Folate Related Pathway Gene Analysis Reveals a Novel Metabolic Variant Associated with Alzheimer’s Disease with a Change in Metabolic Profile

Metabolic disorders may be important potential causative pathways to Alzheimer’s disease (AD). Cerebrospinal fluid (CSF) decreasing output, raised intracranial pressure, and ventricular enlargement have all been linked to AD. Cerebral folate metabolism may be a key player since this is significantly affected by such changes in CSF, and genetic susceptibilities may exist in this pathway. In the current study, we aimed to identify whether any single nucleotide polymorphism (SNPs) affecting folate and the associated metabolic pathways were significantly associated with AD. We took a functional nutrigenomics approach to look for SNPs in genes for the linked folate, methylation, and biogenic amine neurotransmitter pathways. Changes in metabolism were found with the SNPs identified. An abnormal SNP in methylene tetrahydrofolate dehydrogenase 1 (MTHFD1) was significantly predictive of AD and associated with an increase in tissue glutathione. Individuals without these SNPs had normal levels of glutathione but significantly raised MTHFD1. Both changes would serve to decrease potentially neurotoxic levels of homocysteine. Seven additional genes were associated with Alzheimer’s and five with normal ageing. MTHFD1 presents a strong prediction of susceptibility and disease among the SNPs associated with AD. Associated physiological changes present potential biomarkers for identifying at-risk individuals.


Introduction
Early-onset, familial Alzheimer's disease (AD), with a prevalence of around 1%, is known to be associated with high-penetrance mutations in the genes coding for amyloid precursor protein (APP), presenilin 1 (PSEN1), and presenilin 2 (PSEN2) [1]. PSEN1 and 2 cause an impairment in γ-secretase activity and lead to an increase in the ratio of the 2 forms of amyloid beta, Aβ1-42: Aβ1-40 leading to AD with an early age of onset spread from 25 to 65 years of age. APP mutations result in early onset AD between 35 and 65 years of age. By contrast, late-onset AD is multifactorial, with many genetic risk factors, including APOE4, the highest risk factor for AD, environmental, nutritional, metabolic, and lifestyle factors. No causative genes have been identified for late-onset AD, but non-coding genetic errors have been suggested [2], as well as heritable and non-heritable epigenetic changes, as potential disease onset mechanisms [3], some of which may be offset through nutrition and diet [4,5]. Metabolic disorders have recently been highlighted as a potential cause and target for AD, including insulin signalling dysfunction and brain glucose metabolic disturbances, the latter suggested as hallmarks of AD and underlie the proposition that AD should be regarded as type III diabetes, specifically affecting the brain. patients [25], while others have reported raised intracranial pressure [26] as well as enlarged ventricles [25], suggesting a CSF drainage issue. Disease severity associated with increased ventricular enlargement in AD [27][28][29][30] supports a view that a CSF drainage obstruction may be operating in these patients and that shunting, in restoring optimal drainage, improves outcomes. Based on our hypothesis, CSF drainage obstruction, including reduced glymphatic pathway capacity [31][32][33][34][35][36][37][38], may produce cerebral folate issues that would add to the pathophysiology associated with AD.
Thus, it seems to us that CSF drainage insufficiency, resulting from multiple different causes, would lead to cerebral folate deficiency/imbalance, while at the same time failing to remove toxic molecules, including amyloid. It is also possible that rising amyloid levels could result in CSF drainage compromise through toxic effects on draining cells in the subarachnoid spaces. In either case, the situation would be greatly exacerbated. These possibilities should be investigated as a potential mechanism for various conditions of cerebral cortical malfunction as well as neurodegeneration leading to dementia, including Alzheimer's disease.
Nutrigenomics is a relatively new field of research that looks at individual susceptibilities based on analyses of genes involved in metabolic pathways and how they might respond to diet and environmental factors [4,71]. As a first step, in a case-control study design, we aimed to investigate whether any single nucleotide polymorphisms affecting folate and associated metabolic pathways were significantly associated with AD, which might have formed a susceptibility to AD triggered by other (environmental) factors acting on the susceptible metabolism. Figure 1 shows the inter-relationships between the folate metabolic cycle, DNA synthesis, methylation pathway, and neurotransmitter and nitric oxide synthesis. From this, it is clear that errors or issues with folate metabolism can have severe consequences for brain function through the synthesis of neurotransmitters, specifically biogenic amines and acetylcholine, the production and maintenance of cells, cardiovascular and neurovascular health, etc. 5-methyl tetrahydrofolate (5mTHF) is the main dietary form of folate and is the usual entry point into folate metabolism from where it feeds 2 pathways. The dihydro-tetrahydrobiopterin (BH2-BH4) cycle produces tetrahydrobiopterin from 5mTHF that is required for nitric oxide synthesis, linked to cardiovascular health, and biogenic amine synthesis, producing some of the key neurotransmitters of brain functions, including cognition, learning, memory, attention, mood, and sleep. 5mTHF also passes through a rate-limiting step, transferring its methyl group to vitamin B12 that then methylates homocysteine to methionine, producing tetrahydrofolate (THF) that feeds into other parts of the metabolic cycle. Methionine is converted to s-adenosyl methionine (SAM), the universal methyl donor involved in most methylation reactions. synthesis, producing some of the key neurotransmitters of brain functions, including cognition, learning, memory, attention, mood, and sleep. 5mTHF also passes through a ratelimiting step, transferring its methyl group to vitamin B12 that then methylates homocysteine to methionine, producing tetrahydrofolate (THF) that feeds into other parts of the metabolic cycle. Methionine is converted to s-adenosyl methionine (SAM), the universal methyl donor involved in most methylation reactions. Figure 1. Diagram of the folate metabolic cycle and links to other pathways 5-methyl-THF is the major species of folate derived from food and forms the recycling point for folate metabolism. It forms the rate-limiting step, through the action of methionine synthase (MTR), in the methylation of vitamin B12 and thus the rate-limiting step for the methylation of homocysteine to methionine. Thus, it is critical to folate metabolism generally and to the production of S-adenosyl methionine (SAM), the universal methyl donor for methylation reactions. MTHFD1 is a multi-role enzyme involved in three reactions in the folate pathway, forming a long route back to 5-methyl-THF. Tetrahydrofolate (THF) can be recycled back through 5, 10-methenyl THF, forming a short route and requiring B6 and serine to glycine reactions. Both long and short routes require methyleneTHF reductase (MTHFR) for the final step to 5-methyl-THF. MTHFD1 is further involved in recycling 10formyl-THF to THF, fuelling purine synthesis. 5, 10-methenyl THF can also be reduced to dihydrofolate, fuelling pyrimidine synthesis. Other pathways include 5-methyl-THF feeding directly into biogenic amine and nitric oxide synthesis through the BH4 cycle, methionine feeding directly into the methylation pathway, and acetylcholine synthesis. Folic acid is an artificial substance that enters the folate cycle without any 1 carbon moiety to supply to the metabolic process and so acts to dilute the 1 carbon pool, as well as having other negative effects at higher doses (see text). Diagram of the folate metabolic cycle and links to other pathways 5-methyl-THF is the major species of folate derived from food and forms the recycling point for folate metabolism. It forms the rate-limiting step, through the action of methionine synthase (MTR), in the methylation of vitamin B12 and thus the rate-limiting step for the methylation of homocysteine to methionine. Thus, it is critical to folate metabolism generally and to the production of S-adenosyl methionine (SAM), the universal methyl donor for methylation reactions. MTHFD1 is a multi-role enzyme involved in three reactions in the folate pathway, forming a long route back to 5-methyl-THF. Tetrahydrofolate (THF) can be recycled back through 5, 10-methenyl THF, forming a short route and requiring B6 and serine to glycine reactions. Both long and short routes require methyleneTHF reductase (MTHFR) for the final step to 5-methyl-THF. MTHFD1 is further involved in recycling 10-formyl-THF to THF, fuelling purine synthesis. 5, 10-methenyl THF can also be reduced to dihydrofolate, fuelling pyrimidine synthesis. Other pathways include 5-methyl-THF feeding directly into biogenic amine and nitric oxide synthesis through the BH4 cycle, methionine feeding directly into the methylation pathway, and acetylcholine synthesis. Folic acid is an artificial substance that enters the folate cycle without any 1 carbon moiety to supply to the metabolic process and so acts to dilute the 1 carbon pool, as well as having other negative effects at higher doses (see text).

Folate Metabolism
THF is a central hub of folate metabolism. It can produce 5, 10-methylene THF through the conversion of serine to glycine. This can cycle back to 5mTHF or convert to dihydrofolate (DHF), giving up its methylene to thymidylate synthase that produces pyrimidines, key nucleotides in DNA, and RNA synthesis. DHF is hydrolysed back to THF. THF can also receive a formyl group through the conversion of formate from blood plasma that also exists in CSF [72], or through the metabolism of 5-formyl-THF. This is mediated by methylene-THF-dehydrogenase 1 (MTHFD1), which also converts the product, 10-formyl THF, to 5, 10-methenyl THF. MTHFD1 also acts to balance the 2 halves of folate metabolism through interconversion of 5, 10-methenyl and 5, 10-methylene THF. These 3 reaction steps, mediated by MTHFD1, are known as the long route back to 5mTHF, while MTHFR mediates the final step from 5, 10-methylene THF back to 5mTHF for both the long route and short route. 10-formyl-THF dehydrogenase (FDH) acts as a buffer to maintain a pool of the reactive THF, as it is known to bind tighter to this product than to the substrate 10-formyl-THF [73]. This would also have the effect of depleting 1-carbon availability in the presence of high levels of FDH that would decrease 10-formyl-THF levels and prevent downstream conversions, including purine synthesis. This would explain cell cycle arrest in cancer and other cells produced by elevated levels of FDH [74]. Figure 1 also shows the entry point of folic acid, a synthetic, stable form of folate. At high doses of folic acid, there would be an overload of the system with increased DHF, perhaps slowing down/inhibiting conversion of 5, 10-methylene-THF to DHF and thereby decreasing the production of pyrimidines. Increasing THF levels through this pathway might also have a knock-on effect in slowing down the methylation of homocysteine to methionine. Further details are provided in the figure. High-dose folic acid supplementation has been linked to abnormal folate metabolism and issues related to folate deficiencies [75][76][77] although some caution is required in the interpretation of these studies and conclusions [78]. In our previous research on neonatal hydrocephalus, we also found a negative effect of folic acid in precipitating hydrocephalus in a susceptible rat model of this congenital condition [12]. Folate, but perhaps not folic acid, is thus a vital metabolite, and its metabolism is essential to many associated pathways involved in the development, maintenance, and healthy functionality of tissues, particularly the brain. Up or downregulation of any of the enzymes involved would have significant effects, as would genetic errors affecting the function of these proteins. Table 1 gives details of the cases/samples used in the study, while Table 2 analyses clinical and phenotypical characteristics, highlighting the differences between the 2 groups of patients. Interestingly, the AD group is significantly younger than the normal ageing group, with a mean age of 75.2 vs. 86.5. Thal phase, CERAD, score, and Braak stage all accurately associate with the AD group [79] with an additional highly significant association of the APOE genotype. Brain weight is significantly lower in AD, with a mean weight of 1133.5 g vs. 1291.9 g. Although the mean postmortem delay is different between groups, this, along with brain pH, does not show significance.
In the multivariable model, three SNPs emerged as strong predictors of AD (COMT -rs4680 omitted due to collinearity) ( Table 1). The contribution of ADRB1 (rs1801253) was observed to be highest in disease status (OR: 9.61). The risk of SNPs in the multivariable analyses was best explained by the dominant model (Table 3).
The highest odds were observed for CBS (rs234706) and MTHFD1 (rs1076991) (OR: 95.12 and 67.38, respectively). The risk of SNPs was most significant under the additive model (Table 4). 25 gene variants were identified that were significant in any of the three tests with all significant at p ≤ 0.05 and many at the much higher significance of p ≤ 0.01 or higher (bold p-values in Table 5). Only 12 were significant using Mann-Whitney U tests (Supplementary Table S4), while more were significant in either of the Chi Squared tests. Some were significant across all tests ( Table 5). The most significant finding was that ApoE4 was associated with AD. Even though ApoE4 is known in the literature to have a 40% risk of the disease, the current finding is surprising in showing a 70% association in a small number of individuals picked for disease severity.  Importantly for our hypothesis, 2 folate-related genes were found to be significantly associated with AD, methylene tetrahydrofolate dehydrogenase 1 (MTHFD1) and methylene tetrahydrofolate reductase (MTHFR), with MTHFD1 significant on all tests and MTHFR significant only on a Chi, Squared test in which positive and neutral variants were grouped and tested against negative variants. Table 5 also shows the direction of the association, i.e., whether associated with AD or with normal ageing. Gene SNPs associated with normal ageing may provide some protection from AD.
Other SNPs associated with AD are involved in monoamine transport at synapses (SLC18A10, SLC6A2) as well as involved in detoxification of xenobiotics and sulphites (CYP2D6, SUOX). Those associated with normal ageing and not AD, which may therefore be protective against AD, are involved in monoamine metabolism, methylation, and signalling (MAOA, ADRB1, COMT), and detox pathways (CYP2D6, SOD2, GSTM1). There are also SNPs involved in thyroid hormone activation and transport (DIO2, SLC01C1) and neurotransmitter receptors (5-HT2A, DRD2). The remaining SNPs, only significant on Chi Squared tests where negative and neutral SNPs are grouped, are associated with AD and are involved in methylation, including betaine homocysteine methyl transferase (BHMT), cysteine beta-synthase (CBS), and glutathione S-transferase P1 (GSTP1).

Changes in the Metabolic Profile Associated with Folate Gene SNP
We measured folate metabolites and enzymes in tissue lysates of normal and Alzheimer's disease individuals with negative SNPs in MTHFD1 and/or MTHFR and compared these to normal and Alzheimer's disease individuals with normal or neutral SNPs in these genes. The individuals, their genotypes, and results of the analysis are shown in Tables 6 and 7, with the data shown in graphical forms in Figures 2 and 3. Table 6. Samples used in folate analyses correlated to gene SNP data.

Normal Ageing Alzheimer's Normal Controls Alzheimer's Controls
SampleIDs Target  molecules  12_11  09_31  14_09  14_11  12_01  13_45  14_50  19_29  19_31  11_25  14_08  17_36  17_34  19_04 A  12_05 B  12_32 C  11_28 D   MTHFD1  (rs1076991)  CT  TT  TT  TT  TT  TT  TT  TT  TT  CC  CC  CT  CC  CT  CT  CT  CT   MTHFD1  (rs2236225)  AA  AG  AA  AA  AA  AA  GG  GG  AA  GG  GG  AG  AG  AG  GG  GG   MTHFR  (rs1801131)  TT  GT  GT  TT  GG  GG  TT  GG  TT  GT  GT  TT  TT  TT  GT  GT   Tests of significance. Bold values were calculated for each measure against that for tissue folate, which was not significantly different between the groups (see Figure 2) and thus was used as a measure of all other elements.
Each metabolite or enzyme was tested against a normal control (C). Glutathione is significantly different in both normal and Alzheimer's with mutant SNPs but not in Alzheimer's cases that do not. MTR is significantly different in both Alzheimer's with or without negative SNPs. MTHFD1 is significantly different only in Alzheimer's without negative SNPs. These data are shown in graphical form in Figures 2 and 3.
There was no significant difference in tissue folate levels ( Figure 2) although the controls had a non-significant reduced level compared to the other groups. We, therefore, used the average folate level to calculate the fold levels of the other metabolites and enzymes (Figure 3). In the severe Alzheimer's cases that have negative SNPs in MTHFD1, there is a significant increase in glutathione that is seen in the severe cases without these SNPs. There was no effect of the negative SNPs on the levels of either MTHFD1 or MTHFR. However, in severe cases with normal or neutral SNPs, there is no rise in glutathione, but there is a significant rise in MTHFD1.
There is also a significant rise in MTR in severe cases, both with and without negative SNPs, relative to controls and a small but non-significant (p ≤ 0.06) increased MTR in severe cases without negative SNPs compared to severe cases with negative SNPs.
There was no significant difference in tissue folate levels ( Figure 2) although the controls had a non-significant reduced level compared to the other groups. We, therefore, used the average folate level to calculate the fold levels of the other metabolites and enzymes (Figure 3). In the severe Alzheimer's cases that have negative SNPs in MTHFD1, there is a significant increase in glutathione that is seen in the severe cases without these SNPs. There was no effect of the negative SNPs on the levels of either MTHFD1 or MTHFR. However, in severe cases with normal or neutral SNPs, there is no rise in glutathione, but there is a significant rise in MTHFD1.   . Metabolic profiles of samples analysed for key metabolites and enzymes plotted as fold of folate levels since this was the only consistent measure between the cases. Glutathione is significantly raised in Severe Alzheimer's with mutant gene SNPs for MTHFD1 and MTHFR. This is not seen in any other group, including Alzheimer's without mutant SNPs. In these latter cases, we see significantly raised MTHD1, which is not a mutant. In both cases, we see potential protective mechanisms against raised homocysteine levels, which are elevated in all cases except normal ageing, although none are significant. MTR is also elevated in both affected and control Alzheimer's cases. This is not mutated, so it is also involved in metabolising 5-methyl-THF to methylate homocysteine to methionine and thus feeding the methylation pathway. Together, the increased glutathione and MTR would act to keep homocysteine levels low. Significance is indicated by stars: ** significant at 0.01, *** significant at 0.001.
There is also a significant rise in MTR in severe cases, both with and without negative SNPs, relative to controls and a small but non-significant (p ≤ 0.06) increased MTR in severe cases without negative SNPs compared to severe cases with negative SNPs.

Discussion
There is an assumption in the genetics and nutrigenomics literature that adverse gene SNPs have a negative effect on protein function and thus a knock-on effect on the processes they are involved in. In this study, we focused on the genes involved in folate me- Figure 3. Metabolic profiles of samples analysed for key metabolites and enzymes plotted as fold of folate levels since this was the only consistent measure between the cases. Glutathione is significantly raised in Severe Alzheimer's with mutant gene SNPs for MTHFD1 and MTHFR. This is not seen in any other group, including Alzheimer's without mutant SNPs. In these latter cases, we see significantly raised MTHD1, which is not a mutant. In both cases, we see potential protective mechanisms against raised homocysteine levels, which are elevated in all cases except normal ageing, although none are significant. MTR is also elevated in both affected and control Alzheimer's cases. This is not mutated, so it is also involved in metabolising 5-methyl-THF to methylate homocysteine to methionine and thus feeding the methylation pathway. Together, the increased glutathione and MTR would act to keep homocysteine levels low. Significance is indicated by stars: ** significant at 0.01, *** significant at 0.001.

Discussion
There is an assumption in the genetics and nutrigenomics literature that adverse gene SNPs have a negative effect on protein function and thus a knock-on effect on the processes they are involved in. In this study, we focused on the genes involved in folate metabolism, methylation, and neurotransmitter synthesis, and also included APOE genotyping. Surprisingly, we found a 70% association of APOE4, rather than the predicted 40%, with AD further reinforcing its high-risk factor status and also indicating potential direct involvement in the condition in the severe cases used in this study. Several recent studies highlight the role of ApoE in several critical processes, including involvement in amyloid plaque and neurofibrillary tangle formation, insulin resistance, decreased amyloid clearance, mitochondrial dysfunction, and autophagy [5,80]. Suggestions have been made for targeted drugs [80] and nutritional and lifestyle changes [5] aimed at ApoE4 processes and pathways to prevent or treat the disease. The second most significant association with AD was MTHFD1 SNP rs1076991. This was highly significant in all tests, and thus we can suggest must be a significant risk factor for late-onset AD. Others have found a weak association between a different MTHFD1 variant, SNP rs2236225, and early-onset AD [81,82]. We found a highly significant association of this variant with AD only in a Chi-Squared test, where negative and neutral variants were grouped together and tested against positive variants. There was no significance in the Mann-Whitney U test or the alternative Chi-Squared test. Therefore, we can agree with the studies demonstrating the weak association of this variant but have found a very significant association with the other variant of MTHFD1, which is a novel finding of this study. MTHFD1 is a critical folate enzyme involved in 3 parts of folate metabolism ( Figure 1) and 4 enzymatic profiles: (i) methylenetetrahydrofolate dehydrogenase (NADP+ dependent) 1, (ii) methenyltetrahydrofolate cyclohydrolase, (iii) formyltetrahydrofolate synthetase, and (iv) C-1-tetrahydrofolate synthase, reflecting its importance to the folate metabolic cycle. Thus, a negative variant of MTHFD1 should have a remarkable effect on folate metabolic balance, decreasing the 5mTHF pool and potentially leading to homocysteine and formate toxicity and errors/reduced DNA synthesis and repair. Similarly, a negative variant of MTHFR should result in raised homocysteine levels that would exacerbate neurodegeneration and increase the risk of developing AD [83]. In the cases studied here, MTHFD1 is very significantly associated with AD, while MTHFR is probably only weakly associated, as already reported for this variant [84] by contrast to other studies finding 1-3 additional variants of this gene associated with AD [83,85,86]. We found a significant increase in glutathione in severe Alzheimer's cases with negative variants in MTHFD1 and/or MTHFR compared to those with positive or neutral variants ( Figure 2). We surmise that raised homocysteine, resulting from failure to regenerate 5-methylTHF, is being shunted to SAM and glutathione to prevent the toxic build-up of homocysteine. Interestingly, this was not seen in severe cases with positive or neutral variants in MTHFD1 and/or MTHFR. In these cases, we found a significant increase in MTHFD1 and MTR perhaps in response to raised homocysteine and to increase methylation. This would also fuel the hypermethylation seen in the Alzheimer's cortex and reported by us and others (Naz et al., 2022, in revision) and may explain the reduced SAM we report here. Additionally, we found no significant difference in tissue levels of folate, indicating that the changes seen are likely to be a response to ineffective gene products and/or to physiological changes in metabolism rather than folate supply. These significant changes in metabolic profile may provide early markers of at-risk individuals who may therefore be protected by appropriate metabolic support, including perhaps folate supplements [12,[87][88][89].
The other gene variants associated with AD are involved in monoamine neurotransmitter delivery to, and reuptake into synapses, and in detoxification from xenobiotics and sulfites. 2 genes involved in the methylation pathway are weakly associated with AD, finding significance in only one of the Chi Squared tests. These are involved in the conversion of homocysteine to methionine (BHMT) and in the generation of cysteine from homocysteine in the pathway to glutathione (CBS). The remaining significant associations are with normal ageing, indicating a possible protective effect of these gene variants. These are involved in receptors for and breakdown of monoamines, as well as in detoxification through breakdown of various drugs and environmental toxins. Interestingly, COMT variant rs4633 has been associated with Alzheimer's [90] while in the current study it is clearly associated with normal ageing. Other variants of COMT have been found to be not associated with Alzheimer's or other psychiatric conditions [91][92][93] indicating that COMT may be associated with other related factors rather than directly to disease aetiology or progression.
We propose that one trigger for the onset and severity of Alzheimer's may be a physiological change associated with a cerebral CSF drainage issue and a consequential cerebral folate issue. This is reflected in the fact that the severity of this, and other cerebral conditions, is associated with increasing fluid accumulation and ventricular dilation. Life events that decrease drainage capacity might include infection, inflammation, and trauma or accelerated cell loss in these susceptible ageing individuals as well as due to toxic levels of amyloid, e.g., as found in sleep deprived individuals. Strategies to maintain drainage, and perhaps increase drainage may therefore present an effective target for prevention and treatment of this condition. Changes in metabolic profiles, as found in this study, may also provide early markers of at-risk individuals.

Material and Methods
Brain tissue: All brain tissues were supplied from the Manchester Brain Bank under their ethical approval (09/H0906/52+5 and 19/NE/0242) for the collection and use of human tissue in ageing research. In this study, we were provided with frozen unfixed tissue from the occipital cortex that included the full thickness of the cortex from the pia to the ventricular ependymal. CSF samples were also provided from the same brains, which were collected postmortem. Information on the cases used is provided in Table 1. Only individuals who were clearly normal, based on both pathology and clinical observations prior to death, or who were clearly suffering severe Alzheimer's disease were included in this study. 25 individuals from each category were analysed.
The cerebral cortical plate was dissected and then sent to a commercial company, LGC Genomics, for quality checking, further processing, and SNP analysis. They extracted DNA from the tissue using their in-house LGC Kleargene extraction chemistry. Genotyping was performed by LGC genomics using their in-house competitive allele specific PCR (KASP) technique. Genotype data were sent to LifecodeGX Ltd. to input them into their bespoke software that matches SNPs to specific metabolic pathways. The software also categorised SNPs according to their functional effect based on literature reviews, namely beneficial, neutral, or harmful (see https://www.lifecodegx.com/, accessed on 19 May 2022, for details and references). These data were then tabulated in Microsoft Excel. Heat maps were generated in Excel to give a pictorial representation of the SNP data (Supplementary data). Statistical analyses were carried out using GraphPad Prism and STATA (11.0). Genotypic and allelic frequencies of SNPs were calculated and compared between normal subjects and AD patients. Categorical variables were given as absolute and relative frequencies (no., %age). The significance of the differences from random distribution was estimated through Chi-square, Yates' corrected Chi-square, and Fisher's exact tests. The quantitative variables were expressed as mean ± SD (standard deviation), and T-test statistics were employed for comparison. At each SNP, odds ratios (OR) and respective 95% confidence intervals (95% CI) and univariable regression were calculated while keeping the most common homozygous genotype as a reference. Multivariable regression was performed to identify the combination of SNPs associated with AD. For this, a stepwise logistic regression model was applied by including all SNPs one by one and eliminating the non-significant ones. To predict the most plausible inheritance model, significant SNPs in multivariable regression were also assessed under dominant, recessive, and additive models. The level of significance for the p-value was <0.05 for all tests.

Supplementary Materials:
The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/metabo12060475/s1, Table S1: Heat maps of the effects of SNP variants of known genes found in the neurotransmitter pathway in normal (a) and severe Alzheimer's disease (b) cases, Table S2: The odds of differences of genotypic allelic frequencies of SNPs related to nervous system in normal subjects and AD patients: outcome of univariate regression depicted as OR and 95% CI, Table S3: The odds of differences of genotypic allelic frequencies of SNPs related to methylation in normal subjects and AD patients: outcome of univariate regression depicted as OR and 95% CI, Table S4: Mann-Whitney U tests of significance of SNPs associated with normal ageing or Alzheimer's disease. Highlighted genes show significant associations.  Informed Consent Statement: Informed consent was obtained as part of the Manchester Brain Bank Ethical Review. This study was covered under the Brain Bank ethical permission for research on collected tissue.

Data Availability Statement:
The data is presented in the paper in tables and supplementary tables.