Insulin-stimulated Diacylglycerol Production Results from the Hydrolysis of a Novel Phosphatidylinositol Glycan*

We recently described the insulin-dependent release of a carbohydrate substance from plasma membranes which regulated certain intracellular enzymes (Saltiel, A. R., and Cuatrecasas, P. (1986) Proc. Natl. Acad. Sei. U. S. A. 83,5793-5797). This enzyme-modulating substance appeared to arise from the phosphodiesterase hydrolysis of a novel inositol-containing glycolipid. This is supported by observations that insulin stimulated the rapid generation of [sH]myristate-labeled di- acylglycerol in cultured BCsHl myocytes. Myristoyl diacylglycerol production in these cells was unaffected by epinephrine, although arachidonate-labeled diac- ylglycerol was rapidly produced in response to stimulation by this a-1 adrenergic agent. The production of distinct species of diacylglycerol was apparently due to hormonally specific hydrolysis of different precur- sors. A novel glycolipid was identified on silica TLC or high pressure liquid chromatography which served as a substrate for the insulin-stimulated phosphodiesterase reaction. This glycolipid was metabolically labeled with radioactive inositol, glucosamine, and myristic acid, suggesting a phosphatidylinositol (PI)-glycan structure. Treatment of this glycolipid Analysis Lipids-Following preincubations minimal essential unlabeled dimyristoyl glycerol, myristic monomyristoyl to each sample. HPLC-The phospholipid was chromatographed on a silica HPLC eluted with a linear 20-min gradient of chloroform/methanol/glacial acetic acid/H20 (65:25:101 40: 45:105) at 1 ml/min. The aqueous products of the hydrolysis reactions were identified on an analytical SAX HPLC column, eluted with a linear 15-min gradient of 60% methanol to 0.5 M triethylamine-formate, pH 4.5, at 1 ml/min, as described previously (3). Bioassay of Water-soluble Hydrolysis Products-The water-soluble products of the hydrolysis reactions were also identified by the ability to stimulate CAMP phosphodiesterase in adipocyte particulate fraction (3). The high affinity enzyme was assayed as described previously phosphodiesterase (PDE) activity in rat adipocyte particulate fraction. Re- sults are the means of triplicate determinations expressed as % basal activity, in which variability was less than 5%.

The molecular mechanisms of transmembrane signaling in insulin action are poorly understood. It has been suggested that some of the metabolic effects of the hormone may be explained by the generation from the plasma membrane of a unique substance or group of substances which acutely regulate certain insulin-sensitive enzymes (1,2). We recently reported (3, 4) that two such activities were apparently produced by the insulin-sensitive phosphodiesterase cleavage of a novel plasma membrane-associated glycolipid. These enzyme-modulating substances were released from liver or mus- cle cell plasma membranes in response to insulin or an exogenously added phosphatidylinositol (PI)'-specific phospholipase C and acutely modulated the activity of the low K , cAMP phosphodiesterase (3) as well as pyruvate dehydrogenase and adenylate cyclase (5). These enzyme-modulating activities apparently resulted from structurally related carbohydrate-phosphate substances which contained inositol, phosphate, glucosamine, and other monosaccharides. This action of insulin was further evaluated in the cultured murine myocyte cell line BC3Hl (4) by following the radioisotopic labeling of the putative glycolipid precursor and the carbohydrate products. Insulin caused the hydrolysis of this glycolipid precursor and the incorporation of [3H]inositol and [3H]glucosamine into purified HPLC fractions which coeluted with the phosphodiesterase-modulating substances.
Insulin causes the increased labeling of several phospholipids, including the phosphoinositides and phosphatidic acid (6-111, and also stimulates the production of diacylglycerol (4, 12, 13). Farese et al. (13) suggested that stimulation of diacylglycerol production in BC3Hl cells by insulin was due to enhanced de novo synthesis, since the hydrolysis of phosphoinositides could not be detected. In this report we propose that insulin-stimulated diacylglycerol production in BC3H1 cells is due to the action of a phospholipase C which selectively cleaves a PI-glycan substrate, generating both diacylglycerol and the inositol-glycan which modulates certain insulin-sensitive enzymes. Moreover, the selective action of this insulinsensitive phospholipase C on the PI-glycan may result in the exclusive generation of myristate-labeled diacylglycerol, which can be distinguished from the arachidonate-containing diacylglycerol derived from the hormone-stimulated hydrolysis of the phosphoinositides in these cells.
Cell Culture-BC3H1 cells were cultured on collagen-coated multiwell plates or Petri dishes in Dulbecco's minimal essential medium supplemented with 20% Nu Serum or 10% fetal calf serum. Cells were grown to maximal density and confluency, at which time they become responsive to insulin (14). Radiolabeled precursors were The abbreviations used are: PI, phosphatidylinositol; HPLC, high pressure liquid chromatography. added for 20 h in serum-free medium. [3H]Myristic and arachidonic acids were complexed 1:l to bovine serum albumin prior to addition.
Extraction and Analysis of Lipids-Following preincubations with the appropriate isotopes, cells were resuspended in Dulbecco's minimal essential medium without serum and treated with 10 nM insulin at the designated intervals. Reactions were terminated by the addition of 1 ml of chloroform, methanol, 1 N HC1(2001001), followed by 0.5 ml of 10 mM formic acid. Following centrifugation at 500 X g for 5 min, the upper phase was discarded. For phospholipid analyses, the lower organic phase was dried under N2, resuspended in chloroform/ methanol/H20 (972) and spotted on oxalate-impregnated Silica Gel G plates. These were sequentially developed in chloroform/acetone/ methanol/glacial acetic acid/HZO (1042:2:1) and chloroform, methanol, 4 N NH,OH (45:3510). One-cm regions were scraped and eluted with chloroform/methanol (21) and counted. Phosphoinositide standards were identified by iodine staining.
For analysis of neutral lipids, the chloroform/methanol organic phases were dried under N2 and resuspended in 1 ml of diethyl ether followed by addition of 1 ml of 50 mM formic acid. The upper ether phase was aspirated, and the lower aqueous phase was re-extracted with 1 ml of diethyl ether. The pooled ether phases were dried under N2, resuspended in chloroform, and spotted on Silica Gel G plates which were preactivated at 60 "C for 1 h. Plates were twice developed in petroleum ether/diethyl ether/glacial acetic acid (70302). Five pg of unlabeled dimyristoyl glycerol, myristic acid, and monomyristoyl glycerol were added as carriers to each sample.
HPLC-The phospholipid precursor was further chromatographed on a silica HPLC column, eluted with a linear 20-min gradient of chloroform/methanol/glacial acetic acid/H20 (65:25:101 to 40: 45:105) at 1 ml/min. The aqueous products of the hydrolysis reactions were identified on an analytical SAX HPLC column, eluted with a linear 15-min gradient of 60% methanol to 0.5 M triethylamineformate, pH 4.5, at 1 ml/min, as described previously (3).
Bioassay of Water-soluble Hydrolysis Products-The water-soluble products of the hydrolysis reactions were also identified by the ability to stimulate CAMP phosphodiesterase in adipocyte particulate fraction (3). The high affinity enzyme was assayed as described previously (15).

RESULTS
Diacylglycerol Production in BC&1 Cells-The time course of production of diacylglycerol was evaluated in response to hormones. Cells were labeled with [3H]myristi~ acid ( Fig. 1) or [3H]arachid~nic acid (Fig. 2) and exposed to insulin or epinephrine, known to act as an a-1 adrenergic agonist in these cells (13). In [3H]myristate-labeled cells ( Fig. 1) insulin caused a rapid increase in labeled diacylglycerol, which was maximal at 1 min, declined by 2 min, and slowly increased thereafter. This biphasic pattern was observed in several experiments. In contrast, exposure to epinephrine resulted in no significant change in [3H]myristoyl diacylglycerol above basal levels. In cells incubated with r3H]arachidonic acid ( Fig.  2), insulin treatment caused a slow increase in labeled diacylglycerol. In contrast, epinephrine produced a rapid increase in [3H]arachidonyl diacylglycerol which was maximal at 1 min, declined by 5 min, and then slowly increased. Insulin had no detectable effect on the generation of [3H]stearyl diacylglycerol (not shown).
Identification of a Phosphatidylinositol-Glycan Precursor-In previous reports (3, 4) a novel PI-glycan was identified which served as a precursor for the insulin-dependent generation of phosphodiesterase-modulating substances in liver and muscle cells. Cultured BC3H1 cells were separately incubated with [3H]inositol, [3H]glucosamine, or [3H]myristi~ acid for 20 h. Lipids were extracted, treated with or without PIphospholipase C purified from Staphylococcus aureus, and chromatographed on thin layer plates (Fig. 3). Extraction of phospholipids from inositol-labeled cells revealed two tritiated spots which were diminished by incubation with PI-phospholipase C (Fig. 3a). The PI-phospholipase C-sensitive spot migrating toward the solvent front coeluted with PI. The remaining PI-phospholipase C-sensitive spot, which migrated myristic acid, extracted with chloroform/methanol/HCl(2OOlOOl) and resuspended in 1 ml of 50 mM ammonium bicarbonate, pH 6.0. Following 2-h incubation with or without 1 pg/ml PI-phospholipase C (PLC), solutions were re-extracted and organic phases were spotted on silica plates, developed as described under "Experimental Procedures." Phosphoinositide standards were visualized by iodine staining. PIP, phosphatidylinositol phosphate. between phosphatidylinositol 4-phosphate and phosphatidylinositol, exhibited chromatographic behavior identical to the glycolipid identified as precursor for the PI-phospholipase Cgenerated phosphodiesterase modulators (3). This spot was tentatively identified as the PI-glycan. Extraction of cells preincubated with [3H]glucosamine revealed one major PIphospholipase C-sensitive spot on TLC which comigrated with the PI-glycan (Fig. 3b). [3H]Myristic acid was incorporated into several lipid species, although only the radioactivity residing in spots corresponding to PI and the PI-glycan was diminished by treatment with PI-phospholipase C (Fig. 3c).
HPLC of the PI-Glycan-The radiolabeled PI-glycan could be purified further by an HPLC silica column. The region of the thin layer plate which contained the PI-glycan was scraped and eluted. The resulting substance was a poor substrate for hydrolysis by PI-phospholipase C. However, when reconstituted by sonication with lipid vesicles containing 1:l mixtures of phosphatidylethanolamine:phosphatidylcholine, complete hydrolysis was observed. The TLC-eluted lipid was reconstituted and incubated with or without PI-phospholipase C for 2 h and then extracted with chloroform/methanol and phase separated. The organic phases were chromatographed on an HPLC silica column (Fig. 4). A predominant peak containing labeled inositol (Fig. 4a) or myristic acid (Fig. 4b) was detected at 22 min which was diminished in PI-phospholipase C-treated extracts. Twenty-four-h incubation of this  . 4. Silica HPLC of the PI-glycan (PI-G). Following thin layer chromatography, the PI-phospholipase C (PLC)-sensitive spot (as detected in Fig. 3) was scraped and eluted, and resuspended with 100 pg of a phosphatidylethanokphosphatidylcholine into 1 ml of 50 mM ammonium bicarbonate, pH 6.0, by sonication. These solutions were treated with (W) or without (0) 1 pg of PI-phospholipase C for 2 h. Following extraction the organic phase was injected into a silica HPLC column, eluted as described under "Experimental Procedures." One-ml fractions were counted. Phosphoinositide standards were identified by TLC of eluted fractions. PIP, phosphatidylinositol phosphate. fraction with PI-phospholipase C resulted in its complete hydrolysis (data not shown).
Product Analysis of PI-Phospholipase C Hydrolysis of the PI-Glycan-After PI-phospholipase C digestion of the PIglycan, both the aqueous and organic products were analyzed. The labeled PI-glycan was purified by HPLC and treated with PI-phospholipase C as described above. For glycolipid extracted from cells labeled with [3H]inositol the aqueous products were resolved by ph&e extraction and chromatographed directly on an HPLC SAX column (Fig. 5). PI-phospholipase C treatment caused the generation of one peak of radioactivity eluting at 15 min. This labeled substance eluted with a retention time distinct from that of cyclic 1,2-inositol monophosphate (9 min) or 1-or 2-inositol monophosphate (24 min). The PI-phospholipase C-generated radiolabeled substance corresponded to radiolabeled Peak I previously detected in insulin-treated BC3Hl cells. No radioactivity was detected in fractions corresponding to Peak I1 (3,4). Previous studies (4) suggested that Peaks I and 11 contained cyclic and monophosphate inositol phosphate derivatives, respectively. Detection of only Peak I by PI-phospholipase C hydrolysis of the purified PI-glycan precursor followed by direct chromatography suggests that the cyclic compound may be the only product generated initially.
The identity of the labeled aqueous product as the previously detected bioactive phosphodiesterase-modulating substance was validated by an assay of the enzyme-modulating activity produced from unlabeled precursor. Unlabeled cells or without (0) l pg/ml PI-phospholipase C for 2 h. Following extraction the aqueous phase was injected directly onto an analytical SAX HPLC column, eluted as described under "Experimental Procedures." One-ml fractions were counted. Cyclic 1,2-inositol monophosphate (ZP) and inositol 2-monophosphate were detected by TLC (30) of collected fractions.
were extracted and the PI-glycan was identified on HPLC by coelution with trace amounts of the labeled purified glycolipid. The resulting silica HPLC-purified PI-glycan was treated with or without PI-phospholipase C, aqueous products were injected into an HPLC SAX column, and fractions were assayed for phosphodiesterase-modulating activity (Fig. 6). A single phosphodiesterase-modulating substance was identified with a retention time (15 min) identical to that of the radiolabeled product. Further analysis of elution behavior on P-2 gel filtration columns and high voltage thin layer electrophoresis (3) revealed that the radioactive and bioactive substances displayed identical chromatographic and electrophoretic properties (data not shown).
To analyze the nonaqueous product of the PI-phospholipase C-catalyzed hydrolysis, the purified [3H]myristic acid-labeled PI-glycan was treated with or without PI-phospholipase C and subsequently extracted with ether. TLC of this neutral lipid fraction (Fig. 7) indicated that PI-phospholipase C treatment caused the generation of diacylglycerol which contained [3H]myristi~ acid.
Insulin Stimulates the Hydrolysis of the PI-Glycan-The effect of insulin on the turnover of the PI-glycan was evaluated (Fig. 8). Cells were prelabeled with [3H]inositol or [3H] myristate, and radioactivity in the PI-glycan spot identified by TLC was determined. As previously reported, a small (20%) but significant decrease in [3H]inositol labeling was observed after 30 s of insulin exposure, followed by a gradual (40%) increase in counts, perhaps reflecting resynthesis (Fig. 8u). In [3H]myristate-labeled cells (Fig. 8b) insulin caused a 60% were extracted with chloroform/methanol/HCl (2001001) and streaked onto a Silica Gel G plate, developed as detailed under "Experimental Procedures." The spot corresponding to an HPLCpurified [3H]inositol-labeled PI-glycan standard was scraped and eluted and reconstituted with phospholipids as described in the legend to Fig. 4. Solutions were treated with (B) or without (0) 1 gg of PIphospholipase C. Following extraction, aqueous phases were chromatographed on a SAX HPLC column as described in Fig. 5.One-ml fractions were lyophilized, resuspended in 0.1 ml of 10 mM formic acid, and assayed for ability to stimulate the low K,,, CAMP phosphodiesterase (PDE) activity in rat adipocyte particulate fraction. Results are the means of triplicate determinations expressed as % basal activity, in which variability was less than 5%. decrease in the labeling of the PI-glycan by 1 min, which did not return to basal levels until 20-30 min. These differences in the rate of relabeling by these two precursors were consistently observed in several experiments and may be due to inability to attain equilibrium during the labeling period.

DISCUSSION
We recently reported (3-5) that insulin stimulated the production of two related complex carbohydrate-phosphate substances in BC3Hl cells containing inositol and glucosamine. The generation of these substances was accompanied by the production of diacylglycerol, followed by phosphatidic acid, and appeared to result from the phosphodiesteratic cleavage of a novel PI-glycan. In this report we characterize this process further by examining the rapid insulin-dependent production of [3H]myristate-labeled diacylglycerol, which appeared to result from the hydrolysis of a novel precursor, rather than from de m u 0 synthesis as suggested previously (12,13). This rapid generation of [3H]myristoyl diacylglycerol could be distinguished from a second slower phase which was also observed in [3H]arachidonic acid-labeled cells. The significance of the biphasic pattern of diacylglycerol production is unclear, although it may result from a phosphorylation and subsequent dephosphorylation reactions, as reflected by a  Fig. 4 and treated with or without 1 pg/ml PI-phospholipase C for 1 h. Following extraction, the lower organic phase was spotted on Silica Gel G plates for analysis of neutral lipids as detailed under "Experimental Procedures." Diacylglycerol (DAG) was identified by iodine staining. transient increase in phosphatidic acid in response to insulin (4). In contrast to insulin, epinephrine, acting through a-1 adrenergic receptors in these cells, caused the rapid generation of [3H]arachidonyl diacylglycerol, but had no discernible effect on generation of [3H]myristoyl diacylglycerol. These results indicate that insulin and epinephrine stimulate the activities of distinct phospholipases C which hydrolyze different substrates. The putative insulin-sensitive phospholipase C appears to catalyze the phosphodiesteratic cleavage of a novel glycolipid which contains PI, glucosamine, and other monosaccharides and is labeled with myristic acid. This action of insulin causes the generation of two potential second messengers: 1) the inositol phosphate-glycan which modulates several insulin-sensitive enzymes including CAMP phosphodiesterase, pyruvate dehydrogenase, and adenylate cyclase (3,5 ) ; and 2) myristoyl diacylglycerol. Significantly, the insulinstimulated hydrolysis of this glycolipid does not produce inositol Tris phosphate (11, 13), whereas epinephrine stimulates a phosphoinositide-specific phospholipase C, resulting in the production of inositol phosphates (13) and arachidonyl diacylglycerol. This proposed substrate specificity for hormonally regulated phosphodiesterase hydrolysis reactions is further supported by the recent identification of a phospholipase C activity with a distinct substrate specificity for the PI-glycan and a related phosphoinositide glycoside.   Fig. 1 and exposed to 10 nM insulin at the designated intervals. Labeling of the PI-glycan was determined by scraping and counting of a spot on silica gel thin layer chromatography corresponding to an HPLC-purified [3H]inositol-labeled PIglycan standard. All reactions were performed in triplicate, and results were reproduced in several experiments.
The possible role of diacylglycerol in insulin action has been considered (4,(16)(17)(18)(19). Diacylglycerol and the structurally related phorbol esters are known to specifically activate a Ca2+-and phospholipid-dependent protein kinase C (20). Phorbol esters or protein kinase C mimic certain actions of insulin (16-19), yet antagonize other effects of the hormone (21)(22)(23). Furthermore, direct activation of protein kinase C or intracellular translocation of the enzyme by insulin have not been observed (24). The discrepancies between the actions of insulin and phorbol esters may be explained in part by the unique pathway of diacylglycerol formation described here. The myristoyl diacylglycerol may produce only a limited activation of protein kinase C or is perhaps directed to an enzyme which is similar but not identical to the kinase. Such an enzyme, which was phospholipid dependent but Ca2+ independent, has recently been described (25). Moreover, the recent identification of a new family of protein kinase Crelated genes provides further evidence for multiple kinase C proteins, perhaps with distinct regulatory domains (26). Alternatively, the production of diacylglycerol in the absence of inositol Tris phosphate-induced calcium mobilization may result in the selective regulation of a kinase, perhaps causing the activation in situ of the membrane-associated form of the enzyme, but ineffective in facilitating the translocation of a cytoplasmic component of the kinase to the plasma membrane. Although we have detected increased labeling in fractions of diacylglycerol in response to insulin, it is possible that the minor changes in mass which occur are insufficient to produce stimulation or translocation of protein kinase C, or that these observed changes in diacylglycerol levels are significant only in certain membrane compartments. In this regard, we are unable to assess changes in mass in this specific diacylglycerol, and thus the precise chemical magnitude of the change is impossible to determine.
The novel PI-glycan described here shares some homology with the glycolipid responsible for the anchoring of several proteins to the plasma membrane (27). The insulin-sensitive hydrolysis of the PI-glycan and the enzymatic release of the variant surface glycoprotein in Trypanosoma brucei exhibit several similarities (28). The trypanosome protein is covalently linked to dimyristoyl PI via a glucosamine-containing glycan, which can be hydrolyzed by the S. aureus PI-phospholipase C (28) and an endogenous phospholipase C (29).
The endogenous enzyme cleaves this PI-glycan anchor but is inactive in hydrolyzing phosphoinositides or other phospholipids (29). In a separate report: we describe the purification and characterization of a phospholipase C in liver plasma membranes with similar specificity. The activation of this enzyme by the insulin receptor complex may be a critical transduction mechanism in the action of the hormone.