A Systematic Review on the Feasibility of Salivary Biomarkers for Alzheimer’s Disease

Early AD diagnosis is critical for ameliorating prognosis and treatment. The analysis of CSF biomarkers yields accurate results, but it necessitates a lumbar puncture procedure. Screening for peripheral biomarkers in saliva is advantageous since this medium is noninvasive and inexpensive to obtain. The objective of this systematic review is to analyze saliva biomarker studies which aim to diagnose AD. Titles, abstracts, and reference lists for publications from January 2004 to February 2020 were screened for by searching Google Scholar and PubMed. The inclusion criteria involved published studies that consisted of both AD and control groups. 88 studies were screened, and 20 publications fulfilled the inclusion criteria. These selected publications were scrutinized and included in this review. Aβ42, tau, certain metabolites, and oral microbiota might serve as reliable biomarkers for AD diagnosis. These results showcase the legitimate feasibility of proteomic, metabolomic, and microbiotic compounds in saliva for AD diagnostics in the near future. Supplemental studies must consider standardizing the analytical methods of measuring salivary biomarkers to establish coherence for the selection of valid AD biomarkers. Validation studies will require a large sample size of biomarker-diagnosed individuals for independent populations. This ensures accuracy and rigidity for receiver operating characteristic (ROC) curves that can be set for the most optimal salivary biomarkers in future clinical settings.


Background
A lzheimer's Disease (AD) accounts for roughly 70% of all cases pertaining to dementia (1). The characteristic hallmarks of AD include the presence of amyloid plaques and tau tangles, but the direct cause of AD is unclear. These compounds can be present many years before clinical symptoms become visible. The amyloid hypothesis suggests that the accumulation of amyloid aggregates serves as the primary catalyzer for AD pathogenesis, but a multitude of failed clinical trials attempting to "plaque-bust" have posed serious questions concerning the legitimacy of the amyloid hypothesis.
Eli Lilly developed Solanezumab, a monoclonal antibody that was investigated for its potency against soluble Aβ oligomers (2). 400 mg of this drug was given to patients with mild Alzheimer 's every four weeks. 1,057 patients were administered Solanezumab and 1,072 patients were administered a placebo. After undergoing phase III trials, Solanezumab was not significantly effective at slowing cognitive decline. Roche developed Gantenerumab, another monoclonal antibody investigated for its ability to clear Aβ clumps in people with familial AD. Fewer than 1% of the AD demographic consists of individuals with familial AD (3). The potency of Gantenerumab was also assessed along with Solanezumab as a phase 2/3 clinical trial by the Dominantly Inherited Alzheimer Network Trials Unit (DIAN-TU). A total of 194 participants participated in this DIAN-TU clinical trial, of whom 52 participants were administered Gantenerumab. Study subjects participated for an average of five years. After undergoing phase 2/3 trials, Gantenerumab did not prove to be significantly potent against cognitive decline (4).
T h e o b s e r v e d re s u l t s o f S o l a n e z u m a b a n d Gantenerumab in these clinical trials is probable justification for revisiting the authenticity of the amyloid hypothesis, but a more fundamental observation must be made. Treatment may yield no significant effect if it is administered during the symptomatic phase of AD. Therefore, there is a pressing need to develop a simple and noninvasive test which can pinpoint the presence of AD in its presymptomatic phase so that treatment can more successfully mitigate its neurodegenerative effect.
Biomarkers in saliva are being explored as an alternative diagnostic approach. The autonomic nervous system (ANS) regulates the innervation of both CN VII (facial nerve) and CN IX (glossopharyngeal nerve), which triggers saliva secretion by salivary glands. AD compromises ANS function, which may be related to a modified salivary composition for individuals with AD (5). The ability to conduct an early diagnosis in a noninvasive manner remains elusive. Although an array of clinical approaches are implemented to determine the presence of AD, utilizing a biomarker-driven test can aid in establishing an early diagnosis. Viable biomarkers in CSF are used to detect AD, but this method is both costly and invasive.
Significantly different levels of amyloid and tau in blood have been reported for AD subjects in comparison to controls (6)(7)(8). Saliva also contains a plethora of biomarkers, as about 40% of diagnostic blood proteins are also found in saliva (9). It features many advantages for diagnostic purposes compared to other bodily fluids. Saliva is both cheap and easy to obtain, as it can be collected in a noninvasive manner. These advantages facilitate the sampling process of this medium, which is useful for rapid disease screening. The objective of this review is to provide an overview of the literature pertaining to the utility of saliva as a medium for analyzing biomarkers that are specifically associated with AD.

Methods
A standard protocol was implemented for selection of publications in this review. PubMed and Google Scholar were used to conduct a literature search for publications from January 2004 to February 2020. Searches were conducted with the following keywords: Alzheimer, biomarker, dementia, saliva. Titles, abstracts, and reference lists were used to select pertinent publications. Studies included in this systematic review are original publications that analyze potential salivary AD biomarker candidates. Each study consisted of saliva samples from both AD subjects and control subjects. Subject metrics such as age, gender, and sample size for both AD and control groups were considered, along with biomarker type, technique of biomarker quantification, and statistical analysis.

Results
Of the 88 screened studies, 20 were selected, scrutinized, and included in this review. Several studies determined the potential utility of Aβ42 and tau as salivary biomarkers, but other compounds including acetylcholinesterase, metabolites, and oral microbiota were investigated as well.

Acetylcholinesterase
Salivary acetylcholinesterase (AchE) levels were analyzed in two studies. Ellman's colorimetric method was implemented for both of the studies. Bakhtiari et al. tested saliva samples from 15 AD subjects and 15 control subjects. Higher levels of AchE were reported, but statistical significance was not established. Sayer et al. tested saliva samples from 47 volunteers (22 AD cases, 14 AD nonresponder cases, and 11 control cases). They found an overall negative correlation between age and AchE levels, as the r-value was -0.768 (with p<0.001). It was also reported that AD subjects had 73% lower levels of AchE in comparison to control subjects (with p<0.005).

Aβ42
Salivary Aβ42 levels were analyzed in six studies, and all of them utilized an enzyme-linked immunosorbenttype assay (ELISA) with the exception of two. Bermejo-Pareja et al. tested 126 saliva samples from both AD and control cases, in addition to 51 saliva samples from Parkinson's patients. They concluded that salivary Aβ42

Metabolites
AD progression damages the autonomic nervous system, which maintains saliva secretion. Compromising saliva secretion may affect salivary metabolite composition. Levels of various salivary metabolites were analyzed in three studies, which totaled 310 AD subjects, 60 MCI subjects, and 288 healthy control subjects. Huan et al. used liquid chromatography mass spectrometry (LC-MS) to assess the following metabolites: alanylphenylalanine, aminobytyric acid + H2, amino-dihydroxybenzene, choline-cytidine, glucosylgalactosyl-hydroxylysine * (H2O), histidylphenylalanine, m e t h y l g u a n o s i n e , p h e n y l a l a n y l p h e n y l a l a n i n e , phenylalanylproline, and urocanic acid. Their work featured two clinical studies to further confirm their findings. Between AD and control subjects, there was a significant difference for the following metabolites (with p<0.01): choline-cytidine, histidylphenylalanine, m e t h y l g u a n o s i n e , p h e n y l a l a n y l p h e n y l a l a n i n e , phenylalanylproline, and urocanic acid. Between AD and aMCI subjects, there was a significant difference for the following metabolites (with p<0.01): alanylphenylalanine, aminobytyric acid + H2, aminodihydroxybenzene, glucosyl-galactosyl-hydroxylysine* (H2O), and phenylalanylproline. Liang et al.
implemented ultraperformance liquid chromatography mass spectrometry (UPLC-MS) to assess the following metabolites: inosine, ornithine, phenyllactic acid, and spinganine-1-phosphate. They concluded significantly elevated levels of spinganine-1-phosphate and ornithine for AD subjects in comparison to control subjects and significantly lower levels of inosine for AD subjects in comparison to control subjects (with p<0.01). Marksteiner et al. utilized a mass spectrometry kit (AbsoluteIDQ® p150) to analyze endogenous metabolites. ANOVA was used to conduct statistical analysis, along with the Dunnett post hoc test (with p<0.05). Between AD and control subjects, they concluded a significant decrease of the following lipids: PCae C34:1-2, PCae C36:1-2-3, PCae C38:1c3, and PCae C40:2-3. For MCI subjects, there was also a reported decrease in PCae C36:1-2-3. However, among all groups in this study, there were no significant differences for the following lipids: diacylphosphatidylcholines, lyso-acyl-phosphatidylcholines, and sphingomyelins.

Salivary Microbiome
Miklossy observed the presence of spirochetes in 451 of 495 AD brains (10). Her research notes the association between these bacteria and AD. Spirochetes play an etiological role in the manifestation of diseases like Lyme disease and syphilis. Spirochetes are a part of the oral microbiome, and further studies suggest that the oral microbiome is involved in AD pathogenesis. There may be species within the oral microbiome that provide insight into dementia progression.
Species within the oral microbiome were investigated in two studies. Liu et al. assessed the abundance of oral microbiota with 39 AD subjects and 39 control subjects. This study also took into consideration the presence or absence of APOEε4 for each subject. Two techniques were implemented: 16S rRNA sequencing was used to examine oral microbiota and Sanger sequencing was conducted to genotype subjects as either APOEε4(+) or APOEε4(-). None of the bacterial species were found to accelerate AD progression in this study. AD subjects had significantly higher levels of Moraxella, Leptotrichia, and Sphaerochaeta in comparison to control subjects. However, AD subjects had significantly lower levels of Rothia compared to control subjects. APOEε4(+) subjects had significantly lower levels of Actinobacillus and Actinomyces, and APOEε4(-) subjects had significantly higher levels of Abiotrophia and Desulfomicrobium.

Tau
Salivary tau levels were analyzed in three studies. Ashton et al. tested 213 saliva samples from both AD and control cases (53 AD and 160 healthy older controls),  (33) in addition to 68 saliva samples from individuals with aMCI. T-tau levels were analyzed in duplicate using the human total tau assay on the HD-1 SIMOA device. There was not a statistically significant relationship between salivary total-tau levels and age (p=0.190, r=0.080) or gender (female median: 9.6 ng/L, male median: 12.3 ng/L, p=0.872). Increased median t-tau levels in AD patients were observed between healthy older controls, aMCI subjects, and AD subjects (9.6 ng/L, 9.8 ng/L, and 12.2 ng/L respectively), but statistical significance was not established. Shi et al. utilized both mass spectrometry and ELISA to measure salivary tau. Mass spectrometry of whole samples was initially done to detect the presence of tau protein. Luminex ELISA assays were utilized to analyze p-tau and t-tau levels. Mann-Whitney U-tests were applied to assess statistical differences in tau levels between the AD and healthy control groups (21 AD cases and 38 control cases). Control subjects were consenting volunteers who also scored above 27 on the Mini-Mental State Examination (MMSE). In comparison to healthy controls, AD subjects had lower t-tau levels, as well as higher p-tau and p-tau/t-tau levels (p<0.05). However, these differences were determined to be statistically insignificant, and there was minimal difference of total protein levels of tau between the AD and control groups (p<0.05). Pekeles et al. obtained unstimulated saliva in order to analyze the p-tau/t-tau ratio at different phosphorylation sites. Tau-4 antibody measured t-tau levels, and antibodies binding to Thr181, Ser396, and Ser404 were utilized to quantify phosphorylation for sites T181, S396, and S404 respectively. The combined antibody Ser400/Thr403/Ser404 was used for the combined phosphorylation site S400/T403/T404. The Western Blot technique was then implemented to analyze p-tau and t-tau levels in saliva. 337 volunteers participated throughout the two clinical studies conducted by Pekeles et al., including 87 AD subjects and 167 control subjects (of which there were 91 normal elderly control subjects and 76 young normal controls). Their first study included 55 aMCI subjects as well, and their second study included 16 FTD subjects and an additional 12 neurological patients that did not suffer from dementia. Nonparametric tests such as the Shapiro-Wilk test and the Mann-Whitney U test were used to assess the expression of salivary tau at each respective phosphorylation site for the first round of the study. Their findings indicated a significantly higher p-tau/t-tau ratio at the S396 and S404 sites, as well as the combined S400/S404/T404 site for AD patients in comparison to the elderly control individuals. The second round of study (using the two-tailed Kruskal-Wallis statistical test) reported higher median p-tau/ttau levels at site S396 for AD subjects versus those of normal elderly controls. However, Pekeles et al. reported no correlation between elevated salivary tau levels and both CSF tau and hippocampal volume. There was also significant variation for salivary tau levels in AD subjects, which may pose a limitation towards implementation of tau as a legitimate AD biomarker.

Trehalose and Lactoferrin
Both trehalose and lactoferrin levels in saliva were analyzed in two studies. Lau et al. utilized an extended gate ion-sensitive field-effect transistor (EG-ISFET) biosensor to analyze salivary trehalose. Trehalose is a salivary sugar which has shown to alter the metabolism of the Amyloid Precursor Protein (APP), meanwhile reducing the aggregation rate of amyloid (11). 60 saliva samples were tested, including 20 AD subjects, 20 PD subjects, and 20 control subjects. Higher salivary trehalose levels were found in the AD subjects, but statistical significance was not established. Carro et al. used an ELISA to detect salivary lactoferrin levels. It has been observed that pathogenic microbes could contribute to the development of AD (12). The presence and effect of antimicrobial peptides to counteract microbes involved in the pathophysiology of AD remain an underexplored area of research. Lactoferrin is a nonenzymatic antimicrobial peptide which is present in various bodily fluids, including saliva. The objective was to determine if decreased lactoferrin levels could serve as an indicator of AD. 365 individuals participated throughout the two clinical studies conducted by Carro et al., including 116 AD subjects, 59 aMCI subjects, and 131 control subjects. Their first study also included 59 aMCI subjects. Mass spectrometry was implemented to confirm that this protein could be detected in saliva before further experimentation. This study also analyzed salivary lactoferrin levels for aMCI and PD subjects. Carro et al. concluded (with p<0.001) significantly lower levels for both AD and aMCI subjects in comparison with the control subjects, but PD subjects had significantly higher levels in comparison with control subjects. 7.43 μg/mL was the established cutoff value between AD/ MCI subjects and controls in this study.

Supplementary Biomarkers
S u p p l e m e n t a r y b i o m a r k e r s w e re a n a l y z e d in two studies. Manni et al. tested 38 saliva samples from both AD and control cases to assess dim light melatonin onset (DLMO) and salivary melatonin levels in order to determine the circadian phase of subjects experiencing early AD. An in-home melatonin salivary test was implemented for this study. It was concluded (with p=0.028) that DLMO ensued later for AD subjects compared to control subjects. Consequently, melatonin levels in AD subjects were significantly lower for control subjects. Ralbovsky et al. utilized a genetic algorithm coupled with an artificial neural network to conduct Hyper-Raman spectroscopy on samples from 39 volunteers (11 AD, 18 MCI, 10 control). They were able to identify several regions within the Raman spectrum which successfully distinguished between the three groups investigated.

Discussion
This systematic review aims at providing a proper assessment on the literature addressing salivary AD biomarker candidates. The studies observing salivary AchE suggest that it may not serve as a reliable biomarker, despite overall decreased AchE with age (13). Many other biological factors play a role in affecting overall AchE levels in both the brain and saliva, but significantly lower salivary AchE levels might prove to serve as a potential method of determining a compromised cholinergic system (14). Salivary Aβ42 seems to be a reliable biomarker, as five studies in this review analyzing salivary Aβ42 detected significant differences between AD subjects and control subjects. One study reported no significant differences when analyzing other isoforms of Aβ42, including Aβ40 (15). Another study conducted multivariate analysis and noted that although there were differences in Aβ42 concentrations between AD and control subjects, the data was neither statistically significant nor correlated to the severity of AD (16).
The disaccharide trehalose was analyzed as well as a multitude of metabolites. There seems to be a correlation between the expression of trehalose and metabolism of the Amyloid Precursor Protein (APP) (11). There were no significant differences in levels of trehalose, but there were significant differences in levels of various metabolites between AD subjects and controls (17)(18)(19). Research observing the relationship between AD and the salivary microbiome has been recent. Both studies reported different salivary microbiome compositions among AD individuals and controls. One study noted that the microbiotic composition begins to change for those who are classified as cognitively normal but are at-risk for acquiring AD (20). Two studies concluded significantly higher levels of p-tau/t-tau for AD subjects (21,22), but one of these studies reported great variance in their data (22). Statistical insignificance in salivary tau levels was determined in one study (23). It was also reported that salivary tau expression was well characterized at the S396 phosphorylation site (22).
Carro et al.'s results show some validation of lactoferrin as a potential biomarker. Lactoferrin is present in several biological fluids and serves as part of the innate immune system. Some studies have suggested that certain pathogens may play a role in AD by compromising the function of the blood-brain barrier, thus enabling accelerated Aβ42 growth. This may justify the reason for lower lactoferrin levels for individuals with AD, but further studies are needed to confirm this. Manni et al. noted that individuals with mild to moderate AD experienced delayed melatonin secretion, thus explaining significantly lower salivary melatonin levels. The machine learning model incorporated by Ralbovsky et al.
sustained very high accuracy, sensitivity, and specificity averages. It was able to successfully identify a multitude of strong differentiating salivary biomarkers among AD, MCI, and control groups.
Several analysis techniques were implemented throughout these studies, so a standardization by which to investigate salivary biomarkers would provide a more coherent method of selecting future AD biomarkers. Many of these clinical studies featured a small sample size, so a large sample needs to be incorporated for future studies in order to establish reliable reference ranges for biomarker expression levels. Saliva production, circadian rhythms, and oral health are important factors which affect saliva composition. This necessitates further research into how these factors may affect the accuracy of saliva as a medium for AD diagnosis. The precise mechanisms by which these biomarkers become secreted in saliva is not understood. There is still a need to acquire insightful knowledge to explain the presence of these biomarkers in saliva. Advancing the understanding of the pathophysiology of AD requires a thorough comprehension of the association between saliva and AD.

Conclusions
This systematic review intends to determine the feasibility of various salivary biomarkers in order to achieve an early diagnosis of AD. Subject metrics, biomarker type, and methods of biomarker analysis were examined to establish a solid answer on the viability of a saliva test. The reported data indicates that certain salivary compounds may serve as valid AD biomarkers, but a large sample size and a standardization of biomarker analysis techniques must be implemented to further assess the reproducibility of the studies included in this systematic review. More studies featuring large biomarker-diagnosed populations can further validate the use of other potential salivary biomarkers in independent populations. Establishing accurate ROC curves for each respective salivary AD biomarker is necessary to shift towards a biomarker-based AD diagnosis.
Ethics Approval and Consent to Participate: Not applicable.

Consent for Publication: Not applicable.
Availability of Data and Materials: All analyzed and generated data in this study, as well as supplemental information are included in this manuscript.