Inhibition of Hepatocytic Autophagy by Adenosine, Aminoimidazole- 4-carboxamide Riboside, and N-Mercaptopurine Riboside EVIDENCE FOR INVOLVEMENT OF AMP-ACTIVATED PROTEIN KINASE*

To examine the role of AMP-activated protein kinase (AMPK; EC 2.7.1.109) in the regulation of autophagy, rat hepatocytes were incubated with the AMPK proactivators, adenosine, 5-amino-4-imidazole carboxamide riboside (AICAR), or N-mercaptopurine riboside. Autophagic activity was inhibited by all three nucleosides, AICAR and N-mercaptopurine riboside being more potent (IC50 5 0.3 mM) than adenosine (IC50 5 1 mM). 2*Deoxycoformycin, an adenosine deaminase (EC 3.5.4.4) inhibitor, increased the potency of adenosine 5-fold, suggesting that the effectiveness of adenosine as an autophagy inhibitor was curtailed by its intracellular deamination. 5-Iodotubercidin, an adenosine kinase (EC 2.7.1.20) inhibitor, abolished the effects of all three nucleosides, indicating that they needed to be phosphorylated to inhibit autophagy. A 5-iodotubercidin-suppressible phosphorylation of AICAR to 5-aminoimidazole4-carboxamide riboside monophosphate was confirmed by chromatographic analysis. AICAR, up to 0.4 mM, had no significant effect on intracellular ATP concentrations. Because activated AMPK phosphorylates and inactivates 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) reductase (EC 1.1.1.88), the rate-limiting enzyme in cholesterol synthesis, the strong inhibition of hepatocytic cholesterol synthesis by all three nucleosides confirmed their ability to activate AMPK under the conditions used. Lovastatin and simvastatin, inhibitors of HMG-CoA reductase, strongly suppressed cholesterol synthesis while having no effect on autophagic activity, suggesting that AMPK inhibits autophagy independently of its effects on HMG-CoA reductase and cholesterol metabolism.

Numerous biochemical processes are involved in the maintenance and support of cell function and cellular growth. These processes need to be coordinated not only relative to each other but also in relation to the metabolic energy available to the cell. The coordination is generally thought to be effectuated in a diffuse fashion by the adenine nucleotides (AMP, ADP, and ATP), which are both carriers of energy (ATP and ADP) and capable of serving as direct regulators of the activity of many enzymes. Although the cellular energy metabolism is too complex to be described by a manageable algorithm, the activities of most metabolic pathways have been found to correlate rea-sonably well, positively or negatively, with the overall cellular energy state as embodied e.g. in the concept of "energy charge" ([ATP ϩ 1 ⁄2ADP]/[ATP ϩ ADP ϩ AMP]) (1).
In recent years, the enzyme AMP-activated protein kinase (AMPK) 1 (EC 2.7.1.109) has been emerging as a "master switch" that mediates at least part of the coordination between energy state and metabolism (2). The activity of this enzyme is regulated in a complex fashion by adenine nucleotides as well as by an upstream kinase (itself activated by AMP) and becomes active when the AMP level is high relative to ATP. Activated AMPK in turn shuts down energy-requiring pathways like fatty acid and cholesterol synthesis through phosphorylation and inactivation of the pertinent key enzymes while indirectly activating an energy-producing pathway like fatty acid oxidation (2). However, so far the documented regulatory effects of AMPK have been largely confined to lipid metabolism. It would obviously be of interest to study other metabolic pathways to assess the generality of this enzyme's role in metabolic regulation.
Autophagy occupies a central position in cellular metabolism, supplying small molecules both for anabolic and catabolic purposes through the bulk degradation of proteins and other cellular macromolecules. By a poorly understood sequestration process, pieces of cytoplasm become encapsulated by cellular membranes, forming autophagic vacuoles that eventually fuse with lysosomes to have their contents degraded. The autophagic-lysosomal pathway is subject to complex regulation at the initial sequestration step, by hormones, growth factors, metabolites, and various signaling molecules (3).
Two previous sets of observations are particularly pertinent to the energy-metabolic control of autophagy: (i) ATP depletion has been shown to reduce the autophagic activity (4), the overall autophagic-lysosomal protein degradation (5), and the fractional volume of autophagosomes (6) in isolated hepatocytes, apparently by inhibiting several steps in the autophagic-lysosomal pathway (7). (ii) Adenosine and, more potently, N-substituted adenosine analogues are capable of suppressing hepatocytic autophagy (8) and protein degradation (9). It is noteworthy that a large increase in hepatocytic AMP levels is a characteristic feature of both ATP depletion (10 -12) and adenosine addition (13). The possibility should, therefore, be considered that AMP, perhaps through activation of AMPK, might be a mediator of autophagy suppression under both conditions. Because autophagy apparently requires energy (4,7), it would seem logical that it be included it among the processes shut down by AMPK under conditions of energy depletion (2).
In the present work we have therefore investigated whether the autophagy-inhibitory effect of adenosine might be mediated by its phosphorylation to AMP. Furthermore, we tested the effect of 5-amino-4-imidazolecarboxamide riboside (AICAR), shown to be a potent activator of AMPK after its intracellular phosphorylation to AICA-ribotide or ZMP (14,15). Both adenosine and AICAR were found to inhibit autophagy strongly. An inhibitor of adenosine kinase (EC 2.7.1.20), 5-iodotubercidin (ITu) (16), suppressed the autophagy-inhibitory effects of both treatments, indicating that formation of AMP and ZMP, respectively, was necessary to inhibit autophagic activity. The results thus support the contention that AMPK may be involved in the regulation of hepatocytic autophagy.  (17). By this method we obtained about 4 g of wet mass of Ͼ90% intact hepatocytes (cells able to exclude trypan blue). The wet mass was measured by weighing sedimented cell pellets, which have a composition corresponding exactly to intact liver tissue (17). The cells were incubated as 0.4-ml aliquots (70 -80 mg cellular wet mass/ml) at 37°C for 2 h in rapidly shaking glass tubes; the shaking ensures adequate oxygenation as indicated by the maintenance of high physiological ATP levels (4,7,9).

Biochemicals
Measurement of Autophagy-Sequestration of a long-lived cytosolic enzyme, LDH, was used to measure autophagy (18). Leupeptin (0.3 mM), a cysteine proteinase inhibitor (5), was added to the cell suspensions during incubation to prevent the lysosomal degradation of LDH (18). After the incubation, the cells were washed twice with 4.0 ml of isotonic sucrose solution (10% w/v) followed by centrifugation (1600 ϫ g, 4 min). The cells were then resuspended in 500 l of 10% sucrose, briefly warmed to 37°C, electrodisrupted by a single high voltage pulse (2 kV/cm), and transferred to a new tube containing 0.5 ml of ice-cold post-disruption buffer (100 mM potassium phosphate, 2 mM EDTA, 2 mM dithiothreitol, and 0.05 mM sucrose, pH 7.5). To separate autophagic vacuoles from the rest of the cellular material, 0.6 ml of the cell suspension was carefully layered on top of a 4-ml density cushion (2.2% sucrose, 8% metrizamide, 50 mM potassium phosphate, 1 mM dithiothreitol, and 1 mM EDTA, pH 7.5) and centrifuged at 3700 ϫ g for 30 min (i.e. approximately 1.1⅐10 5 g⅐min Ϫ1 ). After centrifugation and removal of the supernatant (containing the cytosol), the pellet (containing the autophagic vacuoles) was washed in 10% sucrose and resuspended in resuspension buffer (50 mM potassium phosphate, 1 mM EDTA, and 1 mM dithiothreitol, pH 7.5) and frozen at Ϫ70°C. The remaining 0.4 ml of the cell suspension was also frozen at Ϫ70°C, to be used for measuring the total cellular LDH. LDH activity in resuspended pellets (autophagocytosed LDH) and in the disrupted cells (total LDH) was measured spectrophotometrically with a Technicon RA 1000 autoanalyzer. The amount of sequestered LDH relative to total cellular LDH was taken to indicate the autophagic sequestration rate and was expressed as percent/h. Intracellular ATP Concentration-The intracellular ATP content was measured with a luminometer from LKB-Wallac Oy (Oulu, Finland) using a luciferin/luciferase assay procedure. Immediately after incubation the cells were acid-precipitated by ice-cold perchloric acid (final concentration, 2% w/v), followed by a 10-min incubation on ice and a 15-min centrifugation at 3700 ϫ g. A fixed volume of supernatant was transferred to a new tube, neutralized with freshly made 1 N KOH, and frozen at Ϫ20°C. On the measurement day, the samples were diluted 40 times with Tris acetate buffer (0.1 mM Tris acetate, 2 mM EDTA, pH 7.75), before measuring the intensity of the produced light in each sample relative to an ATP standard.
Measurement of Intracellular Nucleosides and Nucleotides-Intracellular nucleosides and nucleotides were separated and quantified by reverse phase HPLC as described by Shevchuk et al. (19) using a reverse phase Supelcosil LC-18-T column (25 cm ϫ 4.6 mm) from SUPELCO (Park Bellefonte, PA). The HPLC instrument system was from Waters (chromatographic division of Millipore Co., Milford, MA), consisting of a Waters 486 tunable absorbance detector adjusted to 254 nm, a WISP 710B automated injector and pump, and a Waters multisolvent delivery system. Data were analyzed by a Waters Maxima 820 chromatography work station.
After incubation, the samples were immediately treated with 2% perchloric acid and incubated for 15 min at 0°C. The samples were then centrifuged for 15 min at 3700 ϫ g, the supernatants were transferred to a new set of tubes, neutralized with freshly made 1N NaOH, and stored at Ϫ20°C until the measurement day. All samples and solutions used in the HPLC system were filtered through a 0.22-m Millipore filter.
The mobile phase consisted of two buffers: buffer A (100 mM KH 2 PO 4 , pH 5.0, 1.5% acetonitrile, 0.08% tetrabutyl ammonium bromide) and buffer B (150 mM KH 2 PO 4 , pH 5.0, 10% acetonitrile, 0.08% tetrabutyl ammonium bromide). The measurement time for each sample was 45 min at a constant flow rate of 1 ml/min with detection at 254 nm. The chromatographic procedure was divided into three phases: (i) the gradient phase (in which the mobile phase gradually changed from 100% buffer A to 100% buffer B within 2 min), (ii) the constant phase (100% buffer B for the next 30 min), and (iii) the wash phase (100% buffer A for 15 min). Each peak was identified by coelution of the samples with added standards.
Cholesterol Synthesis-Cholesterol synthesis in isolated hepatocytes was measured as the incorporation of [2-14 C]acetate according to Rustan et al. (20). After incubation in the presence of [2-14 C]acetate (100 M; 1 Ci/ml), the cells were washed in 10% isotonic sucrose (4 min at 1600 ϫ g) to remove dead cells and excess [2-14 C]acetate. The cells were then dissolved in 1 ml of 50 mM potassium phosphate (pH 7.5) and frozen at Ϫ70°C until the next day, when extraction was performed.
The cell samples were thawed and homogenized by sonication (60 mA/s), and 400 l of homogenate was transferred to a new tube. Cellular lipids were extracted by mixing with 20 times the volume of chloroform/methanol (2:1, v/v), incubation for 30 min, addition of four times the volume of 0.9% NaCl solution, and incubation for 15 min, which allowed the mixture to separate into two phases. After centrifugation (5000 ϫ g for 5 min), the inorganic phase, and the protein layer was gently removed, and the organic phase was dried under nitrogen at 40°C. The residual lipid extract was redissolved in 200 l of hexane and separated by thin layer chromatography (acid-resistant silica gel TLCfoils, F 1500, from Schleicher & Schuell) using hexane/diethyl ether/ acetic acid (80:20:1, v/v/v) as the developing system. Lipids were recognized by visualizing added standards in iodine vapor and cut out, and their radioactivity was quantified by liquid scintillation counting in an LKB Wallac beta counter (1261 multigamma).

Inhibition of Autophagy by Adenosine: Potentiation by 2Ј-Deoxycoformycin and Suppression by 5-Iodotubercidin-Previ-
ous studies have shown that adenosine and (more potently) some of its N 6 -substituted derivatives, e.g. N 6 -dimethyladenosine and N 6 -methylmercaptopurine riboside, inhibit protein degradation in isolated rat hepatocytes (9). As shown in Fig.  1A, adenosine was able to suppress hepatocytic autophagy (measured as the sequestration of endogenous LDH), which would explain its degradation-inhibitory ability. The effect of adenosine was strongly potentiated by dCF, a specific adenosine deaminase (EC 3.5.4.4) inhibitor (21) that lowered the IC 50 of adenosine from 1 to 0.2 mM. The inhibition of autophagy by adenosine would thus normally seem to be restrained by rapid adenosine deamination.
Incubation in the presence of 10 M ITu, a potent and specific inhibitor of adenosine kinase (16), virtually completely abolished the inhibitory effect of adenosine ϩ dCF on autophagy (Fig. 1B). This absolute requirement for adenosine phosphorylation would suggest that its inhibitory effect on autophagy is mediated by AMP. Measurements of intracellular AMP levels (by HPLC) indeed revealed a 10-fold increase (transiently peaking at 3 mol/g wet mass) after the addition of adenosine ϩ dCF. 2 Inhibition of Autophagy by AICAR and N 6 -Mercaptopurine Riboside-One possible mechanism for the inhibition of autophagy by AMP might be through activation of the AMPK. To investigate this possibility, we incubated hepatocytes with the nucleoside analogue, AICAR, which in its monophosphorylated form (AICA-ribotide, or ZMP) is a potent activator of hepatocytic AMPK (22). As shown in Fig. 2A, AICAR inhibited hepatocytic autophagy in a dose-dependent manner (IC 50 of about 0.3 mM). Its effect on autophagy was completely suppressed by 10 M ITu, consistent with a requirement for phosphorylation to ZMP. N 6 -MPR, an adenosine analogue thiolated at the N 6 position, also inhibited autophagic seqestration in an ITu-sensitive manner suggestive of a phosphorylation requirement (Fig. 2B). N 6 -MPR was more potent than adenosine, perhaps because the thiol group at the N 6 -position prevented its deamination by adenosine deaminase, as previously suggested to explain the potent protein degradation-inhibitory effects of other N 6 -substituted adenosine derivates (9).
The intracellular phosphorylation of AICAR to form ZMP and suppression of this process by ITu were confirmed by reverse phase HPLC analysis of intracellular metabolites in hepatocytes. As shown in Fig. 3A, addition of 0.4 mM AICAR to the hepatocyte suspension resulted in its rapid intracellular accumulation. Accumulated AICAR was effectively phosphorylated to form ZMP, which at 2 h had almost completely replaced AICAR (Fig. 3B). This phosphorylation was suppressed by the addition of 10 M ITu (Fig. 3C), indicating the involvement of adenosine kinase.  Effect of AICAR on the Intracellular ATP Concentration-Autophagic sequestration has previously been shown to correlate positively with intracellular ATP levels (4), suggesting a requirement for energy. To check whether AICAR might inhibit autophagy by altering the intracellular ATP concentration, its effect on hepatocellular ATP levels was measured. As shown in Fig. 4, AICAR concentrations below 0.6 mM had no effect on intracellular ATP levels, although autophagic sequestration was inhibited 80%. At higher AICAR concentrations, some reduction in ATP (about 30% at 1 mM) was observed. Apparently, AICAR does not exert its inhibitory effect on autophagy by altering intracellular ATP.
Effects of Adenosine, AICAR, and N 6 -MPR on Hepatocellular Cholesterol Synthesis-One of the established biological roles of activated AMPK its to phosphorylate and inactivate HMG-CoA reductase (EC 1.1.1.88), the rate-limiting enzyme in cholesterol synthesis (23,24). If adenosine, AICAR, and N 6 -MPR really activate AMPK, they would therefore be expected to inhibit hepatocellular cholesterol synthesis. As shown in Fig. 5, all these compounds inhibited cholesterol synthesis in isolated hepatocytes, with about the same potencies as in their inhibition of autophagy. ITu (10 M) supressed these effects on cholesterol synthesis completely, demonstrating the need for phosphorylated intermediates (AMP, etc.). The results in Fig. 5 thus support the notion that adenosine, AICAR, and N 6 -MPR are able to activate hepatocellular AMPK under the experimental conditions employed.
Effect of Lovastatin and Simvastatin, Inhibitors of HMG-CoA Reductase, on Autophagic Sequestration-The parallel suppression of autophagy and HMG-CoA reductase activity by adenosine and its analogues would be consistent with a role for HMG-CoA reductase in the regulation of autophagy. To examine this possibility, we incubated hepatocytes with lovastatin or simvastatin, two potent and specific drugs that lower cholesterol synthesis by inhibiting HMG-CoA reductase (25). As shown in Fig. 6, neither lovastatin (panel A) nor simvastatin (panel B) exerted any effect on autophagy, although they potently inhibited hepatocellular cholesterol synthesis (IC 50 Ͻ 5 M). HMG-CoA reductase would thus not seem to be involved in regulation of autophagy. DISCUSSION Autophagy, or macroautophagy, is a cellular process that plays a major role in the bulk sequestration and subsequent lysosomal degradation of long-lived cytosolic proteins, ribosomal RNA, and cytoplasmic organelles (3, 26 -28). The initial autophagic sequestration step is highly regulated by growth factors, hormones, and metabolites such as amino acids and adenosine (which can be regarded as negative feedback inhib-itors of autophagy) and by protein phosphorylation (8, 9, 29 -31). A type 2A protein phosphatase (32) as well as the Ca 2ϩ / calmodulin-dependent protein kinase II (31), the cAMPdependent protein kinase (33), and protein tyrosine kinase activity (34) have been implicated in the regulation of autophagy in isolated rat hepatocytes.
Depletion of ATP inhibits autophagic sequestration, suggesting that autophagy is an energy-dependent process (4, 7). However, ATP depletion also leads to increases in intracellular AMP and adenosine (10,12). Because adenine, adenosine, and some of their thiolated or methylated derivates also inhibit autophagic protein degradation (8,9,35,36), the posibility should be considered that at least part of the autophagy-inhibitory effect of ATP depletion might be mediated by adenosine or AMP, perhaps through AMPK.
AMPK is the central component in a multisubstrate protein kinase cascade involved in the regulation of lipid metabolism (37). AMPK is activated in response to elevations of the AMP/ ATP ratio, caused e.g. by hypoxia, environmental stress, star- vation, cell damage, heat, etc. (23,38). Upon its activation, AMPK shuts down major ATP-consuming biosynthetic processes like cholesterol and fatty acid synthesis, thereby preserving ATP for more essential and immediate cellular needs such as the maintenance of ion gradients. It has been suggested than AMPK may serve as a general integrator of metabolic responses to changes in energy availability likely to have regulatory effects extending beyond lipid metabolism (2).
In the present study, the ability of the adenosine kinase inhibitor ITu (16) to abolish the autophagy-suppressive effect of adenosine clearly shows that the suppression requires AMP formation and most likely is mediated by AMP. This effect of ITu would effectively rule out the possibility that adenosine might mediate its effect through purinergic (adenosine) receptors, known to be present at the surface of rat hepatocytes (39,40). The administration of adenosine alone has previously been shown to raise the concentration of AMP in isolated rat hepatocytes (13). Under the conditions used in the present experiments, with the adenosine and adenylate deaminase inhibitor, dCF (21), included to prevent the deamination of adenosine and AMP, a rapid increase in intracellular AMP to very high levels has been demonstrated. 2 The marked synergism between adenosine and dCF in suppressing autophagy would thus be consistent with a mediation by AMP. Both adenosine and dCF are also capable of elevating the hepatocytic levels of other adenosine metabolites, like S-adenosylhomocysteine and (secondarily) S-adenosylmethionine (41,42), but this pathway would be potentiated rather than antagonized by ITu (43). AMP can directly activate or inactivate a number of enzymes, e.g. in glucose/glycogen metabolism (2,13); the ability of AMP to inhibit autophagy does not, therefore, prove that AMPK is involved. However, the equally effective inhibition of autophagy by AICAR would more strongly implicate AMPK. AICAR becomes a potent and specific activator of AMPK after its intracellular phosphorylation to AICA-ribotide or ZMP (14,15,22,44). AICAR was rapidly converted to ZMP under the present conditions, and the autophagy-suppressive effect of AICAR was abolished by ITu, strongly suggesting that ZMP rather than the nucleoside itself was the active autophagy suppressant. Furthermore, the ability of AICAR to virtually completely inhibit hepatocellular cholesterol synthesis, reflecting the inactivating phosphorylation of the rate-limiting enzyme HMG-CoA reductase by AMPK (24,37,45), clearly demonstrated that AMPK was activated by AICAR under our experimental conditions. AICAR, at concentrations that inhibited autophagy by 80%, had no significant effect on hepatocellular ATP levels, in accordance with previous observations (22), nor did it detectably reduce cellular viability. Its effect on autophagy would therefore be likely to be mediated by AMPK activation rather than by energy depletion. N 6 -MPR, an aminothiolated adenosine analogue, inhibited autophagy more potently than did adenosine in our experiments. Like in the case of adenosine and AICAR, its autophagysuppressive effect was abolished by ITu, indicating mediation by a 5Ј-phosphorylated derivative, i.e. an AMP analogue, likely to activate AMPK. However, previous work (15) indicated that N 6 -MPR 5Ј-monophosphate did not activate purified rat liver AMPK, in contrast to AMP and ZMP. To see if the situation might be different in intact cells, we measured the effect of N 6 -MPR on hepatocellular cholesterol synthesis, as an indirect measure of in situ AMPK activity. N 6 -MPR was found to strongly inhibit cholesterol synthesis, and in an ITu-sensitive manner, suggesting that its 5Ј-monophosphate was in fact able to activate AMPK in intact cells. The discrepancy between this observation and the inability of N 6 -MPR 5Ј-monophosphate to activate purified AMPK can perhaps be explained by the com-plex manner in which AMPK is regulated. The direct allosteric activation of AMPK by 5Ј-nucleotides may increase its activity maximally 5-fold (46), whereas its phosphorylation by an upstream AMP-activated protein kinase kinase may cause a more than 20-fold activation (47,48). These two effects can act synergistically to produce, potentially, a more than 50-fold activation of AMPK (48). Because the effects of 5Ј-nucleotides on the two enzymes are independent (49), it is theoretically quite possible for a given nucleotide, like N 6 -MPR, to cause a substantial activation of AMPK by binding to the upstream kinase rather than to AMPK itself.
In the present work, hepatocellular cholesterol synthesis has been used as an indirect measure of AMPK activity. However, the data also demonstrate that all the investigated nucleosides cause a strong suppression of HMG-CoA reductase activity and of cholesterol formation, raising the possibility that the latter effects could be instrumental in mediating the inhibition of autophagy. The effects of lovastatin and simvastatin, two potent and specific inhibitors of HMG-CoA reductase (25), were therefore investigated. Both inhibitors completely blocked cholesterol synthesis while having no effect on autophagy. It would thus seem clear that neither HMG-CoA reductase activity nor de novo cholesterol formation is required for autophagic activity. The effects of adenosine, AICAR, and N 6 -MPR on cholesterol metabolism and on autophagy should probably be regarded as two independent consequences of AMPK activation: HMG-CoA reductase phosphorylation/inhibition and modification of another as yet unknown protein.
In conclusion, the present results suggest that hepatocytic autophagy can be suppressed by AMP through activation of AMPK. The adenosine-antagonistic effect of ITu would rule out a direct regulation by adenosine, making it unlikely that adenosine (an RNA degradation product) can function as a physiological feedback inhibitor of autophagic degradation. Although 0.5 mM added adenosine could elevate intracellular AMP levels dramatically in the presence of dCF, physiological adenosine concentrations would probably be too low to have much regulatory effect in the absence of inhibitors. On the other hand, high levels of AMP can be reached under hypoxia and other conditions of energy depletion (10 -13), which also cause a suppression of autophagy (4 -7). The physiological relevance of the regulation of autophagy by AMP/AMPK may therefore be to shut down this energy-requiring process under conditions of energy depletion.