Normalization of hyperhomocysteinemia improves cognitive deficits and ameliorates brain amyloidosis of a transgenic mouse model of Alzheimer's disease

Abstract

Hyperhomocysteine (HHcy) is a risk factor for developing Alzheimer's disease (AD). Previously, we showed that diet-induced HHcy accelerated the AD-like phenotype of a transgenic mouse model, i.e., Tg2576. In the present work, we tested whether an HHcy-lowering strategy in this model would be beneficial. Tg2576 mice received methionine-rich or regular chow diet for 5 mo. Next, while the chow control group was kept on the same regimen, the other mice were randomized into two groups: one was kept on the methionine-rich diet (Met On), the other switched to chow (Met Off). Compared with controls, 5 mo on the methionine-rich diet resulted in HHcy (plasma Hcy level, treated: 12.7±1.2 μM vs. control: 3.1±0.4 μM) and significant behavioral impairments (% freezing, treated: 2.4±1.4% vs. control: 19.9±6.9%). At the end of the study, while the Met On group kept Hcy level elevated, the Met Off group had these values indistinguishable from the controls. The reduction in Hcy levels resulted in a significant improvement of the fear-conditioning performance, and an amelioration of the brain amyloidosis. Our results demonstrate that lowering HHcy in a transgenic AD-mouse model is beneficial since it significantly improves behavior deficits and brain amyloidosis. Our findings provide new biological insights for future clinical trials aimed at lowering this modifiable risk factor in human AD.—Zhuo, J.-M., Praticò, D. Normalization of hyperhomocysteinemia improves cognitive deficits and ameliorates brain amyloidosis of a transgenic mouse model of Alzheimer's disease.

High circulating levels of homocysteine (Hcy), also known as hyperhomocysteinemia (HHcy), has been linked closely to cognitive dysfunction and Alzheimer's disease (AD) pathogenesis (1, 2). Epidemiological and clinical studies demonstrated that HHcy level is associated with a greater prevalence of cognitive dysfunction and AD (2–4). Conversely, AD patients tend to have higher Hcy levels than their age-matched controls (5). As a result, HHcy has been recognized as an independent risk factor of AD (6, 7).

Homocysteine, a nonprotein sulfur amino acid, is an intermediate product of the methionine cycle. Its formation requires the presence of folate and B vitamins (8). Although we do not have a full understanding of the biological association between HHcy and AD, many possible mechanisms have been proposed. These include microvascular damage (9), oxidative stress (10), diminished brain bleomycin hydrolase activity (8), interference in DNA methylation (11) and repair (12), excitation damage through the glutamate receptors (13, 14), and the direct alteration of the APP metabolic pathways (15–18).

Unlike the AD genetic risk factors (e.g., ApoE), HHcy is a modifiable one, since its levels in the body can be manipulated by altering the dietary levels of methionine, folate, or B vitamins. Based on this knowledge, clinical trials, mainly through folate dietary intervention, have been performed to test the effect of Hcy reduction on the onset of dementia and AD development (19, 20). However, most of the results accumulated so far have been disappointing (19, 21–23). Even though these studies are limited in their relatively small sample size or short treatment period or by recruiting subjects without a real condition of HHcy (i.e., mean plasma Hcy: 9.2μM) (24), they have raised concerns as to whether the Hcy-lowering approach is a useful therapeutic strategy in AD patients.

In the present work, by using an established model of diet-induced HHcy and acceleration of brain pathology in AD transgenic mice (15), we tested whether the reduction of HHcy levels would ameliorate their AD-like phenotype. At the end of the study, we found that HHcy reduction in these mice rescued their cognitive deficits and reduced brain Aβ peptides deposition.

Our findings provide evidence that the acceleration of the AD-like phenotype secondary to elevated Hcy levels in a transgenic mouse model can be ameliorated by normalizing the circulating high Hcy level. They have important clinical implication for future planning of clinical trials aimed at reducing Hcy levels as a therapeutic strategy for AD.

MATERIALS AND METHODS

Tg2576 mice and diet treatments

All animal procedures were approved by the Temple University School of Medicine Animal Care and Usage Committee. Tg2576 transgenic female mice expressing hAPP with the Swedish mutation (K670N/M671L) (25) were treated with a methionine-rich diet to induce HHcy, as previously reported (15). All mice were genotyped by polymerase chain reaction (PCR) analysis using tail DNA and were kept on a 12-h light/dark cycle with access to food and water ad libitum. Starting at 9 mo of age, mice were randomized into two different diet groups: rodent chow enriched with methionine (7.7 g/kg), or vehicle standard chow [control (Ctrl) group]. Diets were custom-made, prepared by a commercial vendor (Harlan Teklad, Madison, WI, USA) and matched for kilocalories (26). After 5 mo consuming these diets, all the mice went through fear-conditioning tests, and their blood was collected for Hcy quantification. After the behavioral test, while the Ctrl group remained on standard chow, mice receiving the methionine-rich diet were randomized into two groups: one was switched to standard chow (Met Off group); the other, kept on the methionine-rich diet (Met On group) for an additional 2 mo (Fig. 1). At randomization, no significant difference was found between the Met On and Met Off groups in terms of plasma Hcy concentration or fear-conditioning performance. At the end of the additional 2 mo, all of the animals underwent a second fear-conditioning test and were then sacrificed (Met On group: n=5; Met Off group: n=5; Ctrl group: n=6). At sacrifice, mice were perfused with ice-cold phosphate buffered saline (PBS) containing 10 mM EDTA. Brains were removed and dissected in two hemibrains by midsagittal dissection. The left hemibrain was used for biochemistry assays; the right hemibrain was fixed in 4% paraformaldehyde in 0.1 M PBS (pH 7.6) overnight for immunohistochemistry studies.

Figure 1.

Experimental design. At the age of 9 mo, Tg2576 mice were randomized to either methionine-rich diet (treated group) or standard chow (control group) for 5 mo. Next, while the control group was kept on chow, the treated group was randomized further into two new groups: the Met On group, which continued the methionine-rich diet feeding, and the Met Off group, which was switched back to standard chow. All 3 groups were followed for an additional 2 mo and then sacrificed at 16 mo.

Behavioral tests

Animals went through the same fear-conditioning test twice: at the age of 14 and 16 mo (Fig. 1). The behavioral test was conducted in a conditioning chamber (19×25×19 cm) equipped with black methacrylate walls, transparent front door, a speaker and grid floor (StartFear System; Panlab/Harvard Apparatus, Holliston, MA, USA). The chamber was placed inside a sound-attenuated box and had a scale underneath the grid floor, which recorded the relative movement of the animals inside to detect the freezing behavior (27). The conditioning context was cleaned with 95% ethanol between each mouse. In the training stage, each animal was put inside the conditioning chamber and allowed free exploration for 2 min in the white noise (65 dB) before the computer software StartFear (Panlab/Harvard apparatus) delivered the conditioned stimulus (CS) tone (30 s, 90 dB, 2000 Hz) paired with a foot-shock unconditioned stimulus (US; 2 s, 0.6 mA) through a grid floor at the end of the tone. A total of 3 pairs of CS-US pairing with a 30 s intertrial interval (ITI) were presented to each animal in the training stage. The mouse was removed from the chamber 1 min after the last foot-shock and placed back in its home cage.

The contextual fear-conditioning stage started 24 h after the training phase when the animal was put back inside the conditioning chamber for 5 min with white noise only (65 dB). The animal's freezing response to the environmental context, but not the tone, were recorded and analyzed by StartFear software (activity threshold set to 10, time threshold to 2 s and gain to 16).

The tone fear-conditioning stage started 2 h after the contextual stage. The animal was placed back to the same chamber with different contextual cues, including white wall, smooth metal floor, lemon extract drops, and dimmed yellow light condition. After 3 min of free exploration, the mouse was presented to the exactly same 3 CS tones with 30 s ITI in the training stage without the foot-shock. The mouse was brought back to the home cage 1 min after the last tone, after its freezing responses to the tones were recorded.

Immunohistochemistry

Immunohistochemistry was performed as described previously by our group (18, 28). Briefly, serial 6-μm-thick coronal brain sections were cut, and every eighth coronal section was analyzed across the area of hippocampus and somatosensory cortex (a total of ∼6 sections/brain), beginning with a randomly selected section from the first 8 sections. The sections were pretreated with formic acid (FA; 88%) and subsequently with 1% H₂O₂ in methanol. The primary antibody 4G8, which targets at Aβ 17–24 aa (4G8, diluted 1:1000) was used to identify Aβ deposits. After incubation with biotinylated anti-mouse secondary antibody, the avidin-biotin complex method (Vector Laboratories, Burlingame, CA, USA) was used to detect the Aβ deposits with 3,3′-diaminobenzidine (DAB) as a chromogen. The software Image-Pro Plus (Media Cybernetics, Inc., Silver Spring, MD, USA) was used to capture the light microscopic images from the stained sections and quantify the Aβ deposits across the hippocampus and somatosensory cortex. The threshold optical density that discriminated staining from background was determined and kept constant for all of the quantifications. The area occupied by Aβ immunoreactivity was divided by the total area of interest to calculate the percentage area of Aβ deposits.

Biochemical analyses

Sequential extractions of brain homogenates in RIPA and then FA were performed as described previously (15, 29). The RIPA fraction contains soluble, whereas the FA fraction contains the insoluble forms of the Aβ peptides. Sensitive sandwich ELISA kits (Wako Chemicals USA Inc., Richmond, VA, USA) were used to measure Aβ1–40 and Aβ1–42 levels. Analyses were always performed in duplicate and in a coded fashion. Blood was collected at the age of 14 and 16 mo old by retro-orbital collection and heart puncture, respectively. Plasma was obtained by spinning down the blood at 2000 RPM for 10 min at 4°C and frozen at −80°C until analysis. Homocysteine levels were assayed by using the Abbott homocysteine assay with a fluorescence polarization immunoassay on the IMx^® analyzer (Abbott Laboratories, Abbott Park, IL, USA) (15). 8-Iso-prostaglandin F_2α level in cortex RIPA fractions was assayed by an enzyme immunoassay kit, following the protocol provided by the company (Assay Designs, Ann Arbor, MI, USA).

Western blot analyses

RIPA fractions of brain homogenates were used for Western blot analyses. Samples were electrophoresed on precasted gels (Bio-Rad Laboratories, Hercules, CA, USA) and transferred to nitrocellulose membranes. Odyssey infrared imager system (Li-COR, Lincoln, NE, USA) was used for detection. Antibodies used in the current study were as follows: anti-APP (22C11; Chemicon International, Temecula, CA, USA), anti-secreted APPβ (sAPPβ) (6A1, IBL America, Minneapolis, MN, USA), anti-β-site APP cleaving enzyme 1 (BACE-1) (IBL America), anti-ADAM10 (Chemicon International), anti-APP C-terminal for C-terminal fragments (CTFs) (EMD Biosciences Inc, La Jolla, CA, USA), anti-presenilin 1 (PS1) (Sigma-Aldrich, St. Louis, MO, USA), anti-presenilin enhancer 2 (Pen-2) (Santa Cruz Biotechnologies, Santa Cruz, CA, USA), anti-Nicastrin (Cell Signaling Technology, Danvers, MA, USA), anti-anterior pharynx-defective 1 (APH1) (Sigma-Aldrich), anti-apolipoprotein E (apoE) (Santa Cruz Biotechnologies), anti-insulin-degrading enzyme (IDE) N-terminal (EMD Biosciences Inc.), anti-neprilysin (NEP) (Santa Cruz Biotechnologies), anti-glial fibrillary acidic protein (GFAP) (Cell Signaling Technology), anti-glycogen synthase kinase 3 α/β (GSK3α/β) (Santa Cruz Biotechnologies), anti-phospho- GSK3α/β (Ser-21/9) (Cell Signaling Technology), and anti-β-actin (1:4000; Santa Cruz Biotechnologies).

Data analysis

Data analyses were performed using analysis of variance (1-way, 2-way ANOVA as appropriate) by SigmaStat with post hoc comparison (Fisher's PLSD) when needed. Values in all figures represent means ± se.

RESULTS

Normalization of HHcy after 2 mo of standard chow feeding

Starting at 9 mo of age, Tg2576 mice were assigned to two groups: one fed with a methionine-rich diet (treated group), the second standard chow (Ctrl group) for 5 mo (Fig. 1). At the age of 14 mo, as expected plasma Hcy concentration in the treated diet group was significantly higher than their controls (treated: 12.7±1.2μM; Ctrl: 3.1±0.4 μM; P<0.01).

Next, while the Ctrl group was kept on the standard chow, the treated group was randomized into two new groups: the Met Off group received vehicle standard chow; the Met On group continued to receive the methionine-enriched diet for an additional 2 mo (Fig. 1). No significant difference in Hcy levels was observed between these two groups at the beginning of the 2 mo (Fig. 2).

Figure 2.

Plasma Hcy levels in Tg2576 mice at 14 and 16 mo of age. After 5 mo on the methionine-rich diet, Hcy plasma levels were increased significantly when compared with controls. However, while the Met On group kept Hcy levels elevated, in the Met Off group plasma Hcy levels were reduced back to normal after 2 mo of standard chow feeding. Values represent means ± se. **P < 0.01.

After 2 mo on the standard chow, the Met Off group had a significant reduction in Hcy levels (14 mo old: 11.9±2.1 μM vs. 16 mo old: 3.6±0.3 μM, P<0.01). The new reduced Hcy levels were in the same range of the control group at 16 mo of age (Met Off: 3.6±0.3 μM; Ctrl: 3.4±0.3 μM; P>0.05). By contrast, the Met On group, which was kept on the methionine-rich diet, had plasma Hcy concentration much higher than either Met Off group or Ctrl group (Met On: 16.8±2.5 μM; Met On vs. Ctrl, P<0.01; Met On vs. Met Off, P<0.01) (Fig. 2).

Effect of HHcy reduction on fear-conditioning test

In the current study, fear-conditioning test was performed twice at the age of 14 and 16 mo to assess the effect of HHcy and then its reduction on cognition (Fig. 1).

When the first test was performed, we detected a significant group effect between the treated and Ctrl groups in the contextual fear-conditioning paradigm [F (2, 12) = 4.324, P<0.05] but not in the tone fear-conditioning stage (P>0.05). Post hoc analysis of the data confirmed that compared to the Ctrl group, the treated group (the combination of Met On and Met Off groups) showed significantly less freezing behavior (Fig. 3).

Figure 3.

Effect of HHcy on the contextual fear-conditioning test in Tg2576 mice at 14 and 16 mo. Mice were tested after 5 mo on a methionine-rich or standard chow diet (14 mo old), then retested after an additional 2 mo. Values represent means ± se. *P < 0.05.

At the age of 16 mo, all 3 groups (i.e., Ctrl, Met Off and Met On) underwent the second fear-conditioning test. One-way ANOVA analysis detected a trend of difference among the 3 groups in the contextual fear-conditioning stage [F (2, 12)=3.25, P=0.075] but not in the tone fear-conditioning stage (P>0.05). Post hoc analysis found that the Met On mice, but not the Met Off mice, remained impaired in this memory test, as demonstrated by their significantly less freezing behavior than the Ctrl mice (Fig. 3). No significant difference was observed between the Met Off and Ctrl groups (Fig. 3). No significant age effect or age X group interaction among these 3 groups was detected in the fear-conditioning test between 14 and 16 mo old.

Effect of HHcy reduction on Aβ accumulation

All three groups of animals were sacrificed after the second fear-conditioning test, and the Aβ1–40/42 levels in the RIPA and FA extractions of their brains were assayed (Fig. 1).

As shown in Fig. 4, compared with the Ctrl group, the Met On group, but not the Met Off, had a statistically significant increase in the RIPA extracted Aβ1–42 levels in the cortex and hippocampus, and in the FA extracted Aβ1–42 in the cortex.

Figure 4.

Effect of HHcy on brain Aβ peptide levels in Tg2576 mice. RIPA-soluble (RIPA; A, C) and FA-extractable (FA; B, D) Aβ1–40 (A, B) and Aβ1–42 (C, D) levels in the cortex (Ctx) and hippocampus (Hippo) were measured by sandwich ELISA. Values represent means ± se. *P < 0.05; **P < 0.01.

However, even though the Met Off group had a smaller elevation than the Met On, both groups showed significant increase in the FA extracted Aβ1–40 in the cortex (Met On: 202%, Met Off: 169%) and hippocampus regions (Met On: 238%, Met Off: 183%), and the FA extracted Aβ1–42 in the hippocampus (Met On: 220%, Met Off: 181%), when compared with Ctrl groups (Fig. 4).

Serial brain sections were immunostained with 4G8, an anti-Aβ antibody reactive to amino acid residues 17–24, and the positive immunoreactive areas were calculated. As shown in Fig. 5, we found that, compared with the Ctrl group, the Met On group had a statistically significant increase in Aβ deposits in the hippocampus (Met On: 1.08±0.24%, Ctrl: 0.51±0.10%, P<0.05), and the somatosensory cortex (Met On: 1.21±0.13%, Ctrl: 0.73±0.14%, P<0.05). However, the Met Off group did not show any significant difference in the amount of Aβ deposits when compared to the Ctrl group (Fig. 5).

Figure 5.

Effect of HHcy on brain Aβ deposition in Tg2576 mice. A) Representative sections of brains from each group immunostained with 4G8 antibody. B) Quantification of the area occupied by Aβ deposition in hippocampus (Hippo) and somatosensory cortex (SSC). Values represent means ± se. *P < 0.05.

Effect of HHcy reduction on APP metabolism

Next, we investigated the effect of HHcy lowering on the main APP metabolic pathways. Compared with brains from the Ctrl group, total APP levels were unaltered in either the Met On, or Met Off group (Supplemental Fig. 1). Similarly, the β-secretase pathway represented by BACE-1, sAPPβ, and CTF-β (C99) did not behave differently among the three groups (Supplemental Fig. 1). No significant alteration of the α-secretase pathway was detected either, as measured by ADAM10 and CTF-α (C83) (Supplemental Fig. 1). Moreover, no change was found for PS1, Pen-2, Nicastrin, and APH1 protein levels, the four components of the γ-secretase complex, among the three groups (Supplemental Fig. 1).

Then we analyzed two of the major proteases involved in the catabolism of Aβ, i.e., IDE and NEP (30). Steady-state protein levels of these proteins measured by Western blot were similar among the three groups of animals, and the same was valid for the ApoE and GFAP (Supplemental Fig. 2).

Effect of HHcy reduction on GSK3

Since we previously reported that diet-induced HHcy associates with an alteration of the GSK3 pathway, and in particular with a reduced Ser-21/9 phosphorylation level on GSK3α/β (15), next we checked whether this was also the case in the current study. As shown in Fig. 6, we found that the total protein levels for GSK3α and GSK3β did not differ significantly among the three groups. By contrast, we observed that compared with Ctrl groups, brain homogenates from Met On group, but not the Met Off group, had statistically significant lower GSK3α/β Ser-21/9 phosphorylated levels (Fig. 6).

Figure 6.

Effect of HHcy on GSK3 levels in Tg2576 mice. A) Representative Western blots of brain homogenates probed with specific antibodies against total GSK3α/β, and phosphorylated GSK3α/β on Ser-21/9, respectively. B) Densitometric analyses of the immunoreactivities to the antibodies shown in (A). Values represent means ± se. *P < 0.05.

Effect of HHcy reduction on brain lipid peroxidation

Brain levels of 8-iso-prostaglandin F_2α, a specific and sensitive marker of lipid peroxidation, were also assayed (31). We found that the Met On group had significantly more 8-iso-prostaglandin F_2α than either the Met Off or Ctrl group, while no significant difference was detected between the Met Off and Ctrl groups (Met On: 2844.4±1170.5 ng/mg tissue, Met Off: 251.6±93.2 ng/mg tissue, Ctrl: 43.5±10.8 ng/mg tissue, P<0.05; post hoc: Met On vs. Ctrl, P<0.05, Met On vs. Met Off, P<0.05).

DISCUSSION

Today more than 26 million people suffer from AD worldwide, and the diagnosed cases are expected to reach more than 100 million by 2050 (32). However, the current AD pharmacotherapy provides only a limited and temporary cognitive benefit (33). Since even modest improvement could have significant effect on the quality of the patient life, caregiver burden, and socioeconomic cost, in recent years an enormous effort has been spent on developing disease-modifying therapeutic strategies.

High circulating levels of Hcy is a known risk factor of AD, which, by contrast with genetic risk factors, can be corrected by dietary intervention. However, conflicting data are available on the issue of whether the Hcy-lowering strategy is beneficial in AD.

Our laboratory has previously established a diet-induced HHcy model in the Tg2576 mice in which the high levels of Hcy result in an acceleration of their AD-like phenotype, including behavioral deficits and brain Aβ accumulation (15). Since a study found that in C57B6 mice, 4 wk normal diet feeding was sufficient to normalize Hcy levels previously increased by a 2-wk high-methionine diet (34), we designed our current experiments to test the biological effects of a similar Hcy-lowering strategy but in a different animal model. Thus, after HHcy was induced in Tg2576 mice by feeding them a methionine-rich diet, we reduced the Hcy levels to a normal range by switching them back to standard chow diet for another 2 mo and then evaluated the effects on cognition and brain amyloidosis. To the best of our knowledge, this is the first study to assess the effect of an HHcy-lowering strategy in an AD-like animal model.

First, we confirmed that 5 mo feeding with the methionine-rich diet was sufficient to induce HHcy, although the actual plasma levels were a little lower than our previous report, most likely due to the difference in the feeding time (5 vs. 7 mo; ref. 15). However, 2 mo of standard chow feeding successfully reduced HHcy level in the Met Off group back to the normal range as found in the Ctrl group, thus validating our Hcy-lowering strategy in the current study.

The effects of Hcy lowering on cognition were assessed in the fear-conditioning test paradigm. First, we detected that after 5 mo of methionine-rich diet, mice had less freezing behavior in the contextual, but not in the tone fear conditioning, when compared with the Ctrl group. In this paradigm, less freezing behavior suggests that mice with HHcy do not remember well the environmental context, which typically depends on the integrity of the hippocampus (35). This deficit is in accordance with previous report that elevated Hcy impairs hippocampus neuronal cells (12). When retested after 2 mo, we found that Met Off group, after Hcy normalization, showed no significant difference in the second fear-conditioning test when compared with the Ctrl group. While it is possible that the sample size could have influenced this result, it is important to note that the Met Off group before the change of the diet had a clear reduction in the freezing behavior when compared with the Ctrl group.

By contrast, the Met On group, whose Hcy levels remained elevated, was still impaired in the contextual fear conditioning (Fig. 3). This observation indicates that Hcy normalization in the Met Off group rescued the HHcy-induced cognitive impairments. Regarding the tone fear-conditioning test, no freezing behavior was observed before the tone (CS), which suggests that animals could not recognize the same foot-shock chamber after disguise such as floor and odor changes. The retest of fear conditioning in these mice was also not contaminated by their experience in the first test, as shown by the absence of freezing at the training stage before the foot-shock. One possible explanation for the negative result in the tone fear conditioning is that the Tg2576 mice after middle age may not be responsive to the high-frequency sound at 2000 Hz, and only a lower frequency sound may reveal the effect of HHcy.

All the mice were sacrificed after the second fear-conditioning test and their brain Aβ levels were assayed. Compared with the Ctrl group, Aβ peptide levels in the Met On group increased in both RIPA and FA extractions in most of the brain regions considered. By contrast, the Met Off group, whose HHcy levels were now normalized, did not show significant Aβ level elevation in some brain extractions, especially in the RIPA fractions. Since the RIPA extraction contains mainly soluble Aβ peptides whose formation precede the insoluble Aβ contained in the FA extraction, it is possible that while the 2 mo switch to the chow diet is sufficient to affect the soluble Aβ (RIPA fraction), a longer time is necessary to affect the insoluble forms (FA extraction).

The beneficial effect of Hcy reduction on amyloidosis was more obvious when we analyzed the brain Aβ deposits in these mice. Immunostaining analyses of the brain sections found that only the Met On, but not Met Off group, had statistically more Aβ deposits than the Ctrl group in both hippocampus and somatosensory cortex. Taken together, the results from these analyses suggest that HHcy reduction in the Met Off group significantly slowed the Aβ accumulation process in the brain.

When we assessed the APP metabolism, no difference was detected in the steady state levels of APP, α, β, γ-secretase and their pathways. Further, we did not observe any changes in the Aβ major catabolic pathways (i.e., IDE, NEP), or apoE, an Aβ chaperon protein, or GFAP, a marker of astrocytosis. However, we cannot rule out the possibility that the activities of these enzymes were changed by Hcy level alterations, and further studies are warranted.

Also, in accordance with our previous observation (15), we found that while total GSK3α/β protein levels remained unaltered in all three groups, Ser-21/9 phosphorylation of the GSK3α/β was significantly reduced in Met On group, when compared with Met Off and Ctrl group. As GSK3 Ser-21/9 phosphorylation reduces GSK3 activity (36), the lower phosphorylation level observed in the Met On group suggests that the GSK3 pathway is activated in this group and ultimately could account for the Aβ elevation (37). This observation is further supported by the fact that the Hcy reduction in the Met Off group, which resulted in less Aβ formation and deposition, prevented the changes in GSK3 phosphorylation levels.

It is known that HHcy can also increase oxidative stress in the CNS, which plays an important role in the pathogenesis of AD (10, 38). In our study, brain lipid peroxidation levels in the Met On group were significantly higher than either Met Off or Ctrl group as shown by the elevation of a lipid peroxidation marker, 8-iso-prostaglandin F_2α levels. By contrast, no difference in this marker was detected between the Met Off and the Ctrl group, further supporting the beneficial effect of our Hcy-lowering strategy. This finding is very interesting and could have potential implications for AD therapy since it would support the concept that brain oxidative stress is a reversible phenomenon.

Our study provides biological evidence that a lowering Hcy strategy in a transgenic mouse model results in multiple beneficial effects on its AD-like phenotype. In particular, this therapeutic approach can rescue the cognitive deficits, reduce the rate of brain Aβ formation and deposition, and suppress brain oxidative stress.

Even though it is always very difficult to translate any finding obtained in an animal model to human disease in general, the beneficial effects of Hcy reduction observed in our study are somewhat in contrast to previous clinical trials, which failed to detect any significant improvement in cognition after Hcy reduction (19). Since dementia is a complex syndrome with many factors involved, HHcy is probably only one of them, and it is possible that a lowering strategy for this risk factor could be effective only in subjects with actual HHcy (i.e., plasma Hcy>14 μM). This could be part of the reason why most clinical trials did not detect beneficial effect from Hcy reduction, because in these studies the researchers recruited only subjects with slightly elevated Hcy levels (<14 μM), rather than subjects with HHcy (21–23, 39). In support of this hypothesis is the fact that a few studies reported significant improvement in cognition after reducing the Hcy levels in subjects with HHcy (40, 41).

In summary, the current study not only confirms that HHcy induces Aβ elevation and cognitive deficits but also provides biological evidence supporting the concept that Hcy reduction ameliorates important AD neuropathological features, including the cognitive deficits, brain Aβ deposits, and oxidative stress in an animal model. These beneficial effects offer strength to the validity of an Hcy-lowering therapeutic approach in AD, but further clinic trials applying this strategy on subjects with actual HHcy are crucial to evaluate the full biological and therapeutic effect of this inexpensive treatment.