Background Frontotemporal dementia (FTD) clinical diagnosis is challenged by the variable correspondence between the clinical syndrome and underlying neuropathological changes, the phenotypic overlap with Alzheimer's disease (AD) and the lack of pathophysiologic diagnostic biomarkers. As synapse degeneration is an early event in pathological frontotemporal lobar degeneration (FTLD), a surrogate marker of synapse loss could aid the early diagnosis of FTD. The aim of this study was to evaluate the diagnostic performance of a panel of 9 synaptic proteins in cerebrospinal fluid (CSF) for FTLD in a neuropathological cohort. Method We included cognitively normal controls (n=35, 69 years+/‐7) and patients with neuropathological confirmation of FTLD (n=49, mean age‐at‐CSF 67 years+/‐10) or AD (n=25, 73 years+/‐6) from the Penn FTD Center. Subsets of the FTLD group included FTLD‐TDP (n=25) and FTLD‐Tau (n=24) neuropathological subtypes and “pure FTLD” (neurofibrillary tangle score of B0/B1 according to the NIA‐AA classification; n=39). We quantified the synaptic panel by targeted mass spectrometry using isotopically‐labeled proteotypic peptides as internal standards. We used linear regression with Dunnett’s post‐hoc tests to compare group differences, receiver‐operating characteristic curves to assess diagnostic accuracy (AUC) and linear regression adjusting for age‐at‐death and sex to assess the association with tau burden. Result Of the 9 synaptic proteins, CSF Calsyntenin‐1 showed the strongest association with disease etiology (r2 =.12, p=.0006). Calsyntenin‐1 was lower in FTLD compared to controls (p=.03) and AD (p=.0008), particularly in “pure FTLD” (p=.02 vs controls, p=.0004 vs AD). Calsyntenin‐1 levels were comparable between AD and controls (p=.33) and between neuropathological FTLD subtypes (p=.66). Calsyntenin‐1 did not correlate with age‐at‐CSF analysis in controls (p=.74), FTLD (p=.07) or AD (p=.40). Calsyntenin‐1 showed the best diagnostic accuracy for differentiating FTLD from AD (AUC=73.4%) and showed similar accuracy in “pure FTLD” (AUC=75.7%) and in FTLD‐TDP (AUC=76.0%) and FTLD‐Tau (AUC=70.4%). Calsyntenin‐1 directly correlated with global pathological tau burden in FTLD (r2=.22; p=0.005) and in “pure FTLD” (r2=.17; p=0.02). Conclusion Low CSF levels of the post‐synaptic modulator, Calsyntenin‐1, is specific to FTLD, particularly in cases with little or no tau pathology and could aid the early diagnosis of FTD and the differential diagnosis from AD and other tauopathies.

BackgroundA key pathological event in frontotemporal lobar degeneration (FTLD) is the alteration of the RNA metabolism. Despite this, no study has characterized the diversity of RNA species using high‐throughput sequencing approaches and correlated them with the main neuropathological hallmarks across FTLD subtypes.MethodTotal and small RNA sequencing was performed in the frontal cortex of patients neuropathologically diagnosed with FTLD‐TDP (including non‐mutation carriers [sFTLD‐TDP;n = 9], and carriers of the C9orf72 repeat expansion [FTLD‐C9;n = 11]), FTLD‐tau (n = 13, six carrying the p.P301L mutation in MAPT) and controls without neuropathological alterations in the same brain region (n = 7). Gene and miRNA co‐expression modules were identified using WGCNA. Cell‐type proportions were estimated through cell‐type deconvolution using MuSiC. Gene ontology enrichment analyses were performed using Metascape. We assessed in the frontal cortex the presence of pTDP43 (in sFTLD‐TDP and FTLD‐C9), dipeptide repeats and RNA foci (in FTLD‐C9), and tau aggregates (in FTLD‐tau) through quantitative immunohistochemistry and correlated their density with transcriptome‐wide RNA alterations.ResultOur results indicate statistically significant correlations between gene and miRNA co‐expression modules, neuropathological changes and cell‐type proportions specific for each FTLD subtype. The most significant findings include: in sFTLD‐TDP, the density of pTDP43 positively correlated with a gene co‐expression module (R = 0.9,p = 0.04) enriched with splicing functions (p<1×10−8), which directly correlated with a miRNA co‐expression module (R = 0.77,p = 5×10−4) and the proportion of a microglial subpopulation (R = 0.69,p = 4×10−3). In FTLD‐C9, the density of poly(GP) repeats inversely correlated with the proportion of a neuronal subpopulation (R = ‐0.62,p = 0.041), and negatively correlated with a gene co‐expression module (R = ‐0.67,p = 0.024) enriched with protein phosphorylation functions (p<1×10−4). In FTLD‐tau, the density of tau aggregates negatively correlated with the proportion of a neuronal subpopulation (Fig.1A;R = ‐0.77,p = 0.003) and positively correlated with a miRNA (Fig.1B;R = 0.88,p = 8.8×10−5) and a gene (Fig.1C;R = 0.65,p = 0.016) co‐expression module enriched in genes with neuron ensheathing functions (Fig.1D;p<1×10−20).ConclusionOur data demonstrate selective vulnerability of cell‐subtypes to neuropathological changes. In addition, we describe striking correlations between the main neuropathological hallmarks of each FTLD subtype and specific gene and miRNA co‐expression modules, including their hub genes and miRNAs which might be used as biomarkers to identify the FTLD neuropathological substrate in vivo, and reveal novel molecular mechanisms amenable for therapeutic intervention in FTLD.