The interaction between mDvl2 and Plcd4a is restrained in the stable SEC14L2 knockdown PC3 cells with or without 400 ng/ml Wnt5a stimulation. Quantification of relative mDvl2 levels from three independent experiments is shown as mean ± SEM on the right (see also Figure 2-source data 1, **p<0.01; ns, non-significant). DOI: 10.7554/eLife.26362.007 The following source data and figure supplements are available for figure 2: Source data 1. Numerical data for relative protein level of mDvl2-Flag in Figure 2I.  Immunofluorescence of PLCd1-PH-mCherry (red, PIP 2 probe), b-catenin (green, PM marker) and DAPI (blue, nucleus marker) shows PIP 2 accumulation in the PM of Msec14l3 mutant cells. The first two whole embryo panels are 3D views of z-stacks (n = 30 for 4 hpf, n = 34 for 6 hpf), while the last panel is enlarged views of single z-stack pictures (z = 8 for 4 hpf, z = 7 for 6 hpf) from regions encompassed by white boxes. Scale bars, 100 mm for whole

The GOLD and the Ga domains of Sec14l3 are crucial for activating Plcd activity
To determine which Sec14l3 domain(s) is/are critical for PLC activation, several truncated forms of Sec14l3 were combined with a PIP 2 probe for transfection in HEK293T cells. PM isolation assays showed that the truncated forms of CARL-TRIO domain (DN) and Sec14 domain (DS) still acted similarly to full-length Sec14l3, where a reduction was observed in the amount of PIP 2 at the PM ( Figure 4A). However, rather than leading to PIP 2 degradation, forms of Sec14l3 lacking the GOLD domain (DG) or Ga subunit (DGa) induced PIP 2 accumulation at the PM, suggesting that these forms are functioning as dominant negatives ( Figure 4A). Similar results were observed with immunostaining ( Figure 4B). To further evaluate this phenotype, we overexpressed the truncated forms of Sec14l3 in human SEC14L2 knockdown stable cells, and found that enrichment of PIP 2 in the PM due to SEC14L2 depletion was partially compromised by transfecting the full-length form of zebrafish Sec14l3, but not DG or DGa forms ( Figure 4C).
In consistent with the above biochemical data, overexpression of DG or DGa mRNA in wild-type embryos led to a broader and shorter embryonic axis. And, neither the morphological epiboly defects nor marker gene-labeled CE defects seen in Msec14l3 mutant embryos were rescued by either of these two truncated mRNAs ( Figure 4D). Moreover, we found that DG or DGa form overexpression caused a significant reduction of p-Erk in wild-type embryos (Figure 4-figure supplement 1A and B) and a much lower calcium responsiveness in PC3 cells ( Figure 4E and F, Figure 4figure supplement 1C), which were quite similar to what happened in sec14l3 deficient situation ( Figure 1F). Taken together, these data strongly suggest that the GOLD domain and the Ga subunit domain of Sec14l3 are required for activating Plcd activity.

Sec14l3 promotes PM translocation and hydrolytic activity of Plcd4a in response to Wnt5a stimulation
To further study the mechanism by which Sec14l3-mediated Wnt/PLC activation, we determined whether Sec14l3 translocated to the plasma membrane upon Wnt stimulation. As shown in Immunofluorescence of PLCd1-PH-mCherry (red) and GFP (green, indicating Wnt5b-expressed cells) is shown. Mosaic expression of 100 pg wnt5b mRNA was created in embryos with even distribution of PLCd1-PH-mCherry mRNA. White polygons outline GFP expressed cells and single z-stack pictures (z = 10) from numbered regions in the whole embryo panels (3D view of z-stacks) are enlarged. Scale bars, 100 mm for whole embryos; 25 mm for the enlarged panels. (E) sec14l3 overexpression inhibits accumulation of PIP 2 in wnt5b morphant embryos. Mosaic expression of sec14l3 by injecting 150 pg mRNA was created in embryos with even distribution of PLCd1-PH-mCherry mRNA in std-MO or wnt5b-MO injected embryos. Single z-stack pictures (z = 11) from numbered regions in the whole embryo panels (3D view of z-stacks) are enlarged. (F) PM PIP 2 quantification of (C-E) by calculating intensity of (PM-Cytosol)/Cytosol PLCd1-PH-mCherry. Data are shown as mean ± SEM. (see also Figure 3-source data 1, **p<0.01; *p<0.05; ns, non-significant; n ! 50 cells from 10 embryos in three independent experiments). DOI: 10.7554/eLife.26362.011 The following source data and figure supplements are available for figure 3: Source data 1. Numerical data for Figure 3A Figure 5A, Figure 5-figure supplement 1A, Wnt5a stimulation induced a rapid translocation of Sec14l3-GFP to the PM, which was similar to the calcium sensor protein STIM1 (Liou et al., 2005) and enhanced by co-transfection of Dvl2. To rule out the possibility that Wnt5a-induced Sec14l3 PM recruitment might be a mere consequence of the calcium release, we tested the effect of hFz5/ mDvl2 stimulation on Sec14 protein location in HEK293T and MCF7 cells that do not respond to hFz5/mDvl2 for calcium release (Mikels and Nusse, 2006;Foldynová-Trantírková et al., 2010). In HEK293T cells, endogenous SEC14L2 protein was enriched in the PM by hFz5 transfection, which was further enhanced by hFz5/mDvl2 coexpression ( Figure 5B). Similar phenomenon was observed in MCF7 cells ( Figure 5-figure supplement 1B). Additionally, we demonstrated that SEC14L2 knockdown had no effect on hFz5-mDvl2 interaction ( Figure 5-figure supplement 1C), which suggests Fz-Dvl interaction might be independent of Sec14-like proteins. Therefore, combining the above data with pairwise biochemical interactions among Sec14l3, Fz and Dvl (Figure 2, Figure 2figure supplement 1), we propose that Sec14l3 is a component of the Fz/Dvl/Sec14l3 complex, and its PM recruitment is promoted directly by Fz/Dvl in response to Wnt signaling stimulation.
Then we hoped to know how Sec14l3 regulates PLC activity. PLC has been proposed to serve as a membrane attachment enzyme, which hydrolyzes many substrates such as PIP 2 without dissociating from the lipid surface (Rhee, 2001). Therefore, we speculated that Sec14l3 might be necessary for recruitment of Plcd4a to the PM for executing function. To test this hypothesis, we analyzed the subcellular localization of Plcd4a upon overexpression of Sec14l3. Immunostaining results in MCF7 cells showed that Sec14l3 actually promoted the PM localization of Plcd4a after co-transfection with hFz5 and mDvl2 ( Figure 5C). To determine which Sec14l3 domain(s) is/are critical for PLC recruitment, PM isolation assay was performed. In contrast to the cytoplasmic enrichment of Plcd4a in rest cells, there was a substantial enrichment of Plcd4a in the PM upon transfection with hFz5 or Rfz2 or both, which was enhanced by co-transfection with DGa or DG forms of Sec14l3, but not with DNG form ( Figure 5D, Figure 5-figure supplement 1D). Furthermore, we found that Rfz2-induced enrichment of Plcd4a at the PM was inhibited in SEC14L2 knockdown cells, which could be restored by overexpressing the full-length, DGa or DG form, but not DNG form, of Sec14l3 ( Figure 5E). Therefore, we speculate that both the TRIO-CARL domain and GOLD domain are required for recruiting Plcd4a to the PM upon Wnt receptors stimulation. Particularly, different from the receptors stimulation, DG form is sufficient to block Wnt5a-induced Plcd4a PM recruitment in PC3 cells, indicating a more important function of the C-terminal GOLD domain of the protein ( Figure 5F), which is consistent with the functional analysis of PIP 2 localization ( Figure 4A). Taken together, although Wnt5a ligand and its Fz receptors trigger Dvl2/Sec14l3-dependent Plcd4a PM recruitment in a slightly different way, possibly due to the diverse functions of the C-terminal GOLD2 domain upon ligand stimulation, Sec14l3 is important for the PM translocation of Plcd4a, which is mainly mediated by the C-terminal GOLD domain rather than the Ga subunit domain.
The next question is whether Sec14-like protein-mediated PM translocation promotes Plcd4a binding to PIP 2 . To address this issue, we performed liposome binding assay. As shown in Figure 5G, interaction between purified Plcd4a protein and liposome-bound PIP 2 was detected following incubation with HEK293T control cell lysate; however, this interaction was significantly weakened when the SEC14L2 knockdown cell lysate was used. This result indicates that Plcd4a accesses PIP 2 in a Sec14l3-dependent manner.

Sec14l3 functions as a GTPase protein in Wnt/PLC activation
We noticed that, although Sec14l3-DGa acts as a dominant negative form ( Figure 4A-C), it works differently with DNG or DG form in mechanism, based on Plcd4a PM recruitment results ( Figure 5D-F, Figure 5-figure supplement 1D). As previously reported, human SEC14-like proteins contain a proposed Ga subunit and possess considerable GTPase activity (Habermehl et al., 2005;Novoselov et al., 1996;Merkulova et al., 1999;Novoselov et al., 1994). We wondered whether zebrafish Sec14l3 has the same property. The high sequence homology in the GTP binding motif and P loop region between zebrafish Sec14l3 and human SEC14L2/SEC14L3/SEC14L4 ( Figure 6A) suggests a GTPase activity of zebrafish Sec14l3. Then we adopted BODIPY-FL-GTPgS conventional assay and MESG-based single-turnover assay to detect the GTP binding and GTP hydrolysis activities of Sec14l3 (Lin et al., 2014;McEwen et al., 2001;Tõ ntson et al., 2012;Webb, 1992). It estimated the K diss and K hydr rate constants for Sec14l3 to be 0.298 ± 0.089 min À1 and 0.151 ± 0.025 min À1 respectively ( Figure 6B-C). These results indicated that Sec14l3 is a genuine GTPase protein with GTP binding and hydrolysis activities.
A hallmark of G proteins is their ability to undergo conformational switches, from the GDP-bound 'off' state to the GTP-bound 'on' state and vice versa (Flock et al., 2015;Vetter and Wittinghofer, 2001). To gain full insights into the dynamic switch between Sec14l3 forms, we examined the binding affinity of its interacting proteins with GTPgS/GDP-loaded Sec14l3 in vitro. Results indicated that hFz5 had no preference for GDP-or GTPgS-bound Sec14l3, while Dvl2 preferred binding to GDPbound Sec14l3 (Figure 6D,E and G). On the contrary, Plcd4a showed much stronger interaction with GTPgS-bound form ( Figure 6F-G). These observations suggest that Dvl2 might participate in the switch of Sec14l3-GDP to Sec14l3-GTP, which then binds to and activates PLC.
Another important question is how active Sec14l3-GTP is cycled back to inactive Sec14l3-GDP. We speculated that Plcd4a, a binding partner of Sec14l3-GTP, might act as the GTPase-activating protein (GAP). To test this hypothesis, we compared the GTP hydrolysis activity of Sec14l3 in the presence and absence of Plcd4a protein. Results showed that the K hydr rate was increased about 2.5fold to 0.315 ± 0.056 min À1 in the presence of 94 nM Plcd4a, showing the GAP activity of Plcd4a ( Figure 6H). Additionally, quantification of the K hydr constants over different concentrations of Plcd4a determined its EC 50 value (50% of maximal effect value) as 21.3 nM ( Figure 6I). On the other hand, our data disclosed that Plcd4a is incapable of stimulating GTP uptake by Sec14l3 ( Figure 6J). Therefore, Plcd4a acts not only as a Sec14l3-GTP effector but as a terminator, a GAP of Sec14l3-GTP.
To verify the importance of the Ga domain for the GTPase activity, full-length Sec14l3 and Sec14l3-DGa were purified from E. coli, and resuspended for steady-state GTPase activity assay. Results showed that full-length Sec14l3 stimulated the hydrolysis of GTP in a dose-dependent manner (data not shown), while Sec14l3-DGa exhibited the relatively low GTPase activity ( Figure 6K), indicating the Ga subunit actually engenders the GTPase activity of Sec14l3. Moreover, incubation with lysates from cells transfected with DGa form led to much higher levels of Plcd4a-PIP 2 association compared to the full-length form of Sec14l3 ( Figure 6L); Ga domain deletion had no effect on interaction with hFz5-CT, but impaired its binding to mDvl2 and Plcd4a ( Figure 6M and Figure 6figure supplement 1A and B). We speculate that the GTPase activity deficient Sec14l3-DGa is unable to bridge Dvl2 and Plcd4a for complex formation so that Plcd4a-bound PIP 2 may not be hydrolyzed due to PLC autoinhibition (Hicks et al., 2008).

Upregulation of PLC activity rescued sec14l3 deficiency-induced CE defects in embryos
To confirm that attenuated PLC/Ca 2+ signaling was responsible for the phenotypes in Msec14l3 mutant embryos, we used U73122, an inhibitor of PLC (Ashworth et al., 2007), to test whether PLC inhibition phenocopies Msec14l3 mutants. Compared to the DMSO control, U73122 (1.5 mM or 3 to the PM in PC3 cells. Statistical data are presented. (G) SEC14L2 depletion perturbs Plcd4a access to PIP 2 . Equal amounts of purified Plcd4a protein and liposomes with or without PIP 2 were incubated with control or SEC14L2 depleted cell lysates. Statistical data are presented. All numerical data represented as a graph in the figure are shown in Figure 5-source data 1. DOI: 10.7554/eLife.26362.018 The following source data and figure supplement are available for figure 5: Source data 1. Numerical data for graphs in Figure 5 and  Figure 7A). Additionally, U73122 treatment caused a significant enhancement of PIP 2 accumulation at the PM at 4 and 6 hpf ( Figure 7B), which phenocopied Msec14l3 mutants.
We explored a possibility to rescue Msec14l3 mutant phenotype by artificially activating PLC in mutants. We constructed an X/Y linker-truncated form of Plcd4a, Plcd4a-D28, which was assumed to enhance the basal activity by relieving its autoinhibition (Hicks et al., 2008). Injection of plcd4a-D28 mRNA into Msec14l3 mutant embryos partially rescued defects in morphology and CE marker gene expression ( Figure 7C). So it is very likely that Plcd4a plays a role in gastrulation cell movements by mediating Sec14l3 effect in the Wnt/Ca 2+ signaling pathway.

Discussion
To date, only a few biochemical studies based on overexpression or inhibitors of proteins have suggested the implication of heterotrimeric G proteins in Wnt/Ca 2+ signaling (Malbon, 2004;Katanaev et al., 2005;Schulte and Bryja, 2007). However, it is unresolved as to how Fz/Dvl couples with G/GTPase proteins in Wnt/Ca 2+ signaling (Sheldahl et al., 1999;Aznar et al., 2015;Schulte and Bryja, 2007). In this study, we show that Sec14-like phosphatidylinositol transfer proteins can function as GTPase proteins in Wnt/Ca 2+ signaling. As modeled for Sec14l3 ( Figure 7D), Sec14l3-GDP can form complexes with Fz and Dvl; in respond to non-canonical Wnt stimulation, activated Sec14l3-GTP associates with and activates PLC at the PM, and promotes PLC-mediated PIP 2 hydrolysis to generate second messengers that propagate the Wnt/Ca 2+ signaling cascade. In zebrafish, depletion of maternal sec14l3 impairs Wnt/Ca 2+ signaling transduction and consequently causes defective gastrulation cell movements. Our findings not only reveal a critical function for Sec14l3 in regulating Wnt/Ca 2+ signaling, but provide a comprehensive view of mechanisms about GTPase proteins involvement during the signaling transduction, breaking the argument whether the 7-TM Fz can directly bind to and activate G proteins. We propose it is the GTPase proteins, Sec14l3/ SEC14L2, other than the classical heterotrimeric G proteins, that can simultaneously bind to upstream Fz/Dvl and activate downstream PLC for signal propagation. Therefore, the function of Sec14-like proteins is not limited to regulate the exchange of membrane lipids.
The direct Fz-Sec14l3 interaction can also be interpreted as Sec14l3-mediated regulation at the receptor level, such as Fz internalization or recycling in Wnt/Ca 2+ signaling (data not shown), which needs to be further investigated. Although our clues so far can't discriminate the transducer as a trimer or multimer, the organizer function of Sec14l3 in Wnt/Ca 2+ is recognizable. What's more, it has shown as mean ± SEM from three separate experiments (**p<0.01; p<0.05; ns, non-significant). (H) Plcd4a functions as the GAP of Sec14l3. Data are representative GTP hydrolysis curves from three independent experiments. The K hydr constant is statistically significant between two treatments with p<0.05. (I) Quantification of the K hydr constants over different concentrations of Plcd4a. Each concentration is plotted as mean ± SEM from three independent experiments. (J) GTP binding activity of 10 mM Sec14l3 in the absence or presence of 160 nM Plcd4a. Data shows an example of BODIPY-GTPgS uptake curve of three experiments. These two K diss constants have no significant difference by the Student's t test. (K) Ga subunit deletion led to a decreased GTPase activity of Sec14l3. Data shown are representative curves out of three replicates and are plotted as mean ± SEM. (L) Sec14l3-DGa protein overexpression disturbs Plcd4a-mediated PIP 2 degradation in vitro. Equal amounts of purified Plcd4a protein and liposomes containing PIP 2 or not were incubated with Sec14l3-transfected cell lysates. Liposomebound Plcd4a was eluted and analyzed with quantification data as mean ± SEM on the right (*p<0.05). (M) Ga subunit mediates Plcd4a interaction with Sec14l3 in HEK293T cells. Quantification data from three independent experiments are shown as mean ± SEM on the right (**p<0.01). All numerical data represented as a graph in the figure are shown in Figure 6-source data 1. DOI: 10.7554/eLife.26362.021 The following source data and figure supplement are available for figure 6: Source data 1. Numerical data for graphs in Figure 6 and  shows the PM accumulation of PIP 2 in zebrafish embryos treated with 1.5 mM DMSO or 1.5 mM U73122 at 4 and 6 hpf. The first two whole embryo panels are 3D views of z-stacks, while the last panel is enlarged views of single z-stack pictures (z = 5 for both 4 hpf and 6 hpf) from regions encompassed by white boxes. Scale bars, 100 mm for the whole embryos; 25 mm for the enlarged columns. (C) Active Plcd4a overcomes the CE defects in Msec14l3 mutants. 100 pg plcd4a-D28 mRNA was used for injection. Lateral views for embryos in the first panel, dorsal views for those in the last panel. Blue and red twoway arrows indicate the width of neural plate and the length of notochord respectively. Scale bars, 100 mm. (D) Hypothetic working model of Sec14l3 participation in Wnt/Ca 2+ signaling. In the absence of Wnt (left panel), Sec14l3 is mainly maintained in the ER and cytoplasm, forming a heterodimer with Dvl in its inactive state, Sec14l3-GDP. Upon Wnt5 stimulation (right panel), Fz/Dvl-mediated Sec14l3 is recruited to the PM and switched to the active state, Sec14l3-GTP, and subsequently promotes Plcd4a localization from cytoplasm to the PM and then the consequent activity at least in two Figure 7 continued on next page been suggested that lipid transfer proteins may not simply function as diffusible vehicles mediating lipid transfer between membranes, but also devices of assigning PI lipids to various enzymatic reactions in a strictly regulated biological context (Mousley et al., 2012). Our findings disclose multifaceted functions of lipid transfer proteins such as Sec14l3 and fill an important gap in our understanding of how Wnt/Fz/Dvl transduce the signal to PLC for Ca 2+ release.
Although Sec14l3 is initially identified as a member of PITP, its depletion does not cause PIP 2 or PIP 3 reduction at the PM both in zebrafish embryos and mammalian cells. Our studies demonstrate that the depletion of Sec14l3 leads to PIP 2 accumulation at the PM due to inefficient activation of PLC. However, we cannot exclude the possibility of the involvement of Sec14l3 PI transfer activity in embryonic development because other PITP family members, such as Sestd1 and Sec14l1, are also highly expressed in zebrafish embryos (data not shown). Therefore, to investigate the intrinsic transfer activity of Sec14l3 during early embryogenesis, genetic analysis of double or triple mutants is necessary.
One of the most important aspects of our work is the discovery that the GTPase activity of Sec14l3 is critical for Plcd4a activation. We find that Plcd4a binds to Sec14l3-GTP with apparently higher affinity and functions as a Sec14l3-GAP protein. However, Sec14l3-GEF proteins remain unknown. Wnt receptor Fz is a kind of G protein-coupled receptors, which can act as GEF proteins for their cognate G proteins upon binding of a ligand (Malbon, 2004;Schulte and Bryja, 2007;Dijksterhuis et al., 2014). We doubt that hFz5 or Rfz2 functions as the Sec14l3-GEF protein, because these Fz proteins show similar binding affinity towards GDP-or GTP-bound Sec14l3 and cannot accelerate the GTP uptake by Sec14l3 ( Figure 6D,G and Figure 6-figure supplement 1C). Considering that Fz receptors can activate Gai proteins and enhance Wnt/PCP signaling via the Dvlbinding protein, Daple, a novel non-receptor GEF (Aznar et al., 2015;Ishida-Takagishi et al., 2012), and Dvl2 does prefer to bind toward Sec14l3-GDP rather than Sec14l3-GTP in our hand ( Figure 6E and G), we tend to believe that Dvl2 serves as a scaffolding protein to recruit an unknown Sec14l3-GEF, thereby enhancing Sec14l3-GTP formation. As for which is the particular GEF for Sec14l3 in Wnt/Ca 2+ signaling transduction, more studies are needed. Besides, with the aid of available configuration-specific Sec14l3-GTP antibody in the future, it will be of great interest to figure out the conformational switches in vivo.
In summary, this study sheds light on a unique feature of the Sec14l3 protein. Through its intrinsic GTPase activity, it is capable of tightly coupling phospholipase activation with the proximity of PLC to its substrate. Moreover, these findings also provide the mechanism by which Dvl promotes calcium signaling.

Materials and methods
Embryos, injection and TALEN mutants generation sec14l3 TALEN mutants were generated in the Tg(flk:EGFP;gata1:dsRed) (PRID:ZFIN_ZDB_-FISH_150901_14755; ZFIN_ZDB_ALT_051223_6) transgenic fish using the FastTALE TALEN Assembly Kit (SiDanSai, Shanghai, China). The target site was near the start codon and included 22 bp both upstream and downstream (Figure 1-figure supplement 1C). To identify the candidate fishes with mutated alleles, genomic DNA was extracted from the tail and amplified using primer pairs as follows: the forward primer 5'-ccagcggcggagataaatc-3' and the reverse primer 5'-acatctatgacagacagcaatg-3'. The amplicons were purified for sequencing to determine the mutation types or digested with NlaIII to distinguish wild type and mutant embryos. The progenies derived from crosses between secl14l3 heterozygotes were raised to adulthood, and homozygous mutant males and females were identified by genotyping. MZsec14l3 or Msec14l3 mutant embryos were obtained by crossing homozygous mutant females to homozygous mutant males or wild-type males, respectively; Zsec14l3 mutant embryos were obtained by crossing heterozygous females to heterozygous Figure 7 continued aspects: its PIP 2 hydrolytic activity to generate second messenger InsP 3 and DAG for signaling propagation (Ca 2+ release and p-Erk activation); and its GAP activity to terminate Sec14l3-GTP. DOI: 10.7554/eLife.26362.024 males. Fishes were handled according to the institutional animal care and use committee (IACUC) protocol (AP#13-MAM1), which was approved and permitted by the Tsinghua University Animal Care and Use Committee. Embryos were staged according to Kimmel et al. (1995).

Whole-mount in situ hybridization and immunofluorescence
Whole-mount in situ hybridization was performed essentially following a standard protocol. Digoxigenin-UTP-labeled antisense RNA probes for detecting dlx3b and ntl mRNA were generated in vitro using a linearized plasmid (Huang et al., 2007). Following in situ hybridization, the embryos were immersed in glycerol and photographed using the Ds-Ri1 CCD camera under a Nikon SMZ1500 stereoscope. Embryonic immunofluorescence was carried out as previously described (Zhang et al., 2012), and the following antibodies were used: rabbit anti-p-Erk (Thr202/Tyr204) (Cell Signaling Technology, Danvers, MA, #9101, RRID:AB_331646), rabbit anti-p-Akt (Ser473) (Cell Signaling Technology, #4060, RRID:AB_2315049), mouse anti-SEC14L2 (ORIGENE, Rockville, MD, TA503723, RRID: AB_11126641), phalloidin (Sigma, St. Louis, MO, p1951, RRID:AB_2315148). Immunostained embryos were imaged under Nikon A1RMPSi lasers scanning confocal microscope using z-stack devices with 3.5 mm interval. And whole embryo images are 3D views of z-stacks, while the magnified images are pictures of a single z plane snap.
Cell culture, transfection and stable cell line establishment HEK293T cells (RRID:CVCL_0063) were cultured in DMEM (Life Technologies) supplemented with 10% FBS (Hyclone, Logan, UT) and 50 mg/ml penicillin/streptomycin (PS) (Invitrogen). PC3 cells (RRID:CVCL_0035) were cultured in F12K with 10% FBS and 50 mg/ml PS. All cell lines were obtained from the Cell Resource Center, Peking Union Medical College (which is the headquarters of National Infrastructure of Cell Line Resource, NSTI). Cell lines were checked for free of mycoplasma contamination by PCR and culture. Its species origin was confirmed with PCR. The identity of the cell line was authenticated with STR profiling (FBI, CODIS). All the results can be viewed on the website (http://cellresource.cn). Transfections were performed using the polyethylenimine method. To establish human SEC14L2 knockdown stable cell line, negative control shRNA (SHC016) and SEC14L2 shRNA (TRCN0000019589, TRCN0000019590) plasmids (ordered from Sigma) were transfected into HEK293T/PC3 cells and then transfected cells were selected with puromycin (1 mg/ml). Following removal of the puromycin, the cells were allowed to recover and expand in regular growth medium and then were screened by protein immunoblotting. For serum starvation, cells were incubated in culture media in the absence of any additional supplements for 16 hr.
Imaging of calcium levels in embryos and cytosolic calcium measurements by flow cytometry To image calcium levels in embryos, pXT7-GCaMP6 plasmid was constructed based on pGP-CMV-GCaMP6 and linearized for GCaMP6 mRNA labelling. Then GCaMP6 mRNA was mixed with Rhodamine and injected into 1 cell stage embryos, which were embed in low-melting agar at the indicated stages for time lapse imaging. Then the resulting images were treated to generate pseudocolor ratio images as previously described (Slusarski et al., 1997b(Slusarski et al., , 1997a. For calcium measurements in PC3 cells, cells were transfected with indicated plasmids for three days and loaded with calcium dye Quest Fluo 8-AM (AAT Bioquest, Sunnyvale, CA, 21083) or Rhod2-AM (AAT 21064) before flow cytometry analysis. During analysis, baseline fluorescence was measured for 50 s and stopped for Wnt5a addition, then measurement was resumed immediately for a total of 200s.

Liposome binding and membrane isolation assays
For the liposome binding assay, 1 mg HA-tagged Plcd4a protein purified from E. coli, 20 ml 1 mM PolyPIPosomes (Echelon Biosciences, Salt Lake City, UT, Y-0000 and Y-P045), none or 5% PIP 2 , 500 ml cell lysates transfected with control shRNA or SEC14L2 shRNA and 500 ml binding buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.05% NonidetP-40) were mixed and rotated for 4 hr at 4˚C and then centrifuged at 13,000 rpm for 10 min. The liposome pellet was then washed with 1 ml of binding buffer (50 mM Tris, pH 7.5, 150 mM NaCl, 0.05% NonidetP-40) for three times. The bound and flowthrough samples were eluted in 2xSDS loading buffer, separated by SDS-PAGE, transferred to nitrocellulose, and then Plcd4a levels were measured by immunoblot with anti-HA antibody.
For the membrane isolation assay, a Minute TM Plasma Membrane Protein Isolation Kit (Invent Biotechnologies, Inc., Plymouth, MN, Catalog number: SM-005) was used according to the manufacturer's instructions. The isolated membrane samples were then dissolved in 2xSDS loading buffer and detected by separated by western blot using appropriate antibodies. GTP binding and hydrolysis assays GTP binding and hydrolysis activity are measured using BODIPY-FL-GTPgS conventional assay and MESG-based single-turnover assay respectively; BODIPY-FL-GTPgS conventional assay is based on the release of the fluorescence quenching of BODIPY-FL-GTPgS (a non-hydrolyzable GTP analog) upon its binding to G proteins. BODIPY-FL-GTPgS (Invitrogen, G22183) binding to recombinant Sec14l3 was determined in 10 mM HEPEs (pH 8.0), 1 mM EDTA and 10 mM MgCl 2 (HEM buffer). The fluorescence (l ex = 490 nm and l em = 510 nm) was monitored for samples at different concentrations of BODIPY-FL-GTPgS, following the addition of 10 mM Sec14l3 protein, in a fluorescence microplate reader (Thermo Scientific VARIOSKAN FLASH). For the kinetic experiments, fluorescence of BODIPY-FL-GTPgS alone was measured in HEM at room temperature for 3 min and then binding was initiated with addition of excess Sec14l3. The change in fluorescence was measured over time and fitted with one phase exponential equation: a*(1-e -kt ) to obtain the K diss constant using Graph-Pad Prism 5.
The hydrolysis of GTP by Sec14l3 was measured by the MESG system monitoring the time course absorbance at 360 nm. The reaction was determined in 100 ml solution containing 50 mM MOPs (pH 7.0), 1 mM EDTA, 200 mM GTP, 1 U/ml purine nucleoside phosphorylase, 0.2 mM MESG and the recombinant Sec14l3 protein (reconstituted in Tris-HCl buffer). Single-turnover GTPase reactions were initiated by adding of MgCl 2 to a final concentration of 5 mM using the injector unit followed by immediate measurement (Thermo Scientific VARIOSKAN FLASH). The hydrolysis data fitted with one phase exponential equation: a*(1-e -kt ) to obtain the K hydr constant using GraphPad Prism 5. The time-course of Pi release from Sec14l3-GTP at four different concentrations was monitored, and finally averaged to determine the K hydr constant of Sec14l3.
For measurements of Plcd4a GAP functions, indicated amounts of GAP proteins were mixed with Sec14l3 before MgCl 2 initiation, and control experiments with indicated GAP proteins were carried out to provide a background of absorbance in each independent measurement to be subtracted from the sample signals.
Steady-state GTPase activity was performed using a QuantiChrom ATPase/GTPase Assay kit (GENTAUR, San Jose, CA, DATG-200) according to the manufacturer's instructions.

Statistical analysis
Quantitative data are presented as mean ± SEM, and comparisons were performed between groups using a two-tailed Student's t-test. For all analyses, *p<0.05; **p<0.01 were considered statistically significant; ns indicated statistical non-significance with p>0.05. Each experiment was carried out at least three times independently.