Introduction

The most prevalent cause of dementia is Alzheimer’s disease (AD), characterized by regional brain atrophy, memory deficits, and deterioration of executive functions [1, 2]. AD causes memory and executive function problems, affects other cognitive domains, and causes neuropsychiatric symptoms. AD is a multifactorial disease, however, the etiopathogenesis not being fully understood. Amyloid precursor protein metabolism, phosphorylated tau aggregation, impaired kynurenic acid pathway, and mitochondrial dysfunction are identified as culprits in disease pathology. Lower nicotinamide adenine dinucleotide (NAD+) secondary to impaired tryptophan-kynurenine metabolic system causes aerobic respiration dysfunction which leads to anaerobic respiration and energy loss in AD [3,4,5,6,7]. With the increase in life expectancy, it is predicted that in 2050, more than 106 million people will be diagnosed with AD worldwide, while in 2006, the number of patients was equal to 26 million [8, 9]. Amyloid and tau pathologies, which have different structural abnormalities, are two distinct pathologies that manifest at various stages of the disease’s progression. It has been proposed that when the amyloid and tau diseases overlap, the sickness may turn into a distinct illness [10,11,12]. Prior to that, specific characteristics may have been linked to the development of just plaques in specific brain regions [13].

The posterior cingulate cortex (PCC) is a paralimbic cortical structure located in the core hub of the default mode network (DMN), with a fundamental role in the integrative functions of the DMN, including memory processing and encoding [14]. During episodic memory recall activity, there is the deactivation of the PCC and hippocampal activation. PCC interactions with the hippocampal gyrus of the medial temporal lobe (MTL) network and prefrontal cortical regions are essential for their role in cognitive function. Prominent Aβ deposition and hypo-metabolism in PCC, despite a low degree of atrophy, support the idea of the vital role of PCC dysfunction in connectivity disruption and memory loss associated with AD [15,16,17].

The measurement of the metabolome is related to a novel field of research with increasing importance called metabolomics [18]. This field can aid us by providing a comprehensive diagnosis of neurodegenerative disorders by characterizing absolute metabolites in a specific sample. Metabolomics is more sensitive to environmental and physical factors than genomics and proteomics [19]. Hence, to evaluate the metabolomics content changes in PCC, it can be helpful to compare AD patients and healthy individuals by the metabolome content in PCC [20].

Early diagnosis of AD has been one of the hot-spot research focus, fluid biomarkers, as well as neuroimaging techniques, have been utilized to elucidate the pathophysiology of early AD [21, 22]. The loss of functional integrity of the frontal and hippocampal-based memory systems in individuals can be utilized as a measure of neurodegenerative processes in individuals with a high risk of dementia or with a diagnosis of AD: evidence from recent studies provided an overview of the anatomical–functional interplay between the prefrontal cortex and heart-related dynamics in human emotional conditioning (learning). It proposed a theoretical model to conceptualize these psychophysiological processes, the neurovisceral integration model of fear (NVI-f), that can be impaired in neurodegenerative disorders [23, 24]. Nuclear magnetic resonance (NMR), positron-emission tomography (PET), and mass spectrometry (MS) with conventional techniques have raised hope of achieving this goal [5].

Proton nuclear magnetic resonance (1H-NMR), also called NMR spectroscopy, helps us measure quantitatively microstructural disruptions. Different electron shielding around the metabolites is the discriminative factor in assessing the signals of each metabolite in a strong magnetic field [25]. NMR spectroscopy-based metabolomics is a non-invasive method that does not require complex manipulation protocols of samples. Although for a precise and comparable measure of brain metabolites, MR spectroscopy needs some (and sometimes many) manipulations of protocols post-processing and is subjected to many artifacts. So, it is not a simple technique. The standardizing methods are the main limitation of applying spectroscopy on a large scale and comparing it in different centers [26, 27]. Therefore, this study aimed to look at the similarities and discrepancies of the studies, critically review the advantages and limitations of using NMR spectroscopy as a marker of AD, and investigate the metabolites changes in PCC as one of the signature AD regions in AD patients compared to HC.

Methods

The systematic review and meta-analysis protocol was registered in the PROSPERO, a prospective international register of systematic reviews under record number CRD42018099325. This study was done with standard guidelines such as the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [28] and the Meta-analysis of Observational Studies in Epidemiology (MOOSE) [29].

Study criteria

The cohort, cross-sectional, and case–control design studies were included and evaluated metabolomics changes in PCC patients with AD through the 1H-NMR technique. Patients will be recruited if they are above 65 with the AD diagnosis by altered biomarkers (CSF or PET-scan) and clinical symptoms. Outcome measures included Population (HC/AD), Age (y), Gender (Male/Female), MRI equipment, Tesla (T), MMSE (mean ± SD), absolute and ratio absolutes metabolites in the PCC. HC group were all age and sex-matched with the AD patients and had no history of neuropsychological disorders.

Search methods

ISI Web of Science, Cochrane Library, PROSPERO, PubMed, Scopus, CINAHL, Science Direct from inception, EMBASE, and any of the included research reference lists.

The search was started in 2000 until May 2019. Searches were done using the medical subject headings (Mesh) keywords (“1H-NMR” OR “MRS” OR “Magnetic Resonance Spectroscopy”) AND (“Alzheimer disease” OR “Dementia”) AND (“Metabolite” OR “Metabolomics”). Any meta-analysis, review studies (narrative or systematic), case series or case reports, commentaries, and letters to the editor were excluded. Studies fulfilling the above criteria, with a cohort, cross-sectional, or case–control design, and full-text information were available and considered in the meta-analysis.

In our data collection, studies were included whether they were about brain 1H-NMR, examined individuals with AD, comprised healthy control groups, and compared absolute metabolites in the PCC region. The importance of knowledge was independently derived from experiments by two scholars (K.SH. and F.R.) If applicable, we have contacted the writers of the qualifying papers for missing details.

Quality assessment

Quality assessment was assessed according to standardized tools for grading cohort studies (Newcastle Ottawa-Scale [30]). Newcastle Ottawa-Scale is a tool to assess risk of bias and quality used in a systematic review study with included non-random studies.

Publication bias

The visual inspection of the funnel plots and the Egger test for each group and each metabolite were used to observe the propensity for publishing bias [31]. A significance level of P < 0.10 defined significant publishing bias based on the Cochrane handbook for systematic reviews [31].

Sensitivity analysis

Sensitivity analysis based on study quality (risk of bias) to investigate possible sources of heterogeneity. The primary decision nodes concluded methodological consistency, sample size, and the impact of missed data. The researchers replicated the meta-analysis, and low-quality findings were omitted. The outcome was contrasted and debated based on the extracted conclusions from other researchers.

Statistical analysis

All meta-analyses were carried out using STATA V.14 tools to provide pooled figures, with a corresponding 95 percent confidence interval (CI) and plots for studies disclosing absolute metabolite and metabolite ratio in the PCC. The mean estimates of each sample were pooled using a model of random effects for meta-analysis [32]. The I2 figures and the Chi-square test have been used to test heterogeneity. It was considered that I2 > 50 percent or P < 0.05 demonstrated substantial heterogeneity. In addition, to determine publishing bias, the funnel plot and the Egger regression test were added.

Results

Study selection

A comprehensive search of the literature conducted a total of 3067 relevant studies. A total of 764 were excluded after duplicates. Title—the abstract screening was excluded, and several 2276 studies and 27 full texts were selected (Fig. 1). At last, 18 original articles with cross-sectional or case–control design, comprising 1647 cases, [658 (40%) males and 941 (60%) females] were selected. Healthy controls (HC) 921 (55.9%) and AD 678 (44.1%) were included in the meta-analysis [33,34,35,36,37,38,39,40,41,42,43,44,45,46,47,48,49,50]. Nine studies were excluded after full-text screening (Additional file 1: Table S1) [51,52,53,54,55,56,57,58,59]. The mean age of the AD and HC participants was 73.33 ± 4.22 and 69.73 ± 7.15, respectively, and the Mini-Mental State Examination (MMSE) of the AD and HC were 19.45 ± 2.63 and 28.36 ± 1.27, respectively. Outlines of the search method and the number of studies excluded during each phase of the search are provided in Fig. 1. The table gives a detailed overview of the study population of each of the 18 reviewed studies (Table 1).

Fig. 1
figure 1

Study flow diagram showing how to extract articles

Full size image
Table 1 Characteristic of included studies on various metabolites between patient with Alzheimer’s disease (AD) and healthy controls (HC)
Full size table

A. Metabolite concentration

Metabolite concentrations were reported in 8 studies. N-Acetyl aspartate (NAA), creatine (Cr), choline (Cho), myo-inositol (mI), glutamine (Glu), and glutamate + glutamine (GLx) concentrations were extracted as the target variables (Table 1). On the one hand, results displayed a significant increase in mI content of the PCC in AD group compared to controls (0.32 [95% CI 0.19, 0.46]) (Table 2) (Additional file 1: Figs. S1–S5); on the other hand, there was significant decrease in NAA, Glu, and Glx levels of the PCC in AD participants compared to HC, (mean difference = 0.91 [95% CI − 1.05, − 0.77], 0.75 [95% CI − 1.00, − 0.50], and 0.67 [95% CI − 1.13, − 0.20], respectively) (Table 2) (Additional file 1: Figs. S6 and S7). No significant difference between AD and HC groups was observed in the concentration of Cr and Cho (Figs. 2, 3, 4).

Table 2 Summary of outcome of subgroup analysis of various metabolites between patient with Alzheimer’s disease (AD) and healthy controls (HC)
Full size table
Fig. 2
figure 2

Compared to healthy controls, forest plots of comparison N-acetyl aspartate (NAA) and creatine (Cr) in AD patients. Data type: continuous; effect size: Hedges’ g; effect model: fixed and random model; CI: confidence interval

Full size image
Fig. 3
figure 3

Forest plots of comparison choline (Cho) and myo-inositol (mI) in AD patients compared to healthy controls. Data type: continuous; effect size: Hedges’ g; effect model: fixed model; CI: confidence interval

Full size image
Fig. 4
figure 4

Compared to healthy controls, forest plots of comparison glutamine (Glu) and glutamate + glutamine (GLX) in AD patients. Data type: continuous; effect size: Hedges’ g; effect model: fixed and random model; CI: confidence interval

Full size image

B. Metabolite ratio

Metabolite ratios reported in 18 studies, NAA/Cr, NAA/mI, Cho/Cr, mI/Cr, scyllo-inositol (sI)/Cr, mI/NAA, Glu/Cr, mI/Cho were extracted (Table 1). The results revealed a significant decrease in NAA/Cr ratio (mean difference = 0.15 [95% CI − 0.20, − 0.09]) (Fig. 5), however, the results for Cho/Cr (0.05 [95% CI 0.03, 0.06]) (Fig. 5) and mI/Cr ratio (0.10 [95% CI 0.08, 0.13]) (Fig. 6) were associated with significant increase in the AD group compared to controls (Table 2; Additional file 1: Figs. S8–S13). No significant differences were found in the PCC mI/NAA ratio between the AD and HC groups (Fig. 6). Due to insufficient studies, we could not conduct a meta-analysis for NAA/mI, SI/Cr, Glu/Cr, and mI/Cho ratios (Fig. 7).

Fig. 5
figure 5

Compared to healthy controls, forest plots of NAA/Cr and Cho/Cr in AD patients. Data type: continuous; effect size: Hedges’ g; effect model: random model; CI: confidence interval

Full size image
Fig. 6
figure 6

Forest plots of mI/Cr and mI/NAA in AD patients compared to healthy controls. Data type: continuous; effect size: Hedges’ g; effect model: random model; CI: confidence interval

Full size image
Fig. 7
figure 7

Forest plot of comparison single metabolites in AD patients compared to healthy controls. Data type: continuous; effect size: Hedges’ g; effect model: random model; CI: confidence interval

Full size image

Discussion

According to this meta-analysis, in PCC absolutes, substantial variations were observed between early detection AD patients and HC of the NAA, mI, Glu, and Glx and metabolite ratios, including NAA/Cr, Cho/Cr, and mI/Cr. There was no significant difference in Cr and Cho levels and mI/NAA ratio in early detection AD compared with HC groups. The hippocampus, crucial for declarative memory, serves as the physical hallmark of AD [60]. Potential indicators for predicting the transition from MCI to AD include hippocampal shrinkage and its rate of atrophy [61]. It has also been shown that functional and structural networks connected to the hippocampus have decreased integrity [62]. Although different studies use different network mapping techniques, the geographical distribution of the discovered synchronized degeneration networks (SDNs) with hippocampal epicenters is remarkably consistent with earlier research [63]. Numerous AD histological and neuroimaging discoveries have suggested that the hippocampus and surrounding areas, generally known as the medial temporal lobe, have close relationships [64, 65]. According to a large body of research, the hippocampus and prefrontal cortex connect via oscillatory synchrony, reflecting bidirectional information flow, and may play a significant role in memory and learning outside of the limbic system [66]. These regional and network-level discoveries highlight the significance of the hippocampus and the related functionally and structurally connected regions in the pathophysiology of AD [67].

NAA is synthesized in the mitochondria of the brain cell and is considered a marker of neuronal integrity, visibility, density, functional mitochondria, and capacity in brain tissues [68]. The remarkable decline in the absolute metabolite in AD patients will represent neuronal loss and mitochondrial function [69]. Therefore, a significant decrease in NAA levels of the PCC would directly reflect early detection of AD pathology in this region. 1H-NMR studies have also shown a widespread reduction in NAA/Cr ratio in AD patients and its association with extensive neuronal dysfunction and loss [54]. NAA/cr is thus one of the sensitive markers for AD, in line with our findings of decreased NAA/cr ratio in the PCC [54, 70, 71]. Cho is mainly present in myelin and cell membranes as Cho-bound membrane phospholipids [72]. Free Cho and acetylcholine absolute cytosol are hardly detectable through 1H-NMR due to their low absolute in brain tissues [73]. An elevated level of Cho detected by 1H-NMR is attributed to the catabolism of phosphatidylcholine in the neuron membrane to provide free choline molecules in the cytosol, increasing it in absolute [51, 73]. Similarly, disinhibited regulation of choline acetyltransferase due to neuronal dysfunction can lead to high absolute Cho, justifying the high Cho/Cr level in early detection AD patients that can be interpreted into changes in membrane metabolism and neuronal membrane disruption. γ-Secretase is a complex of transmembrane proteases which consists of the PEN2, PSEN, NCSTN, and APH-1 monomers, and The breakdown of the amyloid precursor protein into insoluble amyloid β-peptides (Aβ) is responsible for and is regulated by SLC2A13, which was a target in AD for Aβ reduction therapy [74, 75]. SLC2A13 also encodes the (H+) myo-inositol co-transporter and is essential in the metabolic regulation of glial cells. Increased mI in PCC could reflect SLC2A13 downregulation, γ-secretase overactivation, and increased Aβ production deposition in PCC in early AD [75].

Last but not least, glutamate is the primary central nervous system (CNS) excitatory transmitter, which plays a significant part in thought, memory, and plasticity [76]. These metabolites are synthesized from glutamine by glutaminase in neurons [77]. The N-methyl-d-aspartate receptor (NMDAR) is usually the most calcium-permeable (Ca2+) receptor and type of glutamate receptor [78]. The receptor can interact between beta oligomers and glutamine, glutamate, or glutamine + glutamate (Glx) metabolites.

Limitations and future directions

Sample sizes were relatively small, and randomized clinical trial (RCT) studies were not included, requiring more significant diagnostic accuracy. The real added value will come from new studies where metabolomics technologies are added to evaluate changes in metabolism in the brain and peripherally spectroscopy and the interconnections between both. Therefore, what has been done so far is limited insights into metabolism using a small number of markers that authors are reviewing. The combination of imaging and genomics has proved to be a powerful pair, but a critical question may come to mind: could imaging be equally valuable for metabolomics? Some researchers think it might, particularly with the adoption of 1H-NMR. Our meta-analysis showed that metabolite changes in the posterior cingulate cortex could be used as a marker for the early detection of Alzheimer’s disease. NAA and ml, and Cho/Cr ratio biomarkers seem to be substantial metabolites for early detection of AD, which can be of interest to researchers. It is recommended to design studies similar to the studies analyzed here but with a more significant number of participants by age group, as well as taking into account the years that have passed since the early diagnosis.

Conclusions

Our meta-analysis showed that microstructural disruptions in the PCC could be used as a marker for early AD detection. Although NAA, mI, Glu, and (NAA, Cho, and mI)/Cr biomarkers are substantial metabolites for diagnosis and are most sensitive for diagnosis. The critical biomarker can be of interest to researchers. Two susceptible areas involved in the pathophysiology of AD early on are the PCC and the hippocampus. Notably, the PCC-epicentered network predicts AD development, including brain atrophy and cognitive impairment, but not the hippocampus-epicentered network. Our findings lend credence to the network degeneration concept of AD and imply that PCC could be employed as possible disease-progression markers. The findings also shed light on the mechanisms behind network disease in AD.