Homocysteine modulates 5‐lipoxygenase expression level via DNA methylation

Summary Elevated levels of homocysteinemia (Hcy), a risk factor for late‐onset Alzheimer's disease (AD), have been associated with changes in cell methylation. Alzheimer's disease is characterized by an upregulation of the 5‐lipoxygenase (5LO), whose promoter is regulated by methylation. However, whether Hcy activates 5LO enzymatic pathway by influencing the methylation status of its promoter remains unknown. Brains from mice with high Hcy were assessed for the 5LO pathway and neuronal cells exposed to Hcy implemented to study the mechanism(s) regulating 5LO expression levels and the effect on amyloid β formation. Diet‐ and genetically induced high Hcy resulted in 5LO protein and mRNA upregulation, which was associated with a significant increase of the S‐adenosylhomocysteine (SAH)/S‐adenosylmethionine ratio, and reduced DNA methyltrasferases and hypomethylation of 5‐lipoxygenase DNA. In vitro studies confirmed these results and demonstrated that the mechanism involved in the Hcy‐dependent 5LO activation and amyloid β formation is DNA hypomethylation secondary to the elevated levels of SAH. Taken together these findings represent the first demonstration that Hcy directly influences 5LO expression levels and establish a previously unknown cross talk between these two pathways, which is highly relevant for AD pathogenesis. The discovery of such a novel link not only provides new mechanistic insights in the neurobiology of Hcy, but most importantly new therapeutic opportunities for the individuals bearing this risk factor for the disease.


Introduction
Alzheimer's disease (AD) is the most common form of chronic neurodegenerative condition with dementia in the elderly, affecting approximately 6-8% all persons aged > 65. While only a minority of AD cases is secondary to missense mutations in genes involved for either Ab precursor protein (APP) or presenilin-1 and presenilin-2, the cause of sporadic AD remains unclear, and a combination of environmental and genetic risk factors has been implicated . Considering that an effective treatment for AD is still unavailable, interventions aimed at controlling these risk factors can in the long run have a significant impact in reducing the number of cases and associated cost (Beydoun et al., 2014;Alzheimer's Association, 2015). Elevated levels of total plasma homocysteine (Hcy), also termed hyperhomocysteinemia (HHcy), are considered one modifiable risk factor for developing AD; however, the underlying mechanisms remain unknown (Zhuo et al., 2011;Shen & Ji, 2015).
Homocysteine is an intermediate metabolite of the methionine cycle, produced from the hydrolysis of S-adenosylhomocysteine (SAH), which is a by-product of methylation reactions involving the methyl donor Sadenosylmethionine (SAM). As result, HHcy typically associates with elevation of intracellular SAH, a potent inhibitor of methyl-transfer reactions, low SAM and ultimately DNA hypomethylation (Selhub, 1999).
The 5-lipoxygenase (5LO) is an enzyme widely expressed in the central nervous system where its levels increase in an age-dependent manner particularly in cortex and hippocampus, two brain regions highly vulnerable to neurodegenerative insults (Chinnici et al., 2007).
Previous works form our and other groups have shown that 5LO is upregulated in AD brains, and its modulation influences the development of the AD-like phenotype of different transgenic mouse models of the disease (Chu et al., 2012a(Chu et al., , 2013Giannopoulos et al., 2014).
In addition to nucleotide polymorphism in the promoter region of the 5LO gene, other factors could influence its expression and ultimately drive its functional role in AD pathogenesis (Qu et al., 2001). Among them, epigenetic modifications have been implicated as important regulatory mechanism for this enzyme. In particular, 5LO expression seems to be tightly regulated by DNA methylation and de-methylation of its promoter is a prerequisite for 5LO gene expression, protein synthesis and activity (Uhl et al., 2002).
Considering the importance of both genetic and environmental risk factors for late-onset AD, one can envision a scenario in which 5LO (the gene) and HHcy (environmental factor) interact and together modulate disease pathogenesis. However, although they have been independently associated with AD, a mechanistic link between them has not been explored. We hypothesize that in a condition of HHcy, the increased intracellular SAH results in cellular hypomethylation that modulates 5LO gene expression influencing the neuronal cells phenotype and ultimately neurodegeneration.

Diet-induced HHcy
To start investigating the role that Hcy has on the expression levels of the 5LO enzymatic pathway, we examined brain cortices from 3xTg-AD mice with diet-induced HHcy . Compared with controls, brain cortex homogenates from these mice had a significant increase in the steady state levels of 5LO protein, but no changes were observed for another lipoxygenase, 12/15LO (Fig. 1A,B). This increase was associated with a significant elevation of the leukotriene B4 (LTB4) levels, a 5LO major metabolic product, and 5LO mRNA levels (Fig. 1C,D). Consistent with the dietary HHcy condition, the same samples had higher levels of S-adenosyl-homocysteine (SAH) but lower S-adenosyl-methionine (SAM) (Fig. 1E). The altered SAH/SAM ratio resulted in a significant reduction in the methylation of the 5LO DNA, which was associated with a significant decrease in all three major DNA methyl-transferase enzymes, DNMT1, DNMT3a, and DNMT3b ( Fig. 1F-H).

Genetic-induced HHcy
Compared with wild-type mice, brain cortices from cystathionine bsynthase deficient mice (cbsÀ/À), which spontaneously develop HHcy (Sapna et al., 2009), had a significant increase in 5LO protein levels, which was associated with higher LTB4 and 5LO mRNA levels ( Fig. 2A-D). By contrast, no changes were observed for the steady state protein levels of 12/15LO ( Fig. 2A,B). Moreover, the same mice had significant reduction in 5LO DNA methylation levels, which was associated with a significant decrease in the levels of DNMT1, DNMT3a, and DNMT3b ( Fig. 2E-G).

In vitro studies
For these studies, we used the Neuro 2A (N2A) cells, a mouse neuroblastoma cell line stably expressing human APP carrying the K670 N, M671L Swedish mutation (N2A-APPswe). To confirm the effect of Hcy on 5LO expression levels and its DNA methylation, we treated the N2A-APPswe cells with 50 lM DL-homocysteine for 24 h, and then, supernatant and cells lysates harvested for biochemistry analyses. As shown in Figure 3, we observed that compared with controls, cells exposed to Hcy had a significant increase in the amount of Ab1-40 and Ab1-42 levels ( Fig. 3A) and 5LO protein levels, but no changes were detected for steady state levels of 12/15LO (Fig. 3B,C). These changes were associated with a significant increase in LTB4 as well as 5LO mRNA levels ( Fig. 3D,E). In the same samples, we also observed a significant increase in cellular levels of SAH and a decrease in SAM (Fig. 3F), which was associated with a reduction in 5LO DNA methylation and steady state levels of DNMT1, DNMT3a, and DNMT3b (Fig. 3,G-I).
To prove a causative role for the elevation of S-Adenosyl-homocysteine Hydrolase (SAH) secondary to HHcy in the demethylation of 5LO DNA and upregulation of the 5LO pathway at the message and protein levels and Ab formation, next we transfected the neuronal cells with SHA hydrolase (AHCY), which catalyzes the breakdown of SAH (Turner et al. 2000). Over-expression of AHCY alone did not influence 5LO pathway or Ab levels ( Fig. 4A-C). By contrast, over-expression of AHCY in the presence of Hcy prevented the upregulation of the 5LO pathway (protein and mRNA levels), the increase in Ab formation ( Fig. 4A-C), and the reduction in DNMT1, DNMT3a, DNMT3b (Fig. 4A,B), and 5LO DNA methylation (Fig. 4E).
In another set of experiments, N2A-APPswe cells exposed to high Hcy levels were also incubated with actinomycin-D, inhibitor of mRNA translation, and cycloheximide, an inhibitor of protein synthesis, and the effect on 5LO and Ab investigated. As shown in Figure 5, we found that Finally, to further demonstrate the susceptibility of 5LO mRNA transcription level to its DNA methylation state, N2A-APPswe cells were incubated with 5-Aza-2dc, a potent de-methylating agent. As shown in Figure 6, incubation with this agent by inducing a significantly reduction in 5LO DNA methylation resulted in a significant increase in its mRNA levels, which was associated with an elevation of Ab1-40 and Ab1-42 levels in the supernatant.

Discussion
In the current article, we provide the first experimental evidence that high level of Hcy, also known as hyper-Hcy or HHcy, by inducing an increase in intracellular SAH levels, and subsequent reduction in the methyltransferases activity results in 5LO DNA hypomethylation and ultimately upregulation of 5LO enzymatic pathway and Ab formation.
Despite some conflicting data, most of the epidemiological and clinical studies have revealed that HHcy doubles the risk for developing AD independently of several other major factors (Ravaglia et al., 2003;Clarke et al., 2014). As HHcy occurs more frequently in old age, it has been proposed as a potential metabolic link for the frequent coexistence of aging and neurodegenerative diseases (Parnetti et al., 1997;Ravaglia et al., 2004) and several mechanisms have been proposed to explain this link. These include oxidative stress, excitotoxicity, cerebrovascular damage, and endoplasmic reticulum stress (Perna et al., 2003;Boldyrev & Johnson, 2007;Kim et al., 2008;Kuszczyk et al., 2009). However, the exact biochemical basis by which Hcy modulates neurodegenerative mechanisms relevant to AD remains unknown.
In search for a novel molecular mechanism linking HHcy with AD pathogenesis, in recent years, a methylation hypothesis has emerged in which Hcy elevation leads to a reduced ratio of SAM/SAH and decreased activity of methyltransferases which then may regulate gene expression (Caudill et al., 2001;Castro et al., 2005). In support of this hypothesis, our work shows that SAM levels are reduced and SAH levels increased both in the brains of mice with diet-and genetically induced HHcy and that this change is associated with an upregulation of the 5LO enzymatic pathway both at the protein and message levels. Interestingly, we found that HHcy has no influence on a distinct LO, the 12/15LO, suggesting a selectivity of its biological effect within the central nervous system. It is possible that the lack of effect for HHcy on the expression levels of the 12/15LO is secondary to the fact that, by contrast with the 5LO promoter, the former one is not affected by methylation changes in the central nervous system. Supporting this hypothesis, there are no data in the literature showing that the 12/15LO gene expression is modulated by epigenetic modifications in this particular system.
Previous work showed that folate deficiency, a condition associated with HHcy, increases the production of Ab by affecting SAM-dependent DNA methylation of genes involved in APP metabolism, such as PS1 (Fuso et al., 2005). Our finding add an extra facet to the neurobiology of HHcy and AD pathogenesis by showing that lowering SAM/SAH ratio influences the methylation of 5LO DNA, which results in its activation and increased Ab formation. Considering the importance of both genetic and environmental risk factors in particular for late-onset AD, our study implicates 5LO as the gene and HHcy as the environmental factor, which by interacting together modulates cellular events such as amyloidosis very relevant to the disease processes.
Because we showed that HHcy associates with SAH elevation, lower methyltrasferases, hypomethylation of 5LO DNA, its gene and protein upregulation, and increase in Ab formation, next we wanted to prove a causative role for the elevation of SAH in these inter-related biological effects.
To this end, we setup experiment with neuronal cells transiently overexpressing SAH hydrolase, the only known enzyme to catalyze the breakdown of SAH and efficiently reducing its intracellular levels . These studies showed that SAH hydrolase over-expressing cells were protected from the HHcy-dependent 5LO upregulation and Ab elevation whereby rescuing the HHcy-dependent reduction in all 3 methyltransferase enzymes and 5LO hypomethylation. Additionally, we demonstrated that the effect of HHcy on the 5LO enzymatic pathway was secondary to an increase in transcription of 5LO mRNA rather than an increase in its stability as it was blocked by actinomycin-D. Similarly, blockade of de novo protein synthesis by cycloheximide prevented the HHcy-dependent upregulation of 5LO and increase in Ab formation.
In summary, our studies reveal an unknown biological link between HHcy and 5LO enzyme and elucidate a novel pathway involving Hcydependent SAH elevation, 5LO DNA hypomethylation, and Ab formation. Their significance lies in the establishment of a mechanistic relationship between Hcy (environmental risk factor) and 5LO (genetic risk factor) with the pathogenesis of AD via an epigenetic mechanism. Taken together, they support the hypothesis that impaired Hcy metabolism and dysregulation of important methylation reactions can trigger the activation of an enzymatic pathway which ultimately favors amyloidogenesis and AD onset.
In conclusion, our findings provide novel insights into the molecular mechanisms by which elevated circulating Hcy levels may promote the development of AD-like neuropathology in individuals carrying this environmental risk factor and ultimately afford us with useful information for novel therapeutic opportunities for them.

Experimental procedures Mice and treatments
Animal procedures were approved by Temple University and Fox Chase Cancer Center Animal Care and Usage Committee and in accordance with the Guide for the Care and Use of Laboratory Animals of the National Institute of Health. The 3xTg-AD mice harboring a human mutant APP (KM670/671NL), a human mutant PS1 (M146V) knock-in, and tau (P301L) transgene (Oddo et al., 2003) used in this article were previously described . Briefly, 3xTg-AD mice were randomized into two groups: control group mice [n = 6 (three males and three females)] received standard rodent chow, whereas the Hcydiet group mice [n = 7 (two males and five females)] received a standard rodent chow deficient in folate (< 0.2 mg kg À1 ), vitamin B6 (< 0.1 mg kg À1 ), and B12 (< 0.001 mg kg À1 ), which is known to induce HHcy in mice, starting at 5 months of age for 7 months .

(B) Western blot analysis of 5LO and 12/15LO in lysates from neuronal cells incubated with Hcy or vehicle (ctr). (C) Densitometric analysis of the immunoreactivity is shown in the previous panel. (D) Levels of LTB4 were assayed by ELISA in supernatant from neuronal cells incubated with Hcy (closed bars) or vehicle (ctr) (open bars). (E) Quantitative real-time reverse transcription-polymerase chain reaction (qRT-PCR) analysis of 5LO mRNA in lysates from N2A-APPswe cells incubated with Hcy (closed bars) or vehicle (ctr) (open bars). (F) Levels of SAH and SAM were assayed by HPLC in neuronal cells incubated with Hcy (closed bars) or vehicle (ctr) (open bars). (G) 5LO DNA methylation state in N2A-APPswe cells incubated with Hcy (closed bars) or vehicle (ctr) (open bars). (H) Western blot analyses for DNMT1, DNMT3a, and DNMT3b in lysates from neuronal cells incubated with Hcy (closed bars) or vehicle (ctr) (open bars). (I)
Densitometric analyses of the immuno-reactivity are shown in the previous panel. Data presented are mean AE SEM (*P < 0.05, n = 3). HHcy, hyperhomocysteinemia; 5LO, 5-lipoxygenase.

Biochemical analyses
Levels of SAH and SAM were assayed as previously described (Skelly et al., 2003;Moncada et al., 2008). Briefly, tissues were frozen in liquid nitrogen and reduced into a fine powder using mortar and pestle. The powder was re-suspended in 100 lL of NKPD buffer (2.68 mM KCl, 1.47 mM KH 2 PO 4 , 51.10 mM Na 2 HPO 4 , 7.43 mM NaH 2 PO 4 , 62 mM NaCl, 1 mM EDTA, and 1 mM dithiothreitol) and sonicated at 40 W and 70% duty cycle for about 2 min, and then clarified by centrifugation at 10 000 g for 15 min. Measurements were performed by HPLC analysis using Waters AccQ.Fluor derivatizing reagents (Waters Corp., Milford, MA, USA). The limit of detection is 0.02 nmol, and linearity extends to 5000 nmol. All of the samples were analyzed in triplicate, and the amount of the analyte in the samples normalized by mg protein. Precolumn derivatization of samples and HPLC analysis: an internal standard (10 lL of 2 lM 1,7-diaminoheptane) was added to 10-60 lL of mixtures of deproteinized biological extracts; and borate buffer (0.2 M sodium borate, 1 mM EDTA, pH 8.8) was added to produce final volumes of 80 lL. The AccQ-Fluor reagent (20 lL) was added and the sample mixed by vortexing and analyzed within 24 h. The HPLC analysis and gradient were obtained as previously described (Skelly et al., 2003;Moncada et al., 2008). LTB4 levels were assayed by a specific and sensitive enzyme-linked immunosorbent assay (ELISA) kit (Assay Designs Inc., Ann Arbor, MI, USA), following the manufacturer's instruction and as previously described (Chu & Pratic o, 2011). Ab1-40 and Ab1-42 levels were assayed by a sensitive sandwich ELISA kits (WAKO Chem., Richmond, VA, USA), as previously described (Chu et al., 2012a). Analyses were performed in duplicate and in a coded fashion.
5LO mRNA and Quantitative Real-time RT-PCR RNA was extracted and purified using the RNeasy mini-kit (Qiagen, Valencia, CA, USA), as previously described (Chu et al., 2012a,b). Briefly, 1 lg of total RNA was used to synthesize cDNA in a 20 lL reaction using the RT2 First Strand Kit for RT-PCR (SuperArray Bioscience, Frederick, MD, USA). 5-Lipoxygenase genes were amplified using the proper primers obtained from SuperArray Bioscience. b-Actin was used as an internal control gene to normalize for the amount of RNA. Quantitative real-time RT-PCR (qRT-PCR) was performed using StepOnePlus Real-Time PCR Systems (Applied Biosystems, Foster City, CA, USA). Two microliters of cDNA was added to 10 lL of SYBR Green PCR Master Mix (Applied Biosystems). Each sample was run in duplicate, and analysis of relative gene expression was carried out by StepOne software v2.1 (Applied Biosystems, Foster City, CA, USA).

5LO DNA Methylation assay
To measure the 5-LO DNA methylation levels, we used the restriction digest-qPCR assay as previously described (Dzitoyeva et al., 2009). Briefly, the method utilizes the ability of methylation-sensitive endonucleases to digest only un-methylated recognition sites and their inability to act on sites with methylated cytosine. Thus, if the targeted * CpG is methylated, the site is blocked for the enzyme's endonuclease activity, and as a result, greater amounts of templates are available for the action of the Taq DNA polymerase. The gene of the mouse 5-LO promoter to the 5-untranslated region (UTR) contains the highest C and G densities which have respective recognition sequences cutting sites of methylation-sensitive endonucleases, such as BstUI (CG ↓ CG) and HpaII (C ↓ CGG) (their respective recognition sequences cutting sites are shown in parentheses). The assay was performed as follows. Genomic DNA was extracted from the mouse brain cortex or N2A-APPswe cells using PureLink Genomic DNA Mini kit (Invitrogen) according to the manufacturer's protocol. One microgram of genomic DNA was used in a restriction digest reaction for two methylation-sensitive endonucleases, BstUI, and HpaII. Digested DNA samples were diluted with water, and an aliquot (100 ng DNA) was used for qPCR with SYBR Green PCR Master Mix (Applied Biosystems) as described in the method of quantitative realtime RT-PCR.

Data analysis
Data analyses were performed using SigmaStat for Windows version 3.00 (La Jolla, CA, USA). Statistical comparisons were performed by unpaired Student's t-test or the Mann-Whitney rank sum test when a normal distribution could not be assumed. Values represent mean AE SEM. Significance was set at P < 0.05.