Sustained kidney biochemical derangement in treated experimental diabetes: a clue to metabolic memory

The occurrence of biochemical alterations that last for a long period of time in diabetic individuals even after adequate handling of glycemia is an intriguing phenomenon named metabolic memory. In this study, we show that a kidney pathway is gradually altered during the course of diabetes and remains persistently changed after late glycemic control in streptozotocin-induced diabetic rats. This pathway comprises an early decline of uric acid clearance and pAMPK expression followed by fumarate accumulation, increased TGF-β expression, reduced PGC-1α expression, and downregulation of methylation and hydroxymethylation of mitochondrial DNA. The sustained decrease of uric acid clearance in treated diabetes may support the prolonged kidney biochemical alterations observed after tight glycemic control, and this regulation is likely mediated by the sustained decrease of AMPK activity and the induction of inflammation. This manuscript proposes the first consideration of the possible role of hyperuricemia and the underlying biochemical changes as part of metabolic memory in diabetic nephropathy development after glycemic control.


Results
Streptozotocin-induced diabetic rats were maintained at a hyperglycemic status for either four (short period) or 12 weeks (long period). The glycemic levels were then controlled by treatment with insulin alone or insulin combined with metformin (100 mg/kg/day) for four (D4INS and D4MET) or 12 weeks (D12INS and D12MET). The control groups consisted of non-diabetic (ND) and non-treated diabetic (D) animals maintained for eight (ND8 and D8), 12 (ND12 and D12), and 24 weeks (ND24 and D24). The diabetic rats achieved glycemic control seven to 14 days after treatment initiation, showing glycemic levels comparable to those of the non-diabetic control group (Fig. 1). Accordingly, the glycated hemoglobin (HbA1c) levels were increased in the diabetic animals and were reduced after treatment was administered ( Table 1). The diabetic animals in the short-and long-period groups presented increased proteinuria after eight, 12, 18 and 24 weeks of hyperglycemia. The treatment of the animals in the short-period group was initiated prior to the onset of proteinuria 24 and prevented its development. In contrast, the treatment of the animals in the long-period group was initiated after the onset of proteinuria and had completely reversed the observed proteinuria at the end of the 24-week period. Tubular injury was detected after 12 weeks of diabetes by measuring the KIM-1 levels in the urine, and these levels returned to normal following treatment. Hyperglycemia led to augmented kidney/body weight ratios, which also returned to normal once glycemia control was achieved ( Table 1).
The active phosphorylated AMPK (pAMPK) levels in the kidney were decreased in hyperglycemic animals at all time points evaluated (eight, 12, and 24 weeks) and were not restored four (short period) or 12 weeks (long period) after glycemic and renal function control was achieved (Fig. 2a,b and c). Accordingly, kidney TGF-β protein expression was increased after eight, 12, and 24 weeks of hyperglycemia. Treatments initiated after four weeks of hyperglycemia were able to normalize the TGF-β levels within 8 weeks (short period), but treatments initiated after 12 weeks of hyperglycemia did not reduce the TGF-β protein expression levels until the end of a 24-week period, even though the glycemic levels and renal function were normal (Fig. 2d,e and 2f). We also examined kidney TGF-β 1 gene expression and found that it was increased after 12 and 24 weeks of hyperglycemia and that glycemic control reversed this increase ( Fig. 2g and h). The persistent reduction in the levels of pAMPK could be due to several factors that have already been reported, such as oxidative stress, decreased sirtuin 1 (SIRT1) activity, and/or hyperuricemia [25][26][27] . We observed that the kidney malonaldehyde levels, which serve as a measure of oxidative stress, were increased in hyperglycemic animals. In the short-period group, treatment with insulin alone reversed this increase, but this reversal was not achieved by treatment with insulin combined with metformin ( Fig. 3a). In the long-period group, however, malonaldehyde returned to the normal levels after glycemic control was achieved by either treatment (Fig. 3b and c). Kidney SIRT1 activity in the short-period group was only slightly decreased in the hyperglycemic animals but increased in the diabetic animals that received the metformin treatment (Fig. 3d). No change in SIRT1 activity was observed in kidney samples from the long-period group ( Fig. 3e and f). Uric acid clearance, however, accompanied the trend obtained for the kidney pAMPK levels ( Fig. 3g and h), and a good positive correlation was noted between these two factors (Fig. 3i). An inverse significant correlation was also observed between plasma uric acid levels and kidney pAMPK expression (r = − 0.3856, p = 0.0033).
The energy status was assessed by quantification of kidney AMP, ADP and ATP (Table 2). No difference in AMP/ATP and ADP/ATP ratios was observed between the groups of the short period experiment (8 weeks). However, both treatments (D12INS, D12MET) led to increased kidney AMP/ATP and ADP/ATP ratios in the animals of the long period experiment (24 weeks), not justifying the decreased pAMPK levels.
Because the persistent decrease of pAMPK could affect mitochondrial biogenesis and activity 20 , we examined the expression of PGC-1α and some markers of intermediate metabolism. The kidney PGC-1α levels did not change in the short-period group (Fig. 4a), whereas in the long-period group, the reduced PGC-1α expression levels observed after 24 weeks of hyperglycemia were not normalized by the combined insulin/metformin treatment. Animals treated with insulin alone still presented slightly decreased levels of PGC-1α , although significant differences were not observed compared with either the hyperglycemic or non-diabetic control groups ( Fig. 4b and c). To examine the kidney intermediate metabolism, pyruvate, lactate, malate, succinate, fumarate, glutamine, and glutamate were quantified (Table 3 and Fig. 4d,e and f). After eight weeks, no significant difference was observed between non-diabetic and hyperglycemic animals ( Table 3, Fig. 4d), but a trend towards increased metabolite levels was observed after 24 weeks of hyperglycemia (Table 3). Among the metabolites, only fumarate appeared to have similar levels in the both treated and diabetic groups, which were significantly higher than in the non-diabetic group (Table 3 and Fig. 4f).
PGC-1α activity was linked to the expression of a mitochondrial isoform of DNA methyltransferase 1 (mtD-NMT1) 28 . So, we examined kidney mtDNA 5-mC and 5-hmC levels. Although no difference in either the 5-mC  Table 1. Early renal function impairment is recovered by glycemic control after both short and long periods of hyperglycemia. The body weight and blood glucose were assessed weekly, but the values in the table refer to the measurement at the specified time points. For the short-period group, renal function parameters and HbA1c were assessed at two time points (four and eight weeks). For the long-period group, these parameters were assessed at three time points (six, 18 and 24 weeks). The data are presented as the means ± SEM. *p < 0.05 compared with the respective non-diabetic group + p < 0.01 compared with the respective diabetic group. N = 5 to 6 animals in the short-period group, N = 6 to 10 animals in the long-period group.
or 5-hmC levels was observed in the mtDNA of the animals belonging to the short-period group ( Fig. 4g and j), decreased 5-hmC (but not 5-mC) levels were found after 12 weeks of hyperglycemia ( Fig. 4h and k), and the levels of both 5-mC and 5-hmC were decreased in the mtDNA of the diabetic and treated diabetic animals belonging to the long-period group ( Fig. 4i and l). The sustained decrease in PGC-1α expression, concomitant with fumarate accumulation and the downregulation of mtDNA epigenetic marks, only after late glycemic control, indicate that changes in mitochondrial function during the course of diabetes serve as a contributing factor for hyperglycemic memory.

Discussion
The early tight management of glycemia in diabetic individuals is of utmost importance for slowing down the progression of DKD 16 . Amazingly, a reversal of DKD has been observed 10 years but not five years after pancreas transplantation in eight long-term type 1 diabetes patients with mild to advanced diabetic nephropathy at the time of transplantation 29 . Notwithstanding the clear possibility of the reversal of diabetic nephropathy after the restoration of euglycemia, the reported studies also demonstrate a long delay between these events 29 , which ultimately shows that the biochemical alterations experienced by the kidney over a long period of loose glycemic control are not promptly reversed by tight glycemic management. Different molecular components and mechanisms have been proposed to explain the metabolic memory sensed by kidney cells after exposure to hyperglycemia. Persistent oxidative stress 30 , glomerular basement membrane accumulation 31 , expression of extracellular matrix proteins 15,32 , and increased p38 MAPK activity 33 have  12, and 24 weeks, expressed as fold changes relative to the control groups, and representative immunoblots obtained for TGF-β analysis at each time point. (g and h) mRNA levels of TGF-β (ratio of C T values of TGF-β to those of β -actin) after 12 and 24 weeks analyzed by real-time PCR and expressed as fold changes relative to the control groups. The data are expressed as the means ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001 compared with the respective non-diabetic group. ++ p < 0.01 compared with the respective diabetic group. N = 5 or 6 animals in the short-period group, N = 6 to 10 animals in the long-period group. The western blot experiments included N = 4 or 5 animals in the short-period group and N = 5 or 6 animals in the long-period group. Full-length blots are presented in Supplemental  and were normalized in the animals after the achievement of glycemic control, with the exception of the animals in the short-period group treated with insulin plus metformin. (d,e and f) Sirt-1 activity in kidney samples from the short-and long-period groups. Uric acid clearance was reduced after eight (g) and 24 weeks (h) of diabetes and was not recovered after glycemic control. A correlation between pAMPK expression and uric acid clearance was observed (i). The data are expressed as the means ± SEM. *p < 0.05, **p < 0.01, and ***p < 0.001 compared with the respective non-diabetic group. ++ p < 0.001 and +++ p < 0.001 compared with the respective diabetic group. N = 4 to 6 animals in short-period group, N = 6 to 10 animals in the long-period group. For the Pearson correlation analysis, paired samples of both periods were included (N = 45).  Table 2. Levels of AMP, ADP, ATP, and the ratios AMP/ATP and ADP/ATP in kidney tissue. The data are presented as the means ± SEM in units of pmol/μ g of protein. *p < 0.05, **p < 0.01 and ***p < 0.001 compared with the respective non-diabetic group. + p < 0.05, ++ p < 0.01 and +++ p < 0.001 compared with the respective diabetic group. N = 5 to 6 animals in the short-period group, N = 6 to 10 animals in the long-period group.
been confirmed after the normalization of glucose levels in different in vitro and in vivo models. Additionally, epigenetic mechanisms operating in the pathogenesis of DKD are emphasized as key events underlying metabolic memory, particularly due to their long-term persistence 34 . Indeed, apart from oxidative stress 10,35 , epigenetic alterations regulating the expression of proinflammatory genes have been shown to persist in vascular endothelial cells in several experimental models of hyperglycemic memory 11,12,36,37 .
To the best of our knowledge, there is no information regarding a hyperglycemic memory of sequential components of a fibrogenic pathway implicated in DKD development. In this study, we identified a pathway that is gradually altered in the diabetic rat kidney after the restoration of normal glycemia and that depends on the previous period of hyperglycemia experienced by the animal. First, we found that a four-week period of hyperglycemia is sufficiently long to initiate a metabolic memory of reduced pAMPK expression in the rat kidney, and this reduced pAMPK expression had not returned to the baseline level four weeks after the achievement of glycemic control. The same finding was observed in the animals who underwent 12 weeks of hyperglycemia followed by 12 weeks of glycemic control (Fig. 2a-c). Although it is well known that hyperglycemia triggers a reduced expression of kidney pAMPK 20 , this study provides the first line of evidence demonstrating that the reinstatement of normoglycemia is not sufficient to promptly reverse the kidney pAMPK levels. This effect of hyperglycemia on the sustained downregulation of AMPK activity after glycemic control was previously observed only in bovine retinal capillary endothelial cells (BRECs) in vitro and retinas of streptozotocin-induced diabetic rats 26 . Possible drivers of AMPK activation are an increased AMP/ATP ratio and increased activities of the upstream kinases liver kinase B1 (LKB1) and Ca 2+ /calmodulin-dependent protein kinase kinase 26 . The class III NAD + -dependent deacetylase sirtuin 1 (SIRT1) has been shown to be a positive upstream regulator of the activity of the LKB1/AMPK pathway in human embryonic kidney 293T cells and other cells 25 . SIRT1 downregulation in response to poly(ADP-ribose)polymerase (PARP) activation in BRECs and rat retinas exposed to high glucose or oxidative stress was crucial to understanding the persistent high levels of molecules related to inflammation (NF-κ B), apoptosis (Bax), and damage (PAR) responses observed after restoration of the normal glucose levels, which were effects mediated by the persistently decreased activity of the LKB1/AMPK pathway and increased ROS generation 26 .
Conversely, we observed here that hyperglycemia did not significantly affect kidney SIRT1 activity (Fig. 3d,e and f). This observation does not preclude, however, the possible decrease in SIRT1 activity in some specific cell types, such as proximal tubular cells and glomerular podocytes 38 , an effect that may be diluted in the kidney homogenates used in this study. It is possible to observe a slight, albeit insignificant, decrease in kidney SIRT1 activity after eight weeks of hyperglycemia (Fig. 3d). Nevertheless, the increased kidney SIRT1 activity detected in the diabetic animals that received the metformin treatment (Fig. 3d) and the decreased oxidative stress detected in the treated diabetic animals ( Fig. 3a and c) were not accompanied by a recovery of the kidney pAMPK levels. Indeed, contrary to expectations 26,[39][40][41] , metformin was unable to restore the kidney pAMPK levels ( Fig. 2a and c).
Metformin acts in part by activating AMPK 42 . Different mechanisms, including a mild inhibition of complex 1 of the mitochondrial respiratory chain followed by an increase in the AMP/ATP ratio, activation of LKB1, and inhibition of AMP deaminase, an enzyme in the AMP degradation pathway to uric acid, have been proposed for this activation 43,44 . Given its polarity, metformin enters cells through membrane-bound organic cation transporters (OCT) 45 . The human OCT1 and OCT2 isoforms are important for metformin hepatic uptake and secretory renal clearance, respectively 45 . Both Oct1 and Oct2 are expressed in rodent kidneys 45 , with Oct1 and Oct2 constituting ~23% and ~75% of all renal Oct in rats 24 . It has been estimated that 76% of metformin hepatic uptake and 60% of metformin renal uptake are mediated by Oct1 and Oct1/Oct2, respectively, in mice 45 . However, Oct expression on the basolateral surface of the proximal tubules is downregulated in experimental diabetes 24 . In fact, a downregulation of approximately 50% of renal Oct1 and Oct2 protein expression was observed four weeks after the induction of diabetes by streptozotocin in male Sprague-Dawley rats, and this downregulation was not prevented by administration of a low-dose insulin treatment 24 . In a rat model of type 2 diabetes, the renal expression of Oct2 was also decreased to 50% of that found in the control rat kidney 46 . Thus, it is likely that the early reduction in the expression of Oct in the diabetic rat kidney reduces the renal uptake of metformin and its ability to activate kidney AMPK. Higher doses of metformin than that used in this study (100 mg/kg) would likely be necessary to reach the same kidney AMPK activation effect observed in non-diabetic rodent kidney injury models 40,41 . We did not find a study that quantified pAMPK in the diabetic rodent kidney when metformin was administered after at least four weeks of hyperglycemia.
Independently of hyperglycemia, hyperuricemia is a factor that also leads to a pronounced downregulation of Oct2 expression in rat kidneys 47 . Decreased renal clearance of uric acid with resultant increases in the plasma or serum uric acid levels have been observed in rats after three or seven weeks of diabetes induced by streptozotocin 48,49 . Therefore, we assessed the plasma and urine levels of uric acid (Supplemental Table 1) and found that the decreased renal clearance of uric acid in the hyperglycemic rats was not improved in the treated diabetic animals in both the short-and long-period groups (Fig. 3g and h). In addition to aiding the lack of a metformin effect  Table 3. The fumarate levels are not normalized after late glycemic control. The data are presented as the means ± SEM in units of pmol/μ g of protein. *p < 0.05 and **p < 0.01 compared with the respective nondiabetic group. + p < 0.05 and + + + p < 0.001 compared with the respective diabetic group. N = 5 to 6 animals in the short-period group, N = 6 to 10 animals in the long-period group.
Scientific RepoRts | 7:40544 | DOI: 10.1038/srep40544 on kidney pAMPK levels even after 12 weeks of glycemic control, the sustained decrease in uric acid clearance resembled the sustained decrease in kidney pAMPK expression (Fig. 2a-c). Although we cannot establish a link between these events, a good correlation could be demonstrated (Fig. 3i). At least in the liver of mice and in human HepG2 cells, it has been shown that an increased activity of AMP deaminase and a consequent increase in uric acid production, which are observed in diabetes, mediate hepatic gluconeogenesis via pAMPK downregulation 27 . The effect of excess plasma uric acid on the kidney pAMPK levels still needs to be investigated. An elevation of serum uric acid in type 1 diabetes has been found to be an independent predictor for overt diabetic nephropathy development 50 and it has been indicated as an important and novel player in this process 51,52 . Mild hyperuricemia in rats, corresponding to a 1.5-to 2-fold increase in serum uric acid induced by the uricase inhibitor oxonic acid, causes renal interstitial fibrosis after seven weeks, with increased interstitial collagen deposition and macrophage infiltration but no deposition of uric acid crystals. Fibrosis was prevented by allopurinol administration concomitant with the oxonic acid-containing diet 53 . The role of asymptomatic hyperuricemia in inducing renal fibrosis in patients with DKD has been evidenced in these patients by an increase in urinary TGF-β 1 after allopurinol withdrawal 54 . The persistent decrease in uric acid clearance observed in this study in diabetic rats with adequate glycemic control provides the first line of evidence showing that this effect might be an important factor in the metabolic memory of diabetes leading to DKD.
A crucial role for AMPK activity in regulating mitochondrial function and matrix accumulation in diabetic kidneys has been demonstrated in different mouse models of diabetes 20 . Reduced AMPK activity in the mouse diabetic kidney (16 weeks of hyperglycemia) has been linked to reduced mitochondrial biogenesis, reduced pyruvate input into the Krebs cycle, and reduced mitochondrial superoxide radical generation concomitant with increased expression of glomerular fibronectin, type IV collagen, and TGF-β 20 . The decreased pAMPK expression was also localized to the glomeruli, as shown by immunofluorescence 20 . All of these deleterious effects observed in diabetic animals were found to be reversed by the activation of AMPK through the i.p. administration of 5-aminoimidazole-4-carboxamide-1-β -D-ribofuranoside (AICAR) 20 . Here we did not perform immunostaining for pAMPK localization in the rat kidney. Other researchers have demonstrated, by immunohistochemistry, increased TGF-β expression in both tubular and glomerular compartments in rodent models after 8, 12 and 24 weeks of diabetes [55][56][57] . Since TGF-β is downstream to the other components of the fibrogenic pathway we studied, it is likely that the changes we observed are also present in glomerular and tubular compartments. Indeed, we consider that once we were able to detect the changes in the whole tissue lysate, they may not be limited to a single kidney compartment.
In this study, we observed that the sustained reduction in kidney pAMPK expression was accompanied by a sustained increase in TGF-β protein expression in the long-period group, despite the normalization of TGF-β 1 gene expression, the achievement of glycemic control and the optimization of renal function. In contrast, the normalization of glycemia in the short-period group reversed the increase in TGF-β protein expression induced by hyperglycemia (Fig. 2d,e and f). Thus, other elements in addition to the reduction in AMPK activity appear to be necessary for the sustained high expression of the TGF-β protein, as considered below.
The upregulation of TGF-β under high-glucose conditions is well described [58][59][60][61] . Regarding its role in hyperglycemic memory, male Lewis rats maintained in a hyperglycemic state for two weeks or four months (~16 weeks) and thereafter subjected to strict glycemic control by pancreatic islet transplantation (for an additional 12 months or four months, respectively) exhibited normal glomerular TGF-β 1 gene expression, although increased TGF-β 1 gene expression was observed after four, eight or 12 months of hyperglycemia. However, a point of no return was reached by islet transplantation after eight months (~32 weeks) of hyperglycemia, and glomerular TGF-β 1 gene expression remained increased until the end of the study (12 months following the onset of diabetes), accompanied by increases in the gene expression of fibronectin and collagen IV, urinary albumin excretion, proteinuria, and mesangial expansion 62 . The researchers did not assess the level of TGF-β protein expression.
In vitro experiments with human proximal tubular epithelial cells have demonstrated that a short period of exposure to high-glucose conditions stimulates TGF-β 1 transcription, but the mRNA was poorly translated 63 . Subsequent stimulation of the cells with macrophage-derived cytokine platelet-derived growth factor 63 or interleukin-1β 64 stabilized the TGF-β 1 mRNA and synergistically increased its translational efficiency, leading to sustained de novo synthesis of TGF-β 1 protein 63,64 . One possible explanation for this effect is that glucose primes the kidney for TGF-β 1 protein synthesis through an outside stimulus 63 . Renal macrophage infiltration is an early event in streptozotocin-induced diabetic rats 65 and has been demonstrated to occur in patients with type 2 diabetes 66 . Hyperuricemia appears to contribute to kidney inflammation mediated by the NOD-like receptor protein 3 (NLRP3) inflammasome, resulting in increased interleukin-1β and interleukin-18 expression in STZ-induced diabetic rats 49 . Urate-lowering agents, such as allopurinol and quercetin, suppress renal NLRP3 inflammasome activation, thereby increasing the protection of diabetic rats against kidney injury 49 . Therefore, we hypothesize that the persistent hyperuricemia observed in this study may play a role in the persistent elevation of the TGF-β protein levels in the kidneys of the treated diabetic animals of the long-period group.
Because the reduced kidney AMPK activity is associated with reduced mitochondrial biogenesis and function 20 , we assessed whether PGC-1α expression, epigenetic marks (5-mC and 5-hmC) in mitochondrial DNA, and some intermediate metabolites (pyruvate, lactate, malate, succinate, fumarate, glutamine, and glutamate) are persistently altered in the kidneys of the treated diabetic rats. As observed for the kidney TGF-β protein expression, persistent changes occurred only after late glycemic control ( Fig. 4 and Table 3).
PGC-1α is a co-activator of the forkhead box O (FoxO) family of transcription factors, acting as a master regulator of oxidative metabolism and mitochondrial biogenesis 67 . PGC-1α activity has also been linked to the expression of a mitochondrial isoform of DNA methyltransferase 1 (mtDNMT1) 28 . In this study, we observed a persistent decrease in kidney PGC-1α expression concomitant with a sustained downregulation of mtDNA 5-mC after late glycemic control (Fig. 4c and i). The methylation of mtDNA is poorly understood, and levels of mtDNA 5-mC in the range of 2 to 5% have been reported in rodent and human fibroblasts 68,69 . Shock and coworkers 28 showed that mtDNMT1 translocates to mitochondria, binds to mtDNA proportionally to the density of CpG dinucleotides, and regulates mitochondrial gene expression 28 . The hypermethylation of mtDNA has been observed in retinal endothelial cells exposed to high-glucose conditions and in the retinal microvasculature from human donors with diabetic retinopathy 70 . The hypomethylation of mtDNA in the liver of high fructose-fed rats has been proposed as a novel underlying mechanism of metabolic syndrome 71 . In this study, we assessed the global content of kidney mtDNA 5-mC, and to the best of our knowledge, this manuscript provides the first report of mtDNA methylation in the diabetic kidney.
Mitochondrial DNA hydroxymethylation (5-hmC) was first described in 2011 by Shock and coworkers 28 , and its function in the mitochondrial genome is not yet clear. A proposed pathway for mtDNA hydroxymethylation is the direct addition of 5-hydroxymethyl groups to cytosine residues driven by mtDNMT1 28,72 . In this case, our observation of a persistent downregulation of mtDNA 5-hmC ( Fig. 4k and l) could also be a consequence of a reduction in the mtDNMT1 levels due to a decrease in PGC-1α expression. Yamazaki and co-workers 71 reported reduced levels of mitochondrial 5-hmC and 5-mC in the liver of high fructose-fed rats compared with control rats 71 . Further research on the implications of decreases in mtDNA 5-mC and 5-hmC for DKD development is warranted.
Among the intermediate metabolites analyzed, fumarate was noted as the only one to maintain its same high level in diabetic rats without and with adequate late glycemic control for 12 weeks. During the period of insulin insufficiency, an increase in amino acid catabolism, as well as increased activity of the urea cycle in the liver and rapid arginine consumption, is expected. The kidney plays a key role in the synthesis of arginine from citrulline to maintain the body arginine pool for non-urea cycle functions. The synthesis pathway consumes ATP and aspartate and releases inorganic pyrophosphate, AMP and fumarate 73 ; therefore, this pathway is a possible source of the high kidney fumarate levels observed in the hyperglycemic animals. Early glycemic control, but not late glycemic control, apparently restored the normal metabolism. According to recent data [21][22][23] , the pathway comprising high glucose → reduced AMPK activity → increased Nox4 → reduced fumarate hydratase activity → increased fumarate levels → increased TGF-β expression plays an important role in the development of diabetic kidney disease. We found that components of this pathway (reduced AMPK activity, increased fumarate levels, and increased TGF-β expression) constitute part of the diabetes metabolic memory (Fig. 5). Overall, the present study showed that a short period of hyperglycemia (four weeks) induced persistent decreases in uric acid clearance and kidney pAMPK levels during the subsequent four-week period of glycemic control. The rats subjected to a longer period of hyperglycemia (12 weeks) experienced further kidney metabolic alterations that were not reversed in the subsequent 12 weeks of glycemic control, such as fumarate accumulation, increased TGF-β protein expression, decreased PGC-1α expression, and downregulation of mtDNA methylation and hydroxymethylation, and these effects were concomitant with sustained decreases in uric acid clearance and pAMPK levels. The observed changes were grouped in the pathway shown in Fig. 5. Because hyperuricemia is an independent risk factor for diabetic nephropathy development [50][51][52][53][54]74 , the observed sustained decrease in uric acid clearance in treated diabetes may feed the prolonged kidney biochemical alterations observed after tight glycemic control is achieved. Further interventional research to assess the role of excess plasma uric acid in sustaining the biochemical alterations found in the diabetic kidney in the present study using, for example, allopurinol is needed. Another question raised by the findings obtained in this study is the possible interaction between hyperglycemia and poor uric acid clearance in the induction of metabolic alterations that lead to diabetic complications, which would increase the risk of complications in diabetic subjects with loose glycemic control. The monitoring of the plasma uric acid levels in large clinical trials aimed at evaluating the risk of complications under different targets of glycemic control may aid the understanding of some gaps in the phenomenon of diabetic metabolic memory.

Methods
Animals. Eight-week-old male Wistar rats were maintained under a 12-h light/12-h dark cycle and had access to standard laboratory diet and water ad libitum. For diabetes induction, streptozotocin (40 mg/kg) freshly dissolved in saline was injected into the penile vein. The control animals received saline intravenously. The diabetic rats were maintained in their hyperglycemic state for either four (short period) or 12 weeks (long period). The animals were then treated with insulin alone or insulin combined with 100 mg/kg metformin (Alcon Biosciences, Mumbai, India) administered by gavage for the subsequent four weeks (D4INS and D4MET, respectively) or in drinking water for the subsequent 12 weeks (D12INS and D12MET, respectively) and were then euthanized. The daily dose of metformin was selected based on the results reported by Zheng and coworkers 26 . Age-matched non-diabetic (ND) and non-treated diabetic (D) controls were euthanized after eight (ND8 and D8), 12 (ND12 and D12) and 24 (ND24 and D24) weeks. The insulin treatment of the animals in the short-period group consisted of 5 to 6 UI of NPH insulin (Lilly, Indianapolis, IN, USA) administered subcutaneously in two daily doses. The insulin treatment of the animals in the long-period group consisted of 2 UI of NPH insulin (Lilly, Indianapolis, IN, USA) in the morning and 2 or 3 UI of glargine insulin (Sanofi, Gentilly, France) at the end of the day. The non-treated diabetic animals of the long-period group received 1 to 3 UI of NPH insulin on alternate days to maintain their glycemic levels under 500 mg/dL and avoid suffering and death before 24 weeks. The body weight and glycemia, measured in the tail vein blood with a Breeze TM 2 glucometer (Bayer, Leverkusen, Germany), were monitored weekly for the duration of the study. Twenty-four-hour urine collections were performed at different time points to assess renal function. At the end of the experiments, the animals were euthanized, and the kidneys were collected and immediately frozen at − 80 °C for further analyses. Malonaldehyde Quantitation. Kidney homogenates were prepared by mixing 50 mg of kidney tissue with 500 μ L of phosphate buffer saline and 72 μ L of butylhydroxytoluene 0.2% in a mechanically driven homogenizer. A 100 μ L aliquot of kidney homogenate was used for MDA quantitation. Protein-bound MDA was released by the addition of 10 μ L of 4 M sodium hydroxide to the homogenate, and the mixture was incubated at 60 °C for 1 h. The proteins were then precipitated by the addition of 150 μ L of 1% sulfuric acid, and the samples were centrifuged at 9,300 g for 10 min. Finally, 25 μ L of dinitrophenylhydrazine (1 mg/mL in 2 M hydrochloric acid) were added to 175 μ L of the supernatant, and the mixture was incubated at room temperature and protected from light for 30 min. Aliquots of 100 μ L were injected into a HPLC-DAD system (Shimadzu, Kyoto, Japan). The chromatographic conditions used consisted of a 250 × 4.6 mm i.d., 5.0 μ m Luna C18(2) column (Phenomenex, Torrance, CA, USA) that was eluted with a gradient of water (solution A) and acetonitrile (solution B), both containing 0.2% acetic acid, at a flow rate of 1 mL/min and a temperature of 30  TGF-β Gene Expression. The total RNA from kidney tissues was isolated using the RNeasy ® Mini Kit (Qiagen, Hilden, Germany), and its purity and integrity were confirmed using a Bioanalyzer 2100 system (Agilent Technologies, Santa Clara, CA, USA). Two μ g of RNA was converted to cDNA using a High-capacity cDNA Reverse Transcription kit (Applied Biosystems, NJ, USA). For real-time PCR, the cDNA was diluted to 50 ng/μ L, and forward and reverse primers, Taqman ® Gene Expression Master Mix (Life Technologies, Carlsbad, CA, USA), and RNAse-free water were then added to the mixture. The following primers obtained from Life Technologies were used: Rn00572010_m1 for Tgf-β1 and Rn00667869_m1 for actin as an endogenous control. The reaction was performed in an ABI Prism ® 7500 Sequence Detection System (Applied Biosystems, NJ, USA) under the following conditions: 50 °C for 2 min, 95 °C for 10 min, and 40 cycles of 15 s at 95 °C and 60 °C for 1 min. The gene expression levels were determined by comparing the C T values for Tgf-β1 relative to those of the actin gene.
pAMPK Expression. Phospho-AMPKα (Thr172) expression in kidney samples was analyzed using a sandwich ELISA kit (Cell Signaling, Danvers, MA, USA). Kidney homogenates were prepared by mixing 50 mg of renal tissue and 500 μ L of lysis buffer with 1 mM phenylmethanesulfonyl fluoride. Aliquots containing 250 μ g proteins were diluted in sample buffer, and the assay was conducted following the manufacturer's recommended protocol.
Sirtuin 1 Activity. Sirtuin 1 activity in kidney homogenates was measured using a fluorometric assay kit (Fluor-de-Lys, Enzo Life Sciences, Farmingdale, NY, USA) following the manufacturer's recommended instructions.