AMP-activated Protein Kinase α2 Subunit Is Required for the Preservation of Hepatic Insulin Sensitivity by n-3 Polyunsaturated Fatty Acids

Four-month-old whole-body AMPKα2^−/− mice (29) backcrossed to C57BL/6J mice for nine generations, and wild-type littermate controls were fed on either Chow, cHF, or cHF+F diet for nine weeks. Body weight and food consumption were recorded, and EDTA-plasma and tissues were collected for various analyses as described in the online appendix, available at http://diabetes.diabetesjournals.org/cgi/content/full/db09-1716/DC1. Male mice were used for all the experiments, except for the measurements of hepatic AMPK activity, which were performed on female mice. The experiments were conducted under the guidelines for the use and care of laboratory animals of the Institute of Physiology and followed the “Principles of laboratory animal care” (National Institutes of Health publication no. 85-23, revised 1985).

Blood glucose was measured using calibrated glucometers (LifeScan, Milpitas, CA). Nonesterified fatty acids (NEFAs), triglycerides, and total cholesterol were determined in plasma using the following enzymatic photometric tests: NEFA-C (Wako Chemicals, Neuss, Germany), triacylglycerols liquid, and cholesterol liquid (Pliva-Lachema Diagnostika, Brno, Czech Republic), respectively. Plasma insulin was measured using the Sensitive Rat Insulin RIA Kit (LINCO Research, St. Charles, MO). Total adiponectin levels and adiponectin multimeric complexes were determined using Western blotting (31).

The tissue content of triglycerides was estimated in ethanolic KOH tissue lysates as described before (8). The content and fatty acid composition of the phospholipid, diacylglycerol, triglyceride, and ceramide fractions were assessed in tissue lipid extracts; gene expression was evaluated using real-time RT-PCR (see online appendix).

Livers were collected by freeze-clamping, AMPK was immunoprecipitated from tissue extracts, and the activity was assayed using a peptide substrate (32); see the online appendix.

Five days before the experiment, an indwelling catheter was placed into the left femoral vein under anesthesia (33). Mice were allowed to recover for 5–7 days, followed by a 6-h fast (8:00 a.m.–2:00 p.m.) prior to the experiment. The whole-body glucose turnover was determined under basal (nonstimulated) and insulin-stimulated conditions (hyperinsulinemic-euglycemic clamp), using separate groups of mice. Insulin (Actrapid, Novo Nordisk Pharma, Denmark) was infused at a constant rate of 4.8 mU/kg·min for 3 h, while d-[3-³H]glucose (Perkin Elmer, Boston, MA) was infused at a rate of 15.9 kBq/min. Throughout the infusion, glucose concentration and d-[3-³H]glucose specific activity (during the last hour of infusion) were determined in tail blood. Euglycemia (∼5.55 mmol/l) was maintained by periodically adjusting a variable infusion of 33% glucose (33). At the end of a 3-h infusion period, mice were first anesthetized by diethylether, exsanguinated through the cervical incision, and then killed by cervical dislocation, and tissues (liver and quadriceps muscle) and EDTA-plasma were collected for biochemical analyses (see supplementary Table 1 and Research Design and Methods of the online appendix for details on basic clamp parameters and methodology).

Hepatocytes were isolated from livers of fed mice by a modification of the collagenase method (34) and seeded at a density of 0.5 × 10⁶ cells per each 35-mm Petri dish. Rates of basal and insulin-stimulated de novo lipogenesis and AICAR-stimulated fatty acid oxidation were measured using [1-¹⁴C] acetate and [1-¹⁴C] palmitate, respectively (see online appendix).

All values are presented as means ± SE. Data were analyzed by two-way ANOVA. Comparisons were judged to be significant at P ≤ 0.05.

Specific activities of AMPKα1 and AMPKα2 were evaluated in the liver of ad libitum-fed mice after nine weeks of the differential dietary treatment (Fig. 1). No significant effect of either diet (Chow, cHF, and cHF+F) or genotype (wild-type versus AMPKα2^−/−) on AMPKα1-specific activity was observed, although the AMPKα1 activity tended to be higher in the AMPKα2^−/− mice (Fig. 1,A). In contrast, AMPKα2 activity was stimulated by n-3 LC-PUFAs (cHF+F diet; Fig. 1,B). AMPKα2 activity was not detected in the AMPKα2^−/− mice (Fig. 1 B). No changes were detected in the activity of AMPKα1 and AMPKα2 in the quadriceps muscle in response to n-3 LC-PUFAs (not shown).

At four months of age, at the beginning of dietary treatments, wild-type and AMPKα2^−/− mice fed the Chow diet exhibited similar body weights (Table 1). In mice of both genotypes, cHF-feeding for nine weeks resulted in greater body weight gain compared with the Chow-fed mice. However, this effect was less pronounced in AMPKα2^−/− mice (Table 1). In both wild-type and AMPKα2^−/− mice, the cHF+F diet induced smaller body weight gain than the cHF diet (Table 1 and supplementary Figure 1). None of the differences in body weight gain could be explained by caloric intake, which was similar in all experimental groups (Table 1). The weight of fat depots increased in response to cHF feeding, while the cHF+F diet partially prevented this increase (Table 1). Triglycerides and NEFA levels in plasma of ad libitum-fed mice were similar in the Chow- and cHF-fed mice, while cholesterol levels were markedly and significantly elevated by the cHF diet. n-3 LC-PUFAs lowered plasma lipid levels independently of AMPKα2. Triglycerides and NEFA levels were strongly reduced, even below the levels observed in the Chow-fed mice (Table 1).

After nine weeks of dietary treatment, no change was observed in blood glucose, but elevations were observed in plasma insulin levels in response to the cHF diet in ad libitum-fed mice of both genotypes. However, the increase in plasma insulin levels was less pronounced in AMPKα2^−/− mice, closely reflecting the genotype-dependent differences in body weight gain (Table 1). A similar pattern of changes in insulin levels was also observed in fasted mice, in which n-3 LC-PUFAs significantly reduced insulin levels only in wild-type animals (Table 1). As expected, plasma levels of total as well as high molecular weight form of adiponectin, an adipokine associated with increased insulin sensitivity (35), were increased ∼1.4- and ∼1.2-fold, respectively, in wild-type mice in response to n-3 LC-PUFA supplementation (Table 1 and supplementary Fig. 2, available in an online appendix); however, no significant increase of plasma adiponectin by n-3 LC-PUFAs was observed in AMPKα2^−/− mice.

In further experiments, hyperinsulinemic-euglycemic clamps were performed to evaluate whole-body insulin sensitivity. Under basal conditions, glucose turnover rate (GTO; i.e., glucose uptake in peripheral tissues) was similar in all groups of mice (supplementary Table 1). Under insulin-stimulated conditions (Fig. 2,A–F), the amount of exogenous glucose required to maintain euglycemia during the clamp, i.e., the glucose infusion rate (GIR), was ∼1.3-fold lower in AMPKα2^−/− than in wild-type mice fed the Chow diet (Fig. 2,A). On the other hand, GIR was decreased by the cHF diet to a similar level in mice of both genotypes, manifesting diet-induced insulin resistance. This was attributed to a decreased GTO and, in particular, to an impaired suppression of hepatic glucose production (HGP) by insulin, with HGP being ∼8.5-fold higher in the cHF-fed compared with the Chow-fed wild-type mice (Fig. 2,C). In wild-type mice, cHF+F diet feeding increased GIR and GTO (∼1.9- and ∼1.2-fold increase, respectively) as compared with cHF-fed mice, while HGP was lowered to a similar level as in the Chow-fed mice. These results document the protective effects of n-3 LC-PUFAs from high-fat diet-induced insulin resistance in wild-type mice, namely, at the level of HGP. In contrast, none of these beneficial effects of n-3 LC-PUFAs was observed in AMPKα2^−/− mice, in which neither the GIR (Fig. 2,A) nor the GTO (Fig. 2,B) differed between the cHF+F-fed and the cHF-fed mice, whereas the rate of HGP was even higher in the cHF+F-fed than in the cHF-fed AMPKα2^−/− mice (Fig. 2,C). Although whole-body glycolysis was similar in all the groups (Fig. 2,D), the rate of whole-body glycogen synthesis, which reflects insulin sensitivity of muscle glucose metabolism, was dependent on both diet and genotype (Fig. 2,E). In the Chow-fed mice, the rate of whole-body glycogen synthesis tended to be higher in wild-type mice than in AMPKα2^−/− mice. Only in the former mice was it significantly affected by dietary treatment. Thus, in wild-type mice, glycogen synthesis was decreased ∼2.4-fold in response to the cHF diet, while n-3 LC-PUFAs provided a partial protection from this decrease (Fig. 2,E). A similar pattern of changes in the glycogen synthesis rate in response to n-3 LC-PUFAs was observed when measured directly in the skeletal muscle (Fig. 2 F). Thus, in accordance with the previous study (26), the results of clamp studies suggested impairment of insulin sensitivity in response to whole-body ablation of AMPKα2 in Chow-fed mice. However, AMPKα2^−/− mice seemed to be partially protected against cHF-induced insulin resistance, while AMPKα2 was required for preservation of insulin sensitivity in the skeletal muscle and especially in the liver in response to n-3 LC-PUFA feeding.

In addition to the ad libitum-fed mice (Table 1), plasma lipid levels were also measured in fasted mice, as well as in mice subjected to hyperinsulinemic-euglycemic clamp (supplementary Table 2). In contrast to the ad libitum-fed state, cHF+F diet did not affect either triglyceride or NEFA levels under fasting conditions. Under the hyperinsulinemic-euglycemic conditions, both triglyceride and NEFA levels were lower in the cHF+F-fed than in the cHF-fed wild-type mice (∼1.6-fold and ∼1.4-fold difference, respectively), but no such difference between the diets was observed in AMPKα2^−/− mice. Cholesterol levels were consistently decreased by n-3 LC-PUFAs independently of both the metabolic state and genotype (supplementary Table 2).

In ad libitum-fed mice of both genotypes, the hepatic triglyceride content was increased ∼twofold by cHF compared with the Chow diet, while triglyceride accumulation was increased only ∼1.3-fold by cHF+F diet in both genotypes, documenting a protection against hepatic triglyceride accumulation by n-3 LC-PUFAs (Fig. 3,A). Under hyperinsulinemic-euglycemic conditions, n-3 LC-PUFAs also protected livers of wild-type mice against the cHF-induced accumulation of triglycerides. However, this effect was absent in AMPKα2^−/− mice (Fig. 3 B). Moreover, a strong correlation was found between plasma NEFA levels and hepatic triglyceride content assessed under the clamp conditions in the cHF+F-fed AMPKα2^−/− mice (R² = 0.43, P < 0.05) but not in wild-type mice (R² = 0.08, P = 0.40).

We sought to determine whether the differential effect of n-3 LC-PUFAs on accumulation of liver triglycerides in wild-type and AMPKα2^−/− mice under the clamp conditions could be explained by hepatic lipid metabolism. In cultured hepatocytes isolated from mice following the different dietary treatments, activities of both fatty acid oxidation and de novo fatty acid synthesis were evaluated. The stimulatory effects of 5-aminoimidazole-4-carboxamide-1-β-d-ribofuranoside (AICAR), an AMPK activator, and insulin on fatty acid oxidation and synthesis are shown in Fig. 4,A and B, respectively (for corresponding basal metabolic activities, see supplementary Table 3, available in an online appendix). In hepatocytes from both Chow and cHF diet-fed mice, the absence of AMPKα2 was associated with a trend for lower AICAR-stimulated fatty acid oxidation. Although hepatocytes from cHF-fed mice showed reduced stimulatory effect of AICAR irrespective of the genotype, cHF+F feeding normalized this defect in wild-type but not in AMPKα2^−/− hepatocytes (Fig. 5,A), suggesting AMPK-dependent induction of capacity for fatty acid oxidation by n-3 LC-PUFAs in the liver. The stimulatory effect of insulin on de novo fatty acid synthesis was reduced in hepatocytes from cHF-fed wild-type mice, whereas it was retained in the hepatocytes from cHF-fed AMPKα2^−/− mice (Fig. 4,B). cHF+F feeding tended to restore the stimulatory effect of insulin only in wild-type hepatocytes (Fig. 4 B).

To further characterize hepatic effects of differential dietary treatment, the expression of selected genes was quantified in total RNA isolated from the livers of mice subjected to hyperinsulinemic-euglycemic clamp (Fig. 4,C and D). Feeding cHF diet suppressed expression of lipogenic genes stearoyl-CoA desaturase (SCD-1) and SREBP-1c in all groups (except for SREBP-1c in AMPKα2^−/− mice). This suppression was partially counteracted by cHF+F diet in wild-type but not AMPKα2^−/− mice (Fig. 4 C and D). Together with the de novo fatty acid synthesis data, these results further support the AMPKα2-dependent improvement of liver insulin sensitivity by n-3 LC-PUFAs.

To identify factors predisposing animals to insulin resistance in an AMPKα2-dependent manner, a detailed analysis of hepatic lipids in ad libitum-fed mice was performed. No major genotype-dependent differences in the contents of either ceramides or phospholipids were observed (supplementary Table 4). In contrast, hepatic content of diacylglycerols was affected in a genotype- and diet-dependent manner (Fig. 5,A). Wild-type mice fed the cHF+F diet had lower diacylglycerol content than genotype-matched cHF diet-fed mice, while this effect of the cHF+F diet was not observed in AMPKα2^−/− mice. Moreover, the analysis of fatty acid composition of the diacylglycerol fraction in the liver revealed that wild-type as well as AMPKα2^−/− mice fed cHF diet were characterized by marked increase in the level of PUFA but not monounsaturated or saturated fatty acids (Fig. 5,B, C, and D and supplementary Table 4). The increase in the PUFA content tended to be smaller in the wild-type compared with AMPKα2^−/− mice (∼1.7-fold and ∼2.2-fold, respectively). Administration of n-3 LC-PUFAs completely prevented accumulation of hepatic polyunsaturated diacylglycerols in wild-type mice, whereas their level in the AMPKα2^−/− animals, although decreased, was still significantly higher compared with genotype-matched Chow-fed mice (Fig. 5,B). Regarding polyunsaturated diacylglycerols, α-linolenic acid (18:3n-3) appeared to be by far the most differentially regulated PUFA in the diacylglycerol fraction in the two genotypes (supplementary Table 4). In addition, cHF+F diet markedly reduced hepatic content of monounsaturated diacylglycerols in wild-type but not in knockout animals (Fig. 5 C). Hepatic diacylglycerol levels and their fatty acid composition were also analyzed in mice subjected to hyperinsulinemic-euglycemic clamp (supplementary Table 5 and supplementary Fig. 3). No significant differences among the groups were observed in total diacylglycerols content or in their saturated or monounsaturated fatty acid fractions (supplementary Fig. 3).

Previous animal studies demonstrated that n-3 LC-PUFAs could counteract the development of both hepatic steatosis (8,18,36,37) and hepatic insulin resistance (8,9,16), while suppressing lipogenesis and augmenting lipid catabolism in the liver (8,13,19,21). Using mice with a whole-body deletion of AMPKα2 and high-fat feeding, we show for the first time that AMPKα2 is required for the effect of n-3 LC-PUFAs to preserve whole-body, muscle, and especially hepatic insulin sensitivity, as well as to suppress hepatic and plasma triglycerides as well as NEFA levels under hyperinsulinemic-euglycemic clamp conditions. In contrast, AMPKα2 was not required for protection by n-3 LC-PUFAs from hepatic lipid accumulation and dyslipidemia in ad libitum-fed mice.

In addition to AMPKα2, PPARα was previously identified as an important determinant of n-3 LC-PUFA's effect on lipid metabolism, especially short-term modulation of hepatic gene expression (14) and insulin sensitivity (16). Thus, the reduction in hepatic triglyceride concentrations by fish oil feeding did not rescue insulin action in PPARα-null mice, while hepatic diacylglycerol concentrations were decreased by fish oil in a PPARα-dependent manner and were associated with a preserved hepatic insulin sensitivity (16). It is generally accepted that 1) diacylglycerols rather than triglycerides or ceramides mediate hepatic insulin resistance in mice fed a high-fat diet (19,38,39), 2) diacylglycerol-induced insulin resistance depends on activation of protein kinase C, and 3) that polyunsaturated diacylglycerols in particular are better protein kinase C activators than saturated diacylglycerol species [reviewed in refs (38,39)]. Also our results showed that cHF diet-induced insulin resistance was associated primarily with the accumulation of PUFA in hepatic diacylglycerols and that n-3 LC-PUFA completely prevented cHF diet-induced increase in PUFA diacylglycerols in wild-type mice, whereas in AMPKα2^−/− animals, the content of these lipids was still significantly higher compared with the control. Moreover, it was only in ad libitum-fed mice but not in mice subjected to hyperinsulinemic-euglycemic clamps that the levels of hepatic diacylglycerols and their fatty acid compositions were associated with hepatic insulin sensitivity. It is possible that under clamp conditions AMPKα2-dependent effects of n-3 LC-PUFAs on liver diacylglycerols were masked by metabolic changes occurring during a 3-h infusion of insulin and glucose.

The failure of n-3 LC-PUFAs to decrease hepatic lipids in AMPKα2^−/− mice under clamp conditions could be due to primary alterations in metabolic fluxes in the liver, reflecting 1) increased de novo fatty acid synthesis, 2) decreased secretion of VLDL triglycerides, or 3) reduced fatty acid oxidation. De novo fatty acid synthesis was not the responsible factor, because hepatocytes of the n-3 LC-PUFA-fed AMPKα2^−/− mice showed decreased insulin-stimulated de novo fatty acid synthesis and reduced expression of SREBP-1c and SCD-1, as compared with hepatocytes isolated from n-3 LC-PUFA-fed wild-type mice, reflecting probably low insulin sensitivity of the liver in AMPKα2^−/− mice. Moreover, AICAR-stimulated fatty acid oxidation in hepatocytes from n-3 LC-PUFA-fed AMPKα2^−/− mice was markedly reduced as compared with those from wild-type mice, suggesting decreased hepatic capacity for fatty acid oxidation in the absence of AMPKα2, which could contribute to enhanced lipid accumulation. Therefore, these experiments supported a major role of hepatic AMPKα2 in the regulation of both insulin sensitivity and lipid metabolism by n-3 LC-PUFAs. However, the differential modulation of lipid accumulation by n-3 LC-PUFAs in the livers of wild-type and AMPKα2^−/− mice under clamp conditions could also be secondary to the AMPKα2-dependent effects of n-3 LC-PUFAs in other tissues, resulting in a relatively high hepatic uptake of circulating NEFA in AMPKα2^−/− mice. This is supported by persistently elevated plasma levels of NEFA in AMPKα2^−/− mice, as well as by a significant correlation between plasma NEFA levels and hepatic triglyceride content observed under clamp conditions in AMPKα2^−/− but not wild-type mice fed n-3 LC-PUFA-containing diet. Moreover, it has been shown in humans with nonalcoholic fatty liver disease that most of hepatic triglycerides arise from circulating NEFA (40). That plasma NEFA levels under clamp conditions were reduced only in wild-type but not in AMPKα2^−/− mice fed n-3 LC-PUFAs may reflect a role of AMPKα2 in muscle lipid uptake mediated by lipoprotein lipase (41), as well as the antilipolytic effect of AMPK in adipose tissue, documented for AMPKα1 (42). In any case, decreased fatty acid oxidation in situ in the liver and, possibly even more importantly, abundant supply of circulating NEFA could be responsible for the lack of the antisteatotic effect of n-3 LC-PUFAs in AMPKα2^−/− mice under clamp conditions (Fig. 6).

Previous studies reported contradictory results, showing either 1) activation of AMPK in rat liver (22) and murine adipose tissue (24) or 2) no changes in AMPK activity in the liver, skeletal muscle, and heart of mice (43) in response to dietary n-3 LC-PUFAs. These discrepancies could be related to differences in dietary n-3 LC-PUFA intake, nutritional state of animals, and other parameters. In accordance with the involvement of AMPKα2 in various effects of n-3 LC-PUFAs, our results document activation of AMPKα2 (but not AMPKα1) in the liver of mice by long-term n-3 LC-PUFA treatment, in the absence of significant changes in either the AMP to ATP ratio assessed in whole liver extracts [not shown and ref (44)] or the phosphorylation status of LKB1, an upstream kinase for AMPK [not shown and ref (45)]. In addition, no effect on AMPK activity in either cultured hepatocytes (21) or embryonic kidney cells (not shown) of n-3 LC-PUFAs added to the cell culture medium could be detected. Therefore, the activation of AMPKα2 by n-3 LC-PUFAs probably does not depend on a direct interaction between n-3 LC-PUFAs and AMPK. On the other hand, induction of adiponectin by n-3 LC-PUFAs [results of this study and refs (46,47)] could be involved, because adiponectin activates AMPK in both the liver and skeletal muscle (35). Adiponectin is also required for the activation of AMPK upon administration of PPARγ agonists thiazolidinediones, whereas mice lacking adiponectin show decreased hepatic insulin sensitivity and reduced responsiveness to these compounds (48). Thus, absence of AMPKα2 may blunt adiponectin-mediated effects of n-3 LC-PUFAs. In accordance with the previous study (49), plasma adiponectin levels tended to be reduced in AMPKα2^−/− mice. Moreover, the induction of adiponectin by n-3 LC-PUFAs in AMPKα2^−/− mice was compromised (Table 1 and supplementary Figure 2).

AMPKα1 and AMPKα2 contribute equally to total AMPK activity in the liver (50). In mice with liver-specific ablation of AMPKα2 (27), hepatic AMPKα2 was essential for suppressing hepatic glucose production and maintaining fasting blood glucose levels; however, the absence of AMPKα2 did not affect inhibitory action of insulin on hepatic glucose production. In our study, although fasting blood glucose levels were unaltered by whole-body ablation of AMPKα2, the beneficial effect of dietary n-3 LC-PUFAs on hepatic insulin sensitivity was clearly AMPKα2-dependent. Differential regulation of glucose homeostasis in the above transgenic models likely reflects the complexity of whole-body (26) versus liver-specific (27) deletion of AMPKα2. In contrast to the previous report, showing induction of adiposity and adipocyte hypertrophy in AMPKα2^−/− mice fed a lard-based high-fat diet (49), our study documented a relatively low weight gain, low adiposity, and smaller fat cells in AMPKα2^−/− mice fed a corn-oil based high-fat diet. This discrepancy could be related to the differences in the composition of experimental high-fat diets; however, our results are consistent with the elevated sympathetic tonus of AMPKα2^−/− mice (26), which may stimulate energy dissipation in these animals. In any case, lower body weight of cHF-fed AMPKα2^−/− mice as compared with their wild-type counterparts could be related to better insulin sensitivity of the former mice, as suggested by the differences in insulinemia, results of hyperinsulinemic-euglycemic clamp, stimulatory effect of insulin on lipogenesis, and expression of lipogenic genes in the liver.

In conclusion, the preservation of hepatic insulin sensitivity by n-3 LC-PUFAs in mice fed a high-fat diet depends on AMPKα2. The accumulation of diacylglycerols, which is regulated in an AMPKα2-dependent manner, could contribute to the modulation of hepatic insulin sensitivity in response to dietary n-3 LC-PUFAs. On the other hand, the AMPKα2-dependent acute changes in lipid metabolism and hepatic triglyceride accumulation, which are unmasked under insulin-stimulated conditions such as during hyperinsulinemic-euglycemic clamp, largely reflect the extrahepatic action of n-3 LC-PUFAs. Our results are relevant for the development of novel strategies for prevention and treatment of the metabolic syndrome.

This study was supported by grants from the Czech Science Foundation (301/10/1420) and the Ministry of Education, Youth and Sports (1M6837805002, COST BMB0602-OC08007) of the Czech Republic. Further support included grants from the European Commission (LSHM-CT-2004-005272, EXGENESIS) and from EPAX AS (Norway), as well as the research project AV0Z50110509. No other potential conflicts of interest relevant to this article were reported.

T.J. and M.R. performed experiments and wrote the manuscript. O.K., Z.M.J., D.M., V.K., M.H., and P.J. performed experiments. I.M. quantified tissue nucleotides. M.B. performed analysis of hepatic lipids and contributed to discussion. J.G. performed analysis of hepatic lipids. S.H. contributed to experiments on hepatocytes. T.E.J. contributed to analysis of muscle metabolism. P.F. introduced AMPKα2^−/− mouse model to the Prague laboratory and contributed to discussion. S.H. performed experiments on kidney cells. B.V. developed AMPKα2^−/− mice and contributed to discussion. J.K. wrote the manuscript.

The authors thank D. Grahame Hardie for the sheep AMPKα1 and AMPKα2 antibodies, Z. Szulc for the gift of N-palmitoyl-D-erythro-sphingosine (C17 base), and J. Jones (NMR Research Unit, Department of Biochemistry and Center for Neurosciences and Cell Biology, University of Coimbra, Coimbra, Portugal) for critical reading of the manuscript.