[3H]myoinositol incorporation into phospholipids in liver microsomes from humans with and without type II diabetes. The lack of synthesis of glycosylphosphatidylinositol, precursor of the insulin mediator inositol phosphate glycan.

A class of inositol phosphate-containing oligosaccharides (IPG) derived from a membrane glycan-phosphatidylinositol precursor (GPI) has been identified as a possible mediator of insulin action. Saltiel's laboratory has recently communicated an in vitro assay for the synthesis of GPI in rat liver microsomes. Herein we have established this method in rat and human liver microsomes, it being our end point to evaluate if the pool of GPI was normal in diabetes and if failure of insulin to generate IPG from GPI could be involved in the mechanism of insulin resistance in Type II diabetes. However, subsequent to the detailed study of [3H]myoinositol incorporation into phospholipids in liver microsomes from our study subjects, we demonstrated by gas chromatography/mass spectrometry analysis that the material reported to be GPI is a mixture of lysophospholipids that does not contain hexosamine, ethanolamine, or amino acids.

A class of inositol phosphate-containing oligosaccharides (IPG) derived from a membrane glycan-phosphatidylinositol precursor (GPI) has been identified as a possible mediator of insulin action. Saltiel The concept of chemical mediators in the mechanism of insulin action was first proposed by Larner's laboratory in 1974 (1). Jarett's laboratory (2), working independently and following different experimental approaches, also concluded that insulin elicits at least some of its biological effects by promoting the generation and/or release of unique intracellular mediators. The glycopeptide nature of the mediator(s) was suggested by Larner et al. (3); however, little progress was made in the elucidation of its structure until the discovery in 1986 by Saltiel and Cuatrecasas (4) that a class of inositol phosphate-containing oligosaccharides (IPG)' mimicked insulin action in the regulation of several enzymes. These data have been confirmed independently by several laboratories (5-9). These results suggest that insulin stimulates a plasma membrane phospholipase C with substrate specificity for a glycosyl-phosphatidylinositol (GPI) resulting in the release of IPG and diacylglycerol.
In fact, an insulin-sensitive GPI- phospholipase C has been recently described by Fox et al. (10) in rat liver. Saltiel's laboratory (11)(12)(13)(14) has recently developed an in vitro assay in rat liver microsomes for the synthesis of the GPI precursor from which IPG is released in response to insulin. Furthermore, using this method, the authors studied the structure of IPG and suggested the presence of 4-5 sugar residues (12, 13). Herein, we have characterized this method in human liver because of our objective to explore if a defect(s) in this insulin messenger system is involved in the mechanism of insulin resistance in liver from patients with Type II diabetes mellitus. This method offers several advantages over previously published methods: (a) using a very small amount of liver, it is possible to test if the pool of GPI is normal in Type II diabetes; and (b) partially purified GPI from human liver can be used as a substrate of GPI-specific phospholipase C. Therefore, it would be possible to examine the effect of insulin on this specific phospholipase C in the membranes prepared from normal human liver and to determine if this enzyme is resistant to insulin in Type II diabetes. the plates were dried, cut into l-cm pieces, and 0.5 ml of ethanol was added to each fraction and counted for radioactivity using Aquasol.
In some experiments the chromatography plates were scanned by a TLC radioisotope scanner (TM Analytic, Inc.). Purification of Peak II Phospholipid-Following reverse phase thin layer chromatography, the radioactivity was located on the chromatogram either as described above by cutting and counting of the reference chromatogram or by a thin layer chromatography scanner. The band corresponding to peak II phospholipid was excised from the plate and the material extracted by chloroform/methanol (2:l). The organic solvent was evaporated under NP, and purified peak II phospholipid was suspended in an appropriate volume of chloroform/ methanol (1:2) and stored at -70 "C until further use. GPI-specific Phospholipase C Assuys-[3H]Myoinositol labeling, extraction, and purification of peak II phospholipid from human liver microsomes were performed as described above. The final preparation was stored in 2:l chloroform/methanol at the concentration of 20,000 dpm/50 ~1. The phospholipase C assays were performed with some modification as described by Fox et al. (10). An appropriate amount of the partially purified [3H]myoinositol-labeled peak II phospholipid (approximately 20,000 dpm/assay) was taken in a glass vial, and organic solvent was evaporated under Nz. Then the dried phospholipid was suspended in 50 mM Tris-Cl, pH 7.5, buffer and sonicated three times (5-min intervals) at the power setting 3. The sonication was for 30 s each time at 4 "C. In some experiments, a 1:l mixture (w/w) of phosphatidylethanolamine/phosphatidylcholine (PE/PC) was added to peak II phospholipid as described by Saltiel et al. (20) prior to sonication.
In this way the substrate was incorporated in PE/PC micelles.
In either case, the final substrate was incubated with human liver plasma membranes in a final volume of 0.2 ml at 37 "C for 1 h. The assay was terminated by the addition of chloroform/methanol/l N HCl (100:50:1) and processed as described above.
The enzyme activity was evaluated by following the decrease in peak II phospholipid radioactivity following reverse phase TLC.
In some initial experiments, termination of the assay with toluene and extraction of diacylglycerol by the method of Fox et al. (10) was used, which did not result in any changes in the data. PI-specific Phospholipuse C Assays-PI-specific phospholipase C (generously supplied by Dr. Martin Low) was appropriately diluted with 50 mM Tris-Cl, pH 7.5, and added to the assay medium. The assays were performed essentially as described for GPI-specific phospholipase C except that peak I as well as peak II phospholipids were used as substrates rather than peak II phospholipid. Termination and extraction of phospholipids were performed as described in GPIspecific phospholipase C assays. VSG Lipuse Assays-These assays were performed as described by Hereld et al. (21). The assay was performed in a 25-~1 volume containing 2 rg (about 6000 cpm) of [3H]myristate-labeled membrane binding form VSG. 0.04% sodium dodecvl sulfate. 1% Nonidet P-40. 5 mM EDTA, and 50 mM Tris-Cl, pH 8.6 The enzyme (VSG lipase) was diluted in 1% Nonidet P-40, 5 mM EDTA, and 50 mM Tris-Cl, pH 8.0. After addition of the enzyme and incubation at 37 "C for 30 min, the mixture was thoroughly mixed with 0.5 ml of H20-saturated n-butyl alcohol. The phases were separated by brief centrifugation, and 0.4 ml of the upper phase was counted for radioactivity using Aquasol.
Phospholipase D Assays-These assays were performed as described by Davitz et al. (22). In a typical experiment 10 ~1 of bovine serum was incubated with 2 +g of 3HIlabeledmembrane binding form VSG ( with-a 0.25-mm inner diameter and a flow rate of 11 n.s.i. The temnerature nrogram consisted of a 70 "C initial 2-min hold, followed by a 10 Y?/rn& ramp to 250 "C. The final temperature was held for 10 min. The mass spectrometer was operated in the electron impact mode with a 70.eV input.

AND DISCUSSION
The right panel of Fig. 1 shows that when rat liver microsomes were incubated with [3H]myoinositol and CDP-diglyceride under the identical conditions described by Saltiel's laboratories, the major [3H]inositol incorporation occurred into a phospholipid that stayed at the origin (peak I) and in a phospholipid with an RF value of 0.4 (peak II). Approximately 90% of the total radioactivity incorporated was localized in peak I, which consists of a mixture of phosphoinositides, e.g. PI, PIP, and PIP2. The second peak (peak II) has been identified by  as GPI, the precursor of IPG. Peak II in Fig. 1 has the same R,s value as that reported by  and also as shown in the left panel of the same figure. Peak II is readily hydrolyzed by PI-specific phospholipase C, one of the criteria used (11-14) to identify peak II as GPI. Fig. 2  One lane was cut mto l-cm pieces and counted for radioactivity following the addition of O.,j ml of ethanol (unset) or was sprayed with fluoroenhancer and exposed to Kodak X-Omat film for 'i days to obtain the autoradmgram detailed characterization of the biosynthesis of peak II phospholipid in normal human liver microsomes.
The middle and upper panels of Fig. 3 show that the incorporation of ["Hlmyoinositol into peaks I and II was linear with respect to time and protein concentration.
The lower panel of Fig. 3  optimal incorporation at pH 7.5. At low pH, there was no incorporation of ["Hlinositol into phospholipids. Fig. 4 shows that ["Hlmyoinositol incorporation into phospholipids is absolutely dependent on the presence of CDPdiglyceride. Also, ["Hlmyoinositol incorporation was dependent on the presence of specific metal ions. Mn"+ and Mg" supported the incorporation, Mn"+ being more potent than Mg" (Fig. 5). In contrast to these metal ions, Ca'+ was inhibitory.
Without adding any metal ions, the basal incorporation of ["Hlinositol was higher in the presence of EDTA/ EGTA than in the absence (data not shown). The reason for the inhibitory effect of Ca"+ on ["Hlmyoinositol incorporation is not known but could be due to Ca'+-dependent stimulation of the breakdown of the phospholipid, or the presence of endogenous Ca')+ in the microsome preparations may be in- Liver biopsies were obtained from the patients and microsomes prepared as described under "Experimental Procedures." Phosphoinositide biosynthesis, extraction, and analysis were performed as described in the legend to Fig. 2 Although these data cannot be extrapolated to in vivo conditions, under the optimal in vitro experimental conditions there are no statistically significant differences in the biosynthesis of peak II phospholipid in human liver microsome preparations from normal patients and obese patients with type II diabetes.
The next group of experiments was directed to attempt to answer our second question. Is GPI-phospholipase C stimulated by insulin in human liver plasma membranes?
If so, is insulin stimulation of GPI-phospholipase C altered in Type II diabetes? For this purpose, we needed to partially purify large quantities of peak II phospholipid from human liver to be used as a substrate and liver plasma membranes as a source of GPI-phospholipase C; in an analogous way we (19) and others (24) have evaluated the effect(s) of hormones on phosphatidylinositol 4,Sbisphosphate (PIP,)-specific phospholipase C using exogenous ["H]PIP2. To this end, stability of the phospholipids during the purification and extraction procedure needs to be demonstrated. Fig. 7 shows that when peak I and II phospholipids were extracted from TLC plates and rechromatographed in an identical way, the purified lipids show identical migration properties. These data indicate the stability of peak I and II phospholipids during the extraction Human liver microsomes were incubated with ["Hlinositol and processed as described under "Experimental Procedures" (upper panel). The region corresponding to peak I and II was scraped from plates, extracted with chloroform/methanol as described under "Experimental Procedures," and equal amounts of the purified phospholipids rechromatographed in an identical way as the first chromatography (middle and lower) panel.
procedure. Central to this study are the data shown in Table  I which demonstrate, under several experimental conditions, the failure of insulin and liver plasma membranes to stimulate hydrolysis of peak II phospholipid.
The membranes used in this study contain a catalytically competent insulin receptor kinase capable of phosphorylating the P-subunit of insulin receptor and three membrane-associated endogenous substrates (25). When peak II phospholipid was used as a substrate and liver plasma membranes as a source of GPI-phospholipase C, no breakdown of peak II phospholipid was observed under the assay conditions, i.e. up to 200 pg of membrane protein and 1 h of incubation.
These experiments were performed by suspending the peak II phospholipid in an appropriate buffer followed by brief sonication. Saltiel et al. (20) reported that substrate works optimally if incorporated into micelles prepared from phosphatidylethanolamine/phosphatidylcholine (PE/PC) mixture. Therefore, we repeated the above set of experiments by incorporating the purified peak II phospholipid into PE/PC micelles. Under these conditions, hydrolysis of peak II phospholipid was not observed either. Also, the presence of Ca2+ may have a significant effect on the enzymatic activity. When GPI-phospholipase C assays were performed in the presence of 0.1-5.0 mM Ca'+, no hydrolysis of the peak II phospholipid was observed. Furthermore, when insulin was added in the above assays, no effect of insulin was observed.
If peak II phospholipid is a true substrate of GPI-phospholipase C, insulin might cause the breakdown of newly synthesized GPI, i.e. in the presence of insulin, the level of GPI will be decreased in biosynthesis assays. As shown in Table II, when the biosynthesis of peak II phospholipid was examined in the presence and absence of insulin ( 10m7 M), there was no difference in the level of peak II phospholipid, indicating the failure of insulin to hydrolyze peak II phospholipid.
There are at least three possible explanations for these negative results: (a) in spite of our several protocols, we might have failed to physically couple the exogenous substrate with the plasma membrane enzyme; (b) an intermediary between the activation of the insulin receptor and the activation of GPI-phospholipase C might be missing in the plasma membrane preparation; (c) peak II is not GPI. Clearly, of these three possibilities, the most feasible to test is the last one.
The ability of bacterial phospholipase C to release cell surface protein with a GPI membrane anchor has been widely used to identify these unique proteins, of which over 30 have already been identified (26). As previously stated, this enzyme has been frequently used to identify GPI, the precursor of insulin mediator . The lower panel of Fig. 8 demonstrates that peak II from human liver microsomes like that from rat liver microsomes (Fig. 1) is readily hydrolyzed by bacterial phospholipase C. However, the upper panel of the same figure also demonstrates the complete hydrolysis of peak I. Although hydrolysis of peak II phospholipid occurred by PI-phospholipase C, the fact is that the phospholipase C characterized by Low's laboratory is specific for PI rather than GPI (27). Thus, this enzyme is not specific to elucidate the structure of GPI or to study the functions of GPI in insulin action. In contrast, a GPI-specific phospholipase C from T. brucei (21) and a phospholipase D with a specificity for GPI from human and bovine plasma (22) have been recently identified.
As shown in Table III, neither of these enzymes hydrolyzed to any extent peak II phospholipid under conditions that were able to hydrolyze [3H]mannose-labeled GPI that have been purified from T. brucei (21). The presence of these enzymes during biosynthesis of peak II had no effect on [3H]myoinositol incorporation into peak II phospholipid (Table III). Also, nitrous acid deamination has been used to demonstrate the presence of glucosamine residues in this glycolipid (28). When the purified phospholipid from peak II was subjected to nitrous acid deamination and the resulting product was isolated, it co-migrated with the original peak II phospholipid indicating the absence of glucosamine residue (Table III). Finally, when liver microsomes were incubated as before but with UDP glucosamine, we were unable to demonstrate any incorporation of labeled glucosamine into the peak II (not shown). Gaulton et al. (29) showed a distinct migration of glucosamine-labeled glycosyl-PI from lyso-PI on silica gel TLC plates. Using a similar method, we compared the relative mobility of peak II phospholipid with GPI and lyso-PI. As can be seen in Fig. 9, by two different chromatography methods, peak II co-migrated with lyso-PI and peak I with PI. Thus, the data strongly suggest that peak II phospholipid is not GPI but an inositol-containing phospholipid that co-migrates with lyso-PI.
To further demonstrate that peak II phospholipid is not GPI and exclude the possibility that our negative results in a human liver do not represent species-dependent differences between man and rat, we studied peak II phospholipid from human and rat liver by gas chromatography/mass spectrometry analysis.
The samples were analyzed with the goal of determining if a GPI anchor was present or not. Traditional carbohydrate analysis techniques were used for this purpose. Pertrimethylsilyl ethers were formed on all available hydroxyl groups to afford volatile materials. Rat and human biopsy samples were analyzed. The results were the same for both species. Each of the samples were analyzed for any trace amounts of GPI constituents such as amino acids, ethanolamine, hexosamine, neutral carbohydrates, and inositols. Amino acid analysis failed to identify amino acids. None of the nitrogen-containing species of the GPI (i.e. amino acids, ethanolamine, or hexosamine) were identified in the samples by the gas chro-  Human liver microsomes (100 pg of protein) were incubated with ["Hlmyoinositol as described under "Experimental Procedures" in the presence or absence of insulin (lo-' M) for 30 min at 37 "C. The radioactivity in the peak II phospholipid was analyzed as described in the legend to Fig. 1 The partially purified peak I and II phospholipids were incubated with the indicated concentrations of PI-phospholipase C. The reactions were stopped by adding chloroform/methanol/6 N HCl and processed as described in the legend to  Human liver microsomes were incubated with ["Hlinositol in the presence and absence of specific phospholipases (PL) as indicated, and incorporation of ["Hlmyoinositol into peak II phosphoinositide was examined as described in the legend to Fig. 1. In other sets of experiments, purified peak II phospholipid was incubated with appropriate phospholipase, and the reactions were processed as described above. Nitrous acid deamination experiments with purified peak II phospholipid were performed as described under "Experimental Procedures." The data are mean ? SE. from three different experiment.s in duplicate. Synthesis of peak I and II phospholipids from human liver microsomes was performed as shown in the legend to Fig. 1. Relative mobilities of these lipids were compared with the authentic samples of several phospholipids run in parallel on the same TLC plate. The spots of authentic samples were visualized by exposure to iodine and radioactive bands by counting the radioactivity as described under "Experimental Procedures." A, comparison using reverse phase TLC used in this study. B, comparison using silica gel plates and chloroform/methanol/NH40H/H20 (45:45:3.5:10, v/v/v/ v) solvent system as described by Mato et al. (5). The open bars represent the chromatography of peak I and the closed bars, peak II. procedure since thin layer chromatography plates use a binder for the silica which contains cellulose, and upon hydrolysis, glucose. The materials isolated were a mixture of lysophospholipids.
Of four samples analyzed, all contained lysophosphatidylserine, lyso-PC, and lyso-PI. The myoinositol was identified from both its retention time in comparison with standards and from the characteristic mass fragments of the pertrimethylsilylated compound, m/e 305, 318, 507. With the characteristic fragments of the inositols, chiroinositol was also identified in the samples from both rat and human biopsies.
In summary, our data does not exclude the presence of GPI as the precursor of the insulin mediator IGP nor question the novel hypothesis of insulin action put forward by Saltiel and Cuatrecacas (4) and supported by experimental data from several laboratories (J-8,30). Our data only demonstrate that under the experimental in vitro conditions described by Saltiel's laboratories, GPI is not synthesized. However, it is important to recognize that Saltiel's laboratory originally reported this method in abstract form (11) and although it has been enthusiastically communicated in several scientific meetings and review articles (12-14) it has not been published yet in a peer review journal. However, the general acceptance that a phosphatidylinositol glycan anchor of membrane protein could be the precursor of an insulin mediator has made us and probably other laboratories recognize the several potential advantages of an in vitro method to study the metabolism and structure of this precursor in normal and pathological states. However, we demonstrate here that GPI is not synthesized in vitro.