Neurochemical biomarkers in Alzheimer’s disease and related disorders

Abstract

Neurochemical biomarkers for diagnosing dementias are currently under intensive investigation and the field is rapidly expanding. The main protagonists and the best defined among them are cerebrospinal fluid levels of Aβ42, tau and its phosphorylated forms (p-tau). In addition, novel cerebrospinal fluid biomarkers are emerging and their multiparametric assessment seems most promising for increasing the accuracy in neurochemical dementia diagnostics. The combined assessment of Aβ42 and p-tau has recently shown value for diagnosing prodromal states of Alzheimer’s dementia, that is, mild cognitive impairment. Disease-specific biomarkers for other degenerative dementias are still missing, but some progress has recently been made. As lumbar puncture is an additional burden for the patient, blood-based neurochemical biomarkers are definitely warranted and promising new discoveries have been made in this direction. These diagnostic developments have implicit therapeutic consequences and give rise to new requirements for future neurochemical dementia diagnostics.

Introduction

Neurochemical biomarkers for diagnosing dementias principally rely on the soluble correlates of well known neuropathological features that characterize the underlying neurodegenerative diseases. This provides the unique chance to detect and track the disease even though clinical signs may not be observed. The probable diagnosis of neurodegenerative disorders is mainly based on clinical criteria, while definite diagnosis can only be made by neuropathological examination. Misdiagnosis is a frequent problem of clinical dementia diagnostics during the patient’s lifetime; consequently neurochemical biomarkers for diagnosing dementias have gained enormous importance within the last decade.

The studies of biomarkers in Alzheimer’s disease (AD) demonstrate how neurochemical dementia diagnostics (NDD) are successful in the diagnosis of neurodegenerative disorders.

The neuropathological hallmarks of AD are senile plaques and neurofibrillary tangles [Braak and Braak, 1997]. Since their first microscopic description, proteomic analyses from postmortem brain tissue have increasingly clarified the molecular composition of these structures. Aggregated forms of amyloid-β peptides and hyperphosphorylated tau protein form senile plaques and neuro fibrillar tangles (NFTs) respectively [Glenner and Wong, 1984; Kosik et al. 1986].

Three biomarkers for the cerebrospinal fluid (CSF)-based NDD of AD have been established: the tau proteins in general (tau) and especially their phosphorylated forms (p-tau) showed increased concentrations, whereas amyloid-β1-42 (Aβ1-42) was decreased.

The diagnostic value of these biomarkers for AD, especially regarding separation from nondemented controls (NDCs), has been proven in numerous studies with sensitivities and specificities of 80–90% [Blennow, 2004]. Accordingly, the characteristic concentrations of these biomarkers have been added as supportive attributes to the revised diagnostic criteria for AD [Dubois et al. 2007].

The amyloid-β peptides, amyloid precursor proteins and β-amyloid cleaving enzyme

Aggregated Aβ peptides, Aβ42 in particular, form the main part of the extracellular deposited amyloid plaques which are regarded as the central neuropathological attribute of AD [Glenner and Wong, 1984]. Aβ peptides emerge from the enzymatic processing of the β-amyloid precursor proteins (APPs) [Kang et al. 1987] through β and g secretases [Haass and Selkoe, 1993]. One of the possible physiological functions of the APP under discussion is a participation in the cell–cell and matrix interaction. The resulting Aβ peptides vary between 28 and up to 43 peptides amino acid length in their carboxyterminal appearance, while the aminoterminus can be shortened by two to ten amino acids. The aggregative potential of the Aβ peptides positively correlates with the length of their carboxyterminus and negatively with the length of the aminoterminus. A detailed analysis of amyloid-plaques’ composition revealed an extensive rate of aminoterminally shortened Aβ peptides as well as the carboxyterminal endings 40 (x-40) and 42 (x-42). Moreover, the degree of posttranslational oxidation influences the aggregation characteristics of the peptides [Güntert et al. 2006; Bibl et al. 2006a].

Using a special gel-electrophoretic method with subsequent immunoblot (Aβ-sodiumdodecylsulphate-polyacrylgel electrophoresis [SDS-PAGE]/immunoblot) we were able to demonstrate the regular abundance of the peptides Aβ1-37, Aβ1-38 and Aβ1-39 as well as the oxidized form of Aβ1-40 (Aβ1-40^ox) in CSF, beside Aβ1-40 and Aβ1-42 in CSF [Wiltfang et al. 2002; Bibl et al. 2006b]. Aβ1-40 has the highest relevant peptide rate with 60%, followed by Aβ1-38 and Aβ1-42 in CSF. In 1995, Motter and colleagues proved for the first time that a selective decrease of Aβ1-42 can be found in the CSF of patients with AD [Motter et al. 1995]. The overall concentration of CSF Aβ peptides was virtually unchanged, although the Aβ peptides 1-38 and 1-40 showed unaltered or slightly increased concentrations in the CSF of patients with AD [Wiltfang et al. 2002; Bibl et al. 2006b]. The decrease of raw CSF Aβ42 concentrations confirmed the clinical diagnosis of AD (n = 660) among NDCs (n = 541) with a cumulative sensitivity of 86% and a specificity of 89% in 16 independent studies [Blennow, 2004]. According to a newer, autopsy-controlled study, the reliability of CSF-based NDD for AD is at least on a par with the clinical diagnosis. In an early state of the disease, CSF-based NDD is likely to be even superior to the state of the art clinical work up [Engelborghs et al. 2008]. Nonetheless, other neurodegenerative dementias [e.g. dementia with Lewy bodies (DLB), frontotemporal dementias (FTDs) or Creutzfeldt-Jakob disease (CJD)] also show decreased CSF levels of absolute Aβ42, although less pronounced, diminishing the test accuracy of Aβ42 in differential dementia diagnostics [Mollenhauer et al. 2005; Riemenschneider et al. 2002; Wiltfang et al. 2003]. Defining only CSF Aβ1-42 resulted in a maximum sensitivity and specificity of 65% in the separation of AD from other dementias [Hulstaert et al. 1999]. In contrast, the selective decrease of the Aβ1-42 concentration compared with constant Aβ overall concentrations is more specific for AD. Other dementias more often display the decrease of CSF Aβ1-42 levels in the wake of an overall drop of CSF Aβ peptides [Bibl et al. 2006c, 2007b]. Thus, the diagnostic accuracy between AD and other dementias could be clearly improved by a relative reference of Aβ1-42 to Aβ1-40 or Aβ1-38 (ratio Aβ1-42/Aβ1-40 or Aβ1-42/Aβ1-38) [Kanai et al. 1998; Shoij et al. 1998, 2000; Shoij and Kanai 2001; Bibl et al. 2006c, 2010; Welge et al. 2009; Mulugeta et al. 2011]. The accuracy of the test could also be enhanced by scaling Aβ42 as a percentage portion of the sum of all investigated Aβ peptides. This became particularly evident in identifying AD in prodromal states or in distinguishing from dementias other than AD [Bibl et al. 2007b; Wiltfang et al. 2007].

Early and differential diagnosis of AD and its prodromal states [mild cognitive impairment (MCI)-AD] in comparison to dementias and MCI of another kind has also been investigated by measurement of truncated APP forms sAPPα and sAPPβ after cleavage by α- and β-demonstrated secretase respectively. A total of 188 patients from the German Competence Network were analysed by a multiplexing method based on electrochemiluminescence to assess CSF levels of sAPPα and sAPPβ. Diagnosis of AD and MCI-AD was supported by AD-typical CSF biomarker constellations of elevated tau, p-tau and decreased Aβ42. Both sAPPα and sAPPβ displayed a significant increase in AD and MCI-AD, allowing a detection sensitivity of 75% and an exclusion of MCI and other kinds of early dementia with a specificity of 85% [Lewczuk et al. 2010].

Consistent with these data, CSF β-amyloid-cleaving enzyme (BACE) activity has been found to be increased in AD, but was most pronounced in MCI, which converted to AD during a follow up of 3–6 years compared with cognitively healthy controls and stable MCI respectively [Zetterberg et al. 2008]. Similar results have been reported by Zhong and colleagues, who additionally measured the CSF amount of BACE protein, and total Aβ as being markedly increased in MCI [Zhong et al. 2007]. They found positive correlations between these markers and combined them with total tau to a multiparametric diagnostic approach resulting in classification accuracies of up to 83.5% for differentiating MCI versus controls.

A mass spectrometric approach to investigate the APP metabolism in AD revealed six novel CSF-abundant N-terminal APP fragments. By a combined evaluation of five of these fragments the majority of patients with AD (n = 6) could be differentiated from a control group (n = 5) [Portelius et al. 2009]. A subsequent Immuno-MS analysis identified another 11 APP fragments. Thereof, seven were significantly upregulated in AD [Portelius et al. 2010]. Further studies on larger patient cohorts will have to prove the diagnostic meaning of these findings for dementia diagnostics.

Total tau, phospho-tau and kinases

The intraneuronally deposited neurofibrillary tangles in AD primarily consist of aggregated p-tau. Tau essentially builds the microtubular system of neurons [Buee et al. 2000]. The degree of phosphorylation substantially determines the affinity of tau to the microtubular system, whereas an excessive phosphorylation affects the physiological reciprocation of tau with other microtubular proteins [Billingsley and Kincaid, 1997]. The pathological hyperphosphorylation of tau proteins might also be a precondition for their aggregation to double helix filament pairs in AD.

Increased levels of tau proteins in the CSF of patients with AD were initially measured in 1995 and have been replicated numerous times [Blennow, 2004; Jensen et al. 1995; Vigo-Pelfrey et al. 1995]. The diagnostic relevance of this finding in the diagnosis of AD (n = 2022) against NDCs (n = 1005) could be proved in 20 independent studies, yielding cumulative sensitivities and specificities of 81% and 92%, respectively. However, rises of CSF tau levels have also been detected in dementias other than AD [vascular dementia (VAD), Dementia with lewy bodies (DLB), FTD and especially CJD] as well as after acute cerebral ischemias [Blennow, 2004; Hesse et al. 2001; Bibl et al. 2008b]. Thus, it can be assumed that CSF tau reflects the dimension and dynamic of neuronal loss, regardless of the underlying disease, being a sensitive biomarker for neurodestruction, but unspecific for AD. The CSF concentration of tau is presumably defined by the following factors: nearness of the neuronal decay to the ventricle system; the absolute amount of degenerating synapses or neurones; and how many neurones/synapses degenerate per time unit. Hence tau levels in CSF may be especially noteworthy in tracking the disease progress of cortically located neurodestruction using repeated spinal taps. In diagnostic terms, the measurement of CSF total tau is particularly useful for detecting rapidly progressive neurodegenerative processes, such as CJD. Excessively high tau levels (>1300 pg/ml) should prompt the patient’s clinical work up in accordance with the valid consensus criteria, which includes the determination of 14-3-3 proteins in the CSF [Otto et al. 2002]. The appearance of intraneuronal 14-3-3 proteins in CSF indicates an immense neuronal decay, usually found with CJD.

In contrast to total tau, CSF p-tau levels are more closely related to the specific AD pathology. Extensive tau phosphorylation and subsequent fibrillation are acknowledged as pathogenic processes for neurodegeneration caused by AD. Up to now, 30 out of 79 possible phosphorylation sites have been identified [Buee et al. 2000; Billingsley and Kincaid, 1997] and specific enzyme-linked immunosorbent assays (ELISAs) have been developed against some of them (p-tau199, p-tau181, p-tau231 and p-tau 396/ 404) [Itoh et al. 2001; Vanmechelen et al. 2000; Kohnken et al. 2000; Hu et al. 2002]. Concordantly, all these are subjected to hyperphosphorylation in AD and thus all forms of p-tau have been shown to be elevated in AD [Itoh et al. 2001; Vanmechelen et al. 2000; Kohnken et al. 2000; Hu et al. 2002]. The cumulative test accuracy from 16 independent studies for diagnosing AD (n = 1084) among NDCs (n = 504) reached a sensitivity and specificity of 81% and 91% respectively. In contrast to total tau and Aβ42, these values did not markedly decline when p-tau was applied to the differential diagnosis of AD among dementias of another kind. A study of p-tau199 by Itoh and colleagues in 570 patients (236 with AD and 122 with other dementias) showed a sensitivity and specificity of 85% for detecting AD (n = 236) and disclosing other dementive and nondementive disorders (n = 343) [Itoh et al. 2001]. The test performances of ELISA p-tau199, p-tau181 or p-tau231 were equivalent to each other in supporting the differential diagnosis of AD [Hampel et al. 2004]. In another study, the measurement of p-tau 396/404 reached an extraordinarily good accuracy (96% sensitivity and 94% specificity) in the differentiation of AD from VADs and NDCs [Hu et al. 2002]. However, since then, no one has tried to independently retrace these results. Taken together, the discussed results underline the importance of CSF p-tau values for the differentiation of AD from other kinds of dementia, for which elevated p-tau levels remain restricted to singular cases. Vanmechelen and colleagues demonstrated that CSF p-tau levels in patients with corticobasal degeneration largely overlap with those of patients with AD, but patients with DLB, FTD, Parkinson’s disease with dementia, multiple system atrophy or progressive supernuclear palsy showed considerably lower concentrations [Vanmechelen et al. 2001]. Phosphorylation and dephosphorylation are enzymatically regulated by kinases and phosphatases, respectively, the imbalance of which in AD is well documented [Iqbal and Grundke-Iqbal, 2008]. This prompted us to develop an assay for the extracellularly regulated kinases (ERK 1/2) [Klafki et al. 2009]. Similarly to investigating CSF levels of the enzymatic key players (e.g. BACE) that regulate the expression of CSF Aβ patterns, we aimed for a closer look at the CSF levels of tau phosphorylating kinases. Likewise, we discovered a positive correlation between the enzymes ERK 1/2 and their substrate tau and the product p-tau [Klafki et al. 2009]. Interestingly, the measurement of CSF ERK 1/2 in patients with CJD revealed an increase of ERK2 compared with CSF levels of patients with other dementias, including AD and other neurological disorders. The results suggest ERK 1/2 to be a potential biomarker for CJD [Steinacker et al. 2010]. However, further research will have to evaluate the assay for its diagnostic value also in AD.

Early and differential diagnosis of Alzheimer’s disease: multiparametric cerebrospinal fluid-based neurochemical dementia diagnostics

Numerous studies have examined the simultaneous definition of total tau and Aβ42 for the diagnosis of AD. The cumulative accuracy of 12 independent studies for diagnosing AD (n = 767) against NDCs (n = 428) gave a sensitivity of 89% and a specificity of 90% [Blennow, 2004]. The combined measurement of Aβ1-42 and total tau improved the value of each biomarker alone to a diagnostic sensitivity of 85% for AD, but the specificity to exclude other dementias remained unsatisfactory at 58% [Hulstaert et al. 1999]. In 1998, Shoji and colleagues introduced the AD-index combining tau with the ratio of Aβ40 and Aβ42 (tau × Aβ40/ Aβ42), by which the studies Gunma, Tottori, and Tohoku study 1998 (GTT1) (sensitivity 71% and specificity 83%) and Gunma, Tottori, and Tohoku study 2000 (GTT2) (sensitivity 81% and specificity 87%) [Shoij et al. 1998, 2000] could be evaluated successfully. Both Japanese multicentre studies examined the diagnosis of AD from other dementias and various other neuropsychiatric disorders. Two research groups examined the combined measurement of p-tau181 and Aβ42 for differentiating AD from other dementias [Maddalena et al. 2003; De Jong et al. 2006]. The combination of p-tau181 and Aβ42 (ratio p-tau/ Aβ42) was superior to the measurement of each respective biomarker alone and both studies revealed higher specificities than from a combined tau/Aβ42 definition. In a more recent study, the combination of p-tau/ Aβ42 was hardly better than the individual definition of p-tau for differentiating AD from a larger panel of other dementias [Welge et al. 2009]. Otherwise, diagnostic specificity of Aβ1-42 for differentiating AD from other dementias could be clearly improved by measuring Aβ1-42 relative to other Aβ peptide species (e.g. Aβ1-38 or Aβ1-40) or to the sum of all investigated CSF Aβ peptides (i.e. Aβ1-42%) [Wiltfang et al. 2007]. Disease-specific alterations of tau, p-tau and Aβ42 (i.e. increase in tau and p-tau, and decrease in Aβ42) predicted the development of AD in patients with MCI at 85% 4–6 years before the clinical onset of dementia [Hansson et al. 2006]. Also in the diagnosis of prodromal AD in patients with MCI relative measuring of Aβ1-42/Aβ1-38 gave higher accuracies than absolute Aβ1-42 CSF concentrations [Höglund et al. 2008]. Referencing Aβ1-42 to Aβ1-40 avoids false-positive and -negative AD diagnosis in patients with constitutionally low and high CSF Aβ42 levels, respectively. Moreover, dementias with low Aβ42 levels in the course of an overall decrease of CSF Aβ peptides will be sorted out from the diagnosis of AD [Wiltfang et al. 2002; Bibl et al. 2007b]. Nowadays, some neurochemical laboratories already routinely determine CSF Aβ1-42 concentrations as referenced to Aβ1-40 [Wiltfang et al. 2007]. Even more accurate diagnosis may be achieved by combining Aβ peptide ratios with p-tau (e.g. Aβ1-42/Aβ1-40 or Aβ1-42/Aβ1-38) [Welge et al. 2009]. These results impressively show that a CSF-based multiparametric NDD is superior to the definition of the concentration of a single parameter in various appliances. In order to integrate multiparametric NDD in the routine, inexpensive assay formats that are reliable and easily executed are warranted. The so-called multiplex assays (X-MAP technology) can simultaneously measure up to 100 different analytes (theoretical limit) in one single biological sample. Currently, first examinations have reported good correlations for tau, p-tau and Aβ42 as comparatively determined by ELISA and X-MAP technology. Moreover, promising results could be achieved for the early diagnosis of AD in MCI, when these parameters were defined by the X-MAP technology [Herukka et al. 2005; Olsson et al. 2005; Vanderstichele et al. 2005; Hansson et al. 2006; Lewczuk et al. 2008; Reijn et al. 2007]. Despite these promising results in single centre studies, the biomarkers show less accuracy in multicentre approaches, mainly due to larger intercentre variations. In a broad-scale longitudinal study of 750 patients with MCI the combined evaluation of CSF tau, p-tau and Aβ42 for the prediction conversion to AD within a 2-year follow-up period yielded a sensitivity and specificity of 83% and 72%, respectively [Mattsson et al. 2009]. Thus for further biomarker development, the factors for intercentre variations need to be strictly controlled in future studies [Mattsson et al. 2010b].

Nevertheless, the results raise hope for a future use of multiparametric NDD in routine analysis as a highly efficient screening procedure. An extension of the biomarkers analyzed in the multiplex assay to the Aβ peptides 1-40 and 1-38 is reasonable, for example, diagnostically relevant results can be expected from a combination of the ratios Aβ1-42/Aβ1-38 or Aβ1-42/Aβ1-40 with p-tau [Welge et al. 2009]. Additionally, a differentiated analysis of single p-tau epitopes in the CSF will probably be of interest taking the state of the disease into account, since the preferred positions of the hyperphosphorylation at the tau protein apparently change over the course.

The allele variant of apolipoprotein E4 (APO-e4) is currently the major genetic risk factor of AD and its use as a biomarker for AD has been discussed in the literature, but does not qualify as a sole biomarker due to insufficient specificities of 55% [Mayeux, 1998]. Heterozygote and homozygote carriers of the Apo E epsilon 4-allele have a three- to fourfold and six- to eightfold higher risk of developing AD, respectively [Mulder et al. 2000]. The detailed molecular mechanism of this phenomenon has yet to be clarified. However, APO-e4 has a high affinity to bind Aβ1-42 [Strittmatter et al. 1993] and a reciprocal correlation between the occurrence of the ϵ4 allele and the overall concentration of CSF Aβ could be shown [Hulstaert et al. 1999; Galasko et al. 1998; Riemenschneider et al. 2000], indicating that APO-e4 is involved in the Aβ metabolism. Consequently, reference values for the concentration of CSF Aβ peptides should be investigated in relation to the individual APO-e allele status, which may further increase the diagnostic accuracy of Aβ1-42. Exemplarily, the diagnostic sensitivity of the combined definition of CSF tau and Aβ1-42 for AD could be increased from 88% to 100% when the allele state of APO-e was also taken into account [Andreasen et al. 2001]. More recent data suggest a rapid progression from MCI to AD in homozygote carriers of the Apo E epsilon 4-allele with elevated CSF tau levels [Blom et al. 2009].

The multiparametric panel of CSF biomarker candidates for AD diagnosis may be completed by the use of magnetic resonance imaging and amyloid imaging biomarkers to further enhance diagnostic accuracy [Shaw et al. 2011].

Other dementias and cerebrospinal fluid-based neurochemical dementia diagnostics

The development of specific biomarkers for dementias other than AD that are also frequent is still in the early stages. Major differential diagnoses for AD include FTD, DLB and VAD. Studies on the AD biomarkers tau, p-tau and Aβ1-42 revealed a rather unspecific pattern for each of the three dementive disorders [Blennow et al. 2006]. In FTD, tau protein has been found in a normal range [Hulstaert et al. 1999; Sjögren et al. 2000a, 2000b, 2001] or slightly increased [Riemenschneider et al. 2002; Blennow et al. 1995] paralleled by mildly to moderately decreased CSF Aβ1-42 levels [Hulstaert et al. 1999; Riemenschneider et al. 2002; Sjögren et al. 2000a]. Similar results have been reported for DLB [Mollenhauer et al. 2005; Bibl et al. 2006c; Kanemaru et al. 2000; Gomez-Tortosa et al. 2003] or VAD [Hulstaert et al. 1999; Bibl et al. 2008b; Andreasen et al. 2001]. With regard to p-tau, considerably lower concentrations than in AD have been reported for FTD, DLB and VAD, which enables a more accurate test [Hampel et al. 2004; Sjögren et al. 2001; Parnetti et al. 2001; Galasko and Marder, 2002]. However, the described pattern from normal to slightly increased factors for tau and p-tau with mildly decreased Aβ1-42 appears in all of the three disorders and thus cannot be considered as a disease-specific biomarker.

The use of a urea-based SDS-PAGE followed by Western immunoblot (Aβ-SDS-PAGE/Immunoblot) enabled the quantification of six different CSF Aβ-peptides –37/38/39/40/40^ox/42 in the same sample. Consequently, their amount could be expressed in absolute concentration as well as in the ratios (e.g. Aβ1-38/Aβ1-40 or Aβ1-42/Aβ1-40) or as a percental portion of the sum of all examined Aβ peptides (Aβ1-38% or Aβ1-42%) [Wiltfang et al. 2002; Bibl et al. 2007b].

Applied to the three most commonly occurring degenerative dementias AD, DLB and FTD, this approach revealed disease-specific differences between them. The most distinct are a selective decrease of Aβ1-42% in AD and Aβ1-38% in FTD, and an increase in Aβ1-40^ox% in DLB. The sensitivities and specificities were over 85% for the diagnosis of FTD (Aβ1-38%) and AD (Aβ1-42%), with a validation of these potential biomarkers in over 300 patients. The diagnosis of a DLB could be confirmed by means of the increase of Aβ1-40^ox% with a sensitivity of 88% and a specificity of 73%. These accuracies fall within the range of the requirements or come closest to it. The further progress of Aβ-peptide patterns requires validation in independent studies and investigation of neuropathologically confirmed cases. With respect to the latter, we recently showed that Aβ1-40^ox% does not differ among clinically and neuropathologically defined cases of DLB [Mollenhauer et al. 2011a]. Next, the identified biomarker candidates will have to be integrated in conventional assay formats (e.g. ELISA, multiplex platforms). Importantly, we have already demonstrated high correlations between the concentrations of Aβ1-38 and Aβ1-42 as measured by Aβ-SDS-PAGE/immunoblot and electrochemiluminescence or ELISA, respectively. Although integration into ELISA formats still requires optimized preanalytical handling procedures [Bibl et al. 2008a], we could retrace the selective decrease of CSF Aβ1-38 in patients with FTD by the combined use of electrochemiluminescence and ELISA [Bibl et al. 2011]. Disease-specific biomarkers for DLB and FTD, along with improved differential diagnosis, may also support the identification of prodromal states of these dementias.

The pathophysiological backgrounds of disease-specifically altered Aβ-peptide patterns in the CSF in AD, FTD and DLB remain subject to speculation.

In AD, the selective decrease of Aβ1-42 in the CSF is usually explained by the absorption of the peptide out of the CSF into the intracerebral plaques [Motter et al. 1995]. With respect to decreased concentrations of Aβ1-42 in the CSF in diseases that unusually build plaques (e.g. CJD), alternative mechanisms must be considered, such as dysfunctional links to carrier proteins [Wiltfang et al. 2003; Bibl et al. 2004]. The increase of Aβ1-40^ox% in DLB probably mirrors a complex interplay of interactions between α-synuclein and Aβ peptides on the one hand, and increased oxidative stress, on the other. This probably amounts to a form of Aβ toxicity being typical for the disease process in DLB [Bibl et al. 2006b]. The most surprising finding was the decrease of Aβ1-38% in FTD. On the one hand, plaque pathology is a rare feature in FTD and, on the other, Aβ1-38 does not play a significant role in forming AD senile plaques either. However, presenilin mutations have emerged that are associated with the clinical phenotype of FTD, but lack plaque pathology [Dermaut et al. 2004; Zekanowski et al. 2006]. These presenilins are an integral part of the γ-secretase that mutually defines the c-terminal length of the Aβ peptides. Genetic mutations of presenilins are at the same time the major cause of familial AD.

By use of ELISA, two other groups investigated CSF Aβ peptides in FTD and frontotemporal lobar degeneration (FTLD), respectively [Andersen et al. 2000]. One group found a negative correlation of CSF Aβ40 levels and frontal lobe atrophy in FTD. A more recent study found decreased CSF Aβ40 levels in FTLD compared with AD and controls. The authors have additionally related the CSF Aβ40 to CSF Aβ42 levels (Aβ42/Aβ40), which revealed a highly significant increase in the ratio in FTLD compared with both AD and controls [Pijnenburg et al. 2007]. Our own investigations gave similar results using ELISA, but not by Aβ-SDS-PAGE/immunoblot [Bibl et al. 2008a]. The latter method employs denaturizing sample prehandling. This indicates a pool of Aβ1-40 that is not accessible to antibodies during ELISA, but is sensitive to denaturizing. Interestingly, in FTD, approximately 65% of SDS-accessible Aβ1-40 was detected by ELISA too. In contrast, about 75% of the SDS-accessible Aβ1-40 could be detected by ELISA in controls, AD and other dementias [Bibl et al. 2008a]. Taken together, Aβ peptides carboxyterminally shorter than Aβ1-42 seem to play a special role in the pathogenesis of FTD/FTLD, rather than in AD or other dementias. However, sub-specific analyses of the heterogeneous diagnostic groups FTD/FTLD are still warranted and must be addressed by future research. For the group of FTLD bearing ubiquitin-positive inclusion, the TAR DNA-binding protein-43 (TDP-43) has been identified as a central protein for ubiquitin inclusions [Neumann et al. 2006]. Consecutive CSF studies revealed the elevated TDP-43 levels in patients with FTLD along with reduced Aβ42 concentrations compared with controls without dementia. Otherwise sAPPα and sAPPβ levels remained unchanged between the two groups [Steinacker et al. 2008, 2009]. However, due to a considerable overlap of values, CSF TDP-43 concentrations are still in the early stages of biomarker development.

The neuropathological hallmarks of DLB are Lewy bodies that mainly consist of insoluble and aggregated α-synuclein. α-Synuclein has been detected in CSF and a decrease has been recorded in Parkinson’s disease and DLB compared with AD and controls respectively [Mollenhauer et al. 2008, 2011b]. This marker appears promising for exploring its diagnostic potential in larger validation studies, possibly in combination with the aforementioned Aβ1-40^ox%. Moreover, the measurement of CSF tau in combination with serum heart type fatty acid binding protein has been reported as a possible biomarker for DLB, especially for the differentiation of DLB and AD [Steinacker et al. 2004; Mollenhauer et al. 2007].

Blood-based neurochemical dementia diagnostics: a realistic vision?

Lumbar puncture is a relatively low-risk and easily conducted diagnostic procedure. With an incidence of 2–4%, a postpuncture headache is the most common complication, especially in younger patients. The risk of more severe complications, such as internal bleeding or dyspnoea, can be reduced to a minimum by simple precautions (clotting state, cranial computed tomography or magnetic resonance tomography). Its diagnostic benefits for clarifying the cause of a dementia syndrome are discussed above. Nevertheless, the CSF puncture remains an invasive intervention, which is an additional burden for the patients and makes CSF-based NDD less suitable for assessing the course of the disease or its attenuation upon therapeutic interventions. Thus, blood-based biomarkers are desirable for diagnosing and tracking the disease and relevant research is on the way: lipoproteins, homocysteine, oxidative stress markers (e.g. isopostanes, nitrotyrosines, interleukines) and inflammation markers (e.g. C-reactive protein, C1q complementary systems) displayed distinct patterns between patients with dementia and controls, but were all lacking sufficient accuracy for an applicable diagnostic test. The well known CSF-based biomarkers Aβ1-40 and Aβ1-42 have been investigated in plasma, too, but did not show conclusive results in cross-sectional studies [Solfrizzi et al. 2006; Jellinger et al. 2008; Spitzer et al. 2010].

A novel gene/protein called Alzheimer associated protein (ALZAS) has recently been discovered to be expressed within the APP region of chromosome 21 [Jellinger et al. 2008]. This 79-amino-acid long protein is cleaved from APP and contains the Aβ1-42 fragment, the membrane spanning part of APP and a unique 12-amino-acid cytosolic c-terminus, which is not present in any known isoform of APP. Implicitly, it has the ability to compete for APP’s position in the cell membrane by its transmembrane part [Kienzl et al. 2006]; to support the Aβ aggregation by the hydrophobic Aβ fragment; and to induce inflammation and neurodegeneration by its c-terminus [Jellinger et al. 2008]. Pilot studies in serum of patients with probable AD have detected an up to 10-fold increase in the ALZAS antibody titre directed against the carboxyterminus of this protein. Serum ELISA studies have revealed the highest titres in early stages of the disease, that is, in patients with presymptomatic AD or MCI, but moderately increased titres in fully developed and autopsy-confirmed AD [Kienzl et al. 2002, 2006, 2007].

Otherwise, longitudinal studies for estimating the risk of getting AD in asymptomatic test subjects or in patients with MCI succeeded in defining an appointing biomarker pattern in plasma. One study was population based and included 1756 test subjects with an average postexamination time of 8.6 years. Baseline plasma levels of Aβ40 and Aβ42 were determined and people who displayed increased Aβ40 and decreased Aβ42 in the plasma were at higher risk of developing dementia in the further course [Van Oijen et al. 2006]. More recent studies of others revealed similar results for the association of low or decreasing plasma Aβ42 and cognitive decline [Lambert et al. 2009; Lewczuk et al. 2010; Seppälä et al. 2010]. Thus, serial measurement of plasma Aβ42 may be useful in the detection of the subjects who are at risk for cognitive decline [Seppälä et al. 2010].

Another remarkable study in the field of predictive biomarkers for AD investigated the plasma of patients with MCI, who developed AD in the course of 2–6 years and compared it with healthy test subjects. The outcome was that patients with MCI who later converted into AD had a distinctive expression pattern of 18 plasma proteins compared with the test subjects. The respective proteins are mainly involved in haematopoiesis, the immune defence and apoptosis [Ray et al. 2007]. However, these results could not be confirmed using Luminex-based technology [Soares et al. 2009]. Another study using multiplex ELISA technology confirmed five out of 18 of the initially described proteins to be increased in MCI or AD compared with healthy controls. However, test accuracy appeared to be insufficient for individual diagnosis [Marksteiner et al. 2011].

One recent study in the field of blood-based biomarkers for AD found an intriguing model of combined analysis of Aβ peptide levels along with three novel mass spectrometric identified potential biomarker proteins in the cellular blood fraction to distinguish AD from healthy controls. Moreover, the model was able to separate patients with MCI who converted to AD from MCI nonconverters at a reasonable accuracy [Watt et al. 2010].

The differential diagnosis of VAD could be supported by increased concentrations of Aβ1-40 and decreased concentrations of Aβ1-38 (Aβ1-38/ Aβ1-40) in comparison to AD, dementia in Parkinson’s disease and controls with depression [Bibl et al. 2007a]. These results have been shown in a small-scale pilot study and await confirmation in a larger cohort.

The future role of blood-based biomarkers in NDD is probably to serve as a highly sensitive, but not necessarily specific screening test for neurodegenerative processes that, in case of conspicuous results, may prompt further examinations such as CSF-based NDD or molecular imaging techniques.

Neurochemical dementia diagnosis: therapeutic implications

Approved medical treatment options of cognitive deficits in dementias are still limited at present. Hence, the individual benefit and the therapeutic consequences of an improved dementia diagnosis are debatable. Three different acetylcholinesterase inhibitors and the N-methyl-D-aspartate receptor antagonist memantine are approved for the treatment of AD. One of the three acetylcholinesterase inhibitors has the additional approval for dementia within the scope of idiopathic Parkinson’s disease. There is no approved therapy of cognitive deficits in other forms of dementias or MCI to date [Klafki et al. 2006].

This is initially a bleak perspective, although some aspects of the results are certainly useful: first of all, the majority of the patients would prefer an exact diagnosis at least in order to plan and prepare for their future life. Secondly, patients ethically deserve an exact clarification of their state of health and the expected course of the disease. The basic requirement for this is an as precise as possible diagnosis. Thirdly, the transparent discussion of diagnostic considerations is a vital element of the physician–patient relationship. Patients with a dementia diagnosis must consider far-reaching decisions, and it is the physician’s role to accompany them and their families through the disease process. This includes primary healthcare, sociomedical aspects and psychotherapeutic interventions [Brodaty, 2006; Mattsson et al. 2010a]. Fourthly, the differential pharmacological therapy of cognitive and noncognitive behavioural symptoms relies on an accurate diagnosis: for example, risperidone is approved for the treatment of agitation in AD, yet may have fatal consequences due to the increased sensitivity of neuroleptics in DLB. Both diseases are characterized by a severe cholinergic deficit, making treatment with acetylcholinersterase inhibitors reasonable, which is acknowledged to enhance cognitive functions and effectively decrease behavioural problems over the course of the disease. The same rationale may not apply to a dementia syndrome with behavioural alterations caused by FTD. Acetylcholinersterase inhibitors can also considerably amplify the agitation in FTD since cholinergic deficits are lacking here. Taken together, the different elements of dementia therapy should be integrated into a general therapeutic concept, which is individually adapted and appropriate to the diagnosis [Förstl, 2003].

Last, but not least, therapeutic concepts that appoint pathogenic disease mechanisms are already available for patients with AD within the context of clinical studies. Such approaches include active/passive Aβ immunisation, specific g-secretase inhibitors or venous infusion of immunoglobulin concentrates [Klafki et al. 2006]. The accurate diagnostic decision is essential for including patients in promising studies like these.

Future requirements on neurochemical diagnostics of dementia

The rising hope for disease-modifying therapeutic concepts in AD calls for reliable biomarkers of the disease in the near future. In 1998, a group of experts published nine criteria to which a biomarker for AD should ideally conform: (1) a central aspect of the molecular pathology of dementia should be reflected; (2) it should be validated in neuropathologically confirmed cases and detect a fundamental feature of AD pathology; (3) it should be precise, (4) reliable, (5) easy to accomplish, (6) inexpensive and (7) noninvasive. The test should detect AD with a sensitivity of (8) 85% at minimum and provide a reasonable specificity for exclusion of other neurodegenerative disorders commonly understood as a figure of (9) 75–85 % [The Ronald and Nancy Reagan Research Institute of the Alzheimer’s Association and the National Institute on Aging Working Group, 1998].

Except for the invasiveness of lumbar puncture, CSF-based NDD may fulfil these requirements for AD, but not for other dementias. In blood-based NDD, which would be far less invasive, the validation of promising results is still missing. In particular, neuropathological confirmation of clinical diagnoses is needed. At present, we can only speculate on the test precision and reliability of blood-based biomarkers. However, biomarkers for a prognostic assessment, process evaluation and the evaluation of therapeutic response are not currently available, either in CSF or in blood.

Two items are exceptionally important in order to advance biomarker-based diagnosis. First, technologically highly developed proteomic methods for the measurement of new biomarkers are often too expensive as well as too time consuming and require specially trained personnel. A transmission to easily manageable and highly effective assay forms allowing a routine measurement of biomarkers is necessary. The development of multiplex assays (X-MAP technology) is a fundamental step towards it and the results of initial studies are promising. Second, biological samples should ideally be acquired, processed and stored according to a standardised protocol in future databases for biomaterial. Databases should also acquire complete clinical and paraclinical data for making an exact clinical diagnosis according to the valid consensus criteria. All diagnoses should be set up by a committee of experts. In all cases, a neuropathalogical confirmation of the diagnosis should be aimed for. Highly innovative techniques and easily manageable methods of routine diagnostics should be run comparatively to validate results. Measurements should undergo regular quality surveys. Reliable reference values of identified biomarkers should be established. Thus, biomaterial databases should ideally comprise samples from patients with dementias, neurodegenerative and other differential diagnostically relevant diseases, as well as from healthy volunteers. These reference values will have to be adapted to potential confounders, like age, sex, fasting state, diet, circadian rhythms.

Standardised maintenance of biomaterial databases is the basis for comparable analysis of biomarkers in neurodegenerative diseases and dementias by independent investigators [Bibl and Wiltfang, 2008].