Zinc transport via ZNT5-6 and ZNT7 is critical for cell surface glycosylphosphatidylinositol-anchored protein expression

Glycosylphosphatidylinositol (GPI)-anchored proteins play crucial roles in various enzyme activities, cell signaling and adhesion, and immune responses. While the molecular mechanism underlying GPI-anchored protein biosynthesis has been well studied, the role of zinc transport in this process has not yet been elucidated. Zn transporter (ZNT) proteins mobilize cytosolic zinc to the extracellular space and to intracellular compartments. Here, we report that the early secretory pathway ZNTs (ZNT5–ZNT6 heterodimers [ZNT5-6] and ZNT7–ZNT7 homodimers [ZNT7]), which supply zinc to the lumen of the early secretory pathway compartments are essential for GPI-anchored protein expression on the cell surface. We show, using overexpression and gene disruption/re-expression strategies in cultured human cells, that loss of ZNT5-6 and ZNT7 zinc transport functions results in significant reduction in GPI-anchored protein levels similar to that in mutant cells lacking phosphatidylinositol glycan anchor biosynthesis (PIG) genes. Furthermore, medaka fish with disrupted Znt5 and Znt7 genes show touch-insensitive phenotypes similar to zebrafish Pig mutants. These findings provide a previously unappreciated insight into the regulation of GPI-anchored protein expression and protein quality control in the early secretory pathway.

Approximately one-third of all cellular proteins encounter the endoplasmic reticulum (ER) secretory pathway (1), resulting in posttranslational modifications such as glycosylation, disulfidation, and glycosylphosphatidylinositol (GPI) modification at the C terminus producing GPI-anchored protein. There are more than 150 characterized GPIanchored proteins in humans (2)(3)(4), which function as enzymes, receptors, adhesion proteins, and complement regulatory proteins on the cell surface (5). The biosynthesis of GPIanchored proteins requires more than 20 phosphatidylinositol glycan anchor biosynthesis (PIG) proteins involving sequential additions of sugar and other components to phosphatidylinositol in multiple reaction steps. Thus, mutations in multiple PIG genes are associated with defects in GPI-anchored proteins (inherited GPI deficiency) resulting in prominent human diseases including intellectual disability, hypotonia, facial dysmorphism, seizures, and dystonia (6). The effect of reduced GPI-anchored protein expression requires further investigation.
In this study, the ZNT5-6 and ZNT7 function was analyzed via mutational manipulation studies. The GPI-anchored protein expression was impaired in mutant Z5Z7-double KO (DKO) cells where ZNT5-6 and ZNT7 are nonfunctional, as confirmed via sequential window acquisition of all theoretical fragment ion spectra mass spectrometry (SWATH-MS) analysis (19) compared with the WT. Moreover, disruption of both ZNT5 and ZNT7 led to similar phenotypes to those caused by PIG gene disruptions in the cells and in medaka fish (Oryzias latipes). The mutant medaka exhibited touch-insensitive phenotypes, which are similar to those of zebrafish Pigk and Pigu mutants (20,21). Our results provide insights into the biological functions of zinc mediated by ZNT5-6 and ZNT7 within the compartments of the early secretory pathway and improve our understanding of protein quality control in the ER and the secretory pathway via GPI-anchored protein synthesis.
This was examined in more detail using chimeric mutants of PLAP (PLAP-TM(H) and PLAP-TM(G)) wherein the C-terminal GPI anchor attachment portion of PLAP was substituted with a single membrane-spanning polypeptide-anchored portion of the cytosolic portion of influenza hemagglutinin (HA) or vesicular stomatitis virus glycoprotein since both mutants were located on the cell surface (24). PLAP-TM(H) and PLAP-TM(G) expression in A549-Z5Z7-DKO cells was comparable with that in WT A549 cells, although their activity was significantly decreased ( Fig. 2A). Both proteins were localized to the cell surface when expressed in SK-Z5Z7-DKO and WT SK-MEL-2 (Fig. 2B). For comparison, chimeric sACE and ACE2 mutants were constructed by substituting their membrane-spanning polypeptide portion with the GPI attachment of PLAP to form ACE-GPI and ACE2-GPI, respectively. The activity and protein expression of both ACE-GPI and ACE2-GPI was substantially decreased in A549-Z5Z7-DKO cells compared with those in WT A549 cells (Fig. 2C). ACE-GPI and ACE2-GPI failed to localize to the cell surface when expressed in SK-Z5Z7-DKO cells (Fig. 2D). These results indicate that the GPI anchor is involved in the impairment of the activation and protein expression of GPIanchored zinc ectoenzymes and that ZNT5-6 and ZNT7 play pivotal roles during this process.
The secretory forms of PLAP (secPLAP-HA) and ACE (secACE-HA) were constructed using an HA-tag in place of the C-terminal portion containing the GPI anchor attachment site and C-terminal single membrane-spanning polypeptide anchor, respectively. Both zinc ectoenzymes were detected in the spent medium of transfected WT A549 cells and A549-Z5Z7-DKO cells. However, secPLAP activity substantially decreased, while secACE activity was partially activated in A549-Z5Z7-DKO cells, compared with that in WT A549 cells (Fig. S1, A and B). These results indicate that the GPI anchor is responsible for the impairment of GPI-anchored zinc ectoenzyme expression in Z5Z7-DKO cells.

GPI-anchored protein expression decreased in Z5Z7-DKO cells
We then examined whether the GPI anchor is responsible for the decreases in expression of GPI-anchored proteins that do not bind zinc in Z5Z7-DKO cells using CD55, CD59, and ceruloplasmin (CP) as model GPI-anchored proteins. CP refers to the GPI anchor splicing variant form, which differs from the secretory form of CP (secCP) in the serum. Their expression substantially decreased in A549-Z5Z7-DKO cells compared with WT A549 cells detected by immunoblotting (Fig. 3A). The loss of CD55, CD59, and CP expression on the SK-Z5Z7- Zn mediated by ZNTs is critical to GPI-AP expression DKO cell surface was confirmed via immunofluorescence staining (Fig. 3B). Cell surface expression of CD55 and CD59 was restored by coexpression with ZNT5 or ZNT7 but not by zinc-transport incompetent ZNT5 or ZNT7 mutants (ZNT5 H451A and ZNT7 H70A ) (15,25) when cultured even in the presence of 75 μM ZnSO 4 (Fig. 3C). Cell surface expression of chimeric CD55-TM and CP-TM (GPI anchor attachment was substituted with the membrane-spanning polypeptide portion of ACE2 as described previously) was observed following transfection into SK-Z5Z7-DKO cells coexpressing ZNT5 or ZNT7 at the same level as WT cells (Fig. 3, D and E). Moreover, the expression of secCP was comparable between A549-Z5Z7-DKO and WT A549 cells (Fig. S1C). These results suggest that GPI-anchored proteins require ZNT5-6 or ZNT7 for cell surface expression.

Impairment of endogenous GPI-anchored protein cell surface expression in Z5Z7-DKO cells
The change in expression of endogenous GPI-anchored proteins was examined using WT HAP1 and HAP-Z5Z7-DKO cells because TNAP GPI-anchored protein is substantially lower in HAP-Z5Z7-DKO cells than that in WT HAP1 cells (16). SWATH-MS analysis (26) of membrane fractions identified 2882 proteins and revealed the relative expression differences between HAP-Z5Z7-DKO cells and WT HAP1 cells (Table S1). Gene ontology term enrichment analysis of cellular components using the top 15 proteins revealed that the GPI anchor component was the most enriched protein type whose expression decreased ( Table 2) and included BST2 (first), TNAP (fifth), glypican-4 (GPC4) (sixth), and CD55 (seventh, shown as DAF) as GPI anchor components A C D B Figure 1. Properties of transiently expressed GPI-anchored and single membrane-spanning polypeptide-anchored zinc ectoenzymes in Z5Z7-DKO cells. A, the activity and protein expression of GPI-anchored zinc ectoenzymes (TNAP, PLAP, and CD73). B, immunofluorescence staining of TNAP, PLAP, and CD73, detected in red fluorescence. C, the activity and protein expression of single membrane-spanning polypeptide-anchored zinc ectoenzymes (sACE, gACE, and ACE2). gACE, detected in the lower left panel for the blotting of β-gal, is indicated with an asterisk (*). D, immunofluorescence staining of sACE, gACE, and ACE2. In (A) and (C), expression plasmids were transiently transfected in WT A549 cells and A549-Z5Z7-DKO cells. β-galactosidase (β-gal) was used as the internal control. All activities are expressed as the mean ± SD of triplicate experiments (upper graphs). Ten micrograms of total cellular lysate prepared from transiently transfected cells was subjected to immunoblot analysis (lower panels). In (B) and (D), WT SK-MEL-2 and SK-Z5Z7-DKO cells were transiently transfected with expression plasmid harboring each cDNA in IRES-GFP plasmid, and transfected cells were discriminated by GFP fluorescence. Merged images with GFP and DAPI are shown. Each experiment was performed at least three times and representative results from independent experiments are shown. ACE, angiotensin-converting enzyme; cDNA, complementary DNA; DAPI, 4,6-diamino-2-phenylindole; DKO, double KO; gACE, germline specific ACE; GPI, glycosylphosphatidylinositol; IRES, internal ribosome entry site; PLAP, placental ALP; sACE, somatic ACE; TNAP, tissue nonspecific ALP.
( Table S1). Immunoblotting showed decreased expression of endogenous CD55, BST2, TNAP, and CD59 (unidentified in SWATH-MS analysis) in HAP-Z5Z7-DKO cells compared with WT HAP1 cells (Fig. 4A). Moreover, similar markedly decreased expression patterns were confirmed in WT SK-MEL-2 cells and SK-Z5Z7-DKO cells (CD73 was used instead of TNAP in SK-MEL-2 cells) (Fig. 4B). Immunofluorescence staining and immunoblotting confirmed the loss of CD55 and CD59 on the cell surface in HAP-Z5Z7-DKO cells (Fig. 4, C and D), which was restored by transfecting cells with plasmid expressing ZNT5 (Fig. 4, C-E). These results clearly showed that ZNT5-6 and ZNT7 play pivotal roles in the maturation and cell surface localization of endogenous GPI-anchored proteins.

Examination of GPI-anchored protein expression in Z5Z7-DKO cells with impaired PIG protein function
The GPI anchor is biosynthesized and assembled in sequential reactions by more than 20 PIG proteins ( Fig. S2) (2,3,27). GPI-anchored protein expression is substantially decreased in cells deficient in most PIG genes (28). GPI ethanolamine phosphate transferases PIGN, PIGO, and PIGG are suggested to be zinc enzymes (29)(30)(31).  consistent with a previous study (28). Meanwhile, BST2, CD55, CD59, and TNAP expression in HAP-Z5Z7PIGG-triple KO (TKO) cells and HAP-Z5Z7PIGN-TKO cells was almost identical to that in HAP-Z5Z7-DKO cells without any significant synthetic defects such as growth impairment (Fig. 5B). This suggested that the PIGG or PIGN defects merged with those of both ZNT5 and ZNT7 and that ZNT5-6 and ZNT7 supply zinc to these PIGs. PIGA is an essential component of an enzyme complex involved in the first step of GPI anchor biosynthesis while PIGT functions in the final step of generating GPI-anchored proteins by transferring the biosynthesized GPI anchor to the C terminus (Fig. S2); thus, the impairment of both components results in loss of cell surface expression of GPIanchored proteins (28,32,33). HAP-PIGA-KO and HAP-PIGT-KO cells showed substantially decreased expression of BST2, CD55, CD59, and TNAP (Fig. 5C). Moreover, substantial decreases were maintained in HAP-Z5Z7PIGA-TKO cells and HAP-Z5Z7PIGT-TKO cells, both of which grew in a similar manner to the respective KO cells and HAP-Z5Z7-DKO cells without any apparent synthetic defects. Thus, defects caused by the loss of PIGA or PIGT merged with those caused by deficiency of ZNT5-6 and ZNT7 functions in the GPI-anchored protein expression. These results strongly suggest that ZNT5-6 and ZNT7 contribute to GPI anchor biosynthesis conducted by PIG proteins in the ER, probably by supplying zinc to GPI ethanolamine phosphate transferases such as PIGN, PIGO, and PIGG.
Several possibilities exist to their loss of function. The active sites of PIGN, PIGO, and PIGG are homologous to ALPs (29), which are easily impaired in Z5Z7-DKO cells, as shown in reduced PLAP-TM(G) and PLAP-TM(H) activities ( Fig. 2A) and secPLAP activity (Fig. S1A). Alternatively, we suggest that compared with ectoenzymes, intracellular zinc enzymes may be more easily impaired in Z5Z7-DKO cells because our recent study revealed that the activity of lysosomal-localized sphingomyelin phosphodiesterase 1 (SMPD1) (34,35) is impaired in HAP-Z5Z7-DKO cells (36), although it is not a GPI-anchored zinc enzyme. Since GPI anchor biosynthesis and its transfer to proteins occurs in the ER (Fig. S2) (2,3,27,28), our present study raises a fundamental question of how primarily Golgi apparatus localized ZNT5-6 and ZNT7 contribute to GPIanchored protein maturation in the ER through zinc supply. It is suggested that either PIG proteins including PIGN, PIGO, and PIGG are retrogradely trafficked from the Golgi apparatus to the ER after zinc metalation in the Golgi apparatus or transient localization of ZNT5-6 and ZNT7 to the ER may contribute to supplying zinc to PIG proteins before their trafficking to the Golgi apparatus. We are currently investigating these mechanisms.
In Z5Z7-DKO cells, GPI-anchored protein expression was substantially decreased. Intracellular protein degradation may be a primary underlying cause for this observation, as reported for cells with loss-of-function mutations in most PIG genes (37). They may be degraded by the proteasome through ERassociated protein degradation (ERAD). GPI anchor proteins are known to be retrotranslocated for degradation by ERAD if GPI-anchored biosynthesis is prevented (38,39) and misfolded GPI-anchored proteins accumulate in the presence of proteasome inhibitors (40,41). Another possibility is that they may be degraded via the rapid ER stress-induced export degradation pathway. In this pathway, GPI-anchored proteins are eventually degraded in lysosomes (39,42). Our assumption of these possibilities is derived from our previous results showing that TNAP expression was partially increased by the treatment of inhibitors of both proteasomal and lysosomal degradations (16,25). However, degradation via the rapid ER stress-induced export pathway is unlikely because it is operative for misfolded GPI-anchored proteins after attachment of the GPI (3). In addition to intracellular degradation of GPIanchored proteins, another possibility is that their proproteins lacking the GPI anchor attachment may be secreted into the extracellular space through escaping ERAD degradation in Z5Z7-DKO cells considering that a similar secretion is known to occur in loss-of-function mutations in some PIG genes (PIGO and two other PIG genes) (2, 37, 43). We detected very weak ALP activity in the cultured medium of HAP-Z5Z7-DKO cells but not in WT HAP1 cells (data not shown), which may support this possibility. We are also currently investigating the regulation of GPI-anchored protein expression in Z5Z7-DKO cells from these views. In this study, Znt5 −/− ;Znt7 +/− medaka showed more severe phenotypes than zebrafish Pigu or Pigk mutants (20,21). We speculate that this is because the defects caused by homozygous Znt5 and heterozygous Znt7 are broader and exacerbate other cellular functions, which may cause synthetic defects in GPI-anchored protein expression. Moreover, different phenotypes of Znt5 −/− ;Znt7 +/− medaka from that of Znt5 +/− ; Znt7 −/− indicate that ZNT5-6 and ZNT7 have different functions in vivo, although their contributions to zinc ectoenzyme activation are similar. Clarifying physiological differences between ZNT5-6 and ZNT7 is important because the ZNT5/ SLC30A5 mutation is responsible for perinatal lethal cardiomyopathy (44). A series of medaka mutants would be a useful tool to dissect the overlapping or specific functions of ZNT5-6 and ZNT7 in vivo.
GPI-anchored protein expression was the most significantly decreased protein type in HAP-Z5Z7-DKO cells compared with WT HAP1 cells according to Gene ontology term enrichment of the top 15 proteins in SWATH-MS data. Lysosomes were enriched when analyzing the top 500 proteins with reduced expression in HAP-Z5Z7-DKO cells compared with WT HAP1 cells, suggesting that ZNT5-6 and ZNT7 have broad biological functions in addition to GPI anchor biosynthesis. This is consistent with our recent finding that the lysosomal enzyme SMPD1 requires zinc mediated by ZNT5-6 and ZNT7 (36). Meanwhile, ER-Golgi transport was enriched when analyzing increased expression in HAP-Z5Z7-DKO cells compared with WT HAP1 cells, which may reflect the alleviation of GPI-anchored protein expression defects. Further experiments are required to determine whether these defects were directly affected by the loss of ZNT5-6 and ZNT7 functions or indirectly affected by the loss of GPI-anchored protein expression.
In conclusion, our results revealed a previously unappreciated insight into GPI-anchored protein expression, which requires zinc transport via ZNT5-6 and ZNT7. The results support previous literature showing that both complexes are intimately involved in protein quality control in the early secretory pathway. Considering that GPI-anchored proteins operate as enzymes, receptors, adhesion proteins, and complement regulatory proteins, the present results provide an important basis for understanding the physiological and molecular significance of zinc in biological processes that are dynamically mobilized by zinc transporters. Further studies are required to determine the molecular mechanism underlying GPI-anchored protein expression and dissect the critical role of ZNT5-6 and ZNT7 to improve our understanding of the crucial functions of zinc on protein quality control in the ER and the secretory pathway.

Plasmid construction
Plasmid preparation for the expression of N-terminal HAtagged or FLAG-tagged human ZNT5 (HA-ZNT5 or FLAG-ZNT5), C-terminal HA-tagged human ZNT7 (ZNT7-HA), and their zinc transport-incompetent mutants (HA-ZNT5 H451A or ZNT7 H70A -HA) was performed as previously described (15,16). Plasmids expressing TNAP, PLAP, CD73, sACE, ACE2, gACE, CD55, CP, and CD59 were constructed by inserting full-length TNAP, PLAP, and CD73 complementary DNAs (cDNAs) (17), sACE cDNA (provided by Dr Eric Clauser, Collège de France) (45), ACE2 cDNA (provided by Dr Shuetsu Fukushi, National Institute of Infectious Diseases) (46), or other cDNAs into pcDNA3 or IRES-GFP plasmids (provided by Dr Hirohide Saito, Kyoto University). IRES-GFP plasmid can discriminate cells expressing the subcloned cDNA using GFP fluorescence, when used in transfection studies. The gACE cDNA was constructed by replacing the 5 0 region of the sACE gene coding the N-terminal domain of sACE with the fragment of the 5 0 region of Zn mediated by ZNTs is critical to GPI-AP expression the gACE gene (containing the initiation codon ATG [A is referred to as +1] to a PpuMI restriction site [+452]) (gBlock synthetic gene fragment, Integrated DNA Technologies). SecCP expression plasmid was purchased from DNAFORM, which was fused in frame with a cDNA fragment encoding its GPI anchor to generate its GPI-anchored form (CP) as described elsewhere (47). CD55 and CD59 cDNAs were amplified by PCR using human neural cDNA (DV Biologics) or human pancreatic islet cDNA (Cosmo Bio Co, Ltd) as a template. ACE-GPI and ACE2-GPI fusion plasmids were constructed by replacing the 3 0 regions of ACE and ACE2 encoding the C-terminal transmembrane region with the 3 0 region of the PLAP gene encoding the GPI anchor attached site using two-step PCR using KOD-PLUS Taq polymerase (Toyobo). The cycle conditions were denatured at 95 C for 5 min and amplified for 35 reaction cycles with denaturation at 95 C for 10 s, annealing at 60 C for 30 s, extension at 68 C for 2 min per cycle, and a final extension step at 68 C for 10 min. CP-TM and CD55-TM fusion plasmids were constructed by replacing the 3 0 regions of CP or CD55 encoding the portions of their GPI attachment with that of the membrane-spanning polypeptide portion of ACE2. The secretory form of ACE (secACE-HA) was constructed by replacing the 3 0 region encoding the C-terminal transmembrane region with the fragment encoding the HA tag. The secretory form of PLAP (secPLAP-HA) has been previously described (16). PLAP-TM(H) or PLAP-TM(G) expression plasmids were constructed by inserting cDNA (provided by Dr Deborah Brown, Stony Brook University) (24) into the pcDNA3 plasmid. Plasmids expressing β-galactosidase (pβactβgal) were described elsewhere (48).
Disruption of ZNT5 and ZNT7 or PIGN, PIGO, PIGG, PIGA, and PIGT genes ZNT5 and ZNT7 genes were simultaneously disrupted by CRISPR/Cas9-mediated genome editing using single guide RNA expression plasmids as previously described (16). The constructed plasmids (4 μg) and one-tenth quantity of pApuro vector (25) containing the puromycin resistance gene for A549 cells or pcDNA3 containing the neomycin resistance gene for SK-MEL-2 cells were cotransfected into 80% confluent cells using 4 μl Lipofectamine 2000 (Invitrogen). After culturing for 1 day, the cells were transferred to a 10 cm cell culture dish and cultured in the presence of 4 μg/ml puromycin (InvivoGen) for A549 cells or 2 mg/ml G418 (Nacalai Tesque) for SK-MEL-2 cells, respectively, to establish stable clones. HAP-Z5Z7-DKO cells were established in a previous study (16). HAP1 cells deficient in PIGN, PIGO, PIGG, PIGA, or PIGT genes were established in the same manner as previously described (16). To generate HAP-DKO or HAP-TKO cells, the cells were cultured in the presence of 20 μg/ml blasticidin S (InvivoGen), 0.75 μg/ml puromycin or 1.5 mg/ml G418. Oligonucleotides used for the generation of single guide RNA expression plasmids are listed in Table S2. Gene-edited cells were confirmed by sequencing the genomic DNA amplified PCR fragments using the primers listed in Table S3.

Transient and stable transfection
Cells were seeded into 12-well plates (1.0 × 10 5 cells/well for A549 and 1.2 × 10 5 cells/well for SK-MEL-2) and cultured for 24 h. A549 cells were transfected with 1 μg of empty pcDNA3 or pcDNA3 harboring each cDNA with 0.2 μg of pβactβgal plasmid (for transfection efficiency normalization) in Opti-MEM reduced serum media (Thermo Fisher Scientific) using 1.5 μl Lipofectamine 2000. The same strategy was used for transient transfection of secretory enzymes except that 0.2 μg of pCMV-Gaussia Luc plasmid (Thermo Fisher Scientific) was used to express secretory Gaussia luciferase instead of the pβactβgal plasmid. SK-MEL-2 cells were transfected with 0.5 μg of empty IRES-GFP plasmid or IRES-GFP plasmid harboring each cDNA. The transfection medium was replaced after 8 h and 4 h for A549 cells and SK-MEL-2 cells, respectively, with the corresponding culture medium, and cells were cultured for an additional 16 h and 24 h for A549 cells and SK-MEL-2 cells, respectively, prior to the experiments. Stable transfection in HAP1 cells (seeded at 5 × 10 5 cells/ml in a 6 cm dish) used 4 μg of the respective expression plasmids in Opti-MEM using 4 μl Lipofectamine 2000. Transfected cells were transferred to a 10 cm cell culture dish and cultured in the presence of 0.75 μg/ml puromycin as previously described (16). More than three stable independent clones were established per transfectant in all the experiments.

Immunofluorescence staining
Immunostaining was mainly performed in SK-MEL-2 cells because little nonspecific staining was observed when the primary antibodies (Table S4) were applied to the cells. The cells were cultured on coverslips coated with 0.01% poly-Llysine (Sigma-Aldrich) and fixed with 10% formaldehyde neutral buffer solution (Nacalai Tesque). Meanwhile immunodetection of CD55 and CD59 was performed in HAP1 cells fixed with methanol. Cells were incubated with primary antibodies followed by secondary antibodies (Alexa 594-conjugated goat anti-mouse immunoglobulin G [IgG], Alexa 594-conjugated goat anti-rat IgG, Alexa 488-conjugated donkey anti-rabbit IgG, or Alexa 488-conjugated goat antirabbit IgG) and tertiary antibodies (Alexa 594-conjugated donkey anti-goat IgG or Alexa 594-conjugated rabbit anti-goat IgG) (Thermo Fisher Scientific) without permeabilization. Alexa 488-conjugated donkey anti-rabbit IgG and Alexa 594-conjugated donkey anti-goat IgG (Abcam) were also used as secondary and tertiary antibodies. The antibodies were applied for 1 h at room temperature or at 4 C overnight, and 5 μg/ml 4,6-diamino-2-phenylindole (Thermo Fisher Scientific) was added during the second or third antibody staining to label nuclei. After three PBS washes, the coverslips were mounted onto the glass slides using SlowFade Diamond Antifade Mountant reagent (Thermo Fisher Scientific). The stained cells were examined using an Olympus FSX100 fluorescence microscope. Identical exposure settings and times were used for the corresponding images in each Figure.

Measurement of PLAP and TNAP activity
Total cellular lysates prepared from transfected cells lysed in ALP lysis buffer (10 mM Tris-HCl, 0.5 mM MgCl 2 , 0.1% Triton X-100，pH 7.5) were used to measure TNAP or PLAP activity as previously described (15). Lysates were preincubated for 10 min at room temperature (RT), followed by the addition of 100 μl substrate solution (2 mg/ml disodium p-nitrophenylphosphate hexahydrate or pNPP; Wako Pure Chemicals, in 1 M diethanolamine buffer pH 9.8 containing 0.5 mM MgCl 2 ), and the released p-nitrophenol product was quantified by measuring the absorbance at 405 nm using a Synergy H1 Hybrid multimode microplate reader (BioTek). Calf intestinal ALP (Promega) was used to generate a standard curve. The samples were incubated at 65 C for 30 min to measure PLAP activity to discriminate between other activities.

Measurement of CD73 activity
CD73 activity was evaluated using the malachite green assay as previously described (17). Briefly, 5 μg of total cell lysate prepared from transfected cells in 200 μl of CD73 lysis buffer (10 mM Tris-HCl pH 7.5, 0.1% Nonidet P-40) was incubated at 37 C for 10 min, followed by incubation with 2 mM AMP (Oriental Yeast Co, Ltd) at 37 C for 30 min. Fifty microliter of the reaction solution was mixed with 100 μl of Biomol Green (Enzo Life Sciences) and incubated for 10 min at RT. The amount of inorganic phosphate released from AMP was quantified by measuring the absorbance at 595 nm using a Synergy H1 Hybrid multimode microplate reader. A standard curve was generated using 0 to 200 μM Na 2 HPO 4 solutions dissolved in CD73 lysis buffer.

Measurement of β-gal activity
Total cellular lysate prepared from transfected cells was incubated at 37 C for 10 min, and 120 μl of buffer A (100 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgSO 4 , 50 mM β-mercaptoethanol, pH 7.5) was added to each well. The plate was incubated at 37 C for 5 min followed by the addition of 50 μl 0.1 mM o-nitrophenyl β-D-galactopyranoside (FUJIFILM Wako Pure Chemical) and incubation at 37 C for 60 min. The released o-nitrophenol was quantified by measuring the absorbance at 405 nm. β-D-galactosidase purified from Escherichia coli (FUJIFILM Wako Pure Chemical) was used to generate a standard curve.

SWATH-MS analysis
Isolation of the membrane fraction was described elsewhere (51) with minor modifications. HAP1 cells cultured on 4 × 10 cm dishes were collected, washed with cold PBS, and homogenized in homogenization buffer (225 mM mannitol, 75 mM sucrose, 30 mM Tris-HCl, and 100 μM EGTA, pH 7.4) using a Dounce homogenizer. The homogenate was centrifuged twice at 600g for 5 min at 4 C to eliminate the pellet containing nuclei, 7000g for 10 min at 4 C to remove mitochondria, and the supernatant containing membrane fractions consisting of microsomes, plasma membranes, and lysosomes was centrifuged at 100,000g for 60 min at 4 C. The supernatant was discarded, while the purified membrane fractions were stored at −80 C.
SWATH-MS analysis was performed as described previously (52). Fifty micrograms of membrane protein fractions prepared from WT HAP1 cells and HAP-Z5Z7-DKO cells were solubilized in denaturing buffer (7 M guanidium hydrochloride, 0.5 M Tris-HCl (pH 8.5), and 10 mM EDTA). The solubilized proteins were reduced by DTT for 1 h at 25 C, and, subsequently, S-carboxymethylated with iodoacetamide for 1 h at 25 C. The alkylated proteins were precipitated with methanol-chloroform-water mixture. The precipitates were solubilized in 6 M urea in 0.1 M Tris-HCl (pH 8.5) and diluted fivefold with 0.1 M Tris-HCl (pH 8.5) containing 0.05% Pro-teaseMax surfactant (Promega). The dilutions were reacted with lysyl endopeptidase (Lys-C; Wako Pure Chemical) at an enzyme/substrate ratio of 1:100 for 3 h at 30 C. Subsequently, Lys-C digested proteins were treated with TPCK-treated trypsin (Promega) at an enzyme/substrate ratio of 1:100 for 16 h at 37 C. After C18 clean up, the peptide samples were injected into a NanoLC Ultra system (Eksigent Technologies) coupled with an electrospray-ionization TripleTOF 5600 mass spectrometer (SCIEX), which was set up for a single direct injection and analyzed by SWATH-MS acquisition. The peptides were directly loaded onto a self-packed C18 analytical column, prepared by packing ProntoSIL 200-3-C18 AQ beads (3 μm, 120 Å; BISCHOFF Chromatography) in a PicoFrit tip (ID 75 μm, PF360-75-10-N5; New Objective) of 20 cm length. After sample loading, the peptides were separated and eluted with a linear gradient; 98% A: 2% B to 65% A: 35% B (0-120 min), increase to 0% A: 100% B (120-121 min), maintained at 0% A: 100% B (121-125 min), reduced to 98% A: 2% B (125-126 min), and then maintained at 98% A: 2% B (126-155 min). Mobile phase A composition was 0.1% formic acid in water, and mobile phase B contained 0.1% formic acid in acetonitrile. The flow rate was 300 nl/min. The eluted peptides were positively ionized and measured in the SWATH mode. The measurement parameters are described as follows: SWATH window, 64 variable windows from 400 m/z to 1200 m/z; product ion scan range, 50 to 2000 m/z; declustering potential, 100; rolling collision energy value, 0.0625 × [m/z of each SWATH window] − 3.5; collision energy spread, 15; accumulation time, 0.05 s for each SWATH window.
Spectral alignment and data extraction from SWATH data were performed with the SWATH Processing Micro App in Peakview version 2.2 (SCIEX) using two spectral libraries, an in-house spectral library and a publicly available pan human library (PHL), for increasing the identification number of expressed proteins. The parameters for peak data extraction by Peakview were described as follows: number of peptide per protein, 999; number of transitions per peptide, 6; peptide confidence threshold, 99%; false discovery rate threshold, 1.0%; XIC extraction window, ±4.0 min; XIC width (ppm), 50. The detail of data analysis is described in Fig. S4. This data analysis improves the accuracy of protein identification and quantification by selecting only those peptides that meet the in silico peptide selection criteria (53). The peptide selection criteria are stringent, which increases the reliability of quantification, but reduces the number of peptides used to quantify individual proteins; as shown in Table S1, there are proteins that are quantified with a single peptide. The four GPI anchor components (BST2, TNAP, GPC4, and CD55), which are the most important in this study, were also analyzed by another data analysis procedure (Fig. S5). Quantification by two or more peptides ensures reliability (Table S5). In addition to the quantification results at the protein level, Table S5 also shows the peptide sequences used in the quantification analysis for individual proteins.

Establishment and care of Znt-deficient medaka
All medaka strains including WT were maintained in an aquarium with recirculating water in a 14/10 h light/dark cycle at 26 C, which meets the Regulation for Animal Experiments in Kyoto University approved by the Animal Research-Animal Care Committee of Kyoto University (R3-45 and Lif-K21023).
The injected G0 embryos were cultured and bred into adults. The G0 fish were mated with WT counterparts, and F1 individuals with red fluorescence for Znt5 or green fluorescence for Znt7 in the eyes were selected. The F1 individual was mated with the WT counterpart, and the resulting heterozygous F2 individuals deficient in Znt5 or Znt7 were selected. In addition, continuous mating of F1 individuals with red or green eye fluorescence with their descendants produced individuals with various genotype combinations (Znt5 +/− : Znt7 +/− , Znt5 +/− :Znt7 −/− , Znt5 −/− :Znt7 +/− , Znt5 −/− :Znt7 −/− ). Each genotype was confirmed via PCR using KOD-FX (Toyobo) in the PCR cycling conditions: denaturation for 2 min followed by 30 cycles at 98 C for 10 s, annealing at 55 C for 10 s and extension at 68 C for 30 s for Znt5 and Znt7 alleles. Primers used for PCR amplification and DNA sequencing are listed in Table S8.

Statistical analyses
All data are expressed as the mean ± SD of triplicate experiments. Statistical significance was determined by twotailed Student's t test at p < 0.01 (**).

Data availability
All data generated or analyzed during this study are included in this published article and its supporting information file or are available from the corresponding author (Taiho Kambe, Kyoto University, E-mail: kambe.taiho.7z@kyoto-u.ac. jp) upon reasonable request. Full-length immunoblots corresponding to images in the main text and supplementary figures are shown in Figs. S7-S13.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE (56) partner repository with the dataset identifier PXD032172.