High-glucose toxicity is mediated by AICAR-transformylase/IMP cyclohydrolase and mitigated by AMP-activated protein kinase in Caenorhabditis elegans

The enzyme AICAR-transformylase/IMP cyclohydrolase (ATIC) catalyzes the last two steps of purine de novo synthesis. It metabolizes 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR), which is an AMP analogue, leading to activation of AMP-activated kinase (AMPK). We investigated whether the AICAR–ATIC pathway plays a role in the high glucose (HG)–mediated DNA damage response and AICAR-mediated AMPK activation, explaining the detrimental effects of glucose on neuronal damage and shortening of the lifespan. HG up-regulated the expression and activity of the Caenorhabditis elegans homologue of ATIC, C55F2.1 (atic-1), and increased the levels of reactive oxygen species and methylglyoxal-derived advanced glycation end products. Overexpression of atic-1 decreased the lifespan and head motility and increased neuronal damage under both standard and HG conditions. Inhibition of atic-1 expression, by RNAi, under HG was associated with increased lifespan and head motility and reduced neuronal damage, reactive oxygen species, and methylglyoxal-derived advanced glycation end product accumulation. This effect was independent of an effect on DNA damage or antioxidant defense pathways, such as superoxide dismutase (sod-3) or glyoxalase-1 (glod-4), but was dependent on AMPK and accumulation of AICAR. Through AMPK, AICAR treatment also reduced the negative effects of HG. The mitochondrial inhibitor rotenone abolished the AICAR/AMPK-induced amelioration of HG effects, pointing to mitochondria as a prime target of the glucotoxic effects in C. elegans. We conclude that atic-1 is involved in glucotoxic effects under HG conditions, either by blocked atic-1 expression or via AICAR and AMPK induction.

The enzyme AICAR-transformylase/IMP cyclohydrolase (ATIC) catalyzes the last two steps of purine de novo synthesis. It metabolizes 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR), which is an AMP analogue, leading to activation of AMP-activated kinase (AMPK). We investigated whether the AICAR-ATIC pathway plays a role in the high glucose (HG)mediated DNA damage response and AICAR-mediated AMPK activation, explaining the detrimental effects of glucose on neuronal damage and shortening of the lifespan. HG up-regulated the expression and activity of the Caenorhabditis elegans homologue of ATIC, C55F2.1 (atic-1), and increased the levels of reactive oxygen species and methylglyoxal-derived advanced glycation end products. Overexpression of atic-1 decreased the lifespan and head motility and increased neuronal damage under both standard and HG conditions. Inhibition of atic-1 expression, by RNAi, under HG was associated with increased lifespan and head motility and reduced neuronal damage, reactive oxygen species, and methylglyoxal-derived advanced glycation end product accumulation. This effect was independent of an effect on DNA damage or antioxidant defense pathways, such as superoxide dismutase (sod-3) or glyoxalase-1 (glod-4), but was dependent on AMPK and accumulation of AICAR. Through AMPK, AICAR treatment also reduced the negative effects of HG. The mitochondrial inhibitor rotenone abolished the AICAR/AMPK-induced amelioration of HG effects, pointing to mitochondria as a prime target of the glucotoxic effects in C. elegans. We conclude that atic-1 is involved in glucotoxic effects under HG conditions, either by blocked atic-1 expression or via AICAR and AMPK induction.
The enzyme AICAR-transformylase/IMP cyclohydrolase (ATIC) 3 catalyzes the last two steps of purine de novo synthesis. ATIC metabolizes 5-aminoimidazole-4-carboxamide ribonucleoside (AICAR) to N-formyl-5-aminoimidazol-4-carboxamide ribonucleotide (FAICAR) and then to inosine monophosphate (IMP), an analog of AMP. IMP can then be phosphorylated by adenosine kinase to become ZMP, which can bind to the cystathionine-␤-synthase domains of the ␥-subunit of AMP-activated protein kinase (AMPK), leading to an allosteric change (1). This change makes AMPK a better substrate for its upstream kinases to phosphorylate it at Thr-172 and inhibits dephosphorylation at this site by the protein phosphatases PP2A and PP2C (2,3). This combined effect significantly increases the activity of AMPK ex vivo (4).
Treatment with AICAR has been shown to prevent and/or reverse metabolic syndrome in animal models. In ob/ob mice, fa/fa rats, as well as rats fed a high-fat diet, AICAR treatment has been shown to improve glucose tolerance and whole-body glucose disposal as well as reduce hepatic glucose output and plasma triglycerides and free fatty acids levels (5)(6)(7)(8). In streptozotocin-induced diabetic mice, treatment with AICAR increased AMPK activity within the kidney and was linked to reduction in the glomerular matrix and albuminuria (9). Exogenous AICAR can therefore be considered to be an activator of AMPK in vitro. However, the regulation of ATIC by endogenous AICAR and its relationship to the activation of AMPK remain unknown, particularly in the context of hyperglycemia and diabetes.
The regulation of AMPK is of great interest in the study of diabetes and metabolic syndrome, as evidence would suggest that loss and/or reduction of AMPK signaling plays an important role in the development of insulin resistance. Upon activa-tion, AMPK signals through its downstream substrates to restore normal energy levels by stimulating metabolic processes that generate ATP, such as fatty acid oxidation, or by inhibiting those that use ATP, such as triglyceride and protein synthesis (10). The formation of reactive metabolites, such as reactive oxygen species (ROS) and methylglyoxal (MG), are closely linked to energy homeostasis. The normalization of energy levels, induced by AMPK activation, would therefore provide an indirect means of defense against reactive metabolites. Indeed, it has been shown that activation of AMPK can decrease the production of ROS by improving mitochondrial dysfunction in vitro and in vivo as well in patients with diabetes (11)(12)(13)(14)(15)(16). Treatment with AICAR was shown to rescue mitochondrial biogenesis, pyruvate dehydrogenase activity, and mitochondrial complex activity in diabetic kidneys (9). Interestingly, treatment with AICAR also normalized the production of superoxide, which was found to be reduced in the diabetic kidney. This would be consistent with the concept of mitochondrial hormesis, or mitohormesis, in which a degree of ROS production is required within a given biological system to improve systemic defenses against such cellular stressors by inducing an adaptive response (11).
Several studies have shown that DNA damage from reactive metabolites (17)(18)(19)(20) plays an important role in the development of late diabetic complications (17,18,(21)(22)(23)(24). Increased damage to the DNA would result in an increased demand on repair processes, in particularly the purine building blocks required for the DNA structure. Within the context of diabetes, this is particularly relevant, as there is evidence to suggest that diabetic patients have inefficient DNA repair (25,26). Within the context of the ATIC-AICAR pathway, it is conceivable that the activation of ATIC would be protective, as it would provide the necessary pool of purines required for DNA repair. However, in doing so, the endogenous AICAR would be depleted, leading to the loss of AMPK activation and the subsequent beneficial effects of reducing the production and/or consequences of reactive metabolites. The imbalance between the two functions of the ATIC-AICAR pathway may therefore represent an important mechanism for development of the cellular dysfunction observed under hyperglycemic conditions.
In this study, the effect of high glucose on ATIC as well as the effect of AICAR on the activation of AMPK and the production of reactive metabolites and DNA damage was studied. As both AMPK and ATIC are conserved throughout evaluation, Caenorhabditis elegans was used as an easily accessible model system.

Results
The C. elegans homolog of atic-1, C55F2.1, was identified by in silico analysis. To verify its enzymatic function, a transgenic nematode (tgC55F2.1b) was created, and the activity of AICAR formyltransferase (AICARFT) was determined. This activity was increased in tgC55F2.1b from 17.2 Ϯ 2.9 mol/min to 34.2 Ϯ 2.1 mol/min (p ϭ 0.003) compared with the WT, which corresponds well to the activity of the recombinant human protein (to 43.5 Ϯ 1.8 mol/min, p ϭ 0.010) (Fig. 1A). A putative methylglyoxal synthase domain was identified within C55F2.1; however, neither the protein extracts of the tg nema-todes nor a commercially available purified enzyme of the human homolog enhanced MG production (Fig. 1B).
Stimulation of WT nematodes with high glucose (HG), for 5 days, lead to an increase in the expression of atic-1 of 2.4-fold, as compared with standard (S) conditions (p Ͻ 0.049) (Fig. 1C). This increase in expression was found to persist up to 12 days under HG conditions; however, the effect was not dose-dependent with respect to glucose and not as a consequence of osmotic effects (Fig. S1). The increased expression observed at 5 days of HG was reflected by an increase in AICARFT activity of approximately 2.7-fold compared with S conditions (Fig. 1D).

Effect of atic-1 overexpression
The effect of atic-1 (C55F2.1b) overexpression on lifespan and neuronal damage was determined under standard (S) and HG conditions. Overexpression of atic-1 decreased the median lifespan from 23.6 Ϯ 0.9 days to 20.6 Ϯ 0.6 days (p ϭ 0.004) compared with control nematodes under S conditions ( Fig. 2A). Under HG, the head motility of the WT nematodes was decreased by 0.03 Ϯ 0.01 mm/s (p ϭ 0.006) compared with S conditions (0.016 Ϯ 0.02 m/s). Overexpression of atic-1 decreased head mobility non-significantly under the S condition by 0.03 Ϯ 0.02 mm/s compared with WT nematodes under S conditions. HG treatment of transgenic nematodes further reduced head motility by 0.08 Ϯ 0.01 mm/s (p ϭ 0.012) compared with WT nematodes under HG conditions (Fig. 2B). Thus, the effect of HG on head motility was exaggerated in atic-1-overexpressing C. elegans.
With respect to neuronal damage, HG increased the neuronal damage score from 0.5 Ϯ 0.13 to 1.3 Ϯ 0.23 (p ϭ 0.010) in WT nematodes. Overexpression of atic-1 increased the damage score to 1.2 Ϯ 0.16 (p ϭ 0.070) compared with WT nematodes under S conditions. The damage score in atic-1 transgenic nematodes was not enhanced by HG (p ϭ 0.761) (Fig. 2, C and D). Thus, atic-1 overexpression under S conditions was able to partially reduce the lifespan, neuronal function, and integrity and increased C55F2.1b mRNA expression (Fig. S2).

Protective effect of atic-1 silencing
The up-regulation of atic-1 in part mimicked the effects of HG. Therefore, protection from the effect of HG would be expected when the glucose-dependent induction of atic-1 is inhibited. RNAi treatment of atic-1 was verified by quantitative PCR and decreased mRNA expression to 24.9% Ϯ 11.3% (p ϭ 0.003) under S conditions and to 35.6% Ϯ 14.3% (p ϭ 0.011) under HG conditions compared with WT nematodes (Fig. S3). Down-regulation of atic-1 increased the lifespan under HG conditions from 11.0 Ϯ 0.5 days to 14.7 Ϯ 0.9 days (p Ͻ 0.0001) (Fig. 3A), whereas there was no effect under S conditions (p ϭ 0.427, data not shown). The head motility of control RNAitreated nematodes was decreased under HG conditions, from 0.15 Ϯ 0.01 mm/s to 0.10 Ϯ 0.01 mm/s (p ϭ 0.019). Under S conditions, down-regulation of atic-1 had no effect on head motility compared with control RNAi-treated nematodes. Under HG conditions, down-regulation of atic-1 prevented the decrease in head motility observed in control RNAi-treated nematodes (Fig. 3B). Consistent with the previous finding, HG

Protective effect of AICAR under high glucose
increased the neuronal damage score from 0.82 Ϯ 0.07 to 1.42 Ϯ 0.06 (p ϭ 0.003) in control RNAi nematodes, which was prevented by down-regulation of atic-1 (Fig. 3C).
With respect to the formation of ROS and dicarbonyls, HG increased ROS formation in control RNAi nematodes from 59 Ϯ 13 AU/pixel to 222 Ϯ 9 AU/pixel (p ϭ 0.0001), which was prevented by down-regulation of atic-1 (Fig. 3D). Similar effects were also observed for the formation of MG-derived AGEs (Fig. 3E).
Because AICAR is the substrate of ATIC, it was hypothesized that the beneficial effect resulting from the down-regulation of atic-1 was mediated by an accumulation of AICAR. There was no significant effect of HG on intracellular AICAR concentration (p ϭ 0.074); however, the amount of AICAR was close to the limit of detection. Interestingly, the down-regulation of atic-1 increased intracellular AICAR by approximately 8-fold under S conditions and approximately 13.1-fold under HG conditions compared with control nematodes. There was no significant difference in intracellular AICAR concentration between the S and HG conditions in atic-1 RNAi-treated nematodes. This might indirectly reflect the basal turnover of AICAR by atic-1, suggesting that the baseline atic-1-dependent AICAR turnover is saturated. This would be consistent with the findings observed with overexpression of atic-1, which did not show any effect on intracellular AICAR under S or HG conditions compared with control nematodes (Fig. 4A).
The inability of HG to reduce AICAR significantly is consistent with the failure of HG to induce DNA double-strand breaks (ds-breaks), as measured by the comet assay. Using bleomycin-treated WT nematodes as a positive control (Fig. 4B), it was found that there was no significant difference in ds-break formation between S (Fig. 4C) and HG (Fig. 4D) conditions (Fig.  4E, third column). Overexpression of atic-1 did not reduce the formation of ds-breaks compared with the control under S conditions (p Ͼ 0.05) (Fig. 4E, fourth column) but did reduce dsbreaks under HG conditions to 5.6% (p ϭ 0.002) (Fig. 4E, fifth column). Under control RNAi conditions, ds-breaks decreased under S conditions (7.1%, p ϭ 0.003) (Fig. 4E, sixth column), whereas HG did not induce ds-breaks (p ϭ 0.600) (Fig. 4E, seventh column). Down-regulation of atic-1 increased the formation of ds-breaks under S conditions by 21.9% (p ϭ 0.003) (Fig. 4E, eighth lane), but for unknown reasons, glucose did not

Protective effect of AICAR under high glucose
increase ds-breaks under HG conditions (14.7%, p Ͼ 0.05) (Fig.  4E, ninth lane). This was consistent with HG not exerting its toxic effect in nematodes via genomic DNA damage. Nevertheless, it can be concluded that blocking atic-1 expression enhanced DNA damage under S conditions, whereas overexpressing atic-1 reduced DNA damage more under HG. However, as there were no significant differences between S and HG in WT nematodes, HG-mediated atic-1 induction and DNA damage do not seem to be important pathways when explaining the atic-1 induction related to the effects of HG.

aak-2-dependent effect of atic-1 silencing
To establish the potential mechanism for the effect of atic-1, the downstream signaling pathway was investigated using aak-2 nematodes, a homolog of the catalytic subunit of AMPK. It was found that the effects resulting from the down-regulation of atic-1 under HG conditions were dependent on aak-2, as there was no effect resulting from down-regulation of atic-1 on lifespan in aak-2 nematodes (Fig. 5A). Furthermore, HG-decreased head motility was not improved (Fig. 5B). With respect to the formation of reactive metabolites, ROS formation was increased under HG conditions, from 60 Ϯ 10 AU/pixel to 221 Ϯ 3 AU/pixel (p Ͻ 0.0001), in aak-2 nematodes and was unaffected by down-regulation of atic-1 (Fig. 5C). Similar results were also found for MG-derived AGEs (Fig. 5D). Thus, lack of AMPK abrogated the positive effects of atic-1 suppression on lifespan, head motility, and reactive metabolite accumulation. Head motility under S conditions is defined as 0. *, p Ͻ 0.05; **, p Ͻ 0.01; n.s., not significant. C, neuronal structure was visualized by a pan-neuronal-specific GFP reporter and scored in a blinded procedure as described under "Experimental procedures." Data represent mean Ϯ S.D. of four independent experiments, each with 12 nematodes per group. **, p Ͻ 0.01, calculated using unpaired Student's t test. D, neuronal structure was identified by a pan-neuronal-specific GFP reporter. Shown are representative pictures of a single nematode.

Protective effect of AICAR under high glucose Protective effect of AICAR
As the suppression of atic-1 increased AICAR levels ( Fig. 4A and Fig. S4), a pharmacological approach was used to better understand the AICAR effect. Stimulation of WT nematodes with HG for 5 days led to a decrease in AMPK activity, an effect that was partially normalized by the exogenous addition of 1 mM AICAR (Fig. 6A).
The HG-mediated decrease in lifespan was also partially normalized by treatment with AICAR, improving the lifespan from 14.0 Ϯ 0.6 days to 20.9 Ϯ 0.8 days (p ϭ 0.0014) (Fig. 6B, first group). This effect was lost when aak-2 was knocked out (p ϭ 0.6728) (Fig. 6B, second group), whereas loss of the MG-and ROS-detoxifying enzymes glod-4 (p ϭ 0.002) (Fig. 6B, third group) and sod-3 (p ϭ 0.0025) (Fig. 6B, fourth group) did not affect prolongation of the lifespan by AICAR. This would suggest that AICAR acts independent of direct detoxification of ROS and dicarbonyls.
Suppression of AMPK by RNAi prevented the AICAR improvement in lifespan under HG conditions (Fig. 6C, first group). Similar effects were also observed in glod-4 (Fig. 6C, second group) and sod-3 knockouts (Fig. 6C, third group). Furthermore, double knockout of glod-4 and sod-3 did not affect the lifespan prolongation by AICAR under HG conditions (Fig.  6D). Similar results were also found for head motility (Fig. 6E). HG decreased head motility in the WT to 0.14 Ϯ 0.01 mm/s (p ϭ 0.0239), which could be almost normalized by additional AICAR treatment to 0.18 Ϯ 0.01 mm/s (p ϭ 0.0398), whereas there was no effect under S conditions (p ϭ 0.3364) (Fig. 6E, first group). This effect of AICAR was lost when aak-2 was knocked out (p ϭ 0.7611) (Fig. 6E, second group), but loss of glod-4 (p ϭ 0.0442) (Fig. 6E, third group) and sod-3 (p ϭ 0.0138) (Fig. 6E, fourth group) did not affect the AICAR effect on head motility under HG conditions. Neuronal damage was also not affected in a similar manner. The HG-mediated increase in the neuronal damage score to 1.29 Ϯ 0.12 (p Ͻ 0.001) was normalized by additional AICAR treatment to 0.59 Ϯ 0.13 (p ϭ 0.0007), whereas there was no effect under S conditions (Fig. 6F, first group). This effect was again lost when aak-2 was down-regulated (p ϭ 0.0758) (Fig.  6F, second group), whereas loss of the MG-and ROS-detoxifying enzymes glod-4 (p ϭ 0.0002) (Fig. 6F, third group) and sod-3 (p ϭ 0.0002) (Fig. 6F, fourth group) did not affect the AICAR action of structural damage under HG. Thus, the positive effects of AICAR on head motility, neuronal damage, or lifespan were affected when either glod-4 or sod-3 was missing.
Similar data were also obtained with respect to the accumulation of reactive metabolites. ROS formation in WT nematodes was increased to 205.9 Ϯ 8.6 AU/pixel (p Ͻ 0.001) under HG conditions, and treatment with AICAR partially normalized it to 49.4 AU/pixel (p Ͻ 0.0001), whereas there was no effect under S conditions (p ϭ 0.9933) (Fig. 6G, first group). The effect of AICAR was lost when aak-2 was knocked out (p ϭ 0.8616) (Fig. 6G, second group), but loss of the MG-and ROS-detoxifying enzymes glod-4 (p Ͻ 0.0001) (Fig.  6G, third group) and sod-3 (p Ͻ 0.0001) (Fig. 6G, fourth group) did down-regulation on lifespan of ctrl (solid line) and atic-1 RNAi (dashed line) feeding plates under HG conditions. One representative experiment of three independent experiments is shown, each with 60 nematodes per group. d, day. B, head motility was determined by video analysis as described under "Experimental procedures" under control RNAi or atic-1 RNAi, each under S and HG conditions. Data represent mean Ϯ S.D. of three independent experiments, each with 15 nematodes per group. C, neuronal structure was visualized by a pan-neuronal-specific GFP reporter and scored in a blinded procedure as described under "Experimental procedures" under S or HG conditions with control or atic-1 RNAi. Data represent mean Ϯ S.D. of three independent experiments, each with 12 nematodes per group. D, ROS formation was detected by confocal laser-scanning microscopy of ethidium-labeled C. elegans as described under "Experimental procedures" under S and HG conditions. Data represent mean Ϯ S.D. of three independent experiments, each with 20 nematodes per group. E, MG-derived AGE accumulation was detected by immunostaining and quantified as described under "Experimental procedures." Data represent mean Ϯ S.D. of three independent experiments, each with 20 nematodes per group. *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; calculated using unpaired Student's t test. n.s., not significant.

Protective effect of AICAR under high glucose
not affect the AICAR action of ROS formation. Similar results were also obtained for MG-derived AGEs (Fig. 6H).
Treatment with AICAR not only affected sod-3 and glod-4 independent of the accumulation of reactive metabolites but also the expression of atic-1 mRNA (Fig. S5). This effect was more pronounced under HG than under S conditions but was most importantly lost in the absence of aak-2. Furthermore, knockout of aak-2 and sod-3 had no effect on the

Protective effect of AICAR under high glucose
formation of ds-breaks under S and HG conditions (Fig. 6), suggesting that there is no effect on purine synthesis. Thus, the effects of AICAR on all parameter studied were aak-2 dependent. Furthermore, as the effect of AICAR treatment on all parameters studied was independent of glod-4 and sod-3, mitochondrial targeting by AICAR was studied to better understand the mechanism underlying the effect of AICAR.

The effect of AICAR is independent of antioxidant defense pathways
HG-mediated lifespan reduction in WT nematodes (Fig.  S7A, first group) was inhibited by AICAR as well as rotenone. However, co-administration of AICAR plus rotenone did not affect the lifespan, even under HG conditions. Thus, rotenone alone can substitute for AICAR under HG conditions. It is likely that the rotenone effect is distal from AICARmediated AMPK activation because, in the absence of aak-2 (Fig. S7A, second group), the protective effect of AICAR was lost, whereas the effect of rotenone was not abrogated. Nevertheless, the effect of rotenone, as was the case with AICAR, was independent of the two major detoxifying enzymes glod-4 (Fig. S7A, third group) and sod-3 (Fig. S7A, fourth  group). The hypothesis that the AICAR-and AMPK-dependent effect on lifespan was mediated by mitochondrial ROS production (and not detoxification) was supported by the effect of the antioxidant BHA, which prolonged the lifespan of the WT (Fig. S7B, first group) and could even, in the absence of aak-2 (Fig. S7B, second group), prolong the lifespan, which was also independent of glod-4 (Fig. S7B, third group) and sod-3 (Fig. S7B, fourth group). The life-prolonging effect of AICAR was independent of glod-4 (p ϭ 0.492) and sod-3 (p ϭ 0.274) but dependent on aak-2 (p ϭ 0,040) (Fig. 7A). Further, the lifespan-prolonging effect of rotenone was independent of aak-2 (p ϭ 0.254), glod-4 (p ϭ 0.390), and sod-3 (p ϭ 0.276) (Fig. 7B). Further treatment of AICARtreated nematodes with rotenone had no effect on the lifespan and was independent of aak-2 (p ϭ 0.276), glod-4 (p ϭ 0.291), and sod-3 (p ϭ 0.237) (Fig. 7C). The effect of BHA was independent of aak-2 (p ϭ 0.299), glod-4 (p ϭ 0.118), and sod-3 (p ϭ 0.334) (Fig. 7D). Combined treatment with AICAR and BHA did not improve the lifespan further and

Protective effect of AICAR under high glucose
was independent of aak-2 (p ϭ 0.067), glod-4 (p ϭ 0.230), and sod-3 (p ϭ 0.174) (Fig. 7E). It is important to note that the effects of rotenone and BHA were exclusively present in HG-treated nematodes, indicating that not basal mitohormesis-enabling ROS but, rather, HG-induced excessive ROS production is responsible for the HG effects observed.

Discussion
This study identified C55F2.1 as the C. elegans homolog for the mammalian enzyme ATIC. In contrast to mice and humans (27,28), where HG is associated with increased genomic DNA damage, especially DNA double strand breaks but also interchain cross-links and DNA-protein cross-links (17), there was no evidence of genomic DNA damage being significantly associated with the glucose effects in the model under HG conditions studied. Thus, the HG induced up-regulation of atic-1 is not needed for improving DNA repair in response to HG, and a possible physiological benefit of an atic-1 overexpression remains unknown. However, only genomic DNA damage was studied, which, in the nematode, is dependent on the RecA homolog (29), which does not exclude that mitochondrial DNA damage might play a role in HG-mediated complications, too. Previous studies have shown, in C. elegans, that reactive metabolites preliminarily target mitochondrial proteins (30 -34) and that mitochondrial DNA repair involving exo-3 is important in the regulation response to increased production of reactive metabolites (19). Future studies are needed to address the role of mitochondrial DNA damage and repair in the HG-exposed nematode.
Nevertheless, inhibiting atic-1 expression resulted, under HG conditions, in unchanged development of DNA double strand breaks compared with S conditions and an increase of ds-breaks compared with ctrl RNAi conditions. This indicates that, even in the nematode, not only the RecA homolog (29) but also purine synthesis is important. Under HG conditions, baseline purine synthesis is sufficient to ensure baseline DNA repair.
The findings of this study are more consistent with an important role of the ATIC-AICAR pathway in AMPK regulation of mitochondrial dysfunction, resulting in the generation of reactive metabolites and a subsequent increase in neuronal dysfunction and/or damage and reduction of the lifespan. This is supported by three lines of evidence. First, the effect of atic-1 inhibition and the effect of AICAR administration were entirely dependent on the action of aak-2 and independent of the classical ROS-scavenging and MG-detoxifying activities of sod-3 and glod-4. Second, rotenone, targeting mitochondrial complex I, could normalize the lifespan (35) independent of aak-2, sod-3, and glod-4. Thus it could be hypothesized that AICAR/AMPK normalizes mitochondrial dysfunction in the presence of HG because AICAR addition could not improve the lifespan in the presence of rotenone. Third, the antioxidant BHA also improved the effects of HG-induced events. In the presence of BHA, neither AICAR nor aak-2, sod-3, or glod-4 were required for defense against the HG effects.
These findings would suggest that rotenone is able to reduce mitochondrial superoxide formation, protecting against the effects of HG-induced metabolite accumulation and neuronal damage. This is in contrast to a study in mice that showed that mitochondrial ROS formation inhibited AMPK and reduced streptozotocin diabetes-induced renal damage (9). A possible explanation for this difference could be the very fine hometic ROS response in different organisms (11).
Although a certain dose of ROS might be protective in one model system, it might be detrimental in another (14). This holds true in C. elegans because the protective ROS signal requires aak-2 (36,37). A potential master regulator for this might be daf-2, which also regulates AMPK (38), consistent with the concept of mitohormesis causing an adaptive, lifeprolonging effect involving crosstalk with L-proline metabolism (37). Future studies are required to determine the relative contribution of this cross-talk to the protective effect observed.
Such cross-talk would also affect the complex cell-cell interactions that occur in protecting against the effects of HG. Down-regulation of atic-1 by RNAi treatment resulted in protection of head motility and structural neuronal damage despite the fact that RNAi treatment does not reach neuronal cells, as shown previously by others and in our model system (39,40). Thus, non-neuronal cells affected by RNAi could send sufficient warning signals to neuronal cells to prepare to defend themselves against glucotoxic effects. Alternatively, HG could change the function of non-neuronal cells, which, after having acquired a dysfunctional phenotype, send signals to neuronal cells, leading to their damage. Future studies would be required to distinguish between these two possibilities and determine the cell type responsible for the RNAi effects that have been described.
The induction of atic-1 by HG did not result in depletion of AICAR. The physiological capacity of atic-1 in C. elegans seems to be sufficient to ensure purine synthesis, and the basal concentration of AICAR was not affected by HG. In addition, the physiological concentration of AICAR measured in the nematodes was at the lower detection limit and far below the AICAR concentrations needed to pharmacologically induce AMPK activity (41). Nevertheless, even though not important for the system studied here, downregulation of atic-1 increased AICAR, consistent with a constant basal turnover of AICAR by atic-1 and the defense against HG seen when pharmacological doses of AICAR were used. This subsequently resulted in a reduction of ROS (21), MG-dependent AGEs, and neuronal dysfunction. It remains unknown whether the AICAR-mediated effects of atic-1 down-regulation are targeting mitochondrial complex I alone or whether the AICAR-mediated reduction of atic-1 mRNA expression is also involved. However, it can be concluded that the major effect of the ATIC-AICAR pathway is toward the correction of mitochondrial dysfunction. Interestingly, loss of sod-3 or glod-4 did not impair the ATIC-AICAR effect in this setting, compatible with a model of protection against ROS and dicarbonyls not on the level of detoxification but, rather, by addressing the mitochondrial dysfunction-dependent generation of these metabolites.

Cloning and bacterial overexpression of atic-1
The ATIC homologue C55F2.1b was cloned using cDNA of wildtype C. elegans (RNeasy Micro Kit, Qiagen, Hilden, Germany; First Strand cDNA Synthesis Kit for RT-PCR, Roche Diagnostics, Mannheim, Germany) with primers carrying a

Protective effect of AICAR under high glucose
BamHI and KpnI restriction site, respectively (forward primer, 5Ј-CGA GGA TCC TGA AAT GAC CGA CGG AAA ATC AC-3Ј; reverse primer, 5Ј-AAA GGT ACC GTG TGG AAG AGA CGA AGT CCA GT-3Ј). The 1955-bp fragment was then ligated in-frame into the vector pVH10.05 (kindly provided by H. Hutter) between the neuronal F25.B.3 promotor and the GFP-tag for neuronal overexpression. For ubiquitous overexpression, the ribosomal promotor Prps-5 was cloned via PstI/ BamHI from the vector L4453 (kindly provided by Andrew Fire, Addgene plasmid 1655) into the plasmid pPB95.85 (Andrew Fire, Addgene plasmid 1498). In a second step, the PCR product was ligated in-frame into the vector between the ribosomal promotor and the GFP tag via BamHI/KpnI. Restriction enzymes and ligase were used following the manufacturer's instructions (New England Biolabs, Frankfurt, Germany). Before further using the plasmids, the sequence was confirmed by DNA sequencing.
C. elegans were cultured on 60-mm nematode growth medium plates on an OP50 lawn (Caenorhabditis Genetics Center, University of Minnesota) and transferred to 5-fluorodesoxyuridine plates (300 g/ml 5-fluorodesoxyuridine, Sigma-Aldrich, Taufkirchen, Germany) on day 1 of adulthood for the respective experiments. The feeding method was used for the evaluation of RNAi-treatment (I/C plates; nematode growth medium plus 1 mM isopropyl-␤-D-1-thiogalactopyranoside and 25 g/ml carbenicillin) with Escherichia coli HT115-expressing RNAi for aak-2, atic-1, glod-4, sod-3, or control RNAi, respectively (provided by the Fire Lab C. elegans vector kit (principal investigator Andrew Fire, Addgene, Cambridge, MA)). The nematodes were transferred to new plates daily in the RNAi experiments.
For high glucose stimulation, the worms were treated daily with 150 l of a 400 mmol/liter glucose solution for 5 days. One hundred fifty microliters of 400 mmol/liter glucose solution achieved an intracellular glucose concentration of 14 mmol/ liter in the nematode (31).

Determination of lifespan
In lifespan assays, the nematodes were cultivated on 5-fluorodesoxyuridine or I/C plates under S or HG conditions (30), leading to glucose concentrations of 5.5 mmol/liter (99 mg/dl) and 13 mmol/liter (234 mg/dl) in whole-worm extracts, respectively (31). Where indicated, 150 l of AICAR, BHA, and rotenone were added at a concentration of 1 mM AICAR, 25 M BHA, and 100 nM rotenone, respectively. The nematodes were regarded as dead when they did not move after repeated stimuli. Animals were excluded when they crawled away from the plate, crept into the agar, or contained internally hatched larvae. Experiments were performed at least in triplicates with 60 animals each.
The lifespan experiments with tgATIC mutant nematodes (VH2190) were performed at 15°C to exclude pha-1-dependent effects. Nematodes without fluorescence were used as controls. The experiments with the GFP-expressing strain NW1229 were performed at 25°C to maintain the selected pressure. All other experiments were performed at 20°C.

Evaluation of neuronal damage
Structural damage was determined using a semiquantitative four-staged classification scheme in a blinded procedure (47). The pan-neuronal GFP-expressing strain NW1229 (45) Protective effect of AICAR under high glucose and the pan-neuronal atic-1::GFP fusion protein overexpressing strain VH2202 were visualized by fluorescence microscopy. At least 20 nematodes were scored (0: healthy, no damage; 1: minor damage; 2: major damage; 3: extended loss of neuronal structure and dead animals, these were excluded). To assess head motility as a functional parameter, video analysis was performed at day 12 of adulthood (Moticam 1000, Beyersdörfer, Mandelbachtal, Germany). Relative head motility was calculated with a worm tracking software (WormTracker 4.0, Thomas Bornhaupt, Neustadt/Weinstraße).

Quantification of ROS and MG-derived AGEs
ROS were detected by oxidation of the O 2 -sensitive hydroethidine to ethidium (48). MG-derived AGEs were detected by immunostaining with a mouse primary antibody (AGE06B, Biologo, Kronshagen, Germany) and visualized by an Alexa Fluor-labeled goat secondary antibody (Invitrogen, A11002, Thermo Fisher, Rockford, IL). ROS and MG-derived AGEs were quantified by confocal laser-scanning microscopy (30) and calculated by mean pixel intensities with ImageJ software (49). ROS measurement by quantification of H 2 O 2 concentration was shown to give similar results before (48).

Quantification of DNA damage
DNA damage was determined by alkaline comet assay (50). Nematodes were treated with HBSS buffer (1ϫ Hanks' balanced salt solution (Sigma-Aldrich), 20 nM EDTA, and 10% DMSO (pH 8)) and triturated with a glass pistil to isolate the C. elegans cells. The isolated cells were mixed with low-melting-point agarose, and the mixture was loaded on an agarcoated microscope slide. The slides were immersed in ice-cold lysis solution (2.3 M NaCl, 100 mM EDTA, 10 mM Trizma base, 1% Triton X-100, and 10% DMSO) at 4°C for 3 h. Electrophoresis was performed for 25 min at 24 V in electrophoresis buffer (33 mM NaoH and 200 mM EDTA). After electrophoresis, the slides were washed with neutralization buffer (0.4 M Tris) and stained with 4Ј,6-diamidino-2-phenylindole (1 g/ml). Quantification (51-53) was performed with a fluorescence microscope.

Determination of AICAR levels in C. elegans
AICAR was determined in whole-worm extracts. A pellet with ϳ800 nematodes was homogenized in M9 medium (22 mM KH 2 PO 4 , 42 mM Na 2 HPO 4 , and 86 mM NaCl (pH 7)) by using a Bullet Blender Blue (Next Advance, Inc., Burden Lake Road, NY). Protein concentration was determined by using bovine serum albumin as standard. Before analysis, the samples were diluted with distilled water to a total protein concentration of 1 mg/ml. To a 180-l aliquot of diluted C. elegans was added 20 l of 50 M Thymin-d4 internal standard (Cambridge Isotopes, Tewksbury, MA); this was filtrated with a centrifuge filter (Merck Millipore, Darmstadt, Germany) with a pore size of 0.1 m. On a Quattro Ultima triple quadrupole mass spectrometer (Micromass, Manchester, UK) equipped with an electrospray ion source and a Micromass MassLynx data system (according to Hartmann et al. (54) with some modifications and the admission of AICAR), we performed liquid chromatogra-phy. The optimized multiple reaction monitoring experiment was performed on the most abundant ion transition (m/z 259 -127), which was identified by direct infusion of AICAR (Sigma-Aldrich). Argon as collision gas (collision with an energy of 14 electron volt) was operated in positive ion mode with a needle voltage of 3.15 kV. The system was equipped with a C18 2.0 ϫ 4 mm precolumn cartridge (Phenomenex, Aschaffenburg, Germany) and a Phenomenex Aqua C18 column (2.0 ϫ 250 mm, 5-m particle size). The chromatographic run was performed at 100 l/min with a gradient profile between 0.05 M acetic acid (pH 2.8) (eluent A) and 0.05 M acetic acid (pH 2.8) and methanol (1:1, v/v) (eluent B). The gradient started at 100% A, held isocratic for 2.0 min, increased to 100% B over 8 min, switched to 100% A over 1.5 min, and re-equilibrated for 8.5 min at 100% A. 20 min was the overall run time, and 20 l was the injection volume. Concentrations were calculated by signal AICAR toward AICAR signal into internal standard signal and a fourpoint calibration curve (0 -0.5 M).

Determination of AMPK activity
AMPK activity was determined in whole-worm extracts by an AMPK activity assay (catalog no. CY-1182, CycLex) (55). A pellet with ϳ1000 nematodes was homogenized in lysis medium (50 mM HEPES, 50 mM KCl, 1 mM EDTA, 1 mM EGTA, 5 mM ␤-glycerol phosphate, protease inhibitor mixture (Complete, Roche), 50 mM NaF, 1 mM sodium orthovanadate, 5 mM sodium pyrophosphate, and 0.2 mM phenylmethylsulfonyl fluoride) by using a Bullet Blender Blue (Next Advance, Inc.). Protein concentration was determined by using bovine serum albumin as standard. A 10-l sample was mixed with 90 l of kinase reaction buffer (50 M ATP) and incubated at 37°C for 45 min. The wells were washed five times with wash buffer. 100 l of anti-phospho-mouse monoclonal antibody was pipetted into the wells and incubated for 30 min at room temperature. The wells were washed five times, 100 l of substrate reagent was added, and then this was incubated for 15 min at room temperature. Then we added 100 l of stop solution to the well and measured absorbance at 450 nm.

Determination of AICAR formyltransferase activity
AICARFT was determined by initial appearance of tetrahydrofolate at 298 nm as described by Black et al. (56), employing a thermostatic spectrometer. The cuvette contained 32.5 M Tris-HCl (pH 7), 5 M ␤-mercaptoethanol, 25 M KCl, 0.101 M 10-formyl-tetrahydrofolate, and enzyme; it was filled up to 0.950 ml under N 2 at 25°C. First the non-enzymatic rate was recorded for 10 min and later abstracted from the initial rate obtained after adding 0.05 ml of 1.01 mM AICAR. All educts were prepared from degassed H 2 O and were N 2 -saturated because the blank rate was very sensitive to the amount of oxygen in the solutions. The reaction mixture for the recombinant C55F2.1 protein was mixed prior to the addition of the human AICARFT/IMPCHase (IMP cyclohydroxylase) enzyme and subsequently mixed again.

Protective effect of AICAR under high glucose
Determination of methylglyoxal production MG concentration was measured in whole C. elegans lysate as described by Tötemeyer et al. (57). The lysate was prepared by using a Bullet Blender Blue (Next Advance, Inc.), transferred to tubes, and stored at Ϫ20°C until the assay was performed. MG was assayed colorimetrically using 2,4-dinitrophenolhydrazine. In the assay, standard (10 l containing 0 -10 nM MG) and sample were added to the wells, and then 70 l of distilled water plus 30 l of 0.1% 2,4-dinitrophenolhydrazine in 2 M HCl was added. After 15 min at room temperature, 140 l of 10% NaOH was added. After a further incubation of 10 min at room temperature, the absorbance was measured at 540 nm. The recombinant ATIC protein (Abnova, Neihu, Taiwan) was mixed with 1 mM dihydroxyacetone phosphate in 50 mM imidazole-HCl (pH 7) and incubated for 30 min at 30°C to produce MG.

Statistical analyses
Statistical analyses were performed with Excel 2010 (Microsoft, Redmond, WA) and StatView 5.0 (SAS Institute, Cary, NC). The difference between two groups was analyzed by unpaired Student's t test, and p Ͻ 0.05 was considered to be significant, p Ͻ 0.01 highly significant, and p Ͻ 0.001 extremely highly significant. Analysis of variance was used for comparisons of multiple groups, and Fisher's protected least significant difference post hoc tests were used for additional betweengroup comparisons.