Blood markers that help tell Alzheimer's disease apart from frontotemporal dementia

Neurodegenerative disorders manifested as dementia are characterized by their pathological hallmarks. Alzheimer’s disease (AD) is defined by the aggregation of amyloid-beta (Aβ) and tau, along with neuronal loss and neuroinflammation, although the diagnosis of probable AD can be supported by clinical criteria [1]. Frontotemporal lobar degenerations are clinically more heterogeneous, with several well-recognized clinical phenotypes, including the behavioral variant (bvFTD), primary progressive aphasias (PPAs), corticobasal degeneration (CBD) and progressive supranuclear palsy (PSP), and with two main pathological groups distinguished by the accumulation of TAR DNA-binding protein 43 (TDP43) or the microtubule-associated protein tau [2].

Despite many efforts to refine the clinical diagnostic criteria, studies have shown that the concordance of clinical and pathological diagnoses is far from perfect in these processes. This is due to their insidious onset and progression, the presence of overlapping and atypical clinical phenotypes, and frequent mixed copathologies [3]. For these reasons, the development of biological markers for the in vivo recognition of underlying neuropathology, either through neuroimaging or fluid analysis, has been a focus of research over the past decades. This goal has become even more important with emerging of potentially disease-modifying therapies for specific diseases. Currently, decreased Aß42 and Aß42/40 ratios along with increased phosphorylated tau-181 (p-tau181) in cerebrospinal fluid (CSF) have been standardized as the core fluid biomarkers of the ATN (amyloid-tau-neurodegeneration) scheme for the biological diagnosis of AD [1], allowing for its distinction from other dementia disorders with similar performance as amyloid PET [4,5,6,7].

More recently, ultrasensitive technologies, such as the single-molecule array (SIMOA) platform, which is able to detect proteins in blood at sub-femtomolar concentrations, have uncovered the potential of plasma biomarkers that are more accessible, less time-consuming, and less costly to use [8,9]. Several reports indicate that AD plasma biomarkers correlate well with CSF biomarkers or PET imaging when discriminating between AD and frontotemporal dementias (FTDs) [10,11,12,13]. Plasma levels of neurofilament light chain (NfL) and glial fibrillary acidic protein (GFAP), as biomarkers of neuronal injury and astrocyte reactivity, respectively, also show different profiles in AD and FTD [12,14]. Furthermore, some studies have found plasma biomarkers useful in predicting dementia risk [15,16]. However, most studies have used samples from selected research cohorts, and further characterization of the biomarkers’ performance in clinical real-world populations is needed before their implementation in clinical settings [17].

In this study, we analyze plasma biomarkers in our clinical cohort of well-characterized dementia cases with three aims: (a) to establish the power of AD core, GFAP and NfL biomarkers to discriminate between underlying AD and FTD; (b) to examine the value of GFAP and NfL in distinguishing FTD cases from controls; and (c) to explore biomarker profiles in different FTD phenotypes, including genetically diagnosed cases. Our data broaden the available knowledge on the usefulness of plasma biomarkers to classify dementia disorders in clinical settings.

The population studied is summarized in Table 1. Figure 1 summarizes the boxplots (median values and quartiles) of the biomarkers for each group, and in Table 2, the pairwise comparisons between groups are shown. The Aß42 levels, Aß42/40 ratio, p-tau181 levels, and GFAP and NfL levels showed significant intergroup differences. T-tau levels and T-tau/Aβ42 ratios were not different between patients with disease and controls.

In our clinical cohort of well-characterized dementia cases with sound diagnosis supported by CSF biomarkers, genetics, and neuroimaging, plasma biomarkers showed distinct patterns for individuals with AD and FTD. Single biomarkers were able to discriminate with high accuracy (AUC > 0.82) between AD and controls (Aß42/40 ratio and GFAP) and between FTD and controls (NfL), but achieved only moderate discriminative power (AUC.72 to 0.78) to distinguish between AD and FTD. To achieve good discrimination between both dementia groups, a combination of two biomarkers was required: NfL combined with p-tau181 (0.87) or the GFAP/NfL ratio (0.86). The best discrimination value (0.90) was obtained by a combination of three biomarkers (NfL and p-tau181, plus either GFAP or Aβ42/40 ratio).

In the AD group, plasma AD core biomarkers reproduced the CSF signature of decreased Aβ42 and Aβ42/40 ratios, along with increased p-tau181. Additionally, a significant increase in GFAP and NfL levels, marked in the former and moderate in the latter, was consistent in AD cases. The FTD profile, on the other hand, was characterized by an outstanding increase in NfL levels, a moderate increase in GFAP concentrations, and similar levels of AD core biomarkers to those of the controls. These AD and FTD patterns are very consistent with previous reports [12,18,19].

In our study, none of the biomarkers individually fulfilled the consensus criteria of the World Federation of Societies of Biological Psychiatry to be considered “valid” (>85% sensitivity and specificity) [4]. However, some individual biomarkers obtained good discriminative power between patients with disease and control cases. First, NfL levels showed good discrimination between FTD cases and controls (0.84). This distinction is highly relevant in clinical settings, as it has been shown to aid in distinguishing primary psychiatric disorders from FTD [20,21,22]. NfL concentrations are considered a marker of nonspecific neuronal injury, independent of amyloid pathology [14]. They increase in secondary neuronal damage, in non-AD degeneration, including FTD, motor neuron disease, and some parkinsonisms, as well as in multiple sclerosis and peripheral neuropathies [14,23,24,25]. In particular, NfL is a strong biomarker for most FTD phenotypes [26,27].

Secondly, the discriminatory power for distinguishing AD patients from controls was high using either the Aß42/40 ratio or GFAP levels. These biomarkers may help to support or rule out the presence of underlying AD in cases with ambiguous memory concerns in clinical settings, before considering further CSF or amyloid PET studies. The Aß42/40 ratio in plasma has been reported as an excellent predictor of β-amyloid status [11] and as the first preclinical change in the AD continuum [28]. GFAP is a marker of astrocyte reactivity associated with neuroinflammation and has also been mostly associated with β-amyloid plaque formation [29]. However, in the oldest-old individuals, GFAP levels are often found to be significantly elevated in the absence of positive AD biomarkers, suggesting that old age itself can be associated with neuroinflammation, independently of Alzheimer’s pathology [30].

On the other hand, none of the individual biomarkers reached good accuracy in differentiating between the dementia groups. This is in agreement with previous studies in real-life clinical cohorts reporting different biomarker patterns according to dementia phenotypes but not good enough performance to support a sound diagnosis [30]. The Aβ42/40 ratio and p-tau181 levels were both highly sensitive (≈92%) but had low specificity (<60%) in distinguishing between AD and FTD. Indeed, the performance of p-tau181 in our cohort (AUC 0.75) was lower than that reported in previous studies (AUCs of 0.88 [31] and 0.96 [12]). Total tau levels were not useful for discriminating between the groups, as their overlap between disease cases and controls was large, as documented in other studies [32]. GFAP and NfL exhibited different patterns of elevation in AD and FTD patients, as has also been reported in previous studies [12,18,19,33], but their discrimination values only reached 0.70 and 0.78, respectively.

Recent studies support that a combination of plasma biomarkers is needed in clinical settings to achieve sufficient discrimination between AD and FTD patients [13,18,34,35,36]. An example in our cohort is that the GFAP/NfL ratio increased the discriminative power to 0.86 (89% sensibility and 79% specificity). The combination of NfL plus p-tau181 also significantly increased the discriminative power (≈0.87). Furthermore, a combination of three biomarkers (NfL with p-tau181, plus either GFAP or the Aβ42/40 ratio) achieved a discriminative power of 0.89–0.90 between AD and FTD. These data agree with the findings from a study of three independent cohorts by Verberk et al. [34], who found an AUC of 0.87–0.89 for the use of a combination of NfL, p-tau181 and GFAP to distinguish AD and FTD.

Finally, we did not find significant differences in biomarker levels when comparing between clinical, genetic and (presumed) pathological FTD variants, except for NfL levels that were higher in PGRN mutation carriers compared with C9orf72 expansion carriers, possibly reflecting more aggressive disease in the former. This pattern in PGRN carriers has also been reported in the analysis of genetic-positive FTD cohorts [37,38,39].

The current information about blood biomarkers in FTD indicates that NfL is the main one and is useful in three approaches: the differentiation of bv-FTD from psychiatric disorders, as mentioned above, the prognosis of the onset of clinical disease in asymptomatic mutation carriers, and the monitoring of disease progression [26,27]. The potential for the in vivo distinction of an underlying neuropathology, either tauopathy or TDP43 accumulation, has been addressed in previous studies, but without consistent and practical findings [26,27,40]. Cousins et al. [41] found that the GFAP/NfL ratio distinguished FTD-tau from FTD-TDP43 with excellent accuracy (AUC = 0.90) in patients with pathology-confirmed FTD. We did not find such differences in this or in any other biomarker, but we acknowledge that the number of cases in our study was small and that it lacked pathological confirmation. Comparing clinical phenotypes, we found p-tau181 levels that were barely higher in the bv-FTD variant compared with Sv-PPA. This difference could have been due to Sv-PPA cases being consistently associated with TDP43 pathology, while both tauopathy and TDP43 accumulation are frequent underlying neuropathologies in bv-FTD. However, p-tau181 levels are not consistently associated with tau pathology in most studies [39].

In summary, we present plasma biomarker data of a dementia series with well-documented clinical phenotypes from a single referral center. We confirm the differing patterns of plasma AD core biomarkers in AD and FTD cases, as well as the opposite pattern of elevation regarding GFAP and NFL levels. Single biomarkers offered good discrimination between patients with disease and control cases, but achieving a distinction between dementia cases required a combination of biomarkers. The biomarker profiles were rather similar across different FTD phenotypes, except for higher levels of neurodegenerative biomarkers in PGRN carriers.

We are grateful to the all the patients who participated in this study. We thank Gema Sánchez and Ana Alvarez for the collection and processing of blood samples; Oliver Shaw for English editing of the manuscript. The Vallecas Project has been funded by the Fundación Reina Sofía.

Conception and design of the study T.d.S., P.S.-J., A.R.-C. and E.G.-T. Acquisition of data P.A.-R., M.J.S., A.R.-G., S.W.-R., A.L.N., R.T. and E.G.-T. Analysis of data I.M., M.J.S., P.A.-R., T.d.S. and E.G.-T. Drafting of a significant portion of the manuscript and figures P.A.-R., T.d.S., I.M. and E.G.-T. All authors have read and agreed to the published version of the manuscript.

This study was conducted in accordance with the Declaration of Helsinki, and approved by the Institutional Ethics Committee of Fundación Jiménez Díaz (PIC194-23_FJD, 11/07/2023).

Informed consent was obtained from all subjects involved in the study.

Raw data are available upon request to the corresponding author.

The authors declare no conflicts of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript; or in the decision to publish the results.

This work was supported by a grant from Fundación hna; Ayuda a Neurociencias de la Fundación Tatiana Pérez de Guzmán el Bueno; and FEDER funds, Spain.