Abstract
To investigate amyloid-β (Aβ) in frontotemporal dementia (FTD), cerebrospinal fluid (CSF) Aβ38, Aβ40, and Aβ42 in frontotemporal lobar degeneration (FTLD; N = 18 genetically and/or pathologically confirmed and N = 8 FTD with concomitant amyotrophic lateral sclerosis) were compared with Alzheimer’s disease (AD; pathological or Pittsburgh-compound-B Positron-emission-tomography (PIB-PET) positive; N = 25) and controls (N = 24). For all the Aβ subtypes, group difference was seen and post-hoc analysis revealed lower levels in FTLD compared to controls (p≤0.05). Aβ42/40 ratio showed no difference between FTLD and controls; however, a difference was seen between AD versus FTLD (p < 0.01). This is an intriguing finding, suggesting a possible role of Aβ in FTLD pathogenesis.
INTRODUCTION
Frontotemporal lobar degeneration (FTLD) is a pathological spectrum of disorders for which an accurate antemortem diagnosis is essential to establish clinical prognosis and tailoring patient management. A poor correlation has been found between the clinical syndromes of FTLD and the specific underlying pathology, often resulting in clinical misdiagnosis [1]. As a consequence, we need reliable biomarkers for detecting accurately underlying FTLD pathology. Recently, cerebrospinal fluid (CSF) analysis in clinically diagnosed frontotemporal dementia (FTD) patients showed lower CSF Aβ38 and Aβ40 concentrations compared to Alzheimer’s disease (AD) patients and controls, indicating that these biomarkers may be useful to discriminate FTD from AD [2–6]. However, this finding is somewhat unexpected as Aβ pathology is not a hallmark of FTLD. Moreover, these results have not yet been confirmed in cases with postmortem or genetic confirmation. The goals of this study were to compare levels of CSF Aβ38, Aβ40, Aβ42, and Aβ42/40 ratio in a cohort with definite FTLD (based on autopsy and/or genetic examination, or clinical FTD with amyotrophic lateral sclerosis) against AD (either Pittsburgh-compound-B Positron Emission Tomography (PIB-PET) positive or were pathologically confirmed) and controls. Additionally, the diagnostic performance for CSF-subtypes to differentiate FTLD from AD was calculated. Finally, the relation between postmortem Aβ-load with in vivo CSF Aβ measurements was investigated.
METHODS
Patient groups
Patients were recruited from the Amsterdam Dementia Cohort [10]. FTD cases were matched with AD and control cases (age and sex). Twenty-six FTD patients were collected; 10 cases had TAR DNA binding protein 43 (TDP-43) positivity at postmortem examination (FTLD-TDP), 3 cases carried the C9orf72 hexanucleotide repeat expansion (C9orf72), 3 cases carried a tau-gene mutation (MAPT), and 2 cases had a progranulin mutation (GRN). The remaining eight cases were patients with clinical FTD with amyotrophic lateral sclerosis (FTD-ALS; N = 8), which is known to be strongly associated with underlying FTLD-TDP. Four postmortem confirmed FTLD-TDP cases were also genetically confirmed (three C9orf72 and one GRN).
The neuropathological diagnosis for FTLD was set using the Cairns classification [11]. Additionally, all Aβ deposits in FTLD patients were scored (by AR, NV, and JH) according the ABC-score [A = Amyloid, B = Braak, C = CERAD (Consortium to Establish a Registry for Alzheimer’s Disease)] model proposed by the National Institute on Aging–Alzheimer’s Association [12]. The neuropathologists were unaware of the CSF biomarker results. The presence of motor neuron disease had been confirmed by electromyography based on the El Escorial criteria [13].
AD cases were either PIB-PET positive (AD-PET; N = 14) or were pathologically confirmed using the criteria of Braak [14] (AD-PA; N = 11). All cases underwent antemortem lumbar CSF.
The control group consisted of 24 patients who presented with subjective complaints and normal clinical investigations without objective cognitive deficits. Additionally, these subjects had normal CSF Tau/Ptau/Aβ42 measured in previous CSF analysis (using Innotest Aβ42, Innotest htau-Ag, and Innotest p-tau (181P) assays, Fujirebio, Ghent, Belgium). The local ethical review board approved the study protocol and all subjects gave written informed consent.
CSF
CSF was obtained by lumbar puncture between the L3/L4 or L4/L5 intervertebral space, using a 25-gauge needle, and collected in 10-mL polypropylene tubes. Within 2 h, CSF samples were centrifuged at 2100 g for 10 min at 4°C. A small amount of CSF was used for routine analysis. Aliquots of each sample were immediately frozen at –80°C until further analysis. Detection of CSF Aβ38, Aβ40, and Aβ42 was performed using multiplex kits from MSD according to the manufacturer’s instructions (Meso-Scale-Discovery, Rockville, Maryland). All samples were run in duplicate (Duplo) within one batch. The Duplo-coefficient of variation (Duplo-CV) was calculate following this formula: [(SD of two measurements/mean of the two measurements)* 100]. The intra-assay coefficients of variation (Intra-CV) was calculated as the mean of the duplo-CV. For Aβ38, the Intra-CV was 1.8%; for Aβ40, it was 3.7%; and for Aβ42, 1.4%. From 2017 till 2019, this method/assay was used 12 times in our laboratory, resulting in an inter-assay CVs for Aβ38, 4.8%; Aβ40, 5.2%; and Aβ42, 6.0%.
Tau and Ptau were measured with a sandwich enzyme linked immunosorbentassay (ELISA) (Innotest htau-Ag, and Innotest p-tau (181P) assays, Fujirebio, Ghent, Belgium).
Statistics
Data were analysed with the SPSS software package (version 24 for Windows SPSS, Chicago IL). Differences between groups were tested using Chi-squared test and analysis of variance (ANOVA) followed by t-tests. Bonferroni tests were used to adjust for multiple comparisons. For correlations the Spearman test was used for a small sample size (FTLD-TDP N = 10). The Area Under the Curve (AUC) was calculated using ROC-curves (for CSF Aβ42 and CSF Aβ42/40). The AUC of Aβ42 and Aβ42/40 were compared using DeLong test calculated with MedCalc Statistical Software version 19.0.5 (MedCalc Software bvba, Ostend, Belgium; https://www.medcalc.org; 2019). Significance was set at p≤0.05.
RESULTS
Clinical characteristics and CSF biomarker levels for FTLD, AD, and controls are shown in Table 1. The groups did not differ in sex and age.
Table1
Significance was set at p≤0.05. CSF levels mean (SD) in pg/ml. FTLD = 10 FTLD-TDP, 3 C9orf72, 3 MAPT, 2 GRN and 8 FTD-ALS. AD = 14 AD-PET and 11 AD-PA. Post-hoc analyses for Aβ38 and Aβ40: FTD versus controls = p≤0.05. Post-hoc analyses for Aβ42: AD versus Controls, AD versus FTLD, Controls versus FTLD = p < 0.01. Post-hoc analysis Aβ42/40 ratio, Tau and Ptau: AD versus Controls, AD versus FTLD = p < 0.01.
Group differences for all five biomarkers were significant p≤0.05 (see Table 1). Post-hoc analysis revealed lower CSF Aβ38 and Aβ40 levels for FTLD compared with controls (p≤0.05). Additionally, intermediate Aβ42 levels were found in FTLD patients compared with controls (p≤0.05) and AD (p < 0.01). As expected, in AD lower Aβ42 was found when compared with controls (p < 0.01). Aβ42/40 ratio was significant lower in AD compared with FTLD and controls p < 0.01. Tau and PTau were higher in AD compared with FTLD and controls (p < 0.01).
When a subdivision was made of the different pathological forms of FTLD, a decreased level of Aβ subtypes was seen in the FTLD-TDP cases (mean Aβ38 = 2192 pg/ml, mean Aβ40 = 5231 pg/ml, mean Aβ42 = 581 pg/ml; all p < 0.01) compared with controls (mean Aβ38 = 2968 pg/ml, mean Aβ40 = 6610 pg/ml, mean Aβ42 = 793 pg/ml). This was also seen for Aβ38 and Aβ40 in clinical FTD-ALS (mean Aβ38 = 2243 pg/ml, p < 0.01; mean Aβ40 = 5421 pg/ml; p < 0.05). Statistical analysis to compare the different genetic forms of FTLD was not performed because of the small number of patients in each group (N = 3 C9orf72, N = 2 GRN, N = 3 MAPT). In these cases, normal levels of Aβ subtypes were seen (Fig. 1).
CSF Aβ subtypes in different FTLD cases.
The AUC for Aβ42 and Aβ42/40 was 0.911 and 0.972, respectively (AD versus FTD; Fig. 2). No difference was found between the two AUC curves (p = 0.07).
CSF Aβ42/40 ratio and CSF Aβ42 was used for diagnostic purposes to distinguish AD from FTLD. The AUC for Aβ42 and Aβ42/40 was 0.911 and 0.972, respectively. No difference was found between the two AUC curves (p = 0.07).
No correlation (r = –0.14; p = 0.7) was found between Aβ load in FTLD-TDP and the CSF levels of Aβ subtypes (see Supplementary Material).
DISCUSSION
We found decreased antemortem CSF levels of Aβ38, Aβ40, and Aβ42 in genetically and/or pathologically confirmed FTLD and in FTD-ALS compared to controls. To our knowledge this has not been reported before. Earlier studies showed lower CSF Aβ38 and Aβ40 concentrations in clinically diagnosed FTD patients compared to controls [2–6, 15]; however, this was never shown in confirmed FTLD patients. For Aβ42, in clinically defined cohorts, different levels in FTD compared controls were seen. Several publications show lower levels CSF Aβ42 in FTD whereas others show comparable results between FTD and controls [2, 5].
The reason for reduced CSF Aβ subtypes in FTLD/FTD-ALS remains unclear. One possible explanation for this finding is that amyloid-β protein precursor (AβPP)/Aβ processing or Aβ availability is modified in FTLD [16–19]. Aβ is formed after cleavage of AβPP by proteolytic enzymes. Eventually, Aβ will diffuse from brains to CSF [9, 12]. In this study, lower CSF Aβ38, Aβ40, and Aβ42 together with a normal CSF Aβ42/40 ratio is found. These results are indicative of a lower total Aβ in CSF of FTLD. This can possibly be caused by a lower production of Aβ within the AβPP cascade. Or relatively more (soluble) Aβ remains trapped in FTLD brains, this as a result of underlying FTLD pathological processes [16–19].
In brains of FTD-ALS and FTLD-TDP, which are pathologically similar, inclusion bodies formed by full-length and truncated TDP-43, are found [20, 21]. TDP-43 is a member of the heterogeneous nuclear ribonucleoproteins (hnRNPs) family, a group of proteins that bind RNAs. This peptide is polyubiquinated and hyperphosphorylated and primarily localized in the nucleus of cells. In FTD-ALS/FTLD-TDP, TDP-43 accumulates abnormally as ubiquitinated inclusions in glia and degenerating neurons. Several different TDP-43 aggregates are found intracellularly, as well different TDP-43 oligomers. It is believed that these TDP-43 oligomers may play a role in the disease mechanism in FTD-ALS and FTLD-TDP [7]. Additionally, TDP-43 can disseminate in a prion-like manner (cross-seeding) in diseased brains [22], and it has been found in both sporadic and familial AD brains [23–27]. Recently, it has been demonstrated that TDP-43 oligomers are capable of cross-seeding Aβ to form Aβ oligomers. Additionally, TDP-43 increases and stimulates Aβ assembly in brain [28]. In turn this can lead to lower total Aβ in CSF of FTD-ALS and FTLD-TDP.
In this study, the brain Aβ load in ten FTLD-TDP cases was scored and no correlation was found between the amount of Aβ in the brain and the levels of Aβ in CSF. Possibly, the relatively low Aβ-load seen in these small cohort, results in little variance, and therefore a correlation was not detected. However, significant reduction of Aβ burden in postmortem FTLD brains has been seen [8]. This is also been confirmed in PIB-PET studies [9]. In this study, Aβ stainings were performed using monoclonal antibodies (3D6; dependent of conformational properties of Aβ), which detect mostly fibrillar Aβ and no Aβ oligomers or other soluble form of Aβ. Possibly, we underestimated the amount of soluble Aβ in brains of FTLD-TDP patients. It would be interesting to stain brain sections of FTLD-TDP patients using different monoclonal antibodies against different forms of Aβ together with TDP-43 antibody (double/triple stainings). Further, studies and research are required to study a possible link between Aβ and TDP-43 pathology and in which way this can possibly influence Aβ in CSF and brains of FTLD patients.
Besides above mentioned, CSF Aβ42/40 ratio was used for diagnostic purposes to distinguish AD from FTLD. A higher diagnostic performance was reached compared to CSF Aβ42. This indicates that CSF Aβ42/40 ratio is more indicative in dementia differential diagnosis [29].
A limitation of this study was a relatively small inhomogeneous cohort; some FTLD subjects were neuropathologically confirmed, others were confirmed genetically. Additionally, most patients had TDP-pathology, and only brain material was available for FTLD-TDP and not for the other FTLD (genetical) forms. This limits the generalizability of the results in the general FTLD population. Further, FTD-ALS patients, in this study, do not follow the research criteria for FTD, although FTD-ALS has a strong correlation with FTLD-TDP [20, 21]. On the other hand, the strengths of the present study are the well-defined FTLD and AD cohort with clinicopathological information and concomitant antemortem CSF results.
In summary, lower CSF Aβ subtypes were found in confirmed FTLD cases and in FTD-ALS. The reason for reduced Aβ subtypes in these cases remains unclear; we speculate a possible modification of AβPP/Aβ processing (or Aβ availability) in FTLD or/and an interaction between TDP-43 and Aβ pathology in FTLD-TDP. Further research is needed to elucidate the neuropathophysiology of FTLD.
