Biomarkers Related to Synaptic Dysfunction to Discriminate Alzheimer’s Disease from Other Neurological Disorders

Recently, the synaptic proteins neurogranin (Ng) and α-synuclein (α-Syn) have attracted scientific interest as potential biomarkers for synaptic dysfunction in neurodegenerative diseases. In this study, we measured the CSF Ng and α-Syn concentrations in patients affected by AD (n = 69), non-AD neurodegenerative disorders (n-AD = 50) and non-degenerative disorders (n-ND, n = 98). The concentrations of CSF Ng and α-Syn were significantly higher in AD than in n-AD and n-ND. Moreover, the Aβ42/Ng and Aβ42/α-Syn ratios showed statistically significant differences between groups and discriminated AD patients from n-AD patients, better than Ng or α-Syn alone. Regression analyses showed an association of higher Ng concentrations with MMSE < 24, pathological Aβ 42/40 ratios, pTau, tTau and the ApoEε4 genotype. Aβ 42/Ng was associated with MMSE < 24, an AD-related FDG-PET pattern, the ApoEε4 genotype, pathological Aβ 42 levels and Aβ 42/40 ratios, pTau, and tTau. Moreover, APO-Eε4 carriers showed higher Ng concentrations than non-carriers. Our results support the idea that the Aβ 42/Ng ratio is a reliable index of synaptic dysfunction/degeneration able to discriminate AD from other neurological conditions.


Introduction
Alzheimer's disease (AD) is a severe neurological disorder, clinically characterized by a progressive loss of memory and cognitive impairment. Typical AD-related hallmarks are the extracellular deposition of plaques with β-amyloid peptides (i.e., Aβ 42) and the presence of intracellular neurofibrillary tangles containing the phosphorylated form of the Tau protein (pTau) [1,2].
To date, the strongest pathogenetic hypothesis has been the so-called amyloid cascade, whose first step is the overexpression of Aβ 42 oligomers and/or the failure of their clearance [3]. Aβ 42 aggregates and precipitates in senile plaques. They can have a toxic effect, resulting in the hyperphosphorylation of the Tau protein and the formation of neurofibrillary tangles. Tau is a microtubule-associated protein with a significant role in maintaining the integrity of neuronal cells, and its hyperphosphorylation causes a gain of function with a toxic effect on brain cells. Both processes (Aβ 42 and pTau deposition) interact with each other, leading to synaptic dysfunction and resulting in neuronal death. not specifically, α-Syn is involved in the pathophysiology of AD. Some authors found high levels of CSF Ng in the early stages of disease, and their correlation with cognitive impairment [44][45][46] and α-Syn seems to interact with Aβ and Tau [47,48]. Moreover, Ng and α-Syn showed a positive correlation with pTau and tTau and negative correlation with Aβ 42 and the Aβ 42/40 ratio [29,49,50].
The apolipoprotein genotype E ε4 (APO E4) is known to be the strongest genetic risk factor for late-onset AD, but also for cognitive impairment in other neurodegenerative diseases, mainly for PD, in a dose-dependent way [51][52][53]. Recently, the interaction between APO E4 and CSF Ng was investigated in a population of AD and MCI patients, and the authors found higher levels of the protein in MCI carrying APOE e4(+) than APOE e4(−), suggesting an effect of APOE in determining the levels of CSF Ng in the early stages of AD [54]. The APOE genotype can also influence pathological changes, progression rates and cognitive impairments in animal models and patients with synucleopathies [53,55].
Recently, the role of the Aβ-42/Ng ratio was studied. It was first introduced to check whether the combination of the two biomarkers improved the differential diagnosis of AD dementia., as well as for the Aβ 42/40 ratio. Janelidzed and colleagues [56] showed that the diagnostic performance of the Aβ-42/Ng ratio did not differ from that of the Aβ 42/40 ratio, when they compared AD to non-AD dementias or cognitively stable MCI from MCI that converted to AD. Subsequently, the Aβ-42/Ng ratio has been demonstrated to be an index of synaptic impairment in PD, as Aβ physiologically regulates synaptic function and Aβ-42 is involved in synaptic toxicity [57]. Although no significant difference was found between the PD and control groups, the Aβ-42/Ng ratio showed a significant correlation with cognitive impairment in PD and is able to discriminate PD patients with cognitive deficits [58].
In the present study, we investigated the role of Ng and α-Syn in AD, as potential biomarkers of synaptic dysfunction. We compared patients affected by AD with patients affected by other neurodegenerative (n-AD) and non-neurodegenerative disorders (n-ND), to better define the role of synaptic dysfunction in their etiopathogenesis and in disease progression, as expressed in terms of cognitive impairment. Moreover, the Aβ-42/Ng ratio and Aβ-42/α-Syn were taken into account as potential indices of synaptic dysfunction in AD pathology.

Participants' Features
We analyzed data from sixty-nine (n = 69) AD patients, fifty (n = 50) n-AD patients, and ninety-eight (n = 98) n-ND patients. As shown in Table 1, the groups did not differ in gender and education, but AD and n-AD differed from n-ND in the age at time of lumbar puncture (LP) and cognitive status (MMSE). The list of n-AD and n-ND diagnoses are reported in Tables S1 and S2, respectively.
ROC analyses (Figure 2c,d) showed that the Ng levels were able to discriminate AD from n-AD patients (AUC = 0.768; C.I. 95% (0.682-0.853); p = 0.004), with a sensitivity of 68% and a specificity of 78%, while the AUC of α-Syn did not reach statistical significance (AUC = 0.689; C.I. 95% (0.514-0.804); p = 0.076).  We next investigated whether CSF Ng and α-Syn improved the differential diagn between AD and n-AD when compared to Aβ 42. Given the fact that the Aβ 42/40 r performs better than Aβ42, we assessed the difference in Aβ42/Ng and Aβ42/α-Syn their ability to discriminate AD from n-AD. We found significantly lower values in than in n-AD (Aβ42/Ng, p < 0.001; Aβ 42/α-Syn = 0.001, respectively), similarly to w We next investigated whether CSF Ng and α-Syn improved the differential diagnosis between AD and n-AD when compared to Aβ 42. Given the fact that the Aβ 42/40 ratio performs better than Aβ42, we assessed the difference in Aβ42/Ng and Aβ42/α-Syn and their ability to discriminate AD from n-AD. We found significantly lower values in AD than in n-AD (Aβ42/Ng, p < 0.001; Aβ 42/α-Syn = 0.001, respectively), similarly to what was seen for Aβ 42/40 (Table 2). Data are expressed as medians with interquartile ranges (IQRs) and were analyzed by Mann-Whitney U tests, also indicating the size effect (η 2 ). p-values were corrected for multiplicity by the Holm-Sidak method and are indicated with bold font.

Biomarkers Related to Synaptic Dysfunctions and AD-Core Biomarkers in AD Subgroups
We also assessed the relationship of CSF proteins and cognitive impairment in AD. With this aim, we stratified AD patients into three groups [59]: MCI due to AD, mild AD and moderate AD.
As shown in Table 4, we found statistically significant differences between MCI and moderate AD, with lower Aβ 42/40 ratios (p = 0.014) and Aβ 42/Ng ratios (p = 0.006), and higher pTau (p = 0.026) in MCI than moderate AD. We found no differences between MCI and mild AD, or between mild and moderate AD.

Correlations of CSF Levels of Synaptic-Related Biomarkers with Clinical Features in AD Group
To investigate the relationship of the CSF Ng and α-Syn levels, and Aβ-42/Ng and Aβ-42/α-Syn ratios with demographic and clinical features in AD, we performed Spearman's correlation analyses. We found a significant positive correlation of the CSF Ng levels with the age at time of LP (rho = 0.250; p = 0.002) and age at onset (rho = 0.187; p = 0.008), and a significant negative relationship with MMSE scores (rho = −0.215; p = 0.037), while CSF α-Syn, although showing a positive correlation with the age at time of LP (rho = 0.280; p < 0.001) and with the age at onset (rho = 0.236; p = 0.003), did not display a significant correlation with MMSE. Moreover, Aβ-42/Ng showed a stronger positive correlation with MMSE (rho = 0.447; p < 0.001) than Aβ-42/α-synuclein (rho = 0.368; p = 0.003) ( Table 5).

Relationship of Synaptic-Related Biomarkers with Clinical Features and AD-Core Biomarkers
To investigate the relationships of synaptic-related biomarkers (i.e., Ng, α-Syn, the Aβ 42/Ng ratio, and the Aβ 42/α-Syn ratio), non-parametric variables were log10-normalized, and then, we performed univariate linear regression analysis to investigate the influence of clinical features and AD-core biomarkers on their values (Table S4).
The predicted value of Ng was higher for patients with MMSE scores less than 24 points (MMSE < 24); for pathological values of the Aβ 42/40 ratio (i.e., <0.066), pTau (i.e., >61 pg/mL), and tTau (i.e., >416 pg/mL); and in the presence of the ApoE ε4 (ApoE4) genotype. We considered the pathological cut-off values for the AD-core biomarkers suggested by the manufacturer. Then, hierarchical linear regression analyses were performed to evaluate how AD-core biomarkers (i.e., Aβ 42/40, pTau and Tau) influenced the variance of synaptic-related biomarkers ( Table 6).  For Ng and α-Syn, considering, in the base model, the age at LP, gender, education and pathological values of the MMSE, we showed that the contribution of the base model explained 0.4% and 29% of the variance in the Ng and α-Syn levels, respectively. The addition of pathological values of the Aβ42/40 ratio explained, respectively, 2.8% and 15.6% of the variance of the Ng and α-Syn levels above and beyond the base model. Considering the Aβ-42/Ng ratio, the base model (Aβ-42/Ng~age at LP + gender + education + pathological MMSE + AD-PET + ApoE4) explained 62% of the variance in its value. Model 1 included a pathological Aβ42/40 ratio as a predictor explaining an additional 17.1% of the variance in the Aβ-42/Ng ratio value above and beyond the base model. The addition of pathological values of pTau explained an additional 3% of the variance above and beyond Model 1. However, the addition of pathological levels of tTau did not contribute to elaborating a model with an increased variance above and beyond Model 2. When we considered Aβ-42/α-Syn, the base model (Aβ-42/α-Syn~age at LP + gender + education + pathological MMSE + AD-PET) explained 49.4% of the variance in its value. In Model 1, the addition of a pathological Aβ42/40 ratio as a predictor explained an additional 22% of the variance above and beyond the base model. Moreover, the addition of pathological levels of pTau and tTau did not contribute to elaborating a model with an increased variance above and beyond Model 1.

Discussion
The purpose of this study was to investigate the role of the synaptic proteins Ng and α-Syn in the CSF as potential additional biomarkers for the diagnosis of AD. We performed a retrospective observational study, in which Ng and α-Syn were analyzed in CSF from patients affected by AD, neurodegenerative disorders (n-AD), and other non-neurodegenerative neurological disorders (n-ND).
According to previous literature [6], we found higher CSF concentrations of Ng and α-Syn in comparison with the two control groups. However, when we considered the ability of Ng and α-Syn to discriminate AD from n-ND by ROC curve analyses, we obtained high sensitivity and low specificity for both proteins. When we analyzed the ROC curves to differentiate AD from n-AD, we observed a fair discriminatory capacity for Ng, with a specificity of 78% and a sensitivity of 68%, while no significant results were obtained for α-Syn. Then, considering previous data reported in the literature [56], we investigated the role of the CSF Aβ42/Ng and Aβ42/α-Syn ratios in improving the differential diagnosis of AD and n-AD, such as Aβ42/40 did. We found significantly lower values of Aβ42/Ng and Aβ42/α-Syn in AD than in n-AD. Moreover, we found that, similarly to Aβ 42, the Aβ 42/Ng and Aβ 42/α-Syn ratios performed better than Ng and α-Syn alone.
When we stratified AD patients into three subgroups according to MMSE scores, we found significant differences between MCI and moderate AD when we evaluated the Aβ 42/40 ratio, pTau and the Aβ 42/Ng ratio. Interestingly, the Aβ 42/Ng ratio was able to discriminate MCI from moderate AD. Although we did not find significant differences in CSF Ng concentrations between MCI and mild AD, as previously reported [35], probably because of the small sample, Aβ42/Ng was found to be higher in MCI than moderate AD, and the same was true for the Aβ 42/40 ratio, while we found no difference for α-Syn and Aβ 42/α-Syn. Our results could reflect the progression of synaptic pathology along the disease.
We also stratified AD patients according to the presence/absence of the ApoE ε4 allele and found higher levels of CSF Ng in ApoE4(+) carriers ( Figure 3). An association between the ApoE genotype and synaptic loss in AD was widely demonstrated [60,61]. However, results regarding the relationship between ApoE and CSF Ng are contrasting [54,56,62]. Fan and colleagues found higher levels of CSF Ng in ApoE 4(+), suggesting a role for the relationship between ApoE and Ng in the pathophysiology of AD [54].
Upon analyzing the relationship of synaptic-related biomarkers with the demographic and clinical features of AD patients, we found that both CSF Ng and α-Syn were positively correlated with the age of the patients at onset, and at the time of diagnosis, Ng correlated negatively with the MMSE, while Aβ 42/Ng and Aβ 42/α-Syn positively correlated with the MMSE. These results confirmed previous findings in which Ng was involved in cognitive deterioration and considered a "cognitive biomarker" [50].
To estimate the contribution of clinical features and AD-core biomarkers in modulating the levels of synaptic-related biomarkers, we firstly performed a univariate regression analysis, separately considering the contribution of Ng, α-Syn, Aβ 42/Ng and Aβ 42/α-Syn. For this purpose, we stratified patients on the basis of pathological MMSE scores (i.e., <24); pathological values of the Aβ 42/40 ratio (i.e., <0.066), pTau (i.e., >61 pg/mL) and tTau (i.e., >416 pg/mL); and the presence/absence of the ApoE4 genotype (both in homo-and in heterozygosis) and AD-related PET pattern. Interestingly, we found that both the Aβ 42/Ng and Aβ 42/α-Syn values are influenced by the presence of the AD-related PET pattern; moreover, the Ng concentration and Aβ 42/Ng value were influenced by the presence of the ApoE4 genotype and pathological values of AD-core biomarkers. Then, we analyzed the influences of different AD-core biomarkers on the levels of synaptic-related biomarkers, elaborating different models of prediction by hierarchical linear regression. Considering a basal model in which the age at LP, education, gender, MMSE, AD-related PET, and presence of ApoE4 were simultaneously considered, we obtained a prediction model in which the serial addition of pathological values of Aβ 42/40 and pTau was able to explain 82.1% of the variance in the Aβ 42/Ng ratio value.
Despite a specific role for Ng in AD pathogenesis still being unclear, its location in the postsynaptic dendritic spines of the hippocampal neurons [63] makes this molecule a potential biomarker of synaptic dysfunction in AD. Indeed, most authors found that the CSF Ng concentration is higher in AD than in healthy controls, but also in MCI due to AD, in preclinical AD and even in elderly people [35,[64][65][66][67]. It positively correlates with hippocampal atrophy [37] and with CSF tTau and pTau, and negatively correlates with cognitive performance, Aβ 42 and Aβ 42/40 [29,35,65]. Ng also seems to predict the progression from MCI to AD [34,37]. However, Ng was found to also be associated with different neurodegenerative conditions such as PD and CJD [35,40], and it is not able to distinguish AD from LBD and FTD [34]. Moreover, in a postmortem study, the Ng in the CSF was not correlated with the Ng load in the brain tissue.
In this study, we wanted to analyze the diagnostic performance of Aβ 42/Ng, which has previously been studied as an index of synaptic dysfunction [28,[56][57][58]. Physiologically, amyloid plays a role in synaptic function, while Aβ 42 is toxic to neurons at the presynaptic level. That is why Sancesario recently tested Aβ42/Ng as an index of synaptic failure in the cognitive dysfunction of PD patients [58], showing that both Ng and Aβ42/Ng correlated with the MMSE and were able to discriminate patients with cognitive decline, with the Aβ42/Ng ratio showing a better performance even when corrected for age and sex. Our study showed that, even in AD patients, Aβ42/Ng performed better than Ng alone in discriminating AD from controls, with higher sensitivity and specificity and a better correlation with the core AD biomarkers and with cognition, measured with MMSE scores. Moreover, Aβ42/Ng showed a slightly better AUC than Aβ42/Aβ40 in differentiating AD from n-AD. Our findings confirm and reinforce previous results obtained by Janelidze, who demonstrated a better diagnostic performance of Aβ42/Ng compared to Ng, but our results show stronger diagnostic value for Aβ42/Ng compared to Aβ42/Aβ40. Therefore, our results support the role of the Aβ 42/Ng ratio as a diagnostic biomarker and lay the groundwork for further studies, which will investigate its role as a prognostic factor.
With a similar reasoning, we also tested Aβ42/α-Syn. α-Syn has already been associated with AD pathology, and aggregates were found in patients affected by AD [29,37,44,45,49,57,68,69], but the ratio has never been tested before. Our results showed that Aβ42/α-Syn performed better than α-Syn alone, but worse than Aβ42/Ng in terms of the correlation with clinical features and diagnostic performance.
Altogether, our findings support the involvement of synaptic dysfunction in AD pathogenesis. Indeed, synaptic dysfunction is considered one of the early events that participates in the etiopathogenesis of AD and may be involved in the rate of the progression of the disease in terms of the impairment of cognitive functions. The mechanisms leading to the synaptic dysfunctions, at both the pre-and postsynaptic levels, include Aβ peptides and Tau proteins, which participate in the incredibly early stages, before the formation of amyloid plaques and neurofibrillary tangles, respectively. In fact, in pathological conditions, Aβ peptides can form soluble oligomers, which accumulate at the presynaptic level, impairing axonal transport, synaptic vesicle cycling, and the release of neurotransmitters. Similar effects are observed in the case of the progressive phosphorylation of Tau proteins [26,27].

Study Population
We present a retrospective observational study, which included sixty-nine patients affected by AD, fifty patients affected by neurodegenerative diseases (n-AD), and ninetyeight patients affected by non-neurodegenerative neurological disorders (n-ND), whose details are reported in Table 1, Tables S1 and S2.
The AD and n-AD patients were attending the Unit of Neurology, Department of Biomedicine, Neuroscience, and Advanced Diagnostic (Bi.N.D.-University of Palermo, Palermo, Italy), while the n-ND patients were attending the ALS Clinical Research Center (Bi.N.D.-University of Palermo, Palermo, Italy). We considered patients from 2018 to 2022.
The AD and n-AD patients underwent complete medical history analysis, clinical examinations, neuropsychological testing, and structural brain (MRI) and metabolic (FDG-PET) neuroimaging. Furthermore, the ApoE genotype was investigated for the AD and n-AD patients.
The AD patients received the diagnosis of "Probable AD dementia with high of evidence of AD pathophysiological process" or "MCI due to AD-High likelihood" according to the published criteria (McKhann et al. [11] and Albert et al. [12]).
All the patients gave their written informed consent, which contained the statement "the biological material may also be used for research purposes", and all the procedures were conducted in accordance with the Declaration of Helsinki and its amendments. The Ethics Committee of the University Hospital of Palermo approved the study protocol.

FDG-PET
The AD and n-AD patients underwent FDG-PET scans during the diagnostic work-up. Brain areas were evaluated on the basis of their metabolic statuses. We define as an "ADrelated PET" pattern the presence of a hypometabolism spatial pattern in areas that are generally involved in AD (i.e., the hippocampus, medial temporal cortex, lateral temporal cortex, superior and inferior parietal gyri, posterior cingulate and precuneus) [71].

CSF Sampling, Processing and Analyses
Lumbar puncture (LP) was performed in fasted patients between 8:00 a.m. and 10:00 a.m. during the diagnostic work-up. CSF samples were collected in polypropylene tubes, centrifuged at 300× g for 5 min to remove cell debris and blood contamination, aliquoted in propylene tubes, and stored at −80 • C until analysis, according to international consensus protocols [72].
The CSF Ng P75 and α-Syn levels were measured using commercially available ELISA kits, according to the manufacturer's instructions [30]. The CSF Aβ42, Aβ40, pTau and tTau levels, which represent AD-core biomarkers, were measured using the chemiluminescence enzyme immunoassay (CLEIA), using commercially available kits [4], according to the manufacturer's instructions. The analyses of AD-core biomarkers were performed only for AD and n-AD patients.
The distribution of the data was assessed with the Shapiro-Wilk method. Normally distributed variables are expressed as the mean ± standard deviation (sd); skewed variables, as median and IQR values; and categorical variables, as absolute and relative frequencies.
For skewed variables, we used the Mann-Whitney U test and Kruskal-Wallis one-way ANOVA on ranks, while the differences in normal variables were analyzed through the one-way ANOVA. In the case of statistical significance, Dunn's post hoc test was used for pairwise comparisons, and p-values were adjusted for multiplicity by the Holm-Sidak method. The differences between groups for categorical variables were estimated using the chi-square test or Fisher's exact test, as appropriate.
The discriminatory ability of the studied biomarkers to correctly allocate the participants to different diagnostic groups was evaluated by receiver operating characteristic curve (ROC) analysis and is reported as the area under the curve (AUC) and 95% confidence interval (C.I.). The best statistical threshold for Ng, α-Syn, the Aβ 42/Ng ratio and the Aβ 42/α-Syn ratio, in the prediction of AD diagnosis, was estimated using the Youden method, selecting the threshold at which the quantity Youden's index = ([sensitivity + specificity] − 1) was maximized. The discriminatory ability of biomarkers on the basis of the AUC was classified as follows: "excellent" (AUC 0.90-1.00), "good" (AUC 0.80-0.89), "fair" (AUC 0.70-0.79), "poor" (AUC 0.60-0.69) or "fail"/no discriminatory capacity (AUC 0.50-0.59). Differences in the area under the ROC curve (AUC) for two ROC curves were compared using the bootstrap method [73].
The statistical correlation between continuous variables was evaluated by Pearson's analyses, while non-parametric data were analyzed using Spearman's rho.
The associations of synaptic-related biomarkers with clinical features and AD-core biomarkers were then specifically assessed by univariate linear regression, as a simple model. For this purpose, non-parametric data were log-10 transformed. Then, we used competitive models leveraging a hierarchical linear regression approach to evaluate the contribution to the variance explained by each significant predictor from the independent analyses. The model selection was aided by Akaike information criterion (AIC) and Bayesian information criterion (BIC) calculations.
A p-value less than 0.05 was considered statistically significant.

Conclusions
Our research strengthens the role of the synaptic proteins Ng and α-Syn as biomarkers of synaptic dysfunction, and highlights the potential of the Aβ42/Ng ratio as a reliable biomarker for the early diagnosis of AD. Synaptic pathology is increasingly considered a central feature in the pathogenesis of AD, and synaptic proteins represent an easily measurable feature in AD. AD biomarkers are very specific and sensitive for early diagnosis, but in the "real world", comorbidities and atypical presentation may limit their interpretation. Larger longitudinal studies are needed to validate the Aβ42/Ng ratio as a biomarker to be added to the panel for AD diagnosis.

Institutional Review Board Statement:
The study was conducted in accordance with the Declaration of Helsinki and approved by the Ethics Committee of University Hospital "Paolo Giaccone" (Institutional Ethic Committee Palermo 1 n • 07/2017).

Informed Consent Statement:
Informed consent was obtained from all the subjects involved in the study. Written informed consent has been obtained from the patients to publish this paper.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author. The data are not publicly available due to current privacy laws.