Preferential use of alkyl-acyl phosphatidylinositol for GPI biosynthesis and diagnostic potential of lipidomics for inherited GPI deficiencies

TL;DR

This study shows that specific lipid structures are preferentially used in making GPI anchors and can help diagnose inherited GPI deficiencies.

Contribution

The study reveals the preferential use of alkyl-acyl phosphatidylinositol in GPI biosynthesis and its diagnostic potential for inherited GPI deficiencies.

Findings

1-alkyl-2-acyl PIs are preferentially used in GPI biosynthesis despite being minor cellular components.

Disruption of GNPAT or AGPS leads to GPI-APs with only diacylglycerol.

Lipidomic profiling can diagnose inherited GPI deficiencies by identifying specific intermediates.

Abstract

Glycosylphosphatidylinositol-anchored proteins (GPI-APs) are attached to the cell surface via a glycolipid anchor, GPI, whose conserved core is synthesized from phosphatidylinositol (PI) in the endoplasmic reticulum through a series of enzymatic reactions. Most PI species in mammalian cells contain diacylglycerol, whereas GPI-APs predominantly possess 1-alkyl-2-acylglycerol. The basis for this characteristic lipid structure has remained unclear. Lipidomic analysis revealed that 1-alkyl-2-acyl PIs, although minor components of cellular PI, are preferentially used by GPI-N-acetylglucosaminyltransferase, which catalyzes the first step of GPI biosynthesis. GPI intermediates containing 1-alkyl-2-acylglycerol were further enriched in subsequent biosynthetic steps, resulting in mature GPIs primarily harboring this lipid species. We demonstrate that a 1-alkyl-containing precursor lipid derived from peroxisomes, likely 1-alkyl-glyceronephosphate, contributes to the formation of 1-alkyl-2-acyl PIs. Disruption of glyceronephosphate O-acyltransferase (GNPAT) or alkylglycerone phosphate synthase (AGPS), the first two enzymes of the peroxisomal ether-lipid pathway, abolished 1-alkyl-2-acyl PI, yielding GPI-APs containing only diacylglycerol. Lipidomic profiling of GPI biosynthetic intermediates in GPI-defective cells revealed accumulation of defective-step-specific intermediates, enabling the use of this approach for diagnosing inherited GPI deficiency (IGD).

Figures

Full text

Glycosylphosphatidylinositol (GPI) serves as a membrane anchor for a class of eukaryotic cell surface proteins through post-translational modification at their C-termini. More than 150 human proteins are known to be GPI-anchored (1). The mammalian GPI anchor consists of a conserved glycan core, variable glycan side chains, and a phosphatidylinositol (PI) lipid moiety with the backbone EtNP-Man-Man-(EtNP)Man-GlcN-PI, where EtNP, Man, and GlcN represent ethanolamine phosphate, mannose, and glucosamine, respectively (1) (Fig. 1A). GPI is assembled in the endoplasmic reticulum (ER) through sequential addition of monosaccharides, and EtNPs to PI, transferred en bloc to the precursor protein, and further matured by side-chain modification and lipid remodeling (Fig. 1B).Figure 1**Structure of GPI-AP and its biosynthesis.**A, structure of GPI-AP with side chain. B, biosynthesis and modification of GPI-AP; C, structure of a-alkyl-2-acyl-PI and diasyl-PI.

Mammalian cells mainly contain diacyl PI, whereas 1-alkyl-2-acyl PI is a minor species (2) (Fig. 1C). In contrast, the PI moiety in mammalian GPI-APs is predominantly 1-alkyl-2-acyl form (3, 4, 5) while the diacyl form is used in particular GPI-APs such as human spleen CD52 (6). While diacyl PI biosynthesis is well understood (7), the origin of 1-alkyl-2-acyl PI remains to be determined. Diacyl PI is synthesized from glycerol-3-phosphate (G3P) via lysophosphatidic acid (lysoPA) and diacyl phosphatidic acid (PA), which is converted with CTP to CDP-diacylglycerol (CDP-DAG) by CDP-DAG synthase (CDS1/2). PI synthase (PIS) then condenses inositol with CDP-DAG to yield diacyl PI (8). If 1-alkyl-2-acyl PI is synthesized similarly, peroxisome-derived 1-alkyl-2-acyl PA must replace diacyl PA. Ether lipid biosynthesis begins in peroxisomes, where glyceronephosphate O-acyltransferase (GNPAT) acylates dihydroxyacetone phosphate (DHAP), and alkylglycerone phosphate synthase (AGPS) substitutes the acyl chain with a long alkyl group (9). The resulting 1-alkyl-DHAP is reduced to 1-alkyl-G3P by acyl/alkyl dihydroxyacetone phosphate reductase (ADHAPR), which is localized on both peroxisomes and ER, and subsequently acylated at the sn-2 position to form 1-alkyl-2-acyl PA (10) (Fig. S1).

Although the 1-alkyl-2-acylglycerol moiety in GPI is known to depend on peroxisomal function (11), the source of the alkyl chain remains unclear. Earlier studies with CHO cell mutants defective in GPI biosynthesis showed that GlcNAc-PI and GlcN-PI intermediates contained mostly diacyl PI, whereas the third intermediate, GlcN-(acyl)PI, contained alkyl-acyl PI as a major structure (12). These findings suggested that GPI biosynthesis begins from diacyl PI, followed by lipid remodeling at the GlcN-(acyl)PI stage (11). With recent technological advances in lipid mass-spectrometry allowing identification of minor lipid species such as 1-alkyl-2-acyl PI, we re-evaluated PI and GPI intermediates using HEK293 knockout cells defective in GPI biosynthesis. Here, we show that 1-alkyl-2-acyl PIs are preferentially utilized for GPI biosynthesis and that previously assumed diacyl to 1-alkyl-2-acyl lipid remodeling of GlcN-(acyl)PI most likely does not occur. Lipidomic profiling also revealed that each GPI biosynthesis gene knockout cell accumulates the immediately preceding intermediate, prompting development of a lipidomics-based diagnostic approach for inherited GPI deficiencies (IGDs).

Results

GPI intermediates with alkyl-acyl PI were enriched during the GPI biosynthesis

Lipidomic analysis confirmed that alkyl-acyl PIs of various chain combinations exist as minor PI species in HEK293 and CHO cells. In wild-type HEK293 cells, <5% of PIs were alkyl-acyl form, with the remainder being diacyl PIs (Fig. 2A, lower panel, Table S1A, sheet PI). To determine the proportion of alkyl-acyl forms in GPI intermediates, we analyzed HEK293 knockout (KO) cells for individual GPI biosynthetic genes. Each mutant accumulated the GPI intermediates generated in the step immediately preceding the defective step, allowing analysis of lipid composition by LC–MS/MS. For example, cells defective in PIGL, GlcNAc-PI deacetylase, accumulated GlcNAc-PI. In two clones of PIGL-KO cells, 15.6 to 32% of GlcNAc-PIs contained alkyl-acyl PI (Figs. 2A and S2). Alkyl-acyl containing fraction increased to ∼70% in GlcN-PIs (PIGW-KO) and >90% in GlcN-(acyl)PIs (PIGM-KO). Nearly all late intermediates contained alkyl-acyl PI (Fig. 2A, upper panel, Table S1A, sheet GPI). Thus, alkyl-acyl PI species existed in the first intermediate GlcNAc-PI and progressively increased to become predominant in later intermediates of GPI biosynthesis.Figure 2**1-alkyl-2-acyl PI is preferentially used for GPI biosynthesis.**A, upper: percentage of alkyl-acyl containing products in total GPI intermediates from various mutants of HEK293 cells. Lower: percentage of alkyl-acyl PI species in total cellular PI from wild-type and various mutants of HEK293 cells. B, upper: percentage of alkyl-acyl containing products in total GPI intermediates from various mutants of CHO cells. Lower: Percentage of alkyl-acyl species in PI from wild-type and various mutants of CHO cells. C, upper: percentage of alkyl-acyl containing products in total GPI intermediates from various mutants of U-251 MG cells. Lower: Percentage of alkyl-acyl species in PI from wild-type and various mutants of U-251 MG cells. A–C, representative data from independent experiments. Blue and gray areas in each bar indicate fractions of the alkyl-acyl form and the diacyl form, respectively. Actual amounts of species of GPI intermediates and PI are listed in Table S1, A–C.

To reassess early GPI intermediates in CHO cells, we analyzed PIGL-, PIGW-, and DPM2-deficient mutants. As in HEK293, wild-type CHO cells contained ∼5% alkyl-acyl PIs (Fig. 2B, lower panel, Table S1B, sheet PI). The proportion of alkyl-acyl forms in GlcNAc-PIs (PIGL-mutant), GlcN-PIs (PIGW-mutant), and GlcN-(acyl)PIs (DPM2-mutant) were ∼40%, 14%, and ∼40%, respectively (Fig. 2B, upper panel, Table S1B, sheet GPI). Although the reason for this fluctuation remains unclear, these data indicate that alkyl-acyl forms were also present in the early GPI intermediates in CHO cells, contrary to a previous report (12).

We further examined U-251 MG cells, derived from human glioblastoma, to determine whether the presence of alkyl-acyl form in early GPI intermediates is common among various cell types. In wild-type cells, only 0.3% of PIs were alkyl-acyl (Fig. 2C, lower panel, Table S1C, sheet PI). In PIGL-, PIGV-, and PIGF-KO cells, alkyl-acyl forms accounted for 9.5%, 31.7%, and 39.7% of GlcNAc-PI, EtNP-Man-GlcN-(acyl)PI, and Man-Man-(EtNP-)Man-GlcN-(acyl)PI, respectively (Fig. 2C, upper panel, Table S1C, sheet GPI). Among the GPI-anchored proteins (GPI-APs) expressed on the U-251 MG cells, CD59 and CD73 contained 69% and 83% of alkyl-acyl PIs, respectively (Fig. 3). Collectively, these results demonstrate that [i] mammalian cultured cells contain alkyl-acyl PI as a very minor PI species, [ii] these species are preferentially utilized in the initial step of GPI biosynthesis, [iii] GPI intermediates become increasingly enriched in alkyl-acyl PI along the pathway, and [iv] the degree of enrichment varies among cell lines—nearly 100% in HEK293 and 70 to 80% in U-251 MG cells. Consistent with previous findings, ∼90% of GPI-APs in CHO cells contained alkyl-acyl PI (11).Figure 3Alkyl-acyl PIs are major lipid species of GPI-anchored proteins in U-251MG cells. Lipidomic analysis of PI from the purified CD59 and CD73 proteins expressed on U-251 MG wild -type cells. Representative data from two independent experiments.

1-Alkyl-2-acyl-PIs are mostly used for GPI biosynthesis

In PIGA-KO HEK293 cells, which cannot generate GlcNAc-PI, the proportion of alkyl-acyl PIs was comparable to wild-type (3.5% vs. 4.8%). In contrast, it dropped to 0.3% in PIGL-KO cells, indicating that nearly all alkyl-acyl PIs were consumed for GlcNAc-PI formation (Fig. 2A, lower panel; Table S1A, sheet PI). The low proportions of alkyl-acyl PIs in PIGW-, PIGM-, PIGV-, and PIGN-KO cells (0.7–1.4%) further support their preferential use for GPI biosynthesis. As intermediates approached the mature GPI form (PIGB-, PIGO-, PIGF-, and PIGT-KO cells), alkyl-acyl PI levels recovered to near wild-type (2.7–4.2%) (Fig. 2A, lower panel; Table S1A, sheet PI), likely suggesting feedback inhibition of early GPI biosynthesis by accumulated late intermediates.

Similarly, in CHO cells, alkyl-acyl PI levels in PIGQ- and PIGH-deficient cells incapable of generating GlcNAc-PI were comparable to wild-type (4.2% and 6.6% vs. 5.3%), but dropped sharply (0.2–0.7%) once GPI synthesis began in PIGL-, PIGW-, and DPM2-deficient cells (Fig. 2B, lower panel;, Table S1B sheet PI). U-251 MG cells also exhibited a similar trend (Fig. 2C, lower panel; Table S1C, sheet PI). Together, these results demonstrate that 1-alkyl-2-acyl PIs serve as the preferred substrates for GPI biosynthesis in HEK293, CHO, and U-251 MG cells.

Chemically synthesized diacyl GlcNAc-PI was not remodeled to 1-alkyl-2-acyl GlcNAc-PI in the ER

To test whether lipid remodeling from diacyl to 1-alkyl-2-acyl GlcN-(acyl)PI occurs as previously postulated (12), we used chemically synthesized diacyl GlcNAc-PI as an exogenous substrate for GPI biosynthesis. We have shown that GlcNAc-PI containing 1-stearoyl-2-oleoyl PI [GlcNAc-PI (18:0/18:1)] can be incorporated into PIGA-KO cells, which lack endogenous GlcNAc-PI, and used for GPI synthesis (13). To accumulate early GPI intermediates for lipid analysis, we disrupted the PIGA gene in PIGW-, PIGM-, and PIGV-KO HEK293 cells, and incubated them with diacyl GlcNAc-PI (18:0/18:1). The fatty chain compositions of GlcN-PI, GlcN-(acyl)PI, and (EtNP)Man-GlcN-(acyl)PI formed from this substrate were determined by LC–MS/MS (Fig. 4). In these synthetic GlcNAc-PI–treated double-knockout (DKO) cells, most of the added GlcNAc-PI remained unchanged, but a significant portion was converted into downstream intermediates (Table S2). In PIGA/PIGW-DKO cells, the predominant product was unremodeled GlcN-PI (18:0/18:1), along with GlcN-PI (18:0/18:0) and a small amount of lyso-GlcN-PI (18:0), suggesting partial fatty acid remodeling at the sn-2 position from 18:1 to 18:0 via a lyso intermediate (Fig. 4A). In PIGA/PIGM-DKO cells, GlcN-(acyl)PIs with diverse acyl chains appeared, but none contained an alkyl chain, again indicating sn-2 remodeling of the fatty acid and some variation in inositol-linked acyl chains (Fig. 4B). In PIGA/PIGV-DKO cells, only two intermediates were detected—(EtNP)Man-GlcN-(acyl)PI (18:0/18:1/g16:0) and (EtNP)Man-GlcN-(acyl)PI (18:0/18:0/g16:0) (Fig. 4C). Despite supportive evidence of acyl chain exchange, no alkyl chain was detected in any GlcN-(acyl)PI or (EtNP)Man-GlcN-(acyl)PI species (Table S2). These results indicate that the 1-alkyl-2-acyl structure of GPI lipids is not generated by remodeling of a diacyl precursor within the ER.Figure 4**Synthetic diacyl GlcNAc-PI was not remodeled to 1-alkyl-2-acyl GlcNAc-PI during GPI biosynthesis.**A, chain compositions of GlcN-PI and GlcN-lysoPI in PIGA/PIGW-DKO HEK293 cells after incubation with synthetic GlcNAc-PI (18:0/18:1). B, chain compositions of GlcN-(acyl)PI in PIGA/PIGM-DKO HEK293 cells after incubation with synthetic GlcNAc-PI (18:0/18:1). C, Chain compositions of Man(EtNP)-GlcN-(acyl)PI in PIGA/PIGV-DKO HEK293 cells after incubation with synthetic GlcNAc-PI (18:0/18:1). A–C, representative data from two independent experiments.

Alkyl lipid moiety in PIs and GPIs was derived from peroxisome

Ether phospholipid biosynthesis begins in the peroxisome lumen and continues on the cytoplasmic face of the ER. We examined whether this pathway also produces 1-alkyl-2-acyl PI. DHAP acyltransferase (GNPAT) catalyzes the first step, generating 1-acyl-DHAP. Comparison of GlcNAc-PIs in PIGL-KO and GNPAT/PIGL-DKO HEK293 cells revealed that loss of GNPAT nearly abolished alkyl-acyl GlcNAc-PIs (Fig. 5A, left, Table S3A), indicating that their formation depends on peroxisome-derived 1-alkyl lipids and thus has a 1-alkyl-2-acyl structure.Figure 5**Peroxisome-dependency of alkyl-acyl-PI and GPI intermediates with alkyl-acyl PI.**A, percentage of alkyl-acyl GlcNAc-PI in GNPAT/PIGL-DKO and AGPS/PIGL-DKO compared with PIGL-KO HEK293 cells. B, percentage of alkyl-acyl GlcNAc-PI in AGPS/PIGL-DKO cells compared with PIGL-KO U-251 MG cells. C, left, percentage of alkyl-acyl GlcNAc-PI in VAP A/B/PIGL-TKO and DHRS7B/PIGL-DKO compared with PIGL-KO HEK293 cells. (two-sided t test, p-value of ∗∗∗: 0.003). Right, Percentage of plasmalogen-type phosphatidylethanolamine (PE) in VAP A/B/PIGL-TKO and DHRS7B/PIGL-DKO compared with PIGL-KO HEK293 cells. (two-sided t test, p-value of ∗∗∗: 0.002, ∗: 0.03). D, percentage of alkyl-acyl GlcNAc-PI in VAP A/B/PIGL-TKO cells rescued by VAPA or VAPB expression plasmid and in ACBD4 or ACBD5 or both knockdown HEK293 cells. Some parts of each experiment were repeated at least twice to confirm reproducibility.

Subsequently, AGPS replaces the acyl chain with an alkyl chain to produce 1-alkyl-DHAP. As expected, AGPS-KO HEK293 cells as well as AGPS-KO U-251 MG cells also lacked alkyl-acyl GlcNAc-PIs (Fig. 5, A and B, Table S3, A and B).

We next examined the role of DHRS7B, an acyl/alkyl-DHAP reductase that generates 1-alkyl-glycerol phosphate from 1-alkyl-DHAP (10). In DHRS7B/PIGL-DKO cells, plasmalogen-type phosphatidylethanolamine (PE) was significantly decreased as reported previously (10) (Fig. 5C right), however, 1-alkyl-2-acyl GlcNAc-PIs was not significantly affected (Figs. 5C left, S3A), indicating a minor role of DHRS7B, if any, and the existence of other reductase(s) contributing to alkyl-acyl GPI biosynthesis.

Transfer of peroxisome-derived alkyl lipids to the ER is thought to involve the VAP–ACBD5 tethering complex (14). In VAPA/B-DKO cells, the proportion of 1-alkyl-2-acyl GlcNAc-PIs decreased by approximately half; however, it was not restored by re-expression of either VAPA or VAPB (Figs. 5, C and D, S3B). In addition, knockdown (KD) of ACBD4, ACBD5, or both in PIGL-KO cells did not significantly alter 1-alkyl-2-acyl GlcNAc-PI levels, despite their expression being decreased (Figs. 5D, S3C and D). The percentage of plasmalogen-type PE was unchanged either (Fig. S4). Thus, the reported VAP-ACBD5-mediated lipid transport does not seem to work for the precursor of 1-alkyl-2-acyl PI, implying the presence of additional, unidentified pathway(s) for alkyl lipid trafficking to the ER. In conclusion, synthesis of alkyl-containing GPIs relies primarily on the peroxisomal ether lipid biosynthetic pathway.

Lipid structure of GPI did not affect the expression of GPI-APs and their remodeling

To assess the physiological significance of the preferential use of alkyl lipids, we examined GPI-AP levels in AGPS-KO cells by flow cytometry. In HEK293 AGPS-KO cell clones, surface levels of CD59, DAF, CD109, MFI (melanotransferrin) and FLAER staining were comparable to those in wild-type cells (Fig. S5A). Similarly, surface levels of CD59, DAF, CD109, CD73 and FLAER staining were unaffected in AGPS-KO U-251 MG cell clones (Fig. S5B), indicating that loss of the alkyl form does not impair the cell surface levels of GPI-APs.

We next tested whether the absence of alkyl lipids influences GPI structure using PI-specific phospholipase C (PI-PLC) sensitivity test. Normally, PGAP1 removes the inositol-linked acyl chain in the ER, rendering GPI-APs sensitive to PI-PLC. In both AGPS-KO HEK293 and U-251 MG cells, CD59 was fully released by PI-PLC treatment, as in wild-type cells, showing that PGAP1 acts efficiently on GPIs containing diacyl PI (Fig. S5C).

Finally, we examined fatty acid remodeling in the Golgi apparatus. Whereas GPI intermediates contained unsaturated fatty acids (e.g., 22:6 or 20:4) at the sn-2 position (Table S1A, sheet GPI), mature GPI-APs with diacyl PI carried saturated 18:0 at sn-2 (Fig. 3), demonstrating that the absence of alkyl chains does not interfere with fatty-acid remodeling during GPI maturation.

Step-specific accumulation of GPI intermediates enables diagnostic lipidomics for IGD

Lipidomic analysis revealed that in HEK293 cells lacking each GPI biosynthetic gene, only the immediately preceding intermediate accumulated (Table S1A, sheet GPI). Specifically, GlcNAc-PI accumulated in PIGL-KO, GlcN-PI in PIGW-KO, GlcN-(acyl)PI in PIGM-KO, (EtNP)Man–GlcN-(acyl)PI in PIGV-KO, Man–Man–GlcN-(acyl)PI in PIGN-KO, Man–(EtNP)Man–GlcN-(acyl)PI in PIGB-KO, and Man–Man–(EtNP)Man–GlcN-(acyl)PI in PIGO-KO and PIGF-KO cells (Table S1A, sheet GPI). Fully assembled GPI, (EtNP)Man–(EtNP)Man–(EtNP)Man–GlcN-(acyl)PI, could not be detected in GPI-transamidase–defective cells, such as PIGT-KO cells, most likely owing to technical limitations of our LC-MS/MS. However, previous work showed that free non-anchored GPIs with GalNAc side chain, presumably having (EtNP)Man–Man–(EtNP)(GalNAc)Man–GlcN-(acyl)PI structure, were detected on the surface of GPI-transamidase–defective cells by the monoclonal antibody T5-4E10, indicating transport of GPI to the cell surface (15). To determine whether GPI intermediates similarly reach the plasma membrane, we used subcellular fractionation prior to LC-MS/MS. Subcellular fractionation of PIGL-KO cells suggested that a significant portion of GlcNAc-PI is localized on the cell surface (Fig. 6A, Table S4A, Fig. S6A). PI-PLC treatment of cells before fractionation markedly reduced GlcNAc-PI levels, confirming its surface localization (Fig. 6B, Table S6B, Fig. S6B). Since GlcNAc-PI is synthesized on the cytosolic face of the ER, these data indicate the presence of a transport mechanism that carries intermediates to the plasma membrane. Consistently, when exogenous synthetic GlcNAc-PI was added to PIGA/PIGV-DKO HEK293 cells, part of it was utilized as substrates to produce downstream intermediates (GlcN-PI, GlcN-(acyl)PI, and Man–(EtNP)–GlcN-(acyl)PI). After overnight incubation with the synthetic GlcNAc-PI, PI-PLC treatment at 10 °C or 37 °C for 1.5 h progressively decreased both GlcNAc-PI and GlcN-PI levels, showing that these intermediates are largely surface-localized, and even ER localized intermediates can move to the cell surface during incubation at physiological temperature (Fig. 6C, Table S4C). As PI-PLC cannot cleave Man–(EtNP)–GlcN-(acyl)PI, its localization could not be determined.Figure 6**The first and the second intermediates, GlcNAcPI and GlcNPI are transported to the cell surface.**A, amount of total GlcNAc-PI in the plasma membrane and ER fractions of PIGL-KO HEK293 cells, fractionated by sucrose density gradient centrifugation. B, amount of total GlcNAc-PI of the purified plasma membrane fraction using the plasma membrane protein extraction kit with or without PIPLC treatment. C, amount of GPI intermediates in the PIGV/PIGA-DKO cells incubated with synthetic GlcNAcPI (18:0/18:1) for overnight and further incubated with PIPLC at 10 °C or 37 °C. Cps, count per second. A and B, representative data from two independent experiments.

The strict step-specific accumulation of GPI intermediates suggested a diagnostic application. Lipidomic analysis of serum or leukocytes from patients with PIGL-, PIGV-, PIGB-, or PIGO-IGD (Table S5A) and PIGW-, PIGV- or PIGO-IGD (Table S5B) revealed characteristic accumulation of the corresponding intermediates (Fig. 7). In PIGW-IGD, GPI intermediates were detected only in blood cells. In other types, it was detectable in blood cells, serum, and plasma, with higher levels observed in serum and plasma. These intermediates had lost the PI moiety, leaving only the acyl group on the inositol residue (Fig. 7, Table S5, A and B). This provides a biochemical signature for identifying both IGD and its defective step. Such lipidomic profiling can potentially enable neonatal diagnosis using dried blood spots, an essential step toward implementing early gene therapy for IGD (16, 17).Figure 7Lipidomic analysis of IGD patient's blood showed the specific biomarkers for each defective step. Lipidomic analysis of blood cells and serum from the PIGL-IGD, PIGW-IGD, PIGV-IGD, PIGB-IGD, and PIGO-IGD patients showed the specific GPI intermediate for each step.

Discussion

Here, the result of lipidomic analysis provided direct evidence to verify that the 1-alkyl-2-acyl PI is preferentially used for GPI synthesis. Although the complete de novo pathway for 1-alkyl-2-acyl PI synthesis still needs to be clarified, we have a reasonable model of alkyl-acyl-PI synthesis pathway based on the generation pathways of other ether lipids and diacyl PIs (Fig. S1). In the third step, 1-alkyl DHAP is converted to 1-alkyl G3P by an acyl/alkyl DHAP reductase. It was reported that DHRS7B, which resides on the cytoplasmic sides of both the peroxisome membrane and the ER membrane, plays an essential role in the third step in ethanolamine plasmalogen synthesis (10). Because generation of alkyl-acyl GlcNAc-PI was not significantly affected by DHRS7B-KO, it seems that other isozyme(s) of 1-alkyl-DHAP reductase plays a role for alkyl-acyl-PI. How alkyl-DHAP, which was generated inside the peroxisome, is translocated on the cytosolic side of the peroxisome membrane is also unclear. Because acylation of 1-alkyl-G3P completes in the ER, 1-alkyl-DHAP or 1-alkyl-G3P must be delivered to the ER. It is reported that the VAP-ACBD5 tether plays a role in lipid transport between peroxisome and the ER (14). In VAPA/B and PIGL triple KO cell, 1-alkyl-2-acyl GlcNAc-PIs was decreased, but this could not be rescued by expression of VAPA or VAPB. Furthermore, knockdown of ACBD4, ACBD5 and both in PIGLKO cells showed no significant decrease in 1-alkyl-2-acyl GlcNAc-PIs and plasmalogens (Figs. 5D, S4). Why we could not reproduce the reported data is unknown but at least we can conclude that other transport pathways of alkyl donors is existing and working together with VAP-ACBD5.

Why does GPI favor the scarce 1-alkyl-2-acyl PI instead of the abundant diacyl PI in the cell? According to our previous report, GPI-GlcNAc transferase (GPI-GnT), the first step enzyme complex, requires ARV1 to use a wider repertoire of PI species both of diacyl and alkyl-acyl forms and to enhance the enzyme activity (18). Without ARV1, production of GlcNAc-PIs was greatly decreased but most of them were 1-alkyl-2-acyl form, suggesting that GPI-GnT preferentially uses 1-alkyl-2-acyl PI (18). In HEK293 cells, alkyl-acyl containing GPI intermediates increased to ∼70% in GlcN-PIs (PIGW-KO) and 100% in Man-Man-GlcN-(acyl)PIs (PIGN-KO). In CHO cells, they increased to 40% in GlcN-(acyl)PIs (DPM2-mutant) and ∼90% in GPI-APs. In U-251 MG cells, they increased to ∼40% in Man-Man-(EtNP-)Man-GlcN-(acyl)PI (PIGF-KO), and 70 to 80% in GPI-APs. These results suggested that alkyl-acyl PI species existed in the first intermediate GlcNAc-PI and progressively increased to become predominant in later intermediates of GPI biosynthesis and that not only the early step enzymes but also the late step enzymes prefer to use alkyl-acyl containing GPI intermediates. Thus, we detected a significant increase in the proportion of the alkyl form during the biosynthesis. So, a reason for alkyl-containing GPIs enrichment was that GPI intermediates with alkyl form are better substrates of the GPI biosynthesis enzymes. Then, what happens to the subsrantial population of diacyl-PI-containing species? As shown in Figure 6, when exogenous GlcNAc-PI was added to PIGA-deficient cells, part of it reached the ER, where it was converted by PIGL to GlcN-PI and further modified to GlcN(acyl)-PI by PIGW. GlcN-PI molecules that failed to proceed to the next step were transported to the cell surface and became PI-PLC–sensitive. Incubation at 37 °C also promoted the movement of intracellular GlcNAc-PI and GlcN-PI to the cell surface, where they were cleaved by PI-PLC. These findings suggest that diacyl-type GPI intermediates, which are abundantly generated in the early steps, are transported to the plasma membrane and subsequently degraded in endosomes. Diacyl-type intermediates produced in the ER lumen are likely transported to the plasma membrane when not utilized further. What is the physiological meaning of preferential usage of alkyl-containing GPIs? We don't have a clear answer for it. GPI-APs on AGPS-KO cells were expressed at normal levels, PI-PLC sensitive and normally underwent fatty acid remodeling similar to wild-type cells. If we have to give a possibility, as alkyl lipids are resistant to phospholipase A, alkyl lipids containing GPI-APs might be more resistant to releasing from the cell surface. There should be a certain physiological meaning for the alkyl-containing GPIs.

In PIGL-KO and PIGW-KO cells, levels of alkyl-acyl-PI greatly dropped compared with wild-type cells (Fig. 2 and PI panels in Table S1, A–C). We speculate that increased consumption by GPI-GnT rather than decreased generation in PIGL-KO and PIGW-KO cells caused depletion of alkyl-acyl-PI. Levels of total PI were rather increased in PIGL-KO and PIGW-KO cells as shown in Table S1, A–C (PI panels). In addition, we previously showed that when GPI biosynthesis was impaired, and precursor proteins did not get GPI, GPI attachment signal peptides of particular GPI-APs, such as DAF and CD48, acted to stimulate GPI biosynthesis by upregulating GPI-GnT (19). This is consistent with the finding that the absolute amount of accumulated GPI intermediates is markedly increased in PIGL-KO cells across all cell types in Table S1, A–C (GPI panels). As GPI-GnT prefers alkyl-acyl-PI, these conditions would lead to depletion of alkyl-acyl-PI.

During lipidomic analysis, we noticed that only the immediately preceding intermediate accumulates in HEK293-KO cells of GPI biosynthesis genes, which was also the case in blood cells from patients with IGD. We are planning to establish the mass screening system using neonate dried blood spots with lipidomic analysis. Furthermore, accumulated GlcNAcPI at the cytoplasmic side of the ER in PIGL-KO cells were transported to the cell surface. Previously, we reported that accumulated free GPI in the cells defective in GPI transamidase (GPIT) is expressed on the cell surface (15, 20). GPIT is the enzyme complex consisting of PIGK, PIGS, PIGT, PIGU, and GPAA1, which cleaves the C-terminal GPI attachment signal of precursor proteins and attaches them to a GPI anchor. These accumulated free GPI or GPI intermediates induced the massive complement activation in case of atypical paroxysmal nocturnal hemoglobinuria caused by PIGT or PIGB deficiencies (20, 21). We speculate that other GPI intermediates may also be transported to the cell surface. Where are their destination? It is known that even wild-type cells and blood cells express free GPI (22). Exogenously added synthetic GlcNAc-PI and GlcN-PI were transported from plasma membrane to the ER and used as the substrates for GPI synthesis. There are dynamic movement of free GPI or GPI intermediates destined to recycle, degradation or maybe being a reservoir for GPI production.

Experimental procedures

Reagents

Polyethylenimine “MAX” (PEI-MAX; Polysciences), and Lipofectamine 2000 (Life Technologies) for transfection and phosphatidylinositol-specific phospholipase C(PI-PLC)from Bacillus cereus (Life Technologies) were purchased. The antibodies used for flow cytometry were anti-MF12 (clone SHM72), -CD109 (W7C5), -CD73 antibodies (TY/11.8), (Biolegend), FLAER (CEDARLANE), anti-CD59(5H8) and -CD55 antibodies (IA10). The antibodies used for Western blotting were rabbit polyclonal antibodies: anti-GNPAT (ab75060 Abcam), anti-DHRS7B (15,530 Proteintech), anti-VAPA (HPA009174 Sigma), anti-VAPB (HPA013144 Sigma), and anti-ACBD5 (HPA012145 Sigma) antibodies.

Derivation of mutant cells and generation of knockout cells

The 3B2A cells, stably expressing human CD59 and CD55 in CHO-K1 cells (ATCC), PIGQ-, PIGH-, PIGL-, PIGW-, and DPM2 - deficient CHO cells were previously established in our Lab (23, 24, 25). Also, the PIGA-KO, PIGL-KO, PIGW-KO, PIGM-KO, PIGV-KO, PIGN-KO, PIGB-KO, PIGO-KO, PIGF-KO, and PIGT-KO HEK293 cells were generated before (13, 26, 27, 28). GNPAT-KO, AGPS-KO, VAPA/B-KO, DHRS7B-KO and those in PIGL-KO HEK293 cells (DKO) were generated by CRISPR-Cas9 system. PIGA/PIGW-DKO, PIGA/PIGM-DKO and PIGA/PIGV-DKO HEK293 cells were also generated by CRISPR-Cas9 system. PIGL-, PIGV-, and PIGF-KO, AGPS-KO and AGPS/PIGL-DKO U-251 MG cells were generated by CRISPR-Cas9 system using lentivirus. All single KO cells were made to clones and confirmed by genome sequencing (Fig. S7). As for DKO cells with PIGL- or PIGA-KO, GPI negative sorted cells were used for the assay. HEK293 and CHO cells were cultured in DMEM/Ham's F-12 medium (Nacalai Tesque) with 10% fetal bovine serum (FBS, Sigma). U-251 MG cells (ATCC) were cultured by using high glucose DMEM medium (Nacalai Tesque) containing 10% FBS. All cells were cultured in an incubator at 37 °C in a 5% CO_2_ atmosphere. Sequences of guide RNAs were listed in Table S6.

Generation of knockdown cells

PIGL-KO HEK293 cells were transfected with siRNA for ACBD4 or ACBD5 or both or control siRNA using Lipofectamine RNAiMAX (Thermofisher). At day 3, cells were transfected again and further incubated for 72 h. Cells were analyzed by western blotting, qPCR and lipidomics. As for qPCR, reverse transcription was carried out using a Super Script VILO cDNA synthesis kit (Invitrogen). Real time PCR were performed using SYBR Premix ExTaqTM II (TaKaRa). Relative mRNA level was calculated by normalizing β–actin. Sequences of siRNAs and primers for qPCR were listed in Table S6.

Genomic DNA extraction

Cells (10^6^) were washed with PBS, suspended in 300 μl nuclei lysis solution (Promega) with 0.6 μl RNase, and incubate at 37 ^o^C for 30 min. After cooling at room temperature for 5 min, 100 μl protein precipitation solution was added to the cell lysates, mixed adequately, and kept on ice for 5 min. After centrifugation at 15,000g for 4 min, supernatant was mixed well gently with 300 μl isopropanol by inverting the tube until a visible mass was formed. DNA was pelleted by centrifugation, the supernatant discarded, and then 300 μl of 70% ethanol added to DNA pellets. After removal of supernatants by centrifugation, DNA pellets were air-dried for 15 min. The genomic DNA pellets were solubilized in TE buffer (Nacalai Tesque) by incubation at 37 °C overnight.

Construction of plasmids

A pair of CRISPR short-guide RNA (sgRNA) oligonucleotides of each targeted gene were designed. All the sgRNA and siRNA are shown in Table S6. Each pair of sgRNAs were annealed to form target sgRNA fragments. The pX330-mEGFP plasmid and Lenti-v2-puro-CRISPR plasmid (Addgene) were used to construct sgRNA-Cas9 expression plasmids. The targeting sequences were integrated into the pX330-mEGFP and Lenti-v2-puro CRISPR vectors which were digested with BbsI. For making expression plasmids for VAPA, VAPB, CD73, we amplified the full length of cDNAs form from human brain cDNA library and inserted into Sal1-Not1 site of pME puro 3HA vector. We chose to add epitope tags at the N-termini of VAPA and VAPB based on the previous reports showing functionalities of the tagged proteins (14, 29).

Virus infection

Lentivirus: Lenti-v2-puro-CRISPR plasmids, with pLC1, pLC2, and pVSVG (Thermo Fisher Scientific), were transfected to Lenti X cells (Takara Bio) for virus generation. Transfected Lenti X cells were cultured at 37 °C overnight. Then the old medium was change to a fresh medium for Lenti X cells in the next morning and culture continued. At 24 h later, the virus was collected, mixed with polybrene (8 mg/ml) and added to cells.

Flow cytometry assay

Cells were stained with the primary antibodies in FACS buffer (phosphate-buffered saline (PBS) containing 1% BSA, 0.1%NaN_3_) on ice for 20 min. Then cells were washed twice with FACS buffer. Subsequently, cells were stained by secondary antibodies on ice for 20 min. After twice more FACS buffer washing, cells were analyzed by the MACSQuant analyzer (Miltenyi Biotec).

PI-PLC treatment

Cells were detached form dishes by treatment with detaching buffer (PBS containing 5 mM EDTA and 0.5% BSA) and harvested. Approximately 5 × 10^5^ cells were treated with or without 1 U/ml PI-PLC in 50 μl reaction buffer (4 volumes of Opti-MEM and 1 volume of detaching buffer) at 37 °C for 90 min. For cell fractionation analysis following PI-PLC treatment, cells were washed thoroughly and subsequently subjected to fractionation. PI-PLC is unable to cross the plasma membrane.

GlcNAc-PI treatment

The method for synthesis of diacyl GlcNAc-PI (18:1, 18:0) is shown in our previous report (13). Chemically synthesized diacyl GlcNAc-PI (18:1, 18:0) were incubated with each DKO cell with the concentration of 10 μM in serum-free DMEM medium for overnight. For further treatment with PIPLC, the cells were harvested and treated with or without PI-PLC for 90 min at 37 °C (to allow transport) or 10 °C (to inhibit transport). Then cells were analyzed by lipidomics. PIPLC is fully active either at 37 °C or 10 °C and PI-PLC is unable to cross the plasma membrane.

Cell fractionation

Cells (1 × 10^8^) were suspended in 1.5 ml of a buffer (9.6% sucrose, 20 mM HEPES-NaOH, pH 7.4, protease inhibitor) and destroyed by 10 strokes of a tight-fitting Dounce pestle. The cell lysate was centrifuged at 10,000 rpm at 4 °C for 10 min and the supernatant was placed on top of a continuous sucrose gradient (20–50%). After centrifugation at 35,000 rpm (SW41 rotor) at 4 °C for 19 h, 1-ml fractions were collected from the top and analyzed by SDS-PAGE and Western blotting. The primary antibodies used were;anti-human Na/K ATPase (PM marker, Cell Signaling #3010 rabbit polyclonal antibody), CD81(PM marker, Biolegend #349502 mouse monoclonal antibody), GM130 (Golgi marker, Abcam #ab52649, rabbit polyclonal antibody), Calnexin (ER marker, ENZO, #ADISPA865D, rabbit polyclonal antibody) whereas the secondary antibodies were horseradish peroxidase (HRP)-conjugated anti-mouse IgG, HRP-conjugated anti-rabbit IgG (all from Amersham Bioscience). PM and ER fractions were analyzed by lipidomics.

We isolated the pure plasma membrane fraction by phase separation using the plasma membrane protein extraction kit (ab65400 Abcam) and followed the company's instruction. Cells were pelleted and homogenized with an ice-cold Dounce apparatus. The resulting whole cell lysate was centrifuged at 700g for 10 min at 4 °C to pellet nuclei and cell debris, and the supernatant was centrifuged at 10,000g for 30 min at 4 °C to yield total cellular membranes (pellet). The total membrane pellet was subjected to phase separation twice with centrifugation at 1000g for 5 min at 4 °C, and the plasma membrane fraction (upper phase1,2) and other membrane fraction (intermediate phase1,2 and lower phase) were separated.

Lipid extraction from GPI-APs

10^8^ of wild type U-251 MG cells, permanently transfected with HA-CD59 or HA-CD73 were lysed by 60 mM of Octyl βglucoside and affinity purified HA-tagged proteins using anti-HA beads (HA-7 Sigma) and applied to SDS-PAGE and transferred to PVDF membrane. Proteins were visualized by staining with Ponceau solution. The protein bands were cut out with a new razor blade and transferred to Eppendorf tube. PVDF membrane strips were washed with 1 ml of methanol four times and with 1 ml of water. Membranes were incubated with a filtered mixture of 500 μl of 0.3 M NaOAc pH 4.0 buffer and 500 μl of freshly dissolved 1 M NaNO_2_ for 3 h at 37 °C for the nitrous acid deamination reaction (30). After removing the solution and washing the PVDF membrane strips with 1 ml of water, they were dried up and sent to lipidomic analysis.

Patient samples

Peripheral whole blood was obtained from the IGD patients. Serum and leukocytes were separated and kept frozen. This study was approved by the Institutional Review Board of Osaka University. Informed consent was obtained from all examined persons or their guardians. This study abided by the principles of the Declaration of Helsinki.

Lipidomic analysis

Cell pellets (2 × 10^6^ cells) were sent to Kazusa DNA Institute for analysis of GPI intermediates by mass spectrometry using an unbiased lipidomic method. Untargeted lipidomics was performed as reported previously with some modifications (31, 32) Briefly, the frozen cell pellets were re-dissolved in 200 μl of methanol (Wako Chemicals) containing EquiSPLASH (Avanti Polar Lipids) as an internal standard. After sonication for 30 s and vortexing for 2 s, 160 μl of the suspension was collected and 80 μl of chloroform (Wako Chemicals) was added, and then vigorously agitated at 750 rpm for 20 min at 20 °C. Subsequently, 16 μl of pure water (Wako Chemicals) was added and vigorously agitated at 750 rpm for 20 min at 20 °C. The supernatant was collected by centrifugation at 1670g for 10 min at 20 °C and transferred to an LC vial.

LC-MS/MS analysis was carried out using quadrupole time-of-flight (Q-TOF)/MS (TripleTOF 6600; SCIEX) coupled with an ACQUITY UPLC system (Waters, MA), with some modifications (33). The LC separation was performed with gradient elution of mobile phase A [methanol/acetonitrile/water (1:1:3, v/v/v) containing 5 mM ammonium acetate (Wako Chemicals) and 10 nM EDTA (Dojindo)] and mobile phase B [isopropanol (Wako Chemicals) containing 5 mM ammonium acetate and 10 nM EDTA]. The flow rate was 300 μl/min at 45 °C using an L-column3 C18 (50 × 2.0 mm i.d., particle size 2.0 μm; Chemicals Evaluation and Research Institute).

The gradient program was as follows: 0 min, 0% (B); 6.5 min, 64% (B); 13.5 min, 76.5% (B); 18 min, 98% (B); 20 min, 98% (B); 20.1 min, 0% (B); and 25 min, 0% (B). MS analysis was performed in high-resolution mode for MS1 (∼37,385 full width at half maximum, FWHM) and in high-sensitivity mode for MS2 (∼29,474 FWHM) using data-dependent acquisition (DDA). The MS parameters were as follows: MS1 scan range, m/z 140 to 1700; MS2 scan range, m/z 75 to 1700; MS1 accumulation time, 250 ms; MS2 accumulation time, 100 ms; collision energy, +45/–42 eV; collision energy spread, 15 eV; cycle time, 1301 ms; curtain gas, 30; ion source gas 1, 40 (+)/42 (−); ion source gas 2, 80 (+)/50 (−); temperature, 250 °C (+)/300 °C (−); ion spray voltage floating, +5.5/–4.5 kV; and declustering potential, 80 V. For quality control of the analytical data, the added internal standards were confirmed to have a coefficient of variation (CV) of less than 15%. Raw data files were converted to MGF format using the Sciex MS Data Converter software and subjected to quantitative analysis with 2DICAL (Mitsui Knowledge Industry, Tokyo, Japan). Molecular species were identified based on retention times and tandem MS spectra obtained under DDA conditions.

The raw data files from Q-TOF/MS were converted to MGF files using the program SCIEX MS converter for quantitative analysis with 2DICAL (Mitsui Knowledge Industry). Identification of the molecular species was accomplished by comparison with retention times and MS/MS spectral data from the information-dependent acquisition (IDA) mode.

Data availability

All data are contained within the manuscript and the supporting information.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.