Opposing Effects of ApoE2 and ApoE4 on Glycolytic Metabolism in Neuronal Aging Supports a Warburg Neuroprotective Cascade against Alzheimer’s Disease

Apolipoprotein E4 (ApoE4) is the most recognized genetic risk factor for late-onset Alzheimer’s disease (LOAD), whereas ApoE2 reduces the risk for LOAD. The underlying mechanisms are unclear but may include effects on brain energy metabolism. Here, we used neuro-2a (N2a) cells that stably express human ApoE isoforms (N2a-hApoE), differentiated N2a-hApoE neuronal cells, and humanized ApoE knock-in mouse models to investigate relationships among ApoE isoforms, glycolytic metabolism, and neuronal health and aging. ApoE2-expressing cells retained robust hexokinase (HK) expression and glycolytic activity, whereas these endpoints progressively declined with aging in ApoE4-expressing cells. These divergent ApoE2 and ApoE4 effects on glycolysis directly correlated with markers of cellular wellness. Moreover, ApoE4-expressing cells upregulated phosphofructokinase and pyruvate kinase with the apparent intent of compensating for the HK-dependent glycolysis reduction. The introduction of ApoE2 increased HK levels and glycolysis flux in ApoE4 cells. PI3K/Akt signaling was distinctively regulated by ApoE isoforms but was only partially responsible for the ApoE-mediated effects on HK. Collectively, our findings indicate that human ApoE isoforms differentially modulate neuronal glycolysis through HK regulation, with ApoE2 upregulating and ApoE4 downregulating, which markedly impacts neuronal health during aging. These findings lend compelling support to the emerging inverse-Warburg theory of AD and highlight a therapeutic opportunity for bolstering brain glycolytic resilience to prevent and treat AD.


Introduction
To date, the molecular pathogenesis for most Alzheimer's disease (AD) cases is unknown, and no effective treatment exists. For over three decades, the accumulation of amyloid beta (Aβ) was considered by many to serve as the initiating trigger. However, the failure of numerous Aβ-targeted interventions questions the validity of such therapeutic approaches and stresses the need to broaden the scope in search of new paths to the prevention and treatment of AD.
The presence of an apolipoprotein E4 (ApoE4) allele increases the risk of developing late-onset AD (LOAD). In the brain, ApoE is predominantly produced by astrocytes, while under certain circumstances, such as traumatic brain injury, neurons will also synthesize and secrete ApoE [1]. ApoE transports lipids, promotes neurite outgrowth, and facilitates neuronal communication [2]. ApoE exists in three major isoforms conferring profoundly distinct influences in AD etiology. ApoE2 is a rare variant present in approximately eight percent of the general population and is viewed as an exceptionally neuroprotective variant against AD. ApoE3, the most common form represented in of 1 × 10 5 /mL in DMEM supplemented with 2% FBS and containing 20 µM all-trans-RA (Sigma-Aldrich, Inc., St. Louis, MO, USA). The medium was replaced with fresh medium containing RA every 2 days. After 4 days, differentiated neuron-like N2a-hApoE cells (dN2a-hApoE) were proceeded to further studies.

Animals
Animal Use Statement (AUS # 220-08) was approved by the IACUC at the University of Kansas on 10 January 2019. Six-month-old ApoE2, ApoE3, and ApoE4 gene-targeted replacement (ApoE2-TR, ApoE3-TR, and ApoE4-TR) female mice were obtained from Taconic [20]. These ApoE-TR mice express human ApoE2, ApoE3 or ApoE4 under the control of the endogenous mouse promoter. Glial cells express human ApoE in a native conformation at physiologically regulated levels, and in the same temporal and spatial pattern as endogenous mouse ApoE [24]. The humanized ApoE knock-in (ApoEKI) mice were purchased from the Jackson Laboratory (stock #029018, #027894) (Bar Harbor, ME, USA). These ApoEKI mice express ApoE isoforms from the endogenous ApoE locus at a physiological level with fewer breeding restrictions [25]. Mice were bred and housed until reaching 6-7 months old. Both female and male ApoEKI mice were used. Mice were euthanized through carbon dioxide inhalation according to protocol. Cortical tissues were harvested and stored at −80 • C.

Hexokinase Activity Assay
The hexokinase activity assay was conducted as previously described [21]. Briefly, cell lysates were harvested as described above. Hexokinase activity was measured as the total glucose phosphorylating capacity of the lysate, based upon the reduction of NAD + through a coupled reaction with glucose-6-phosphate dehydrogenase. Results were measured spectrophotometrically by monitoring the increase in absorbance at 340 nm every 1 min for 30 min under 37 • C. The initial linear slope of curve was used to determine ∆A/min for further calculation. Hexokinase assay solution was prepared with 13.3 mM MgCl2, 0.112 M glucose, 0.55 mM adenosine 5 triphosphate, 0.227 mM NAD + , and 1 IU/mL glucose-6-phosphate dehydrogenase in 0.05 M Tris-HCl buffer, pH 8.0 as described previously [26]. After 6-8 min of incubation at room temperature to reach equilibrium, 15-20 µg cell extracts were added into 150 µL assay solution to initiate the reaction. Results were normalized to total cellular protein content using the BCA assay.

Glycolytic Stress Test
Extracellular flux (XF) analysis (Agilent Technologies) was used to evaluate the glycolytic activity. Basal glycolysis and glycolytic capacity were measured by assessing the extracellular acidification rate (ECAR). Cells expressing human ApoE isoforms were seeded at a density of 8000-10,000 cells/well on Seahorse XF96 culture plates. On the experiment day, glycolytic stress test assay medium was prepared as XF base medium (Agilent Technologies, Santa Clara, CA, USA) containing 2 mM glutamine, pH 7.4. The assay was initiated by replacing the growth medium with pre-warmed assay medium and incubating the cell culture plate at 37 • C without CO 2 at least 1 h prior to the measurement. Glycolytic parameters of human ApoE2/3/4-expressing cells were then measured using the Seahorse XF Analyzer following successive injections of 10 mM glucose, 2 µM oligomycin, and 50 mM 2-Deoxy-D-glucose to each well. Cell culture plates were saved after the assay and used for the normalization of the results by assessing the protein content in each well.

LIVE/DEAD Cell Assay
N2a-hApoE cells were seeded on ibiTreat #1.5 polymer coverslip µ-Slide (Ibidi, Martinsried, Germany) with complete growth medium. The medium was replaced with fresh medium containing G418 every two days. On day 4, the culture medium was removed carefully and replaced with the LIVE/DEAD cell assay reagents (Thermo Fisher Scientific) at the working concentration, together with Hoechst dye. 10× images were obtained under the Leica DMI4000 B inverted microscope with FITC (495-519 nm) and Texas Red (595-605 nm) filters. 10-15 images were obtained for each isoform in each passage with similar cell density. Hoechst staining and dead cells were used as controls.

Phase Contrast Imaging
Cellular appearances were evaluated and compared using phase contrast imaging. N2a-hApoE cells were seeded in 35 mm glass-bottom poly-D-lysine pre-coated culture dishes (MatTek Corporation, Ashland, MA, USA). Cells were maintained for 4 days and were replaced with fresh medium every 2 days. Phase contrast images were acquired using an Olympus IX81 inverted epifluorescence microscope equipped with a cage incubator for maintaining precise temperature, humidity and CO 2 control. Slidebook Software Version 6.0 was used for microscope control, image acquisition, image processing, and data analysis (Intelligent Imaging Innovations, Inc., Denver, CO, USA). Images were acquired using a 40× phase contrast magnification on day 4 under 37 • C and 5% CO 2 .

Metabolic Activity Assay
A cell permeable resazurin-based solution, PrestoBlue ® (Invitrogen Life Technologies, Carlsbad, CA, USA), which measures the reducing power of viable cells, was used as an indicator of cellular metabolism. 10 µL of PrestoBlue reagent was directly added to cells in 90 µL of culture medium in a 96-well plate. The plate was incubated at 37 • C for 1 h and fluorescence was read at Ex/Em 528/20,600/40 nm using a plate reader (BioTek). Wells containing only cell culture media (no cells) were used as background control wells. Results were normalized to protein content of each well. Higher fluorescence units indicate greater metabolic activity.

Transfection of ApoE2 in N2a-ApoE4 Cells
Human ApoE2 cDNA clones expressed in the mammalian vector pCMV6 with C-terminal MycDDK Tag were obtained as previously described [20]. An empty pCMV6 vector was used in the control group. Briefly, 2 µg human ApoE2 cDNA was transfected to N2a cells stably expressing human ApoE4 using jetPRIME ® transfection reagent according to the manufacturer manual. Fresh complete medium was replaced 4 h after transfection. Cells were then continuously incubated for 48-72 h before proceeding to further experiments.

Protein Carbonylation Assay
Protein carbonylation levels in mouse cortex lysates were measured using the OxiSelect Protein Carbonyl Immunoblot Kit (Cell Biolabs Inc., San Diego, CA, USA) according to the manufacturer's instructions with minor modifications. Cortex lysate samples were obtained as described in the Western Immunoblotting method section. Briefly, 20 µg protein of each sample was loaded and separated on NuPAGE TM 4-12% Bis-Tris Gel (Thermo Fisher Scientific), followed by transferring onto 0.2 µm pore-sized PVDF membranes (Bio-Rad). After transferring, 2,4-dinitrophenylhydrazine (DNPH) derivatization was conducted as follows: membranes were immersed in 100% methanol for 15 s, dried out for 5 min at room temperature, equilibrated in 1 × TBS containing 20% methanol for 5 min, washed with 2N HCl for another 5 min, and then incubated in 1 × DNPH for exactly 5 min followed by 2N HCl washing for three times. After DNPH derivatization, membranes were washed five times in 100% methanol, 5 min each. Blocking was performed by incubating the DNPH-treated membranes with 5% non-fat milk in 1 × TBS for 1 h at room temperature. Membranes were washed with 1 × TBST three times before incubating with Rabbit Anti-DNP antibody in 1% non-fat milk overnight at 4 • C (anti-DNP 1:1000, Cell Biolabs Inc.). On the next day, after washing with 1 × TBST three times, the membranes were incubated with the secondary antibody in 5% non-fat milk for one hour at room temperature, followed by another four times of 1 × TBST washing. Densitometry of the immunolabeling was measured using C-Digit Blot Scanner (LI-COR) after ECL application. Results were normalized to internal loading control protein, β-actin.

Statistical Analyses
GraphPad Prism 6 (GraphPad Software, La Jolla, CA, USA) was used in statistical analyses. Data represent the group mean ± SEM. Group comparisons were analyzed by one-way analysis of variance (ANOVA) with Tukey's post hoc test or Student's t-test. A p value lower than 0.05 was considered statistically significant.

HK Expression Remained
Relatively Stable in ApoE2-and ApoE3-Expressing Cells but Exhibited a Gradual Decrease with Increasing Passages in ApoE4-Expressing Cells N2a cells that stably expressed human ApoE2, ApoE3, or ApoE4 were generated as previously described [21]. A Western blot was performed to analyze the ApoE protein expression level in representative cell lines for each of the three isoforms at passage one (P1). Arrows indicate the cell lines of interest (ApoE2 no. 6, ApoE3 no. 16, and ApoE4 no. 13) that showed comparable levels of ApoE expression and were thus selected as the final set for the following studies ( Figure 1a). ApoE protein levels in the three chosen cell lines at P5 remained consistent ( Figure 1b). In addition, expression levels of both HK isoforms, HK1 and HK2, in these lines at P5 showed no significant differences (Figure 1c,d). Based on these results, we chose P5 as the starting point of the following time-course study designed to evaluate the interactive impact of ApoE and aging on HK expression. Cell lysates were harvested at every other passage, beginning at P7, and probed for HK1, HK2, ApoE, and glucose transporter type 4 (GLUT4). The expression of both HK1 and HK2 in ApoE4-expressing cells gradually decreased with increasing passage numbers starting at P11, and the reduction culminated at P15, the final passage studied. In contrast, the expression of HK1 and HK2 remained relatively unchanged from P7 to P15 in ApoE2and ApoE3-expressing cells (Figure 1e,f). To determine whether HK changes could be caused by the fluctuation of cellular ApoE levels, human ApoE expression in the same cell lysates was also examined and was found to be unaltered across all the passages in all the three ApoE lines (Figure 1g). Similarly, GLUT4 expression was unchanged with increasing passages in all three lines ( Figure 1h). Taken together, these results indicate that HK is differentially regulated by human ApoE isoforms, and it appears to be chronically and negatively affected by ApoE4. Moreover, the data indicate that the regulation of HK by ApoE isoforms is independent of protein levels of ApoE and GLUT4 in the cell.

ApoE2 Upregulated Whereas ApoE4 Downregulated HK, Which Occurred Concurrently with Downregulation of PFKP and PKM1 by ApoE2 and Upregulation by ApoE4
Following up on the observation that a significant decrease in HK protein expression started at P11 in ApoE4-expressing cells, we next analyzed the expression and activity of HK in the three ApoE cell lines at P9, P11, and P13. At all three passages, compared to ApoE3-expressing cells, ApoE2-expressing cells exhibited a higher expression of both HK1 and HK2, whereas ApoE4-expressing cells showed a significantly lower expression (Figure 2a,b). Consistent with the expression data, ApoE2-expressing cells exhibited a higher level of HK enzymatic activity, whereas ApoE4-expressing cells demonstrated decreased activity when compared to ApoE3 cells ( Figure 2c). Apart from HK, phosphofructokinase (PFK) and pyruvate kinase (PK) also serve as the key rate-limiting enzymes involved in glycolysis. PFK is subject to allosteric regulation by the energy state of the cell and is stimulated by decreases in the ATP/AMP ratio (i.e., low energy state) [27]. The platelet isoform of PFK (PFKP) is the predominant form expressed in the brain [28,29]. PK, with the PKM1 isoform as the primary variant expressed in neurons, shows the sigmoidal kinetics in the allosteric inactivation by increased ATP and alanine [30,31]. Analyses of PFKP and PKM1 expression in the three ApoE cell lines showed that PFKP levels were significantly increased in ApoE4-expressing cells and decreased in ApoE2-expressing cells at all three passages, P9, P11, and P13 ( Figure 2d). PKM1 exhibited similar changes-increases in ApoE4 cells and decreases in ApoE2 cells-although there was a slight delay starting at P13, and the changes culminated at P15-P17 ( Figure 2e). Significant upregulation of both PFKP and PKM1 reflected the abnormally low energy state in the ApoE4-expressing cells, likely caused by downregulation of HK. Glucose transporters exist in several isoforms, including GLUT1-GLUT4, and GLUT1 and GLUT3 are found to be downregulated in AD brains [32]. In our analyses, all three GLUTs highly expressed in the brain, GLUT1, GLUT3, and GLUT4, showed no differences among the three ApoE cell lines at P11 (Figure 2f). In summary, these results indicate that ApoE2 upregulated whereas ApoE4 downregulated HK, which occurred concurrently with downregulation of PFKP and PKM1 by ApoE2 and upregulation by ApoE4. However, GLUTs are not affected by ApoE isoforms.   , and GLUT4 protein levels showed no significant differences among the three ApoE cell lines at P11. n = 5-8/group. Results were normalized to the ApoE3 group in each passage. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. N.S.: not significant.

ApoE2-Expressing Cells Exhibited Enhanced Glycolytic Activity When Compared to ApoE4-Expressing Cells, Which Correlated with the General State of Cellular Health
Because HK is the first rate-limiting enzyme functioning as the "pacemaker" in glycolysis, glycolytic phenotypes in the three ApoE cell lines were then analyzed using a Seahorse glycolysis stress test kit. Starting at P11, the basal glycolytic rate exhibited the lowest in ApoE4-expressing cells. At P15, when cells reached a more advanced age, a greater decrease in the maximum glycolytic capacity induced by oligomycin occurred in Figure 2. HK was upregulated in ApoE2-expressing cells and downregulated in ApoE4-expressing cells, which occurred concurrently with downregulation of PFKP and PKM1 by ApoE2 and upregulation by ApoE4. (a-c) Protein levels of both isoforms and activity of HK were upregulated in ApoE2-expressing cells and downregulated in ApoE4-expressing cells at all three passages, P9, P11, and P13. n = 5-9/group. (d,e) PFKP and PKM1 protein levels were significantly downregulated in ApoE4-expressing cells compared to ApoE2-and ApoE3-expressing cells. n = 4-7/group. (f) GLUT1, GLUT3, and GLUT4 protein levels showed no significant differences among the three ApoE cell lines at P11. n = 5-8/group. Results were normalized to the ApoE3 group in each passage. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. N.S.: not significant.

ApoE2-Expressing Cells Exhibited Enhanced Glycolytic Activity When Compared to ApoE4-Expressing Cells, Which Correlated with the General State of Cellular Health
Because HK is the first rate-limiting enzyme functioning as the "pacemaker" in glycolysis, glycolytic phenotypes in the three ApoE cell lines were then analyzed using a Seahorse glycolysis stress test kit. Starting at P11, the basal glycolytic rate exhibited the lowest in ApoE4-expressing cells. At P15, when cells reached a more advanced age, a greater decrease in the maximum glycolytic capacity induced by oligomycin occurred in ApoE4-expressing cells; in contrast, ApoE2-expressing cells exhibited a significantly higher readout than both ApoE3-and ApoE4-expressing cells (Figure 3a-d). These data indicate that ApoE2-expressing cells harnessed the most robust glycolytic activity, whereas ApoE4-expressing cells possessed a deficient phenotype that became increasingly severe in aged cells. After having observed the distinctive glycolytic profiles exhibited in the three ApoE cell lines, we next investigated whether these differences influenced the general health of the cells. Phase-contrast images revealed that ApoE2 cells presented with an evidently healthier appearance, whereas ApoE4 cells tended to aggregate into clumps, indicative of cells undergoing deteriorating changes. In line with the temporal development in the glycolytic changes, aggregation of ApoE4 cells started at P11 and worsened at P15 (Figure 3e). Moreover, consistent with the cell appearance, ApoE4 cells exhibited a similar time-dependent decline in the overall cellular metabolism assessed by a Resazurin-based assay that measures the reducing power of viable cells (Figure 3g). Despite the deterioration of general health, as clearly indicated by both cellular appearance and metabolic activity, a LIVE/DEAD cell staining assay did not find significant cell death in any of the three ApoE lines at up to P15 (Figure 3f). Therefore, it is possible that the cell aggregation observed in ApoE4 cultures could be caused by compromises to cell membranes that promoted cell clumps, and this detrimental process could accumulate into and culminate in intracellular damages, ultimately leading to cell death. These results, together with the data presented in Figure 2, indicate a strong and positive correlation between the cell's glycolytic strength and the cell's overall health status.

Differential Regulation of HK and Glycolytic Activity by ApoE Isoforms Was Also Observed in Differentiated Neuron-like Cells
To study ApoE effects on neurons, we differentiated N2a-hApoE cells at P9, P11, and P13 by a 96 h treatment with retinoic acid (RA). Phase-contrast images demonstrated elaboration of cellular processes corresponding to the development of neuron-like morphologies, confirming the neuronal phenotype of RA-induced differentiated human ApoE-expressing cells (Figure 4a). Consistent with the data presented above, ApoE levels were no different among the three isoforms; however, expression of both HK1 and HK2, and HK activity, were upregulated in ApoE2-expressing and downregulated in ApoE4-expressing neuronal cells compared to ApoE3 neuronal cells (Figure 4b-e). Similarly, PFKP levels were considerably increased in ApoE4-expressing neuronal cells, pointing to the low energy state in those cells (Figure 4f). Moreover, consistent with the data generated in non-differentiated cells, PKM1 remained at similar levels among the three groups at P11 (Figure 4g); we speculated that significant differences could emerge when cells reached an advanced age. A Seahorse glycolysis stress test was further conducted. As predicted, at P11 and P13, 96 h RA-induced differentiated ApoE4-expressing neuronal cells exhibited the lowest level on both the basal glycolytic rate and the maximum glycolytic capacity compared to ApoE3 and ApoE4 neuronal cells (Figure 4h-j). In summary, these results demonstrate that HK and glycolytic metabolism are differentially regulated in human ApoE-expressing neuron-like cells in an age-dependent manner, further confirming our earlier findings in postmitotic neurons.

Introduction of ApoE2 Ameliorated HK Deficits and Further Improved Glycolytic Metabolism in ApoE4-Expressing Cells
Our findings of strong correlations between cellular glycolytic status and general wellness led us to hypothesize that introduction of the ApoE2 protein into ApoE4-expressing cells would reverse or ameliorate ApoE4-associated HK deficit and glycolytic hypometabolism. To test this hypothesis, N2a-hApoE4 stable cells were transfected with mammalian expression vector encoding human ApoE2 (ApoE4 + ApoE2) or empty vector (ApoE4 + V). 48-72 h after transfection, cells were analyzed for HK protein expression, HK enzymatic activity, and cellular responses to glycolytic stresses. Transfection induced a nearly three-fold higher expression of ApoE2 protein in ApoE4-expressing cells than in ApoE + V cells (Figure 5a). Both HK1 and HK2 protein levels, as well as HK activity, were significantly increased in ApoE4 + ApoE2 cells (Figure 5b-d). In a Seahorse glycolytic stress test, ApoE4 + ApoE2 cells exhibited drastically increased ECAR readout for both the basal glycolytic rate and the maximum glycolytic capacity when compared to ApoE4 + V cells. However, expression of ApoE3 in ApoE4 cells did not exert a rescue effect, indicating the unique neuroprotective property of ApoE2 (Figure 5f-i). Collectively, these data indicate that introduction of ApoE2 could rescue ApoE4-associated HK expression deficits and improve glycolytic function.

PI3K/Akt Signaling Activity Appeared to Be Responsible for ApoE2-Mediated Upregulation of HK2 but Not of HK1 Expression
PI3K/Akt signaling cascades play important roles in cell metabolism, growth, proliferation, and survival, and have been shown to be differentially affected by ApoE isoforms [33,34]. Additionally, PI3K/Akt-dependent HK2 expression has been shown to contribute to cellular bioenergetics and survival [35]. To understand the regulatory mechanism leading to ApoE-mediated effects on HK expression, we next examined the PI3K/Akt activity in two different human ApoE-expressing mouse models-hApoE-TR mice obtained from Taconic and hApoEKI mice obtained from JAX Laboratories-and in N2a-hApoE stable cell lines at P11. Consistent with the previous report [36], we detected isoformdependent differences in ApoE protein expression in the brains of both ApoE mouse models at 6-7 months of age (Figure 6a,c). The ratio of phosphorylated Akt/total Akt (pAkt/tAkt) protein expression levels also showed isoform-dependent differences, upregulated in ApoE2 mouse brains, and downregulated in ApoE4 mouse brains (Figure 6b,d). Consistent with the mouse data, the ratio of pAkt/tAkt expression levels was the highest in N2a-hApoE2 stable cells and the lowest in N2a-hApoE4 cells (Figure 6e). To further examine the role of the PI3K/Akt singling activity in the mediation of ApoE-dependent HK expression, we treated ApoE2 stable cells at P11 with the PI3K inhibitor, LY294002. Exposure of cells to LY294002 for 24 h resulted in a dose-dependent decrease in the pAkt/tAkt ratio along with HK2 but not HK1 expression (Figure 6f-h). We speculated that the increases in total Akt expression could be a compensatory response of cells to pAkt deficits. Considering the well-established role of PI3K/Akt in oxidative stress, we measured total protein carbonylation in cortical tissue collected from 6-7-month-old ApoE3KI and ApoE4KI mice of both sexes. Protein carbonylation is the most common form of protein oxidation, which can be induced directly by ROS or indirectly by secondary byproducts of oxidative stress [37]. As expected, we observed a slightly increased level of carbonylated proteins in ApoE4 brains compared to that in ApoE3 brains (Figure 6i-j). In summary, these results confirm that ApoE isoforms differentially modulate PI3K/Akt signaling cascades; however, PI3K/Akt activity appears to be only responsible for ApoE2-mediated upregulation of HK2 but not of HK1.   HK1 and HK2, and HK activity, were significantly higher in ApoE4 cells transfected with ApoE2 (ApoE4 + ApoE2) compared to ApoE4 cells transfected with empty vector (ApoE4 + V). n = 2-5/group. (e) Glycolytic readouts from a Seahorse stress test were markedly increased in ApoE4 + ApoE2 cells compared to ApoE4 + V cells. n = 8/group. (f-i) N2a-hApoE4 stable cells were transfected with human ApoE2 or ApoE3 plasmids and maintained for 48-72 h. Expression of ApoE2 but not ApoE3 increased HK expression in ApoE4-expressing cells. n = 3-7/group. Results were normalized to the ApoE4 + V group. * p < 0.05, ** p < 0.01, *** p < 0.001, **** p < 0.0001. Figure 6. ApoE-dependent PI3K/Akt signaling activity modulated HK2 but not HK1 protein expression. (a-d) ApoE and pAkt/tAkt ratio were upregulated in ApoE2 and ApoE3 mouse cortex and downregulated in ApoE4 mice. Human ApoE-TR mice were obtained from Taconic; human ApoEKI mice were obtained from JAX Laboratories; both sets of mice were at ages of 6-7 months. Results were normalized to the ApoE3 mice group. n = 4-7 mice/group. (e) Consistent with the mouse data, pAkt/tAkt ratio was significantly upregulated in N2a-hApoE2 stable cells and downregulated in N2a-hApoE4 cells at P11. Results were normalized to N2a-hApoE3 cells. n = 6/group. (f-h) N2a-hApoE2 cells at P11 were treated with serial concentrations of the PI3 inhibitor, LY294002, for 24 h; cell lysates were analyzed for changes in pAkt/tAkt expression ratio and protein levels of both HK1 and HK2. LY294002 induced a dose-dependent decrease in pAkt/tAkt ratio and HK2 expression; however, HK1 expression was unaltered. Results were normalized to the vehicle-treated group. n = 2-6/group. (i,j) Total protein carbonylation levels were slightly increased in 6-7-month-old ApoE4 mouse cortex compared to ApoE3 mice. n = 6 mice/group. * p < 0.05, ** p < 0.01, **** p < 0.0001. N.S.: not significant. Figure 6. ApoE-dependent PI3K/Akt signaling activity modulated HK2 but not HK1 protein expression. (a-d) ApoE and pAkt/tAkt ratio were upregulated in ApoE2 and ApoE3 mouse cortex and downregulated in ApoE4 mice. Human ApoE-TR mice were obtained from Taconic; human ApoEKI mice were obtained from JAX Laboratories; both sets of mice were at ages of 6-7 months. Results were normalized to the ApoE3 mice group. n = 4-7 mice/group. (e) Consistent with the mouse data, pAkt/tAkt ratio was significantly upregulated in N2a-hApoE2 stable cells and downregulated in N2a-hApoE4 cells at P11. Results were normalized to N2a-hApoE3 cells. n = 6/group. (f-h) N2a-hApoE2 cells at P11 were treated with serial concentrations of the PI3 inhibitor, LY294002, for 24 h; cell lysates were analyzed for changes in pAkt/tAkt expression ratio and protein levels of both HK1 and HK2. LY294002 induced a dose-dependent decrease in pAkt/tAkt ratio and HK2 expression; however, HK1 expression was unaltered. Results were normalized to the vehicle-treated group. n = 2-6/group. (i,j) Total protein carbonylation levels were slightly increased in 6-7-month-old ApoE4 mouse cortex compared to ApoE3 mice. n = 6 mice/group. * p < 0.05, ** p < 0.01, **** p < 0.0001. N.S.: not significant.

Discussion
ApoE ε4 increases LOAD risk, whereas ApoE ε2 reduces the risk [38]. ApoE4 carriers showed bioenergetic deficits preceding the onset of AD symptoms [39,40]. Prior results from our lab revealed that ApoE isoforms influenced cerebral glucose cytosolic metabolism [20]. In this follow-up study, we used human ApoE stable cell lines and neuronal models to further show that ApoE2-and ApoE4-dependent differences in glycolytic metabolism-ApoE2 upregulation and ApoE4 downregulation-impact neuronal health and aging, and that HK may mediate these effects (Figure 7). Moreover, we found that introduction of the ApoE2 protein can mitigate the ApoE4-associated reduction in HK expression and glycolysis flux. Collectively, our findings may help explain ApoE2-mediated neuroprotection and suggest potential AD-relevant therapeutic approaches.

Discussion
ApoE ε4 increases LOAD risk, whereas ApoE ε2 reduces the risk [38]. ApoE4 carriers showed bioenergetic deficits preceding the onset of AD symptoms [39,40]. Prior results from our lab revealed that ApoE isoforms influenced cerebral glucose cytosolic metabolism [20]. In this follow-up study, we used human ApoE stable cell lines and neuronal models to further show that ApoE2-and ApoE4-dependent differences in glycolytic metabolism-ApoE2 upregulation and ApoE4 downregulation-impact neuronal health and aging, and that HK may mediate these effects (Figure 7). Moreover, we found that introduction of the ApoE2 protein can mitigate the ApoE4-associated reduction in HK expression and glycolysis flux. Collectively, our findings may help explain ApoE2-mediated neuroprotection and suggest potential AD-relevant therapeutic approaches. Figure 7. In conclusion, our study demonstrates that human ApoE2 isoform upregulates, whereas ApoE4 variant downregulates, neuronal HK and glycolytic activity. We further show that these divergent ApoE2 and ApoE4 effects on glycolysis significantly impact neuronal health and aging. These findings imply that neuronal glycolytic phenotypes may serve as a key contributor to ApoEmediated brain resilience or vulnerability to AD. Blue: receptor binding region. Pink: lipid binding region.
Recent studies indicate relationships exist between glycolytic activity, AD pathology, and AD symptoms [12,13]. This is, perhaps, not surprising because glycolysis contributes to increased ATP synthesis driven by synaptic activity [41]. Proper functions of ion pumps, such as the V-type H + -ATPase (V-ATPase) and Na + /K + -ATPase, have been linked to membrane-bound glycolytic enzymes, indicating the critical role of glycolysis-derived ATP as an important source of energy in ion transport, which is vital for neurotransmitter release and synaptic transmission [42]. Specifically, the dependence of V-ATPase on glycolysis has been widely indicated by its close interaction with PFK, aldolase, and HK [15]. HK inhibition causes V-ATPase disassembly and functional loss in a PI3K-dependent manner [43,44]. We previously demonstrated that V-ATPase subunit expression is upregulated in ApoE2 brains [45]. Data presented herein confirmed that the ApoE2 genotype retains the most robust glycolytic profile during neuronal aging. Therefore, the enhanced glycolysis and the ensuing synaptic strength could potentially serve as a significant In conclusion, our study demonstrates that human ApoE2 isoform upregulates, whereas ApoE4 variant downregulates, neuronal HK and glycolytic activity. We further show that these divergent ApoE2 and ApoE4 effects on glycolysis significantly impact neuronal health and aging. These findings imply that neuronal glycolytic phenotypes may serve as a key contributor to ApoE-mediated brain resilience or vulnerability to AD. Blue: receptor binding region. Pink: lipid binding region.
Recent studies indicate relationships exist between glycolytic activity, AD pathology, and AD symptoms [12,13]. This is, perhaps, not surprising because glycolysis contributes to increased ATP synthesis driven by synaptic activity [41]. Proper functions of ion pumps, such as the V-type H + -ATPase (V-ATPase) and Na + /K + -ATPase, have been linked to membrane-bound glycolytic enzymes, indicating the critical role of glycolysis-derived ATP as an important source of energy in ion transport, which is vital for neurotransmitter release and synaptic transmission [42]. Specifically, the dependence of V-ATPase on glycolysis has been widely indicated by its close interaction with PFK, aldolase, and HK [15]. HK inhibition causes V-ATPase disassembly and functional loss in a PI3K-dependent manner [43,44]. We previously demonstrated that V-ATPase subunit expression is upregulated in ApoE2 brains [45]. Data presented herein confirmed that the ApoE2 genotype retains the most robust glycolytic profile during neuronal aging. Therefore, the enhanced glycolysis and the ensuing synaptic strength could potentially serve as a significant mechanism conferring cognitive resilience to ApoE2 carriers against aging-associated risks for LOAD.
HK is the first key enzyme in glycolysis responsible for phosphorylating glucose to glucose-6-phosphate (G6P). HK activity is found to be decreased in the brains, skincultured fibroblasts, and leukocytes of AD patients [46,47]. Our findings indicate that ApoE4 progressively impairs HK and that the impact of ApoE4 culminates in aged cells, resulting in a reduction in glycolytic activity. PFK, the second irreversible glycolytic enzyme, is subject to allosteric regulation dependent on cell energy status [48,49]. In our study, PFK upregulation occurred concurrently with HK downregulation in ApoE4 cells, which could reflect a compensatory reversal of the glycolytic deficit caused by HK hypoactivity in the cell. PK, the third major glycolytic enzyme, is found increased in cells exposed to AD plasma [50] and in AD brains [49]. In our study, like the changes to PFK, an increase in PK could be a cellular attempt to stimulate glycolytic metabolism. However, despite the elevated expression of PFK and PK, ApoE4 cells remained low on glycolytic activity, indicating the limited rescuing capability of PFK and PK. Moreover, levels of all three GLUTs-GLUT1, GLUT3, and GLUT4-were no different among the three ApoE cell lines, excluding a possible contribution of glucose availability to the glycolytic differences observed in these cells. Together, these observations highlight the role of HK in shaping the glycolytic phenotype associated with the three ApoE variants.
It should be noted that ApoE-mediated effects on glycolysis appear to be cell-type dependent. In line with our results, Zhao et al. demonstrated that, in Apoe −/− primary neurons, ApoE4 failed to induce insulin-stimulated glycolysis compared to ApoE3 [51]. Studies by Orr et al. showed that under stressed conditions, ApoE4-expressing neurons exhibited 50% less glycolytic reserve capacity [52]. In contrast to these findings, an opposite direction involving an increase in glycolysis by ApoE4 was found in astrocytes [53,54]. A higher dependence and capacity for using glucose by ApoE4 astrocytes was indicated by their enhanced basal respiration and glycolytic rate [55]. Serving in a supportive role, astrocytes possess a remarkably divergent metabolic phenotype from neurons. For example, in the proposed astrocyte-neuron lactate shuttle (ANLS), astrocytes function to provide lactate generated in glycolysis as an energy substrate to neurons [56]. In AD and ApoE4 carriers, increased astrocytic glycolysis potentially acts as a compensation for neuronal energy deficit. Regarding the regulation of glycolysis in microglia, ApoE4-expressing iMGLs (human microglia-like cells) were found to exhibit reduced glycolysis flux under normal conditions, whereas proinflammatory stimulus upregulated it [57]. Moreover, research led by Fang et al. showed that compared with ApoE3, the glycolytic function of ApoE4 astrocytes was not significantly altered in one-month cultures, but exhibited a marked decrease in two-month cultures, suggesting a similar time-dependent impact of ApoE on astrocytic glycolysis [58]. Our findings help fill the gap in understanding the role of ApoE isoforms in neuronal glycolysis and demonstrate how it affects neuronal health in the aging process.
The growing literature accumulated over the last two decades has revealed an inverse association between cancer and AD. Individuals with a history of cancer are less likely to develop AD, whereas people diagnosed with AD exhibit a reduced risk of having cancer, independent of ethnic groups or environmental factors [59][60][61][62][63][64]. The origins of this inverse relationship are unknown, although recent studies suggest that the divergent trajectories of cellular bioenergetics involved in these two distinct diseases may be a contributor. The Warburg effect, the preferential use of glycolysis over oxidative phosphorylation (OXPHOS), is a hallmark of cancer, a disease characterized by a high rate of ATP synthesis to promote cellular division. Conversely, an inverse-Warburg effect involving an increased OXPHOS activity is increasingly recognized in the early stages of neurodegenerative diseases [65][66][67][68][69]. It can be reasonably speculated that neuronal hypometabolism associated with early AD, possibly due to an impaired glycolysis, leads to upregulation of lactate uptake through ANLS, thus giving rise to a compensatory increase in OXPHOS along with reactive oxygen species (ROS). Thereby, over time, accumulative lactate release by astrocytes in the brain can be detrimental and contribute to AD pathogenesis [70,71]. Based on our results, we hypothesize that ApoE isoforms differentially impact AD via an inverse Warburg mechanism. The ApoE2 genotype increases neuronal glycolysis, which translates into a more Warburgian phenotype and less ROS. In contrast, ApoE4 downregulates neuronal glycolysis, which promotes an inverse-Warburg environment that increases neuronal vulnerability to oxidative stress and neurodegeneration.
Apart from bioenergetics, the Warburg effect is also critical in supporting biosynthesis via the pentose phosphate pathway (PPP) and the hexosamine biosynthetic pathway (HBP), two major metabolic pathways diverted from G6P. PPP, which maintains the NADPH level in response to the synaptic stimulation and increased oxidative stress, has been shown to be greatly impaired in AD [72][73][74]. In the HBP, fructose-6-phosphate (F6P) derived from G6P is first converted to glucosamine-6-phosphate (GlcN6P). GlcN6P is then subjected to several steps of enzymatic reactions yielding uridine diphosphate N-acetylglucosamine (UDP-GlcNAc), the end product of the HBP. UDP-GlcNAc serves as the donor in protein post-translational modification (PTM) with O-linked β-N-acetylglucosamine (O-GlcNAc), a PTM process referred to as O-GlcNAcylation. O-GlcNAc is found significantly concentrated at neuronal synapses, and elevation of O-GlcNAcylation has been suggested as a promising strategy for AD [15,75]. UDP-GlcNAc can also be metabolized to N-acetyl-D-mannosamine (ManNAc) and, subsequently, to ManNAc-6-P as the first steps in the biosynthesis of sialic acid, the sugar donor used in sialylation, another very important PTM that plays critical roles in the central nervous system [76]. Indeed, our lab has recently first reported that ApoE isoforms in human brains undergo varying degrees of sialic acid modifications, with ApoE2 most extensively sialylated, which ameliorates Aβ fibrillation, whereas ApoE4 is least sialylated [77]. Further work is needed to determine whether these PTM differences may result from differences in HK and glycolytic profiles associated with the three ApoE isoforms. We would like to point out that because the linker region of the plasmids we used in this study introduced an artificial N-glycosylation site, the exhibited PTM ( Figure 1) does not precisely represent the real glycosylation status of ApoE in those cells.
Recent studies show that glycolysis may also be involved in the regulation of mitochondrial function. HK is found to bind to the outer membrane of mitochondria by interacting with the voltage-dependent anion channel (VDAC). Mitochondrial bound hexokinase (mtHK) has long been recognized for its role in increasing the resistance of cancer cells to mitochondria-initiated apoptotic cell death [35,78]. Evidence indicates that mtHK suppresses cytochrome c release, possibly by interfering with the binding of Bax with VDAC and through further inhibition of the mitochondrial permeability transition pore [79][80][81][82]. Moreover, mtHK has been shown to be involved in the regulation of ADP recycling, contributing to a high efficiency of ATP production, energy preservation, and reduced ROS generation [83]. Recently, Chen et al. reported the colocalization of ApoE and VDAC in H9c2 cells and found that ApoE4 induced greater mitochondrial swelling compared to ApoE3 both in vitro and in vivo; however, the underlying mechanism is unknown [84]. Here, our data demonstrate that the ApoE4 genotype is associated with impaired HK expression and activity, which could potentially weaken the above mentioned anti-apoptotic and antioxidant mechanisms leading to declining cellular health. Notably, cell death was not observed within the passages studied. This observation is, in fact, in line with the trajectory of AD development in humans. AD is a unique disease in that it starts with a very long prodromal phase that can last up to 10-20 years, and during this period, various molecular changes can take place that culminate in neuronal death and the onset of clinical symptoms.
Gene delivery of ApoE2 in AD mouse models has been shown to be beneficial for alleviating brain amyloid pathology and neurotoxicity [85,86]. In our study, we found that introducing the ApoE2 protein by transient transfection in ApoE4-expressing cells can significantly improve HK expression and glycolytic activity, indicating a potential mechanism that underlies the neuroprotective effect by ApoE2. However, we did not observe a dose-dependent rescue effect of ApoE2 and more importantly, ApoE3 failed to induce a significant change in ApoE4 cells. These results confirmed a unique neuroprotective role of ApoE2 compared to ApoE3. With respect to the regulatory mechanism, according to our initial data, in 6-7-month-old human ApoE4-KI mouse brains, PI3K/Akt activity is downregulated with a concurrent reduction in HK and an elevation in total protein carbonylation, a marker of the overall level of oxidative stress. However, PI3K/Akt signaling appears to be only partially responsible for ApoE-mediated HK regulation. Several recent studies demonstrate that ApoE acts as a direct transcription factor in the expression of genes related to aging and AD [87,88]. Further work is needed to fully understand the underlying mechanisms governing the regulation of HK by ApoE.

Conclusions
The present study highlights the potentially crucial role of HK-driven glycolytic phenotypes in conferring neuronal resilience or vulnerability associated with ApoE genetic variants to AD. Our findings lend compelling support to the recently proposed inverse-Warburg origin of AD and present a novel and exciting opportunity in the prevention and treatment of AD [16,66,67,89].  Institutional Review Board Statement: Animal Use Statement (AUS # 220-08) was approved by the IACUC at the University of Kansas on 10 January 2019.

Data Availability Statement:
The data that support the findings of this study are available from the corresponding author upon reasonable request.