Sequential hydrolysis of FAD by ecto-5′ nucleotidase CD73 and alkaline phosphatase is required for uptake of vitamin B2 into cells

Extracellular hydrolysis of flavin-adenine dinucleotide (FAD) and flavin mononucleotide (FMN) to riboflavin is thought to be important for cellular uptake of vitamin B2 because FAD and FMN are hydrophilic and do not pass the plasma membrane. However, it is not clear whether FAD and FMN are hydrolyzed by cell surface enzymes for vitamin B2 uptake. Here, we show that in human cells, FAD, a major form of vitamin B2 in plasma, is hydrolyzed by CD73 (also called ecto-5′ nucleotidase) to FMN. Then, FMN is hydrolyzed by alkaline phosphatase to riboflavin, which is efficiently imported into cells. We determined that this two-step hydrolysis process is impaired on the surface of glycosylphosphatidylinositol (GPI)-deficient cells due to the lack of these GPI-anchored enzymes. During culture of GPI-deficient cells with FAD or FMN, we found that hydrolysis of these forms of vitamin B2 was impaired, and intracellular levels of vitamin B2 were significantly decreased compared with those in GPI-restored cells, leading to decreased formation of vitamin B2-dependent pyridoxal 5′-phosphate and mitochondrial dysfunction. Collectively, these results suggest that inefficient uptake of vitamin B2 might account for mitochondrial dysfunction seen in some cases of inherited GPI deficiency.

Extracellular hydrolysis of flavin-adenine dinucleotide (FAD) and flavin mononucleotide (FMN) to riboflavin is thought to be important for cellular uptake of vitamin B 2 because FAD and FMN are hydrophilic and do not pass the plasma membrane. However, it is not clear whether FAD and FMN are hydrolyzed by cell surface enzymes for vitamin B 2 uptake. Here, we show that in human cells, FAD, a major form of vitamin B 2 in plasma, is hydrolyzed by CD73 (also called ecto-5 0 nucleotidase) to FMN. Then, FMN is hydrolyzed by alkaline phosphatase to riboflavin, which is efficiently imported into cells. We determined that this two-step hydrolysis process is impaired on the surface of glycosylphosphatidylinositol (GPI)deficient cells due to the lack of these GPI-anchored enzymes. During culture of GPI-deficient cells with FAD or FMN, we found that hydrolysis of these forms of vitamin B 2 was impaired, and intracellular levels of vitamin B 2 were significantly decreased compared with those in GPI-restored cells, leading to decreased formation of vitamin B 2 -dependent pyridoxal 5 0 -phosphate and mitochondrial dysfunction. Collectively, these results suggest that inefficient uptake of vitamin B 2 might account for mitochondrial dysfunction seen in some cases of inherited GPI deficiency.
The water-soluble vitamins are essential to mammals and play roles in various metabolic reactions by working as coenzymes. The major forms of vitamins B 1 , B 2 , and B 6 in blood are phosphorylated or nucleotide forms (1)(2)(3)(4)(5)(6) and they need to be hydrolyzed for uptake by cells. Alkaline phosphatase (ALP), a glycosylphosphatidylinositol (GPI)-anchored protein (GPI-AP), dephosphorylates thiamine pyrophosphate and pyridoxal 5 0 -phosphate (PLP) to thiamine and pyridoxal (PL), respectively (7,8). However, it is uncertain whether the active forms of vitamin B 2 , flavin-adenine dinucleotide (FAD) and flavin mononucleotide (FMN) (Fig. 1A), are hydrolyzed by the cell surface enzyme for their uptake. Moreover, enzymes which hydrolyze them are not known (9).
In food, vitamin B 2 is mainly in the form of FAD or FMN. Because they are hydrophilic, they are not imported directly into the intestinal epithelial cells; instead, they must be converted to riboflavin (RF, Fig. 1A) on cell surface for uptake (10,11). RF uptake into cells is mediated by three kinds of RF transporters (SLC52A1, 2, 3) (9). FMN and FAD are then regenerated from RF in the cytoplasm by RF kinase and FAD synthetase (9,12). These active forms of vitamin B 2 work as cofactors of various enzymes involved in redox reactions in many metabolic pathways, such as the tricarboxylic acid cycle, vitamin B 6 metabolism, and mitochondrial electron transport chain. Therefore, vitamin B 2 deficiency causes mitochondrial dysfunction as well as various metabolic disorders (9).
One candidate enzyme involved in the hydrolysis of FMN and FAD is ALP. Daniel et al. (11) reported that FMN and FAD were hydrolyzed by ALP purified from the brush-border membrane of rat jejunum. ALP hydrolyzes compounds with a phosphate moiety, such as pyrophosphate, phosphoethanolamine, and PLP (8). There are four main isoforms of ALP in humans-tissue-nonspecific ALP (TNSALP), intestinal ALP, germ cell ALP, and placental ALP-all of which are GPI-APs (13). Among them, TNSALP is ubiquitously expressed and is the major isoform expressed in liver, bone, kidney, blood, and brain (14).
Although, Daniel et al. (11) reported that both FMN and FAD were hydrolyzed by ALP, there are several reports suggesting that two enzymes are involved in the hydrolysis of FMN and FAD. Akiyama et al. purified two independent enzymes, FAD pyrophosphatase and FMN phosphatase, from rat intestinal brush-border (10). Okuda showed that the inhibitory effect of pyrophosphate was greater against FMN hydrolysis than FAD hydrolysis in dog intestinal mucosa and suggested that the small intestine contains at least two kinds of phosphatase: nucleotide pyrophosphatase with low affinity for pyrophosphate and phosphomonoesterase with high-affinity for pyrophosphate (15). Okuda also reported that gastric juice hydrolyzes only FAD, and bile and pancreatic juice hydrolyze only FMN (16). Lee and Ford showed that an enzyme purified from placental trophoblastic microvilli possessed FAD pyrophosphatase activity and the enzyme did not hydrolyze FMN (17). These reports suggest a two-step hydrolysis of FAD (i.e., FAD-FMN, then FMN-RF).
Here, we show that, in human cells, FAD is hydrolyzed in two steps: FAD to FMN by CD73 and FMN to RF by ALP. CD73, also known as ecto-5 0 nucleotidase, is a GPI-AP encoded by the NT5E gene and catalyzes conversion of adenosine 5 0 -monophosphate (AMP) to adenosine (13,18). To determine the hydrolytic activity of human bone ALP and recombinant human CD73 toward vitamin B 2 analogues, RF formation from FMN and FMN formation from FAD were measured. The contributions of these GPI-APs to vitamin B 2 uptake, vitamin B 2 -dependent vitamin B 6 metabolism, and mitochondrial function were determined using GPI-deficient cells. Because CD73 and ALP are both GPI-APs (13,19), GPI-deficient cells contained a decreased amount of intracellular vitamin B 2 when cultured with FAD as the vitamin B 2 source, leading to changes in vitamin B 6 metabolism and mitochondrial dysfunction. These lines of evidence suggest that the mitochondrial dysfunction seen in some severe cases of inherited GPI deficiency (IGD) might be caused by vitamin B 2 deficiency, which would be prevented by the administration of RF.

RF formation from FMN by ALP and FMN formation from FAD by CD73
To analyze the two-step hydrolysis of FAD, RF and FMN concentrations were measured in vitro by HPLC after incubation of 10 μM FMN and FAD, respectively, with human bone ALP and CD73 containing a C-terminal His tag in solution for 15 min. An increase in the RF concentration was detected after incubation of FMN with ALP but not with CD73. In contrast, the FMN concentration was increased after incubation of FAD with CD73 but not with ALP ( Fig. 1B).
ALP-dependent RF formation from FMN is shown in Figure 2, A-C, including a time course and the dependence on the concentrations of ALP and FMN, respectively. RF production increased linearly for 10 min ( Fig. 2A) in an ALP concentration-dependent manner (Fig. 2B). RF production from FMN by ALP was substrate saturable (Fig. 2C), the Km and Vmax values being 0.309 ± 0.051 μM and 7.47 ± 0.25 nmol/min/mg protein, respectively (Table 1). These results demonstrate that purified bone ALP, that is, TNSALP, hydrolyzes FMN, producing RF.
Recombinant CD73-dependent FMN formation from FAD is shown in Figure 2, D-F. Because FAD consists of RF, diphosphate, and adenosine moieties, AMP (adenosine + phosphate) or adenosine could be produced by FAD hydrolysis (Fig. 1A). To determine the products of FAD hydrolysis by CD73, we simultaneously determined FMN, AMP, and adenosine by HPLC after incubation of FAD with CD73. Increased concentrations of FMN and adenosine were detected, but no increase in AMP was observed. The concentrations of FMN and adenosine increased linearly for 30 min (Fig. 2, D and G) in a CD73 concentration-dependent manner (Fig. 2, E and H). FMN and adenosine production by CD73 was dependent on the concentration of the substrate, FAD, and was saturable (Fig. 2, F and I). CD73 is known to hydrolyze AMP to adenosine; in addition, here, we show evidence that recombinant CD73 also hydrolyzes FAD to produce FMN and adenosine.

Kinetic parameters of ALP and CD73 activities
We determined kinetic parameters of the activities of ALP and CD73 (Table 1). Because AMP is a substrate for both CD73 and ALP (13,18), the results for kinetic analysis of FMN and FAD hydrolysis are compared with those for AMP hydrolysis in Table 1. The Km value for FAD hydrolysis by CD73 was higher than that for AMP, and the Vmax value for AMP hydrolysis was higher than that for FAD hydrolysis, suggesting that CD73 binds AMP with higher affinity than FAD and more efficiently hydrolyzes AMP than FAD. In contrast, the Km values for FMN and AMP hydrolysis by ALP were comparable.

Inhibition studies of ALP and CD73
Several types of inhibitors were used to characterize the hydrolytic properties of CD73 and ALP (Fig. 3). FMN hydrolysis and AMP hydrolysis by ALP showed similar patterns of inhibition (Fig. 3A). FAD hydrolysis and AMP hydrolysis by CD73 also showed similar patterns of inhibition (Fig. 3B). However, the inhibition properties of AMP hydrolysis mediated by ALP and CD73 were different. FMN and nicotinamide mononucleotide, which are phosphorylated vitamins, and A, time course of riboflavin (RF) production from FMN by ALP. The RF concentration was measured after incubation of 10 μM FMN with ALP (10 μg/ml). Data represent means ± SD (n = 3). B, ALP concentration dependence of RF production from FMN. The RF concentration was measured 5 min after incubation of 10 μM FMN with ALP (2.5-10 μg/ml). Data represent means ± SD (n = 3). C, Michaelis-Menten plot of RF production from FMN by human bone ALP. The RF concentration was measured 5 min after incubation of FMN (0.5-5 μM) with ALP (5 μg/ml). The enzyme activity was expressed in product concentration per minute per mg protein. Data represent means ± SD (n = 3). D and G, time course of FMN production (D) or adenosine production (G) from FAD by CD73. The FMN or adenosine concentration was measured after incubation of 10 μM FAD with CD73 (10 ng/ml). Data represent means ± SD (n = 3). E and H, CD73 concentration dependence of FMN production (E) or adenosine production (H) from FAD by recombinant CD73. The FMN or adenosine concentration was measured 15 min after incubation of 10 μM FAD with CD73 (5-20 ng/ml). Data represent means ± SD (n = 3). F and I, Michaelis-Menten plot of FMN production (F) or adenosine production (I) from FAD by recombinant CD73. The FMN or adenosine concentration was measured 15 min after incubation of 10 μM FAD with CD73 (10 ng/ml). The enzyme activity was expressed in product concentration per minute per μg protein. Data represent means ± SD (n = 3). When the SD is smaller than the symbols, it is not shown. levamisole, which is an inhibitor of TNSALP (20), decreased the activity of ALP but not CD73. In contrast, α,β-methylene adenosine 5 0 -diphosphate (APCP), an inhibitor of CD73 (21), inhibited CD73 but not ALP. Guanosine 5 0 -monophosphate, a nucleotide, inhibited both CD73 and ALP, which is consistent with AMP being a common substrate of CD73 and ALP (Fig. 3C). Figures 2 and 3 demonstrate that TNSALP hydrolyzes FMN, producing RF, and CD73 hydrolyzes FAD, producing FMN.
Extracellular hydrolysis and uptake of vitamin B 2 , and vitamin B 2 -dependent PLP and PL production in GPI-deficient cells Both CD73 and ALP are GPI-APs (13,19) and phosphatidylinositol glycan anchor biosynthesis class T (PIGT) is required for GPI-AP generation, and therefore, PIGT-KO SH-SY5Y cells (PIGT− cells) were generated using the CRISPR/Cas9 system to obtain CD73-and ALP-defective cells. SH-SY5Y is the human neuroblastoma cell line. The activities of CD73 and ALP in PIGT− cells were significantly lower than in PIGT rescued cells (PIGT+ cells) (Fig. 4, A and  B). Surface expression levels of CD73, TNSALP, and CD59 (another GPI-AP) on PIGT− and PIGT+ cells were compared by flow cytometric analysis (Fig. 4C). The surface expression of CD73, TNSALP, and CD59 was deficient in PIGT− cells.
PIGT− and PIGT+ cells were cultured in vitamin B 2depleted medium for 5 days, followed by culture for 24 h in a medium containing one of the vitamin B 2 derivatives (FMN, FAD, or RF) or a medium without vitamin B 2 . Vitamin B 2 concentrations in the cell and culture medium were measured by HPLC. FAD was the major form of vitamin B 2 in the PIGT+ and PIGT− cells after cultured in medium containing RF, FMN, or FAD, indicating that imported RF was intracellularly converted to FAD (9). The intracellular total vitamin B 2 concentrations were significantly lower in PIGT− cells than in PIGT+ cells after cultured in FMN-or FAD-containing medium, while they were similar after cultured in RF-containing or vitamin B 2 -depleted medium (Fig. 4D) (because RF can be transported into the cells by RF transporters).
These results indicate that the presence of FAD and FMN in the medium did not lead to efficient uptake of vitamin B 2 into GPI-deficient cells; this is because FAD and FMN were inefficiently hydrolyzed to FMN and RF, respectively, because of production from FMN; AMP hydrolysis activities were measured by adenosine production from AMP. The concentration of the inhibitors was 10 μM, except for α,β-methylene adenosine 5 0 -diphosphate (APCP; 2 μM) and levamisole (1 mM). Data represent means ± SD (n = 3). When the SD is smaller than the symbols, it is not shown. B, comparison of inhibitory effects of various compounds on hydrolysis of 10 μM flavin-adenine dinucleotide (FAD) and 4 μM AMP by CD73. FAD hydrolysis activities were measured by FMN production from FAD; AMP hydrolysis activities were measured by adenosine production from AMP. The concentration of the inhibitors was 10 μM, except for APCP (2 μM) and levamisole (1 mM). Data represent means ± SD (n = 3). When the SD is smaller than the symbols, it is not shown. C, comparison of inhibitory effects of various compounds on hydrolysis of 1 and 4 μM AMP by ALP and CD73, respectively. AMP hydrolysis activities were measured by adenosine production from AMP. The concentration of the inhibitors was 10 μM, except for APCP (2 μM) and levamisole (1 mM). Data represent means ± SD (n = 3). When the SD is smaller than the symbols, it is not shown.
FAD hydrolysis by GPI-anchored enzymes the defective expression of CD73 and ALP; this resulted in intracellular vitamin B 2 deficiency.
To analyze the effect of intracellular vitamin B 2 deficiency on the activity of vitamin B 2 -dependent enzymes, concentrations of vitamin B 6 derivatives in the medium were measured. Pyridoxine (PN, a form of vitamin B 6 ) is imported into cells and phosphorylated to pyridoxine 5 0 -phosphate (PNP). PNP is then converted to PLP by the FMN-dependent enzyme PNP oxidase, and PLP is in turn dephosphorylated to PL; PL and PLP are efficiently exported to the medium (22)(23)(24). Extracellular PL and PLP should be a biomarker for intracellular vitamin B 2 status because sum of PLP and PL is a net produced amount of metabolites by FMN-dependent PNP oxidase from PN which is contained in all precultured media as a vitamin B 6 source. After cultivation of PIGT+ and PIGT− cells in the presence of PN and FAD, the combined PL and PLP concentration in the medium was significantly lower for PIGT− cells than for PIGT+ cells (Fig. 4F). There was a significant positive relationship between the sum of PL and PLP concentrations in the medium and the intracellular FMN concentration (p < 0.001, Fig. 4G). These results suggest that the amount of vitamin B 2 imported into the cells affected the vitamin B 2 -dependent PLP and PL production.

Effect of intracellular vitamin B 2 deficiency on mitochondrial function
FMN and FAD act as coenzymes in the mitochondrial electron transport chain, and a decreased intracellular FAD concentration might cause mitochondrial dysfunction (25). To compare the mitochondrial function between PIGT+ and PIGT− SHSY5Y cells, O 2 consumption of cells was measured using a flux analyzer. After culture in vitamin B 2 -depleted medium for 5 days, cells were cultured in medium containing a vitamin B 2 derivative (FAD, RF, or no vitamin B 2 ) for 24 h and their O 2 consumption was then measured (Fig. 5). PIGT− cells showed significantly lower O 2 consumption than PIGT+ cells when they were cultured in FAD-containing medium, whereas those cultured in RF-containing medium showed a similar level of O 2 consumption to that in PIGT+ cells. These results suggest that the GPI-deficient cells are susceptible to mitochondrial dysfunction. Additionally, non-mitochondrial O 2 consumption, which is the residual O 2 consumption after addition of rotenone/antimycin A, was also decreased in PIGT-cells, suggesting the contribution of some flavoenzyme oxidases.

Discussion
The present study showed that purified human TNSALP from bone hydrolyzed FMN and recombinant human CD73 hydrolyzed FAD. ALP catalyzes the hydrolysis of monoesters of phosphoric acid (14). Here, we showed that ALP hydrolyzes monophosphate vitamin B 2 , FMN, but not dinucleotide-type vitamin B 2 , FAD. Inhibition study showed that guanosine 5 0monophosphate and nicotinamide mononucleotide inhibited ALP activity, but NAD and FAD did not, suggesting that ALP has high affinity for compounds with a phosphomonoester moiety.
We also showed that adenosine and FMN were produced from FAD by CD73. Because AMP hydrolysis had a higher Vmax and lower Km than FAD hydrolysis by CD73, we speculate that adenosine was immediately produced from AMP after AMP and FMN production from FAD, with conversion of FAD to FMN being rate limiting. However, at the moment, we cannot completely eliminate the possibility of flavin-pyrophosphate as an intermediate.
TNSALP from human bone was used in the present study. TNSALP hydrolyzes some phosphate compounds, such as inorganic pyrophosphate, PLP (vitamin B 6 ), and thiamine pyrophosphate (vitamin B 1 ). Hypophosphatasia (HPP) is caused by loss-of-function mutations in TNSALP. Decreased conversion of pyrophosphate to phosphate caused dysosteogenesis (8). Because cell surface hydrolysis of PLP to PL is important for vitamin B 6 uptake into cells, decreased PLP hydrolysis activity causes vitamin B 6 deficiency, leading to dysfunction of various vitamin B 6 -dependent enzymes such as glutamate decarboxylase which results in PN-dependent seizures (8,26). In HPP, lowered levels of thiamine pyrophosphate in red blood cells were reported (7). Adenosine and γ-aminobutyric acid concentrations were lower in brain of Akp2 KO mice than in wild-type mice; this gene encodes TNSALP in mice (27). The present study showed that the phosphorylated form of vitamin B 2 , FMN, is a substrate of human TNSALP. Because hydrolysis by ALP of vitamin B 1 , B 2 , and B 6 is required for their uptake, their uptake would be decreased in HPP, which will be a subject of future investigation.
Analysis using GPI-deficient cells, which are defective in both CD73 and ALP cell surface expression, showed that both FAD and FMN uptake activities were lower than in GPI-rescued cells, leading to intracellular deficiency of vitamin B 2 . Thus, the GPI-deficient cells showed dysfunction of the vitamin B 2dependent mitochondrial respiratory chain complex, as well as of PNP oxidase, and enzyme in vitamin B 6 metabolism. GPIdeficient cells showed significantly lower PLP and PL production and O 2 consumption when they were incubated with FAD than PIGT+ cells, whereas those incubated with RF showed a similar level to that in PIGT+ cells (Figs. 4 and 5F and 5, A and B). In addition, a significant positive relationship was found between the concentration of intracellular FMN and the sum of PLP and PL production (Fig. 4G). These results again suggest that cell surface hydrolysis of FAD to RF by the CD73 and ALP contributed to vitamin B 2 -dependent functions (Fig. 5C). However, it might be still possible that some other GPI-APs contribute to these functions.
IGD is caused by mutations in genes involved in the biosynthesis or modification of GPI-APs. Major symptoms of patients with IGD are intellectual disability, developmental delay, and seizures. Because both ALP and CD73 are GPI-APs, expression of these proteins is decreased in some patients with IGD. Here, we demonstrated that GPI-deficient cells showed decreased intracellular vitamin B 2 levels when cultured with FAD, a major form of vitamin B 2 in blood, which led to mitochondrial dysfunction. This is consistent with reports that some severe IGD cases show mitochondrial dysfunction (28). In future, we are planning to analyze the metabolic conditions in vivo using IGD model mice (29). CD73 expression is decreased in phosphatidylinositol glycan anchor biosynthesis class G (PIGG) KO cells (30) and some cases with null mutation of PIGG also showed decreased expression of CD73 and mitochondrial dysfunction (31), suggesting that CD73 expression is important for uptake of vitamin B 2 in PIGG deficiency. Similar to HPP, some patients with IGD show decreased vitamin B 6 uptake and suffer from PN-dependent seizures (32). High-dose nonphosphorylated vitamin B 6 (i.e., PN) treatment was effective in treatment of seizures in more than half of patients with IGD (33). Two types of nonphosphorylated vitamin B 6 are imported into cells and converted to PLP in the cell. One of them, PL, is contained in food from animal sources. PN and its glycoside are contained in food from plant sources (34); PN is intracellularly converted to PNP, and PNP is converted to PLP by FMN-dependent PNP oxidase (22)(23)(24). In addition to decreased uptake of vitamin B 6 , patients with IGD would show decreased intracellular conversion of PNP to PLP due to dysfunction of this FMN-dependent enzyme caused by the decreased vitamin B 2 uptake. Therefore, high-dose PN and RF treatment might be effective for patients with IGD. However, further study is required concerning vitamin B 6 and B 2 metabolism in patients with IGD.
Adenosine, produced by CD73 from AMP, is an immune inhibitory molecule through its receptor expressed on immune cells (35,36). Some tumors show upregulation of CD73, and adenosine promotes both migration and proliferation (37). Therefore, CD73 is a target for immunotherapy for cancer. Antibody against CD73 has been used in a phase 1 clinical study (36). Here, we show the importance of CD73 for vitamin B 2 uptake. In cancer therapy, the effect of antibody against CD73 on vitamin B 2 metabolism should be studied in future study.

Experimental procedures Materials
Human bone ALP was purchased from Calzyme. Human CD73 His-tag was purchased from BPS Bioscience. All other reagents were of analytical grade.

Measurement of hydrolysis of FAD
For measurement of hydrolysis of FAD, FAD and human bone ALP or human CD73 were incubated in reaction buffer containing 50 mM Tris-HCl (pH 7.4), 5 mM MgCl 2 , and 0.002% bovine serum albumin, after separately preincubation of FAD and the enzyme at 37 C for 3 min. At an appropriate time, the reaction was stopped by addition of ice-cold perchloric acid and the mixture was centrifuged (9000g for 5 min). The supernatant was transferred to a fresh tube, neutralized by KOH, and recentrifuged. The supernatant was filtered through a 0.45-μm membrane and simultaneously analyzed for FMN, RF, AMP, and adenosine concentrations by HPLC. The HPLC conditions were optimized from a previously reported method (38). The HPLC apparatus consisted of a Shimadzu LC-10ADvp system equipped with an RF-10Axl spectrofluorometer and SPD-10Avp UV-VIS detector (Shimadzu). Chromatographic separation was performed on an InertSustain AQ-C18 column (150 × 4.6 mm, i.d. 5 μm; GL Sciences) using a gradient elution mode at a flow rate of 1.0 ml/ min. The column temperature was 25 C. Mobile phase A was 10 mM potassium phosphate containing 5 mM EDTA- Oligomycin is an ATP synthase inhibitor; carbonyl cyanide p-trifluoromethoxyphenylhydrazone (carbonilcyanide p-triflouromethoxyphenylhydrazone) is an uncoupler; rotenone is a complex I inhibitor; and antimycin A is a complex III inhibitor. The data represent means ± SD (n = 2). Representative data from two independent experiments are shown. C, scheme of hydrolysis of flavin-adenine dinucleotide (FAD) and flavin mononucleotide (FMN) by CD73 and alkaline phosphatase for its uptake into cells, vitamin B 6 metabolism, and mitochondria function.
FAD hydrolysis by GPI-anchored enzymes disodium salt, adjusted to pH 6.0; mobile phase B was methanol. We used a linear gradient from 8% to 25% mobile phase B from 3 min to 6 min, followed by holding at 25% B until 25 min after injection. Fluorescence measurements were made with excitation at 440 nm and emission at 560 nm to determine FMN and RF concentrations. Absorbance measurements were made at 260 nm for the determination of AMP and adenosine. Hydrolysis activity was calculated as observed production minus nonenzymatic production, which was determined in negative controls from which enzyme was absent otherwise using the same procedure as described previously.

Measurement of hydrolysis of FMN
For measurement of hydrolysis of FMN, FMN and human bone ALP or human CD73 were incubated in reaction buffer containing 50 mM Tris-HCl (pH 7.4) and 5 mM MgCl 2 at 37 C for 3 min (39). The reaction procedures were the same as for analysis of FAD hydrolysis. The produced RF concentration was determined using an isocratic HPLC method with 75% buffer A and 25% buffer B (other conditions the same as for analysis of FAD hydrolysis).

Measurement of hydrolysis of AMP
For measurement of hydrolysis of AMP, AMP and human bone ALP or human CD73 were incubated in reaction buffer containing 50 mM Tris-HCl (pH 7.4) and 5 mM MgCl 2 at 37 C for 3 min (39). The reaction procedures and HPLC conditions were the same as those for analysis of FAD hydrolysis.

Generation of GPI-deficient cells
PIGT-KO cells were generated from SHSY5Y cells (a human neuroblastoma-derived cell line) using the CRIPR/Cas9 system. Plasmid pX330 for expression of human-codonoptimized Streptococcus pyogenes (Sp) Cas9 and chimeric guide RNA were obtained from Addgene. The seed sequence for the SpCas9 target site in the target gene was tcggtgcagaccacctcccgcgg (underline; PAM sequence). SHSY5Y cells were transfected with pX330 containing the gRNA of the target site using Lipofectamine 2000 (Invitrogen). KO clones were obtained by limiting dilution, and KO was confirmed by sequencing the target sites in the genomic DNA and by flow cytometric analysis of CD59 expression [staining with mouse anti-hCD59 antibody (5H8) followed by phycoerythrinconjugated anti-mouse IgG (Biolegend)]. PIGT-KO clone #3 was rescued by transfecting human PIGT-expressing vector pME-puro-3HA-hPIGT. After puromycin selection, the restored population was sorted to obtain PIGT+ cells. PIGT-KO clone #3 was transfected with an empty vector, pME-puro, and the puromycin-resistant population was selected as PIGT− cells. These cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum containing 3 μg/ml puromycin. For flow cytometric analysis, cells were stained with phycoerythrin-conjugated anti-human TNAP (B4-78; isotype, mouse IgG1; Santa Cruz Biotechnology) or anti-human CD73 (AD2, isotype; mouse IgG1; Biolegend) and anti-human CD59 antibody (isotype; mouse IgG1, clone 5H8) followed by phycoerythrin-conjugated goat anti-mouse IgG. Stained cells were analyzed using a MACS QuantVYB analyzer (Miltenyi Biotec).

Measurement of ALP and CD73 activities in SHSY5Y cell lines
ALP activities were measured using a Great EscAPe SEAP kit (Takara Bio Inc) in lysates of PIGT+ and PIGT− cells. The ALP activity is expressed in terms of the amount of placental ALP in the kit, which was used as a positive control. CD73 activities were measured by adenosine formation from AMP in lysate from PIGT+ and PIGT− cells. The quantitative method for adenosine using HPLC was described previously in the section "Measurement of hydrolysis of AMP." The APCP sensitivity of CD73 activity was calculated by subtracting the activity in the presence of 4 μM APCP from the total activity in the absence of APCP.
Extracellular hydrolysis and uptake of vitamin B 2 , and vitamin B 2 -dependent PLP and PL production in GPI-deficient cells PIGT− and PIGT+ cells were cultured in vitamin B 2 -deficient medium for 5 days, followed by culture in medium containing a vitamin B 2 derivative (0.2 μM FMN, FAD, RF, or no vitamin B 2 ) for 24 h. Vitamin B 2 and B 6 concentrations were measured in medium and cells by HPLC. The HPLC conditions for measurement of FAD, FMN, and RF were the same as described previously for measurement of FMN hydrolysis. For the measurement of PL and PLP, a previously reported HPLC method with a fluorescence detector was used after precolumn derivatization with semicarbazide (40).

Measurement of mitochondrial function
Cellular respiration (oxygen consumption rate [OCR]) was assessed using an XFp Extracellular Flux Analyzer (Seahorse Bioscience). Cells were cultured in vitamin B 2 -depleted medium for 4 days. Then, 10 4 cells per well were incubated in poly-Llysine-coated wells with vitamin B 2 -depleted medium (Sigma-Aldrich Inc) or with that supplemented by RF, FAD, or FMN (0.2 μM) for 24 h. The XF Cell Mito Stress Test (Seahorse Bioscience Inc) was used to measure the key parameters of mitochondrial respiration using specific mitochondrial inhibitors and uncouplers: oligomycin (1 μM), carbonyl cyanide ptrifluoromethoxyphenylhydrazone (2 μM), and a mixture of rotenone/antimycin A (both 0.5 μM) were injected sequentially following the manufacturer's instructions. Before drug addition, basal OCR was measured. Oligomycin was injected to inhibit ATP synthase (complex V), and the OCR was recorded. To determine the maximal respiration, the uncoupler carbonyl cyanide p-trifluoromethoxyphenylhydrazone was injected. Finally, a mixture of rotenone/antimycin A was injected to inhibit the flux of electrons through complexes I and III and to enable calculation of the spare respiratory capacity. Residual O 2 consumption shows mitochondria-independent O 2 consumption.

Statistical analysis
Data are expressed as means ± SD. Statistically significant differences were determined using one-way analysis of variance followed by Tukey post hoc test or Student t test, with p < 0.05 or 0.01 as the criterion. Pearson correlation analysis was performed to analyze correlations. Kinetic analyses were performed using Sigma Plot (Systat Software Inc).

Data availability
All data are contained within the manuscript.