Activation of Glycogen Phosphorylase with 5-Aminoimidazole-4-Carboxamide Riboside (AICAR)

The experimental evaluation of the contribution of glycogen phosphorylase (GP) to biochemical pathways is limited to methods that raise cAMP, activating the cAMP-dependent protein kinase/phosphorylase kinase/GP cascade. Such methods convert the unphosphorylated form, “GPb,” which catalyzes glycogenolysis only in the presence of appropriate allosteric activators such as AMP, to the phosphorylated, constitutively activated form, “GPa.” However, activation of GP in this way is indirect, requires a functional cAMP kinase cascade, and is complicated by other actions of cAMP. Here, we demonstrate a strategy for the experimental manipulation of GP in intact dermal fibroblasts, involving activation by the membrane-permeable adenosine analog 5-aminoimidazole-4-carboxamide riboside (AICAR) and inhibition by caffeine and Pfizer compound CP-91149, which bind to GP at distinct sites. Potential complications because of activation of AMP-activated protein kinase by AICAR were assessed with metformin, which activates this kinase but does not activate GP. Using this strategy, we show that glycogen can be a significant and regulatable precursor of mannosyl units in lipid-linked oligosaccharides and glycoproteins.


From the Department of Pharmacology, University of Texas Southwestern Medical Center, Dallas, Texas 75390-9041
The experimental evaluation of the contribution of glycogen phosphorylase (GP) to biochemical pathways is limited to methods that raise cAMP, activating the cAMP-dependent protein kinase/phosphorylase kinase/GP cascade. Such methods convert the unphosphorylated form, "GPb," which catalyzes glycogenolysis only in the presence of appropriate allosteric activators such as AMP, to the phosphorylated, constitutively activated form, "GPa." However, activation of GP in this way is indirect, requires a functional cAMP kinase cascade, and is complicated by other actions of cAMP. Here, we demonstrate a strategy for the experimental manipulation of GP in intact dermal fibroblasts, involving activation by the membrane-permeable adenosine analog 5-aminoimidazole-4-carboxamide riboside (AICAR) and inhibition by caffeine and Pfizer compound CP-91149, which bind to GP at distinct sites. Potential complications because of activation of AMP-activated protein kinase by AICAR were assessed with metformin, which activates this kinase but does not activate GP. Using this strategy, we show that glycogen can be a significant and regulatable precursor of mannosyl units in lipid-linked oligosaccharides and glycoproteins.
Glycogen, a glucose polymer, is a key store of hexose in mammalian cells. Its formation from UDP-Glc and breakdown to Glc-1-P 1 are catalyzed, respectively, by glycogen synthase and glycogen phosphorylase (GP). GP exists in two forms, depending upon its phosphorylation state (1). "GPb," the unphosphorylated form, can be stimulated by allosteric activators such as AMP. Alternatively, phosphorylase kinase can convert the enzyme into its phosphorylated, covalently activated form "GPa," which no longer requires allosteric activation. The regulation of GP may be complicated further by phosphatases and phosphatase inhibitors and the abilities of enzymes such as AMP-activated protein kinase (AMPK) (2) and protein phosphatase-1 (3) to bind to glycogen.
Recently, interest has been generated in glycogen as a source of hexose units in N-linked glycoproteins (4,5). Protein Nlinked glycosylation requires the lipid-linked oligosaccharide (LLO) Glc 3 Man 9 GlcNAc 2 -P-P-dolichol. The oligosaccharide unit is transferred by oligosaccharyl transferase to asparaginyl residues on nascent proteins (6). LLO synthesis requires contributions of enzymes in both the cytoplasm and the endoplasmic reticulum. In the cytoplasm, precursor nucleotide sugars are synthesized. In contrast, the endoplasmic reticulum is the site of assembly of the LLO, requiring UDP-GlcNAc, GDPmannose, and UDP-glucose as well as the lipids mannose-Pdolichol and glucose-P-dolichol, which are synthesized from GDP-mannose and UDP-glucose, respectively. Two independent approaches have suggested that phosphorolysis of glycogen to form Glc-1-P can be stimulated under particular forms of cellular stress and that the Glc-1-P can be converted sequentially to Glc-6-P, Fru-6-P, Man-6-P, Man-1-P, and GDP-mannose in sufficient quantities to substantially enhance glycoconjugate synthesis. Impaired LLO mannosylation in cultured 3T3-L1 and Chinese hamster ovary-K1 cells correlated with glycogen depletion resulting from glucose starvation (4). In addition, increased glucose phosphate production and glycogenolysis were coincident with endoplasmic reticulum stress and improved LLO mannosylation (5). Although such experiments suggest that glycogen may be a significant precursor of mannosyl units for glycoconjugate synthesis, in each case glycogenolysis was activated indirectly. This problem could be solved by deliberate activation of GP. However, current approaches for experimental stimulation of GP in intact cells require activating the cAMP-dependent protein kinase/phosphorylase kinase cascade by adding an appropriate receptor agonist, an activator of adenylyl cyclase, or a membrane-permeable form of cAMP. Activation of phosphorylase kinase by these methods results in the conversion of GPb to GPa, but other pathways also respond to cAMP, so it would be difficult to attribute any effects on LLO synthesis solely to activation of GP.
Here, we examined 5-aminoimidazole-4-carboxamide riboside (AICAR), which like metformin (7) is widely reported to cause activation of AMPK in cells (8). AICAR enhanced LLO extension as anticipated, but surprisingly its mechanism was inconsistent with that of metformin (9). Through a pharmacological approach, we show that AICAR enhanced LLO synthesis primarily by stimulating GP not AMPK.

EXPERIMENTAL PROCEDURES
Reagents-Compound CP-91149 (10) was a generous gift of Pfizer. AICAR (number A611700) was from Toronto Research Chemicals. AICAR-P (ZMP, number A1393), metformin (number D5035), and caffeine (number C0750) were from Sigma. All of the above reagents were prepared as stock solutions in water. Cell culture media were from Invitrogen, and sera were from Atlanta Biologicals. [2-3 H]D-Mannose (10 -20 Ci/mmol) was from Amersham Biosciences, and * This work was supported by Grant GM38545 from the National Institutes of Health and Grant I-1168 from the Robert Welch Foundation. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

2-deoxy-[G-3 H]D-glucose (10 Ci/mmol) was from American Radiolabeled Chemicals.
Culture of Normal and CDG Type I Dermal Fibroblasts-Dermal fibroblasts were cultured as described earlier (11,12) in RPMI 1640 medium with 10% fetal bovine serum. Cells were generally grown to 80 -90% of confluence for experiments. As described earlier (12) the fibroblasts used were normal CRL-1904 (American Type Culture Collection).

Analysis of [ 3 H]Mannose-labeled LLOs and N-Linked
Glycans by HPLC-Glycans were analyzed as described (11,12) by incubating cells for 20 min in RPMI 1640 medium containing 0.5 mM glucose, 10% dialyzed fetal bovine serum, and 40 Ci/ml (2.5 M) [ 3 H]mannose. [ 3 H]-labeled LLOs and N-glycans were recovered and fractionated by HPLC as described previously (11)(12)(13). In some experiments 0.1 mM unlabeled D-mannose was included during the 20-min labeling period. When indicated, the peak heights of [ 3 H]mannose-labeled glycans detected by HPLC were normalized to their mannose contents to determine their percentages in glycan pools (11).
Hexose Uptake Measurements-Uptake of tritium-labeled mannose and 2-deoxyglucose was done over a 10-min period as described (9).
Assay of Glycogen Phosphorylase in Fibroblast Extracts-Assays were performed essentially as described (5). Extracts were prepared by streptolysin O permeabilization (14). A solution was prepared (0.56 ml) with 31.3 mM sodium phosphate (pH 7.4), 3.13 mM MgSO 4 , 3.57 mM NADP ϩ , and 1.79 mg/ml glycogen (Sigma catalog number G1508) and allowed to stand overnight at room temperature (empirically it was noticed that assay reproducibility was greatly enhanced by keeping the NADP ϩ solution at room temperature for at least 2 h). Water, enzymes (glucose-6-phosphate dehydrogenase to 5.7 g/ml, and rabbit muscle phosphoglucomutase to 4.3 units/ml), glucose 1,6-bisphosphate (to 4.3 M), and any activators or inhibitors were added to a final volume of 0.7 ml. The reaction was initiated by adding 0.3 ml of extract containing 0.02 mg of cytoplasmic protein or the corresponding buffer. GP activity was determined spectrophotometrically at 340 nm. Multiple data points were collected over a period of 4 h at room temperature. Over this period assays were linear, and the resulting slopes were used to determine activity. Values for blank reactions lacking extract were routinely subtracted out.

Stimulation of LLO Extension by AICAR-Potential
enzyme activators and inhibitors were tested with primary dermal fibroblasts refed daily with RPMI 1640 medium containing normal glucose (11 mM). To evaluate the effects on synthesis of LLOs and glycoproteins, cells were then labeled for 20 min with 2.5 M [2-3 H]mannose in medium adjusted to 0.5 mM glucose. In our hands this glucose concentration shifts the distribution of LLOs from mostly Glc 3 Man 9 GlcNAc 2 -P-P-dolichol (i.e. completed LLO) to Man 2-5 GlcNAc 2 -P-P-dolichol intermediates as the predominant species (9,11). The effects of 0.5 mM glucose on LLO synthesis are prevented if the labeling reaction also includes 0.1 mM mannose, which efficiently drives LLO mannosylation (11). Thus, enzyme modulators could be evaluated for stimulatory effects on LLO synthesis by labeling in medium with 0.5 mM glucose, whereas inhibitory effects could be discerned with medium containing 0.5 mM glucose and 0.1 mM mannose.
Because metformin is a positive modulator of LLO synthesis (9) we examined AICAR, as it has been widely reported that treatment of cells with AICAR, like metformin, activates AMPK (8). This requires intracellular conversion of AICAR to the phosphorylated nucleotide 5-aminoimidazole-4-carboxamide ribotide (designated AICAR-P for simplicity) by adenosine kinase (15,16). AICAR-P allosterically activates purified AMPK, although with a substantially lower affinity than AMP (17,18). Depending upon the cell type and the experimental system, AICAR-P in intact cells appears to act primarily either by allosteric activation of AMPK (19) or by making AMPK a better substrate for the upstream-activating kinase complex containing LKB1 (20). In any case, the mechanism of AICAR-P differs from that of metformin, which does not activate the isolated enzyme but can activate AMPK in intact cells (7,21).
As shown in Fig. 1, AICAR enhanced extension of LLO intermediates (a and b) and increased the proportion of Glc 3 Man 9 GlcNAc 2 transferred to protein (e and f). However, if LLO extension was optimized by the inclusion of 0.1 mM mannose (Fig. 1, c and d), AICAR had no effect. Interestingly, the efficacy of AICAR was consistently greater than that of metformin (Glc 3 Man 9 GlcNAc 2 -P-P-dolichol for controls was typically 1.5 versus 13% in metformin-treated cells and 29% in AICAR-treated cells), which stimulates LLO synthesis mainly by enhancing mannose uptake (9). However, the effect of metformin on mannose transport (1.8-fold (9)) was greater than that of AICAR (1.3 Ϯ 0.02-fold (n ϭ 22); Fig. 2b), suggesting different mechanisms of action. Like metformin (9), AICAR did not increase transport of 2-deoxyglucose in fibroblasts (Fig. 2a).
To verify that AICAR was less dependent than metformin upon mannose uptake for its effect on LLO (Fig. 2, d-f), indicating that mannose uptake was not a significant factor for AICAR treatment of fibroblasts.
AICAR-P Activates Glycogen Phosphorylase in Fibroblast Extracts-AICAR/AICAR-P has been reported to activate the AMPdependent form of glycogen phosphorylase, GPb, in skeletal (22) and cardiac (19) muscle systems. In both cases AICAR-P activated GP in muscle homogenates, but distinct mechanisms were suggested. Noting that phosphorylase kinase had been reported to be a possible substrate of AMPK (23) (although a more recent study failed to reach a similar conclusion (24)), in skeletal muscle GPb activation was proposed to involve the binding of AICAR-P to AMPK with a sequential phosphorylation of phosphorylase kinase and phosphorylation of GPb (22) rather than to involve the binding of AICAR-P directly to the AMP activation site on GPb. However, the homogenate experiments did not appear to include ATP or an ATP-generating system and were done under conditions in which all three enzymes in the cascade were diluted 130-fold with respect to the homogenate. In contrast, in cardiac muscle, activation by the effects of AICAR was proposed to involve direct allosteric activation of GP by AICAR-P rather than activation of a kinase cascade. This was supported by the observations that treatment of muscle with AICAR led to increased glycogenolysis, decreased glycogen content, no change in the phosphorylation state of AMPK, and no evidence of covalent activation of AMPK or GP. However, phosphorylation of a substrate of AMPK suggested that AICAR did activate AMPK allosterically.
We found that a commercial preparation of muscle GPb (Sigma) could be activated by AICAR-P as well as AMP in the absence of ATP (data not shown), which is highly consistent with direct allosteric activation of GPb by AICAR-P. Importantly, 5 mM AMP and 5 mM AICAR-P, alone and in combination, stimulated GP activity in fibroblast extracts (Fig. 3, upper  panel, lanes 1, 4, 7, and 10), whereas AICAR and metformin did not (data not shown). In these experiments cytoplasmic components were highly diluted, and reactions were done in the absence of ATP. Therefore, the results are consistent with direct allosteric activation and in discord with a sequential enzyme cascade mechanism.
AICAR-dependent LLO Extension Is Blocked by Inhibitors of Glycogen Phosphorylase-To determine whether AICAR enhanced LLO extension by activating GP in intact fibroblasts, two inhibitors of GP were evaluated. Caffeine, which binds to the nucleoside-inhibitory site of GP (1), had no effect on GP activity by itself (data not shown). However, it blocked the stimulation of GP in fibroblast extracts by AMP and AICAR-P (Fig. 3, upper panel, lanes 3, 6, 9, and 12). Using the system for LLO synthesis introduced in Fig. 1, caffeine by itself did not stimulate mannosylation of LLOs (Fig. 3, lower panel, a, d, f,  and i). However, as shown in Fig. 3, b and g, caffeine eliminated the stimulatory effect of AICAR on LLO extension (the chro-matograms shown are representative of four independent experiments). Caffeine did not inhibit the effect of 0.1 mM mannose on LLO synthesis (Fig. 3, c and h), showing that it did not inhibit the relevant mannosyltransferases nonspecifically. Although both GP and AMPK are potential targets for AICAR-P, metformin is known only to activate the latter. GP in fibroblast extracts was not activated by metformin (data not shown), and the ability of metformin to stimulate LLO extension was not blocked by caffeine (Fig. 3, e and j).
Pfizer compound CP-91149 is a highly specific inhibitor of GP, acting both in vitro and in vivo (10). CP-91149 binds to a novel site on GP (25) distinct from the caffeine binding site. Therefore, CP-91149 was used to verify whether AICAR acted by stimulating GP. Although CP-91149 was identified as an inhibitor of hepatic GPa, it also inhibited commercial preparations of muscle GPa in the absence of AMP and, somewhat unexpectedly, muscle GPb activated by AMP (data not shown). As shown in Fig. 3 (upper panel, lanes 2, 5, 8, and 11), CP-91149 inhibited both AMP-and AICAR-P-activated GP in fibroblasts extracts.
CP-91149 was then tested as an inhibitor of the AICARstimulated extension of LLOs in intact cells (Fig. 4, upper  panel). Like caffeine, 20 g/ml CP-91149 by itself did not promote LLO extension (Fig. 4, a and c), but it blocked the stimulatory effect of AICAR (b and d). This confirmed that AICAR acted by stimulating GP. The inhibitory effects of CP-91149 were dose-dependent, with less inhibition by 10 g/ml. It was noticed that the efficacy of the AICAR treatment for LLO synthesis was variable, and the 10 g/ml dose of CP-91149 was considerably more effective as an inhibitor when the efficacy of AICAR was lower (Fig. 4, lower panel). When AICAR increased Glc 3 Man 9 GlcNAc 2 -P-P-dolichol above 20% of the LLO pool, 10 g/ml CP-91149 had no discernable effect. Consistent with previous results, the effects of metformin on LLO extension were not blocked by CP-91149 (not shown). DISCUSSION AICAR differed from metformin in three important ways, in support of the conclusion that the primarily targets of AICAR and metformin in fibroblasts are GP and AMPK, respectively. (i) The effect of AICAR but not metformin on LLO extension was blocked by caffeine and CP-91149. (ii) The effect of metformin but not AICAR on LLO extension was highly dependent upon its ability to stimulate mannose uptake. (iii) AICAR-P, but not metformin, stimulated GPb in fibroblast extracts. These results were unexpected, because AICAR has been reported to activate AMPK in a number of systems, and the fibroblasts were responsive to metformin. Perhaps, compared with other systems, these cells may have a greater ratio of GP to AMPK. Thus, AICAR may be activating both enzymes, but GP activation may be more readily detected and have a much greater significance. Another possibility is that mannose transport in fibroblasts is controlled by an atypical AMPK cascade that is responsive to metformin but not to AICAR. It is interesting that HeLa cells lack functional LKB1 (20), but mannose transport in HeLa cells is responsive to metformin (9).
There is little doubt that AICAR was converted to AICAR-P in fibroblasts, as it is in many mammalian systems (8), especially because AICAR did not activate GP in extracts. Estimates based upon formation of AICAR-P from AICAR in Chinese hamster ovary cells (16) and muscle (15) are consistent with intracellular concentrations in the millimolar range sufficient to activate GP. A canine plasma AICAR concentration of 0.34 mM resulted in approximate AICAR-P tissue concentrations of 0.14 mM in skeletal muscle and 0.81 mM in cardiac muscle (15); extrapolation to 2 mM AICAR (as used in this study) therefore suggests fibroblast intracellular AICAR-P concentrations of ϳ1-5 mM. Similarly, Chinese hamster ovary-K1 cells treated for 7 h with 0.7 mM AICAR accumulated 11 nmol of AICAR-P/mg of cellular protein (16). In our hands, packed fibroblasts giving a volume of 1 ml would contain roughly 0.5 g of total protein, suggesting that incubation of fibroblasts with 2 mM AICAR for 7 h (as done here) might result in ϳ15 mM intracellular AICAR-P.
In any case, these results show that the combination of an activator (AICAR/AICAR-P) with inhibitors that act by distinct mechanisms (caffeine and CP-91149) can be used for the experimental evaluation of GP. This strategy should be applicable FIG. 3. Activators and inhibitors of glycogen phosphorylase in fibroblasts. Upper panel, cytoplasmic extracts were obtained from normal fibroblasts permeabilized with streptolysin O, and glycogen phosphorylase was measured (1-ml assays containing 0.02 mg of extract protein) by phosphorolysis of glycogen as described (5). Each bar represents the slope Ϯ S.D. of five measurements taken over a 4-h period. The slope obtained in the absence of additives (control, lane 1) was arbitrarily assigned a value of 1. Assays were performed in the absence (lanes 1-3) or presence of 5 mM concentrations of the activators AICAR-P (lanes 4 -6, 10 -12) or AMP (lanes 7-12). 10 g/ml CP-91149 was included in assays 2, 5, 8, and 11, and 5 mM caffeine was included in assays 3, 6, 9, and 12. Lower panel, LLOs were analyzed after fibroblasts were grown as in Fig. 1 without mannosylation enhancers  (a, c, d, f, h, and i) Fig. 1 after fibroblasts were grown with no additions (a), with 2 mM AICAR for 7 h (b), with 20 g/ml CP-91149 for 7 h (c) or both AICAR and CP-91149 (d). Lower panel, results from four experiments testing the effects of either 10 g/ml CP-91149 (squares) or 20 g/ml CP-91149 (circles) on enhancement of LLO synthesis by AICAR are shown. Each data point represents one experiment for which the percentage of Glc 3 Man 9 GlcNAc 2 -P-P-dolichol in the LLO pool with AICAR treatment, in the absence (x axis) or presence (y axis) of CP-91149, was calculated (see "Experimental Procedures"). The dotted line indicates the plot expected if CP-91149 had no effect.
whether AICAR-P activates GP allosterically or stimulates AMPK to phosphorylate phosphorylase kinase, because the two inhibitors are effective for GPb and GPa. These results also demonstrate that the Glc-1-P resulting from glycogenolysis can be a significant source of hexose units for glycoconjugate assembly and provide a proof-of-principle for activation of glycogenolysis by the unfolded protein response as the cause of enhanced LLO extension (11,12). Thus, the GP modulators used here provide glycobiologists with a novel approach for experimentally altering glycoconjugate synthesis. GP and AMPK respond to variations in energy metabolism, which can be caused by transient drops in cytoplasmic hexose concentrations. Such fluctuations are likely to affect LLO synthesis and energy metabolism in tandem. Therefore, GP and AMPK may have important roles in maintaining a consistent supply of hexose for LLOs and presumably for other glycoconjugates, such as glycosylphosphatidylinositol anchors, that require GDP-mannose.