Abstract Mice models of Alzheimer's disease (APP/PS1) typically experience cognitive decline with age. G6PD overexpressing mice (G6PD-Tg) exhibit better protection from age-associated functional decline including improvements in metabolic and muscle functions as well as reduced frailty compared to their wild-type counterparts. Importantly G6PD-Tg mice show diminished accumulation of DNA oxidation in the brain at different ages in both males and females. To further explore the potential benefits of modulating the G6PD activity in neurodegenerative diseases, triple transgenic mice (3xTg G6PD) were generated, overexpressing APP, PSEN1, and G6PD genes. The cognitive decline characteristic of APP/PS1 mice was prevented in 3xTg G6PD mice, despite similar amyloid-β (Aβ) levels in the hippocampus. This challenges the dominant hypothesis in Alzheimer's disease (AD) etiology and the majority of therapeutic efforts in the field, based on the notion that Aβ is pivotal in cognitive preservation. Notably, the antioxidant properties of G6PD led to a decrease in oxidative stress parameters, such as improved GSH/GSSG and GSH/CysSSG ratios, without major changes in oxidative damage markers. Additionally, metabolic changes in 3xTg G6PD mice increased brain energy status, countering the hypometabolism observed in Alzheimer's models. Remarkably, a higher respiratory exchange ratio suggested increased carbohydrate utilization. The relative failures of Aβ-targeted clinical trials have raised significant skepticism on the amyloid cascade hypothesis and whether the development of Alzheimer's drugs has followed the correct path. Our findings highlight the significance of targeting glucose-metabolizing enzymes rather than solely focusing on Aβ in Alzheimer's research, advocating for a deeper exploration of glucose metabolism's role in cognitive preservation. Keywords: Alzheimer's disease, G6PD, Oxidative stress, Glucose metabolism, Amyloid-β, Pentose phosphate pathway 1. Introduction Alzheimer's disease (AD) pathophysiology is extremely complex and heterogeneous, involving the accumulation of senile plaques caused by abnormal amyloid-β (Aβ) metabolism and the accumulation of neurofibrillary tangles caused by tau hyperphosphorylation. There is also an increase in the levels of reactive oxygen species (ROS) which induces the transcription of pro-inflammatory genes and the release of cytokines and chemokines that cause neuroinflammation [[41]1]. Neuronal energy metabolism is also compromised in this disease. In particular, there is a decline in glucose metabolism which is one of the earliest and most common abnormalities observed in AD patients [[42]2]. The Pentose Phosphate Pathway (PPP) provides pentose availability to DNA and RNA synthesis, involves NADPH formation which confers protection against oxidative stress, and is an alternative to obtain glycolytic intermediaries such as glyceraldehyde-3-phosphate. Glucose-6-phosphate dehydrogenase, G6PD, is the enzyme at the crossover between glycolysis and the PPP [[43]3]. The G6PD enzyme is present in a wide range of tissues in mammals, with the greatest concentration found in immune cells and testes [[44]4]. While mice that lack the G6PD enzyme do not survive, hypomorphic G6PD alleles are common in humans, affecting about 5 % of the global population, and approximately 400 million people in the world [[45]5]. G6PD overexpressing mice (G6PD-Tg) are partially protected from some age-associated declines: they have improved glucose tolerance and insulin sensitivity, an increased median lifespan in females [[46]6], and old-G6PD overexpressing mice are less frail than their WT controls [[47]7]. Importantly G6PD-Tg mice show diminished accumulation of DNA oxidation (measured as 8-hydroxyguanosine) in the brain and other tissues such as the liver at different ages in both males and females [[48]6]. The fact that oxidative stress is associated with AD [[49]1,[50]8] prompted us to generate a triple transgenic mouse that overexpresses APP/PS1 and G6PD. The APP/PS1 mice are a double transgenic mice model that has been genetically engineered to express a hybrid mouse/human amyloid precursor protein called Mo/HuAPP695swe, as well as a mutated form of human presenilin 1 known as PS1-dE9, in their central nervous system neurons. These mutations have been linked to the development of an AD phenotype at an early age. We hypothesized that the antioxidant properties of G6PD could alleviate the pathological signs of the experimental disease. Our results show that the overexpression of G6PD rescues the cognitive decline caused by AD pathology in the 2xTg mice (APP/PS1). This protective effect is based on a dual mechanism targeting two hallmarks of AD: oxidative stress and the brain's energy deficit. To the best of our knowledge, this is the first time that overexpressing an enzyme involved in general carbohydrate metabolism shows a clear protective effect in an experimental AD model. 2. Material and methods Generation of triple transgenic (APP/PS-1/G6PD) mice. We crossed heterozygous B6C3-Tg(APPswe, PSEN1dE9)85Dbo/Mmjax, from now on (2xTg) mice with homozygous C57BL6/J-OlaHsd Tg-G6PD (G6PD-Tg) [[51]6]. The animals generated were, thus, triple transgenic for the APP, PSEN1, and G6PD genes from now on (3xTg G6PD), or transgenic only for G6PD. The 3xTg G6PD was used for further experiments. B6C3 wild-type (WT) mice were also used as controls. All the determinations were performed when mice were 12- to 14-month-old because the transgenic mice develop Aβ deposits in the brain by 6–7 months of age so both mutations are associated with early-onset Alzheimer's disease. We used both male and female mice. We did not see any difference between sexes in any of the parameters analyzed as shown in the supplementary materials (see [52]Figs. S1–S3) All the animals used in this study were raised and housed in the animal facility of the Faculty of Medicine of the University of Valencia under temperature conditions of 22 ± 2 °C and 55 ± 5 % humidity. All the experimental protocols have been approved by the Animal Ethics Committee of the University of Valencia (protocol numbers: A1549115230055 and A1549125701978). Functional assessment. Mice from the three groups were functionally evaluated at an age of 12.6 ± 0.5 months. Grip strength, motor coordination, and a fatigue resistance test were performed in these animals. Briefly, to assess grip strength a Grip Strength Meter (Panlab, Harvard Apparatus) consisting of a T-bar connected to a dynamometer was used. Mice were restrained so that they could grasp the bar with their forelimbs and the maximum force was recorded in three attempts [[53]9].To assess motor coordination, the Rotarod Test (Panlab model LE8205, Harvard Apparatus) was performed as previously described [[54]10]. An incremental treadmill test was performed using a treadmill (Panlab model LE8710MTS, Harvard Apparatus) with a 5 % incline to assess fatigue resistance. The test was performed starting with a speed of 10 cm/s for the first 4 min, increasing the speed by 4 cm/s every 2 min. The time it takes for the mice to reach fatigue and the speed reached were recorded [[55]11]. Cognitive assessment. The passive avoidance test was used to evaluate the learning and memory of the mice. A box with a compartment illuminated with a lamp and another in darkness separated by a gate is used. On the first day of the test or training, the mouse is placed in the lighted area with the door closed. After 60 s, the gate is opened and the time it takes for the mouse to move into the dark zone is recorded, at which point it is shocked at 0.5 mA for 3 s, and then the animal is taken back to its cage. The test is carried out 24 h after training, in which the mouse is placed back in the illuminated compartment, recording the time it takes to pass into the dark area; if it has not crossed after 300 s, the test is terminated [[56][12], [57][13], [58][14]]. The test was performed again at 7 days to assess implicit long-term memory. The Hebb–Williams maze was used to assess the learning ability and spatial memory of mice. It consists of a battery of mazes in which the time it takes the mice to find the exit is computed, such that a longer time required to complete them is associated with deterioration in cognitive status. Cold water (about 15 °C) is used as a stimulus to motivate the animal to look for the exit, where dry paper is placed. The test is carried out in 8 days, in each of which a different maze is used, increasing its difficulty or complexity. The first three days of the test are training or habituation days, in which the animal freely explores the environment, and no water is used. For each trial, the maximum time allowed to complete the test is 300 s; once that time has elapsed, the animal is taken out of the maze regardless of whether it has managed to find the exit, and the test is considered finished [[59][15], [60][16], [61][17]]. In vivo metabolic assessment. Respiratory metabolism was assessed by indirect calorimetry with the OxyletPro System (Panlab, Harvard Apparatus) in 13- to 14-month-old 2xTg (n = 5) and 3xTg G6PD (n = 11) mice. They were single-housed with ad libitum access to food and water and maintained at 20–22 °C under a 12:12 h light: dark cycle (light period 08.00–20.00). Oxygen consumption was determined by measuring oxygen concentration in air entering the chamber compared with air leaving the chamber. Measurement in each chamber was recorded for a total of 48 h to obtain the parameters of VO[2], VCO[2], energy expenditure, and respiratory quotient, calculated as VCO[2]/VO[2]. For the analysis of the data, the mean value of each of the parameters was calculated at each hour, and they were represented as a function of time. Glucose 6-P dehydrogenase activity. G6PD activity was determined spectrophotometrically by measuring the absorbance at 340 nm after the addition of NADP as described previously [[62]18] in cortex samples of 14-month-old WT (n = 12), 2xTg (n = 11) and 3xTg G6PD mice. Amyloid-β levels determination. Hippocampal Aβ levels were measured by an Aβ 1–42 enzyme-linked immunosorbent assay (ELISA) with reference KHB3441 (ThermoFisher). Sample preparation, processing, and detection were performed according to the manufacturer's instructions. Carbonylated proteins levels. Immunoblot detection of protein carbonyl groups in cortex samples was assessed using the ‘Oxyblot Protein Oxidation Detection Kit’ (Millipore, USA). Briefly, 20 μg of total protein was loaded into gels before electrophoretic separation and transfer onto PVDF membranes. Total protein carbonyls were quantified as densitometry of the blotting divided by the total density of the ponceau red staining. Specific proteins were visualized by enhanced chemiluminescence using a BioRad scanning densitometer and quantified with ImageJ software. Lipid peroxidation determination by HPLC. Lipid peroxidation was determined in cortex samples as described previously [[63]19]. Briefly, this method is based on the hydrolysis of lipid peroxides and subsequent formation of the adduct thiobarbituric acid (TBA) and malondialdehyde, MDA (TBA-MDA2). This adduct was detected by reverse phase HPLC and quantified at 532 nm. The chromatographic technique was performed in an isocratic mobile phase that is a mixture of 50 mM KH2PO4 (pH 6.8) and acetonitrile (70:30). Total Glutathione levels. Total GSH was determined as previously described [[64]20]. Briefly, the method is based on the catalytic action of GSH or GSSG in the reduction of Ellman reagent (DTNB) by a mixture of TPNH and glutathione reductase. The procedure measures the total glutathione (GSH + GSSG) content of unknown mixtures and is not subject to appreciable interference by the presence of other thiol components. Metabolomic analysis. Samples were prepared following a standard protocol for polar metabolites that is further described elsewhere [[65]21]. For the analysis of cortex samples from 2xTg and 3xTg G6PD mice, liquid chromatography equipment coupled to a high-resolution mass spectrometer with an Orbitrap UPLC-QExactive Plus detector was used. The chromatographic and mass spectrometry conditions were the following: UPLC column: Xbridge BEH amide 2.5 μM (2.1*150 mm) Waters, Chromatogram time: 25 min, Vinj: 5 μL, Column temperature: 25 °C, Autosampler temperature 4 °C, Flow: 105 μL/min, Mobile phase A = H[2]O 10 mM. Orbitrap parameters, Mode: ESI pos and neg, Event 1 m/z range: 70–700Da, Event 2, Range m/z:700–1700Da, Full Scan Resolution: 140,000, AGC:3e6, Maximum IT: 200 ms centroid acquisition. Metabolite feature extraction was carried out using the EI-Maven platform [[66]22]. Metabolite identification was performed using an in-house library based on MS/MS spectra and internal standards, which ensures the highest level of metabolite identification following the recommendations of the Metabolomic SocietySumner et al., 2007[[67]23]. Once annotated, the data sets were analyzed with MetaboAnalyst 5.0 software, including a Pathway enrichment analysis [[68]24]. Citrate synthase activity. Frozen cortex samples were homogenized with Tris 75 mM, EDTA 2 Mm and 0.1 % Triton-X (pH 7.4) buffer. The homogenates were sonicated and centrifuged at 1.500g and the supernatant were diluted and used for total protein and activity determination. Spectrophotometric cuvettes were used with 881.7 μL of previous buffer, 3.3 μL of acetyl-CoA (10 mg/mL), 100 μL of DTNB (1 mM), 10 μL of sample and 5 μL of oxalacetate (50 mM) for absorbance measuring during 5 min at 420 nm. Enzymatic activity results were relativized by amount of total protein, determined with Lowry method. Western blot. Homogenates of mice cortex samples were prepared using Tris 75 mM, 2 % SDS and 10 % glycerol (pH 7.4) containing phosphatases and proteases inhibitor cocktail, to which and Laemmli Buffer was added. Proteins were separated in SDS-polyacrylamide in gel electrophoresis (SDS-PAGE) and transferred to PVDF membranes. Primary antibodies against VDAC (1:1000, Cell Signaling, ref. 4866) and against cytochrome C (1:1000, Santa Cruz, ref. sc-13156) were used. Statistical methods. Values are expressed as the mean ± standard deviation (SD). The normal distribution of the samples was assessed by the Shapiro–Wilk or Kolmogorov-Smirnov tests. To compare two different groups, the unpaired Student's t-test was used, or the Wilcoxon test in case of a non-normal distribution. A volcano plot was also employed to show statistical significance (p-value) versus magnitude of change (fold change) when comparing the two groups. When comparing three different groups a one-way analysis of variance (ANOVA) or the Kruskal-Wallis analysis for non-normal distribution data was used. In the case of studying two independent variables, a two-way ANOVA test was used. 3. Results 3.1. G6PD overexpression rescues the cognitive impairment in Alzheimer's mice To assess cognitive function in mice and to evaluate their memory and learning, we employed two distinct tests: the passive avoidance and the Hebb-Williams tests. Within the Hebb-Williams tests, two mazes were utilized: Maze 1, an easy one, and Maze 5, a difficult one. A visual representation of both assessments is presented in [69]Fig. 1A. Fig. 1. [70]Fig. 1 [71]Open in a new tab G6PD overexpression reduces cognitive impairment in 2xTg mice. A) Cognitive tests design: Passive avoidance test (up), Hebb-Williams Maze 1 (left down), and Hebb-Williams Maze 5 (right down). B) Passive avoidance test results. N = 22–26 mice/group. ***p < 0.001 (3xTg G6PD vs 2xTg), ^+++ p < 0.001 (2xTg vs WT). C) Hebb-Williams easy maze (maze 1) results. N = 22–26 mice/group. *p < 0.05, **p < 0.01, ***p < 0.001 (3xTg G6PD vs 2xTg), ^+++ p < 0.001 (2xTg vs WT), ^##p < 0.01 (3xTg G6PD vs WT). D) Hebb-Williams difficult maze (maze 5) results. N = 22–26 mice/group. **p < 0.01, ***p < 0.001 (3xTg G6PD vs 2xTg),+p< 0.05, ^++ p<0.01, ^+++ p < 0.001 (2xTg vs WT). E) Hebb-Williams Easy vs Difficult Mazes results. N = 22–26 mice/group. ***p < 0.001 (3xTg G6PD vs 2xTg), ^+++ p < 0.001 (2xTg vs WT). Hebb-Williams mazes area under the curve. ***p < 0.001 (3xTg G6PD vs 2xTg), ^+++ p < 0.001 (2xTg vs WT), ^###p < 0.001 (3xTg G6PD vs WT). [72]Fig. 1B shows the results of the passive avoidance test. Our findings show that 2xTg mice exhibit the capacity to learn and avoid the deleterious region of the maze 24 h after the training period. However, a week later, these mice exhibited diminished implicit long-term memory, indicative of a tendency to “forget” the associated danger in the test (the shock). On the contrary, the healthy controls (WT) and the 3xTg G6PD mice consistently recall the imperative avoidance behavior thus showing a higher memory retention. Panels C and D show the outcomes of the Hebb-Williams test, depicting the time taken by mice to navigate either Maze 1 or Maze 5. The disparities between 3xTg G6PD and 2xTg mice are statistically significant, with the latter displaying inferior outcomes. Notably, these differences are more prominent when confronted with higher difficulty levels, particularly in Maze 5, as highlighted in panel E. Panel F elucidates the cumulative cognitive performance, measured as areas under the curve, for the three mouse groups in both Maze 1 and 5. The data indicate that the cognitive decline observed in 2xTg mice is ameliorated by the overexpression of G6PD. This underscores the pivotal role of G6PD in preserving cognitive function in the face of cognitive challenges in our model. 3.2. G6PD overexpression does not decrease the hippocampal Aβ levels In elucidating the observed alterations in cognitive function, we analyzed hippocampal Aβ levels. Our results reveal a marked increase in Aβ content in Alzheimer's transgenic mice compared to WT controls, as illustrated in [73]Fig. 2A. It is noteworthy that WT mice, even in advanced life stages, do not exhibit Aβ plaques. The Aβ levels in the brains of 3xTg G6PD animals were found not to be statistically different from those in 2xTg mice. This seemingly paradoxical result, significant cognitive effects in the absence of discernible changes in Aβ levels, suggests that Aβ may not be a pivotal factor in maintaining cognition, at least in this experimental model. The lack of direct correlation between Aβ levels and cognitive decline has been also reported in preclinical as well as clinical studies [[74][25], [75][26], [76][27], [77][28], [78][29]] Fig. 2. [79]Fig. 2 [80]Open in a new tab Cortical G6PD activity and Aβ42 hippocampal levels in WT, 3 Tg-G6PD, and 2xTg mice. A) Hippocampal Aβ42 levels ***p < 0.001. N = 13–14 mice/group B) Cortex G6PD activity. ***p < 0.001. N = 24–30 mice/group. Furthermore, we investigated the G6PD activity in the cortex across the various mouse cohorts. Notably, the G6PD activity in the WT and 2xTg mice did not exhibit a statistically significant difference. In contrast, the 3xTg G6PD mice, displayed a significantly elevated activity of this enzyme, as depicted in [81]Fig. 2B. 3.3. Enhanced physical performance in G6PD overexpressing mice: comprehensive evaluation of body weight, grip strength, endurance capacity, and motor coordination In our assessment of various physical parameters, including body weight ([82]Fig. 3A), grip strength ([83]Fig. 3B), endurance capacity ([84]Fig. 3C), and motor coordination ([85]Fig. 3D), across all experimental groups, a consistent trend emerges. The overexpression of G6PD in the 2xTg mice consistently leads to improved physical performance in the animals. These findings align with our prior investigations involving ordinary mice that solely overexpressed G6PD, revealing a protective effect against age-associated functional decline and frailty [[86]7]. The collective data underscore the potential benefits of G6PD overexpression in promoting overall physical well-being and performance in murine models. Fig. 3. [87]Fig. 3 [88]Open in a new tab 3xTg G6PD mice exhibit an improved physical performance when compared with WT and 2xTg mice. A) Body weight. B) Grip Strenght. C) Aerobic Resistance. D) Motor coordination. N = 11–35 mice/group. *p < 0.05, **p < 0.01, ***p < 0.001. 3.4. Assessing redox status in 2xTg mice. Insights into G6PD-mediated antioxidant effects Given the well-established association between advanced AD and oxidative stress [[89]30], we examined the redox status in the 3xTg G6PD mice, comparing it with WT and 2xTg counterparts. In cortex samples we measured oxidative stress parameters, GSH/GSSG and GSH/CysSSG, as well as markers of oxidative damage, MDA, and protein carbonyls. Our analysis revealed changes in oxidative stress but not in oxidative damage markers between the different experimental groups. While the total glutathione levels in the brain cortex remained unchanged in both 2xTg and 3xTg G6PD mice, the glutathione/redox ratios showed that the 3xTg G6PD mice exhibited a less oxidized state in the cortex, when compared to their 2xTg counterparts. These results, shown in [90]Fig. 4 (panels C, D, and E), underscore the potential antioxidant effects of G6PD overexpression, offering a glimpse into its role in modulating the redox equilibrium in the context of AD. Fig. 4. [91]Fig. 4 [92]Open in a new tab Cortical oxidative stress and oxidative damage markers in WT, 2xTg, and 3xTg G6PD animals. A) MDA levels. N = 22–37 mice/group B) Carbonylated proteins levels. N = 12 mice/group C) Total GSH. N = 12–18 mice/group D) GSH/GSSG ratio measured from metabolomic determinations. N = 25–27 mice/group E) GSH/CysSSG ratio measured from metabolomic determinations. N = 25–27 mice/group * p < 0.05, **p < 0.01. However, we did not find a significant increase in a marker of lipid peroxidation or protein carbonylation in either 2xTg or 3xTg G6PD mice, as illustrated in [93]Fig. 4, panels A and B. These findings align with our prior observations that oxidative damage typically manifests in advanced stages of the disease [[94]31], suggesting a nuanced temporal relationship between oxidative stress and Alzheimer's progression. 3.5. G6PD overexpression in 2xTg mice modulates energy expenditure and substrate utilization G6PD serves a crucial role by not only generating reducing equivalents in the form of NADPH but also redirecting glucose toward pentose production rather than pyruvate synthesis. To unravel the metabolic impact of G6PD overexpression in the 2xTg mice, we conducted in vivo experiments, evaluating whole-body energy expenditure (EE) and the respiratory exchange ratio (RER), defined as the ratio of CO[2] production to O[2] utilization. [95]Fig. 5 shows that the EE and RER values were lower during the light vs dark cycles in all the animals independently of the genotype. The measurements recorded across the two light and dark cycles were averaged by analyzing the total area under the curve. Panels A and B show that G6PD overexpression leads to a significant reduction in energy expenditure among Alzheimer's mice. This effect, quantified in Panel 5B, is estimated to be approximately 5 % of the value observed in the 2xTg mice without the G6PD overexpression. Fig. 5. [96]Fig. 5 [97]Open in a new tab 3xTg G6PD mice show an enhanced glucose metabolism. A) Energy expenditure (EE) over 48 h (N = 5–11 mice/group), B) Area under the curve. C) Respiratory exchange ratio over 48h (RER) (N = 5–11 mice/group). D) Area under the RER curve. (N = 5–11 mice/group). ***p < 0.001. During the light cycle the 3xTg G6PD mice consistently exhibit a substantial increase in RER, approaching values around 0.9, as depicted in [98]Fig. 5C. In contrast, 2xTg animals maintain values around 0.75, indicative of a predominant reliance on lipid consumption. The fact that the 3xTg G6PD mice display a RER closer to one indicates a heightened utilization of carbohydrates as the primary fuel source [[99]32]. 3.6. Metabolic changes associated with overexpression of G6PD in Alzheimer's mice [100]Fig. 6 shows a metabolomic analysis of cortex samples from both 2xTg and 3xTg G6PD animals. The Volcano plot in [101]Fig. 6A reveals significant elevations in five key metabolites—phosphocreatine, ATP, acetyl CoA, GTP, and 2-hydroxy glutarate—in the 3xTg G6PD when compared with 2xTg. These metabolites, are connected to the Krebs cycle, signifying an augmented availability of carbon units for the cycle and an enrichment of high-energy phosphate-containing molecules, as exemplified in Panel 6B. Fig. 6. [102]Fig. 6 [103]Open in a new tab 3xTg G6PD mice show a boost in brain energy metabolism. A) Volcano plot shows, in blue, the metabolites found at different levels (fold-change>2, p < 0.05) in 3xTg G6PD vs 2xTg mice cortex samples. Blue dots represent the metabolites found at different levels for the comparation of 3xTg G6PD vs 2xTg mice cortex samples (fold-change >2, p < 0.05. B) Representation of the main pathways in which the metabolites selected are involved. C) Adenylate Energy Charge Ratio in 3xTg G6PD and 2xTg mice cortex samples. D) VDAC levels in WT, 2xTg and 3xTg G6PD mice cortex. E) CytC levels in WT, 2xTg and 3xTg G6PD mice cortex samples. F) Citrate synthase activity in WT, 2xTg and 3xTg G6PD mice cortex samples. G) NAD^+/NADH ratio in 3xTg G6PD and 2xTg mice cortex samples. H) NADP^+/NADPH ratio in 3xTg G6PD and 2xTg mice cortex samples. (N = 24–26 mice/group). *p < 0.01, **p < 0.05. (For interpretation of the references to colour in this figure legend, the