Background

Alzheimer’s disease (AD) and frontotemporal dementia (FTD) are two of the most prevalent causes of dementia among the elderly. AD is a slow-progressing neurodegenerative disorder that is primarily characterized by memory and cognitive impairments while FTD is a largely heterogeneous disorder that manifests as behavioral changes, language impairment, motor symptoms, or even psychiatric disorders [1,2,3]. The pathological signature of AD is amyloid beta (Aβ) plaque accumulation in the neocortex of the brain, while in FTD, there is degeneration of the frontal and temporal lobes, including in para-limbic areas [4, 5].

Synaptic dysfunction and loss of synaptic plasticity are fundamental processes that underlie several neurodegenerative disorders from early stages. In fact, synaptic pathology has been identified in pre-clinical stages of AD, FTD, dementia with Lewy bodies, and Parkinson’s disease, as well as certain psychiatric disorders [5,6,1–2 and Supplementary Figs. 1–4.

Table 2 Validation results of immunoassays for VAMP2 and SNAP25
Full size table

VAMP2 ELISA

The VAMP2 CSF sandwich ELISA had an LLOQ of 127 pg/mL. The mean intra-assay and inter-assay %CV were 4.1% and 4.8% respectively; thus, the assay is highly precise. The mean dilutional linearity of the three CSF samples was 110%, which was well within the acceptability criteria. Furthermore, mean parallelism of 103% with a 7% standard deviation (SD) was observed, thereby indicating minimal CSF matrix effects in the assay. Lastly, the CSF samples had a percentage recovery within 80 to 120% up to four freeze–thaw cycles, implying that the analyte remains stable with multiple freeze-thaws (Supplementary Fig. 5A).

SNAP25 Simoa

The SNAP25 CSF Simoa assay had an LLOQ of 2.91 pg/mL. The mean intra-assay and inter-assay %CV were 10.2% and 10.9% respectively, both being well within the accepted limit. Three CSF samples were spiked with 200 pg of calibrator protein and serially diluted up to a dilution factor of 98. The mean dilutional linearity of the samples was found to be 106%, which is within the acceptability criteria. The mean parallelism for five CSF samples was 89% which represents minimal matrix effects in the assay. A minor 0.6% negative drift effect was observed for the CSF signals after 3 h 45 min of sample incubation, indicating potential instability of the analyte upon long incubation periods (Supplementary Fig. 5B). This was mitigated by limiting the sample incubation time to 20 min and placing calibrators and run-validation samples at the beginning and end of the immunoassay well plates.

Levels of synaptic proteins SNAP25, VAMP2, and Ng are elevated in CSF of patients with AD and FTD

As expected, the core neuronal biomarkers in CSF: Tau, PTau, and NfL, were able to differentiate significantly between the clinical groups of AD, FTD, and SCD (Supplementary Fig. 6B–D). CSF Aβ42 was significantly decreased in patients with AD compared to FTD and SCD, but could not differentiate between the clinical groups FTD and SCD (Supplementary Fig. 6A).

Patients in both the AD and FTD groups had significantly higher levels of CSF SNAP25, VAMP2, and Ng compared to the SCD group (Fig. 2A–C). Among these, SNAP25 levels alone were significantly elevated in AD compared to FTD. Furthermore, FTD patients had significantly lower NPTX2 levels than the SCD group (Fig. 2D). However, we did not find a differential expression of the synaptic biomarker GluR4 in any clinical group within this cohort (Fig. 2E). All patients in this cohort had CSF SNAP25, VAMP2, Ng, NPTX2, and GluR4 levels above the detection limit of the immunoassays used.

Fig. 2
figure 2

Scatter plot of the synaptic CSF fluid biomarkers in the three diagnostic groups AD (n = 20), FTD (n = 20), and SCD (n = 20). The biomarker values were log-transformed to fit a normal distribution and the model corrected for age and sex prior to conducting pairwise multiple comparisons. A SNAP25, B VAMP2, C Ng, D NPTX2, and E GluR4. *P < 0.05, **P < 0.01, ***P < 0.001

Full size image

Diagnostic potential of the synaptic biomarkers and selection of panels for differential diagnosis

We have visualized the diagnostic value of each synaptic biomarker for differentiating between the clinical group AD vs SCD, FTD vs SCD, and AD vs FTD (Fig. 3A–C; Table 3). SNAP25 (AUC = 0.99) had the greatest diagnostic value for differentiating AD vs SCD patients, followed by Ng (AUC = 0.87) and VAMP2 (AUC = 0.82). No significant AUC was observed for NPTX2 (AUC = 0.62) or GluR4 (AUC = 0.60). The biomarker panel selection consisted of SNAP25 alone (AUC = 0.99). A similar trend was observed for the biomarkers while comparing the clinical groups FTD vs SCD where the synaptic proteins had significant diagnostic values as follows: SNAP25 (AUC = 0.83), Ng (AUC = 0.79), VAMP2 (AUC = 0.77), NPTX2 (AUC = 0.70). The selected synaptic biomarker panel with the highest AUC consisted of VAMP2 and NPTX2 (AUC = 0.97). Interestingly, between the clinical group AD vs FTD, none of the synaptic biomarkers except SNAP25 (AUC = 0.75) showed significant diagnostic value. The identified biomarker panel consisted of SNAP25 and VAMP2 and had a higher AUC than the individual biomarkers (AUC = 0.84).

Fig. 3
figure 3

ROC curves and AUC values of the synaptic CSF biomarkers among the clinical groups. A AD versus SCD. B FTD versus SCD. C AD versus FTD. ROC, receiver operating characteristics, AUC, area under the curve. *P < 0.05, **P < 0.01, ***P < 0.001

Full size image
Table 3 Diagnostic value of the CSF biomarkers
Full size table

Synaptic biomarkers revealed a differential association with classical fluid biomarkers

The correlation matrix of the CSF biomarkers is presented in Fig. 4. In the SCD group, synaptic proteins were all well correlated (Pearson, r = 0.92–0.49), although no significant correlations were found between GluR4 and Ng or NPTX2. However, the strength of these correlations were reduced (Pearson, r = 0.73–0.58) and often not significant in the AD and FTD clinical groups. Furthermore, we did not find a correlation of the axonal biomarker NfL to any synaptic biomarker within this cohort.

Fig. 4
figure 4

Correlation matrix of the CSF biomarkers. Pearson’s correlation was used as all the biomarkers were log-transformed to fit a normal distribution. The significant correlations are indicated in bold. *P < 0.05, **P < 0.01, ***P < 0.001

Full size image

The presynaptic vesicle proteins, SNAP25 and VAMP2, correlated with each other as well as with Tau, PTau, and Ng (Pearson, r = 0.94–0.55) in all three clinical groups. Moreover, the postsynaptic biomarker Ng correlated with SNAP25, VAMP2, and NPTX2 (Pearson, r = 0.73–0.49) in all the clinical groups. In the SCD group, GluR4 had a significant association with SNAP25 and VAMP2 (Pearson, r = 0.60–0.59), although this association was not present in the AD or FTD clinical groups. Interestingly, all CSF biomarkers (except NfL) had a significant correlation with Aβ42 in the SCD group (Pearson, r = 0.73–0.46), although this correlation was lost in the other clinical groups.

Association of the synaptic biomarkers with cognitive scores of patients

The correlation matrix of the CSF biomarkers with the cognitive (MMSE) scores of patients is presented in Fig. 5, and the scatter plots of these correlations are shown in Supplementary Fig. 8. We found that in the overall cohort, the biomarkers SNAP25 (Spearman, r =  − 0.57, P < 0.001), VAMP2 (Spearman, r =  − 0.48, P < 0.001), and Ng (Spearman, r =  − 0.52, P < 0.001) were negatively correlated with the MMSE scores of the patients, indicating that an increase in these synaptic protein concentrations in CSF was associated with decreasing cognition. Upon stratification for clinical diagnosis, NPTX2 in AD patients correlated moderately with their MMSE scores (Spearman, r = 0.55, P = 0.012). No other associations of the synaptic biomarkers with cognitive scores were detected in the stratified cohort.

Fig. 5
figure 5

Correlation matrix of the CSF biomarkers with MMSE scores in the entire cohort and upon stratification into the clinical groups. Spearman’s correlation was used as MMSE scores were not normally distributed even after log transformation. The significant correlations are indicated in bold. *P < 0.05, **P < 0.01, ***P < 0.001

Full size image

Discussion

We measured the concentration of the synaptic proteins SNAP25, VAMP2, Ng, NPTX2, and GluR4 in the CSF of patients with AD, FTD, and SCD. We have shown that both AD and FTD patients have an elevated expression of SNAP25, VAMP2, and Ng compared to SCD, while NPTX2 levels are lower in FTD compared to SCD. Herein, we also reported the development and clinical validation of two novel immunoassays for the SNARE complex proteins SNAP25 and VAMP2. Synaptic proteins involved in diverse functional aspects and localizations were assessed in this cohort as a first attempt to highlight the diversity in synaptic pathology and overlap** pathologies in patients with AD and FTD. Additionally, we observed a differential degree of correlation among the synaptic proteins in AD, FTD, and SCD, which provides proof of concept that synaptic protein levels and their co-expressions are altered in different disease states. The association of each synaptic protein with the classical fluid biomarkers Aβ42, Tau, PTau, and NfL as well as with the cognitive scores of patients was investigated.

To our knowledge, this is the first clinical cohort to report that the CSF synaptic proteins SNAP25 and VAMP2 are elevated in FTD patients and are not specific to AD. In a recent study consisting of neuropathology-confirmed FTD patients, VAMP2 was not found to be significantly increased for FTD compared to controls, although an increasing trend was detected [29]. Although Ng is primarily known to be an AD-specific biomarker, we found that Ng levels, compared to SCD patients, were also significantly elevated in both AD and FTD patients but failed to significantly distinguish these two clinical groups. Similar results have been reported in a previous cohort consisting of AD and FTD patients [30]. The plausible reasons for the lack of differential CSF profiles of synaptic proteins in the two disease groups could be the heterogeneity of FTD as a disease, its co-pathologies with AD and other dementias, or the sheer underrepresentation of FTD in clinical cohorts due to misclassification [10, 31,32,33]. Furthermore, it has been suggested that the implied AD specificity of Ng could be a result of selection bias in cohorts. This is because clinical samples are often stratified based on their Tau levels, and Ng is a biomarker that consistently correlates to high Tau and PTau levels, as we see in our cohort too [34].

NPTX2 reportedly has predictive value for patients with genetic FTD [33, 34]. In this cohort, the NPTX2 levels were significantly lower in FTD patients compared to SCD. However, the FTD patients were not stratified based on their genetic diversity or disease severity, and the lack of differential expression of NPTX2 between AD and FTD could possibly be due to the heterogeneity of these FTD patients. Another notable outcome was that the CSF GluR4 levels were comparable across the clinical groups in this cohort. GluR4 was recently identified as a biomarker that demonstrates preclinical changes in AD patients before neurodegeneration and Tau pathology, although the diagnostic and prognostic value of this AMPA receptor complex protein remains to be evaluated in depth [5, 8, 9, 11]. Herein, we found proof of concept that synaptic pathology is diverse and individual synaptic proteins reflect different aspects of synaptic dysfunction, which is distinct from broad-spectrum neurodegeneration as reflected by changes in Tau, PTau, and NfL. Although limitations of this cohort could arise from the small sample size of each clinical group as well as the heterogeneity within the FTD group, our findings warrant further explorations of synaptic proteins in larger and more diverse clinical cohorts, using immunoassay technologies.

Conclusions

This pilot study indicated that the synaptic biomarkers SNAP25, VAMP2, and Ng were not specific for AD, as they were also elevated in the heterogeneous FTD group compared to SCD. The stronger correlations of the CSF biomarkers in the SCD group compared to the disease groups implied distinct synaptic pathology that may be characteristic of the AD and FTD disease states. Our study also re-iterated recent findings suggesting that CSF synaptic biomarkers reveal neuropathologies that are possibly distinct from broad-spectrum axonal degeneration as exhibited by NfL. In conclusion, these diverse synaptic biomarkers hold promise for understanding the pathophysiological states of different dementias.