Cerebrospinal Fluid Biomarkers as a Diagnostic Tool of the Underlying Pathology of Primary Progressive Aphasia

Abstract

Background: Primary progressive aphasia (PPA) may present with three main clinical variants, namely nonfluent agrammatic (nfaPPA), semantic (sPPA), and logopenic (lPPA) subtypes. Frontotemporal lobar degenerations (FTLD) or Alzheimer’s disease (AD) are the most common etiologies.

Objective: To study the potential of cerebrospinal fluid (CSF) biomarkers for identifying the underlying pathology in patients with PPA.

Methods: CSF levels of total tau protein (τ_T), amyloid-β peptide (Aβ₄₂), and tau phosphorylated at threonine-181 (τ_P - 181) were measured by double sandwich, enzyme-linked immunosorbent assay (ELISA) in 43 patients with PPA, 26 patients with AD, and 17 healthy controls.

Results: All patients could be classified as compatible with the AD or non-AD biomarker profile, either with the three biomarkers (90.7%) or their ratios, especially the τ_P - 181/Aβ₄₂ ratio (9.3%). An AD-compatible biomarker profile was present in 39.5% of all PPA patients, specifically 22.2%, 35.7%, and 75% of nfaPPA, sPPA, and lPPA, respectively. In PPA patients with a non-AD profile (presumably FTLD), two different clusters could be identified according to the τ_P - 181/τ_T ratio, possibly corresponding to the two major FTLD pathologies (tau and TDP-43).

Conclusion: CSF biomarkers may be a valuable tool for the discrimination between PPA patients with AD and non-AD pathophysiology and possibly between FTLD patients with tau and TDP-43 pathology.

Keywords

Alzheimer’s disease biomarkers cerebrospinal fluid frontotemporal lobar degeneration phospho-tau primary progressive aphasia tau

INTRODUCTION

Primary progressive aphasia (PPA) is a syndrome characterized by a relatively selective language dysfunction in the absence of secondary pathology such as stroke and tumor [1]. It is due to a neurodegenerative process, typically frontotemporal lobar degeneration (FTLD) or Alzheimer’s disease (AD) [2, 3]. Three clinical subtypes or variants have been described, namely nonfluent/agrammatic (nfaPPA), semantic (sPPA), and logopenic (lPPA) [4], while unclassifiable forms also exist [5]. Pathological data suggest that, to some extent, each subtype tends to be statistically associated with a particular type of histopathology. Thus, in nfaPPA, FTLD of the tauopathy type is the most common pathology; in sPPA, FTLD of the TDP-43 type is the most common pathology, and in lPPA, AD is the most common pathology [2, 3]. However, many exceptions exist, and many different pathologies have been described in all PPA variants [3].

Existing and especially emerging potential treatments require correct diagnosis during life and necessitate the use of biomarkers. The established CSF biomarkers for AD, namely tau protein in its total and hyperphosphorylated form and amyloid-β peptide have been incorporated in the recent research diagnostic criteria established by the National Institute on Aging and the Alzheimer’s Association, as a tool in order to increase the diagnostic confidence in establishing the presence (or absence) of AD pathophysiological process [6]. Thus, in patients presenting as PPA, they could be of help in discriminating between FTLD and AD and, possibly, between FTLD with tau or TDP-43 pathology [7, 8]. The aim of the present study was to investigate the potential role of CSF biomarkers in every day clinical approach of PPA patients, by discriminating between AD and non-AD pathophysiology.

MATERIALS AND METHODS

Patients

A total of 102 patients, hospitalized in our clinic (tertiary referral center), consecutively during the years 2012–2014 for progressive language disorder (n = 54) or AD (n = 48) were assessed. Of the former, 11 patients did not fulfill the criteria for PPA [4] and were excluded. Of the latter, 22 were excluded due to a significant vascular component in neuroimaging (n = 19), ventricular enlargement (n = 1), or contraindication for lumbar puncture (oral anticoagulation, n = 1). All subjects underwent a complete physical and neurological examination and a neuropsychological battery, testing global cognitive function, memory, frontal executive and visuoconstructive abilities, praxis, activities of daily living, depression, behavioral symptoms, and language, including the Boston Diagnostic Aphasia Examination [9]. For language, the following parameters were assessed clinically and neuropsychologically: articulation, phonological errors, paraphasias, aggramatism, phonemic and categorical fluency, sentence repetition, single word and complex sentence comprehension, spontaneous and confrontation naming, object knowledge, reading, and writing. All patients had magnetic resonance imaging (3T MRI) with additional visual assessment of four areas: left fronto-insular, anterior temporal, left parietal or posterior perisylvian, and medial temporal (hippocampal) atrophy.

Finally, 86 subjects (69 patients and 17 controls) were included in the study and subdivided in well characterized groups:

43 patients fulfilling the criteria for PPA [4]. They were further subdivided in nfaPPA, sPPA, and lPPA according to the criteria of Gorno-Tempini et al. for imaging-supported variants [4]. Thus, for inclusion in the nfaPPA subgroup, patients should have either agrammatism or apraxia of speech, in addition to at least two of the following three: spared object knowledge, spared single word comprehension, but impaired comprehension of complex sentences, plus predominant left posterior fronto-insular atrophy in MRI. For inclusion in the sPPA subgroup, impairment in both confrontation naming and single-word comprehension was required in addition to at least three of the following four: impaired object knowledge, surface dyslexia or dysgraphia, spared repetition and spared grammatic/motor aspects of speech, plus predominant anterior temporal atrophy. For inclusion in the lPPA, impairment in both single-word retrieval and sentence repetition was required, in addition to at least three of the following four: phonologic errors, absence of agrammatism, spared single word comprehension/object knowledge and spared motor speech, plus predominant left posterior perisylvian or parietal atrophy. Patients with PPA not fulfilling the criteria of the above three classical types were considered unclassified (uPPA).

26 patients with AD, fulfilling the criteria for probable AD with documented decline and, additionally, evidence of the AD pathophysiological process, as suggested by the National Institute on Aging and Alzheimer’s Association workgroups [6]. Since this is a CSF biomarker research study, evidence of the AD pathophysiological process according to the used criteria was limited to imaging data (magnetic resonance imaging showing significant medial temporal atrophy) and neuropsychological data (typical amnestic profile).

17 healthy controls (CTRL) with no cognitive complaints and no neurologic, psychiatric or other major disease, undergoing minor surgery such as hernia repair or knee joint surgery under spinal anesthesia. They all had normal cognitive function as suggested by history, semi-structured interview, and within-normal-limits scores on neuropsychological testing prior to operation.

Assignment of subjects to the diagnostic groups had been performed prior to CSF analysis in order to prevent circular errors.

The study was in accordance with the ethical guidelines of the Declaration of Helsinki (1975) and had the approval of the local committee of our hospital (Scientific and Ethical Board of Eginition Hospital). All subjects and/or relatives gave informed consent for inclusion in the study.

CSF sampling and biomarker determination

Lumbar puncture was performed at 10–11 am, after overnight fasting, at the L4-L5 interspace, according to recently proposed recommendations on standardized operating procedures for CSF biomarkers [10]. In brief, four polypropylene tubes were used for CSF collection. The initial tube (2 ml) was used for routine cytology and biochemistry. The next tube (2 ml) was used for syphilis serology, or any other determination suggested by the clinical presentation. The last two tubes (5 ml each) were immediately centrifuged, aliquoted in polypropylene tubes (750 μl each), and, finally, stored at –80°C until analysis. Samples with more than 500 red blood cells/μl were discarded. Aliquots were thawed only once, just before analysis, which was performed within the 1st year of storage in all subjects.

The CSF levels of total tau protein (τ_T), amyloid-β peptide (1–42) (Aβ₄₂), and tau phosphorylated at threonine-181 (τ_P - 181) were measured in duplicate and blind to clinical diagnosis by double sandwich, enzyme-linked immunosorbent assay (ELISA) as provided by commercially available kits (“InnotesthTau antigen”, “β-amyloid_1–42” and “phospho-tau₁₈₁” respectively, Fujirebio, Gent, Belgium) according to manufacturer’s instructions.

Statistical analysis

All variables were checked for normality and homogeneity of variances (Shapiro-Wilk’s and Leven’s tests, respectively). The levels of CSF biomarker and their various ratios did not follow the normal distribution and their variances were heterogeneous. Logarithmic transformation restored the above violations and permitted the use of parametric tests.

The levels of the three CSF biomarkers were compared among groups by two-way multiple analysis of covariance (2-way MANCOVA), followed by univariate tests for each of the biomarkers, with diagnostic group and sex as co-factors and age as covariate. Two-way analysis of covariance (2-way ANCOVA) was also used for the ratios of the three biomarkers and p values were Bonferroni-corrected for multiple tests. Neumann-Keuls post-hoc tests were performed for individual comparisons of variables among groups.

Nonparametrics, including Kruskal-Wallis test, χ² test, χ² test for trend and k-means clustering were also used as appropriate.

Receiver operating characteristics curve analysis (ROC) was performed in order to calculate the cut-off values of each biomarker and ratio for the discrimination between AD and CTRL, with the optimal combination of sensitivity and specificity. Discriminant analysis was also performed in order to calculate the percentages of correct classification and the posterior probabilities of AD presence or absence, assuming a prior probability equal for AD and CTRL or non-AD PPA (50%). For PPA patients, the CSF profile was considered compatible with AD (“AD profile”) when all three biomarkers were abnormal. In patients with inconclusive profile, the ratios of biomarkers were used.

RESULTS

Demographic and CSF levels of biomarkers and their ratios are summarized in Table 1 and Fig. 1. Follow up data (at least 2 years) is also available in Supplementary Table 1.

Thirty one PPA patients (72%) could be classified into the three classical variants of PPA. The remaining 12 (28%) were unclassifiable. Three of them had mixed features of lPPA and sPPA. Two were pure anomic, three had mixed features of lPPA and nfaPPA, and four had mixed features of nfaPPA and sPPA.

No significant differences in age, sex, and disease duration was observed among groups.

CSF biomarker levels

Two-way MANCOVA for τ_T, Aβ₄₂, and τ_P - 181 with diagnostic group and sex as cofactors and age as covariate revealed a significant effect by diagnostic group (p = 0.00019) and by sex (p = 0.033). Univariate tests revealed significant effects by diagnostic group for all three biomarkers: τ_T (F = 5.87, p = 0.0007), Aβ₄₂ (F = 5.7, p = 0.0008), and τ_P - 181 (F = 2.75, p = 0.039). Univariate tests for the three ratios also revealed significant effects by group for all three ratios: τ_T/Aβ₄₂ (F = 7.5, p = 0.0001), τ_P - 181/Aβ₄₂ (F = 4.7, p = 0.0029), and τ_P - 181/τ_T (F = 7.5, p = 0.0001).

A significant effect by sex was observed only for τ_T, with females presenting with higher levels of τ_T (p = 0.015), higher ratio of τ_T/Aβ₄₂ (p = 0.027), and lower ratio of τ_P - 181/τ_T (p = 0.0028) than males. Age did not affect the models significantly.

As expected, post-hoc Newmann-Keuls tests revealed that the AD group presented with significantly higher τ_T and τ_P - 181 levels and lower Aβ₄₂ levels, as compared to CTRL (p = 0.00017, 0.011, and 0.00022, respectively). They also presented with higher τ_T/Aβ₄₂ and τ_P - 181/Aβ₄₂ and lower τ_P - 181/τ_T ratios as compared to CTRL (p = 0.00013, 0.00025, and 0.00016, respectively).

By visual inspection of Fig. 1, patients with nfaPPA were similar to the CTRL group and patients with lPPA were comparable to AD, with sPPA forming an intermediate group. Indeed, post-hoc tests showed no significant difference between CTRL and nfaPPA and between the lPPA and AD groups for all three biomarkers and the τ_T/Aβ₄₂ and τ_P - 181/Aβ₄₂ ratios. The τ_P - 181/τ_T ratio showed comparable values in AD and lPPA; however, it was significantly lower in nfaPPA as compared to CTRL (p = 0.0075).

Patients with lPPA had higher τ_T and τ_P - 181 levels and lower Aβ₄₂ levels as compared to CTRL (p = 0.00015, 0.011, and 0.05, respectively) and nfaPPA (p = 0.0034, 0.044, and 0.016, respectively). They also showed higher τ_T/Aβ₄₂ and τ_P - 181/Aβ₄₂ and lower τ_P - 181/τ_T ratios as compared to CTRL (p = 0.00018, 0.0023, and 0.00013, respectively) and nfaPPA (p = 0.00035, 0.005, and 0.039, respectively).

The AD group showed higher τ_T and τ_P - 181 levels, higher τ_T/Aβ₄₂ and τ_P - 181/Aβ₄₂ ratios, and lower Aβ₄₂ levels and τ_P - 181/τ_T ratio as compared to nfaPPA (p = 0.004, 0.05, 0.00018, 0.00038, 0.00013, and 0.029, respectively).

Patients with sPPA showed higher τ_T levels, higher τ_T/Aβ₄₂ and τ_P - 181/Aβ₄₂ ratios, and a lower τ_P - 181/τ_T ratio as compared to CTRL (p = 0.0038, 0.0037, 0.031, and 0.0005, respectively). Additionally, they showed higher Aβ₄₂ levels, a lower τ_T/Aβ₄₂ ratio, and a tendency toward lower τ_T levels as compared to the AD group (p = 0.017, 0.04, and 0.089 respectively).

Discrimination between AD and CTRL or non-AD PPA

There were no major differences between the results of ROC and discriminant analysis in the discrimination between AD and CTRL, except for Aβ₄₂ (Table 2). Posterior probabilities for AD presence, as suggested by discriminant analysis were plotted against biomarker levels and, as expected, plots were sigmoidal in shape (Fig. 2). After combined inspection of Table 2 and Fig. 2, interesting observations can be made. For τ_T, the two posterior probability plots (AD versus CTRL and AD versus non-AD) are relatively different. The cut-off value versus CTRL suggested by discriminant analysis is 376 pg/ml (the same is suggested by ROC), but the cut-off versus non-AD PPA is higher (461 pg/ml), since abnormal τ_T levels may be observed in non-AD pathologies as well. Thus the 376–461 pg/ml zone is inconclusive. For Aβ₄₂, posterior probability plots are almost identical and the cut-off of discriminant analysis for both versus CTRL and versus non-AD PPA is ∼580 pg/ml. However, the cut-off suggested by ROC is 100 pg/ml higher, suggesting that the 580–682 pg/ml zone may be inconclusive. For τ_P - 181, the two curves seem almost identical with the cut-off suggested by discriminant analysis being at ∼62.5 pg/ml. The cut-off suggested by ROC is ∼6 pg/ml lower, a difference of lower magnitude than the one for Aβ₄₂. Whatever the cut-off level or the posterior probability, it is a theoretical value. For example, a posterior probability for AD at about 60% although above the cut-off, may be considered as marginal, since there is still a 40% chance for the absence of AD. Thus, the inconclusive zones may have significant extensions to either side, occupying a significant part of the linear, steeply rising section of the sigmoid curve. It is reasonable to conclude that biomarker concentrations may be more diagnostic when they correspond to the two non-linear sections of the sigmoid curve, i.e., approaching or entering the plateaus with posterior probability for AD presence >75–80% or <20–25%. For individual biomarker concentrations, many of the non-AD PPA cases (squares) fall within these “gray zones”. However, the τ_T/Aβ₄₂ and especially the τ_P - 181/Aβ₄₂ ratios placed most of the subjects at the “plateau sections” of the curves, leaving only occasional cases into the “gray zones”.

CSF biomarker profiles

Based on cut-offs for τ_T, Aβ₄₂, and τ_P - 181 shown in Table 2 (AD versus CTRL), almost all PPA patients (90.7%) could be easily classified as compatible with the AD or non-AD biomarker profile. However, one of the nfaPPA, one of the lPPA, and two of the sPPA subgroups (total of four PPA patients, 9.3%) were considered borderline or inconclusive, either because the biomarker levels were very close to the cut-off values, or because of conflicting results. In all four patients, the ratios were helpful and permitted characterization of the profile (Supplementary Table 2). Finally, 17 of the 43 PPA patients (39.5%) were compatible with the AD profile. When comparing the proportions of patients with “AD profile” among PPA variants, a statistical significant trend was noted, with an increasing proportion of patients with “AD profile” as we move from nfaPPA to sPPA and finally to lPPA (χ² test for trend p = 0.0297) (Fig. 3). Eight (66.6%) of the uPPA patients had a non-AD profile; of the remaining four (33.3%) with AD profile, all had a clinical picture mixed with logopenic features.

Subdivision within PPA patients presenting with non-AD profile

Since the most common non-AD pathologies in PPA are tau- and TDP43-related, we used k-means clustering (k = 2) of the τ_P - 181/τ_T ratio in PPA patients with non-AD profile. The two clusters identified, differed significantly between each other (p < 0.000001): one cluster comprised patients with τ_P - 181/τ_T ≤0.158 and the other cluster comprisedpatients with τ_P - 181/τ_T ≥0.169 (Fig. 4). The latter may be compatible with more severe (hyper)phosphorylation (i.e., tauopathic) process.

DISCUSSION

The present study is based on CSF biomarkers. Neither it is a pathological study, nor can pathological data be presented, since all of the patients included are still alive. However, a CSF biomarker profile, characterized by increased τ_T and τ_P - 181 and decreased Aβ₄₂, has been recognized as the biochemical signature of AD neuropathology during life.

By the use of all three biomarkers, most (roughly 90%) of PPA patients could be easily assigned to AD or non-AD profiles, based on calculated cut-off values. For those (∼10%) having marginal or inconclusive profiles, use of the ratios (τ_T/Aβ₄₂ and especially τ_P - 181/Aβ₄₂) proved effective for their final classification. These ratios may be helpful by increasing the certainty of CSF-supported diagnosis. On the contrary, the τ_P - 181/τ_T ratio was not of diagnostic value for the discrimination between AD and nonAD patients in PPA.

In general, we feel that the ratios are helpful in the discrimination of AD versus non-AD pathophysiology and this may be true when the absolute values of some of the individual biomarkers fall within “gray zones”. In this case, the data of ROC and discriminant analysis concerning the ratios is only indicative. However, when all biomarkers fall well within the diagnostic area (positive or negative), they should be interpreted individually. This holds true especially in light of pathological studies showing that pathologies other than AD and FTLD related to tau or TDP-43, may rarely underlie the PPA syndrome. These include FTLD with ubiquitin inclusions [2, 3], FTLD lacking distinctive histopathology [3], dementia with Lewy bodies [3], multiple system atrophy [11], and Creutzfeldt-Jakob disease [12].

Only a few studies on the classical CSF biomarkers across the different subtypes of PPA have been published so far (Table 3). They comprised relatively small numbers of patients and their results suggest that an “AD profile” of CSF biomarkers is very rare or absent in nfaPPA and sPPA, while it is present in most or all of patients with lPPA, which, in turn, have biomarker levels comparable with patients with AD [13, 14]. None of these studies reported an attempt to further subdivide non-AD patients in tau- or TDP43-related biochemistry, based on biomarkers, although in one, 27% of nfaPPA patients carried mutations suggestive of TDP43 pathology [13].

In the present study we have shown that nfaPPA and lPPA are not synonymous with frontotemporal and AD biochemistry, respectively, since 22% of nfaPPA have an AD-compatible CSF profile and 25% of lPPA have a non-AD profile. In sPPA most of the patients showed a non-AD profile, yet ∼1/3 presented with the AD profile. Most of the uPPA patients had a non-AD profile; however, all uPPA patients with AD profile had clinical features mixed with logopenic aspects, and of those mixed with logopenic features 4/6 had a CSF AD profile, indicating that, in unclassifiable patients, when some logopenic features are recognized, AD may be the most common underlying pathology. The number of patients included in the present study is higher compared to previous studies, yet it is still small (actually <10 in 2 of the subgroups). Clearly, a study with larger numbers is needed to validate the results of the present study.

The percentage of various types of pathology varies significantly among publications. In lPPA the presence of AD pathology may vary from as low as 0% [15] to as high as 83.5% or more [3], in nfaPPA may vary from 0% [2 , 16] to 45% [17] and in sPPA may vary from 0% [16, 17] to 44.5% [3]. In a recent pathological study of 30 patients who met PPA criteria, 75% of those with nfaPPA, had FTLD pathology, while AD pathology was present in 54% of patients with lPPA [18]. Such a degree of variation may be due to differences in criteria or neuropsychological methods used to classify patients in the PPA variants. However, since the number of eligible PPA patients studied in most of the above publications is usually no more than 30–40 and further subdivision in at least three subtypes results in groups of no more than 10–15 patients each (and sometimes even less), samples may be relatively small to be representative and selection bias cannot be ruled out. In an effort to reduce the effect of such factors, Grossman performed a literature survey, identified a total of 145 eligible PPA cases with pathologic diagnosis, published from different institutions during the years 2005–2010 and calculated the average percentages across publications [3, 19]. He showed that, on average, in nfaPPA AD pathology was present in 25%, a percentage similar to the one of the present study. The corresponding percentages for sPPA and lPPA were 25% and 50%, respectively. For lPPA, this average percentage of AD is somewhat lower than the one of the present study, yet compatible with the notion that it is probably no more than 80%.

To the best of our knowledge, there is no CSF biomarker study in PPA patients aiming at the discrimination between tau and TDP-43 subtypes. There is only one such study in frontotemporal dementia patients, suggesting that the τ_P - 181/τ_T ratio with a cut-off of 0.136 may be of help [8]. In our study, non-AD PPA patients seemed to cluster in two subgroups by a relatively similar cut-off of 0.163. It is tempting to assume that τ_P - 181/τ_T >0.163 may represent patients with tau pathology, while τ_P - 181/τ_T <0.163 may represent patients with non-tau (usually TDP-43) pathology.

In conclusion, CSF biomarkers seem to be a valuable tool for the discrimination between PPA patients with AD and non-AD pathophysiology and possibly between FTLD patients with tau and TDP-43 pathology.

Footnotes

ACKNOWLEDGMENTS

This research has been co-financed by the European Union (European Regional Development Fund –ERDF) and Greek national funds through the Operational Program “Competitiveness and Entrepreneurship” of the National Strategic Reference Framework (NSRF) –Research Funding Program: Joint Programming Neurodegenerative Disease, “Biomarkers for Alzheimer’s disease and Parkinson’s disease”.

Authors’ disclosures available online ().

Appendix

References

Mesulam

(2003) Primary progressive aphasia –a language-based dementia. N Engl J Med 349, 1535–1542.

Mesulam

, Wicklund

, Johnson

, Rogalski

, Léger

, Rademaker

, Weintraub

, Bigio

(2008) Alzheimer and frontotemporal pathology in subsets of primary progressive aphasia. Ann Neurol 63, 709–719.

Grossman

(2010) Primary progressive aphasia: Clinicopathological correlations. Nat Rev Neurol 6, 88–97.

Gorno-Tempini

, Hillis

, Weintraub

, Kertesz

, Mendez

, Cappa

, Grossman

(2011) Classification of primary progressive aphasia and its variants. Neurology 76, 1006–1014.

Wicklund

, Duffy

, Strand

, Machulda

, Whitwell

, Josephs

(2014) Quantitative application of the primary progressive aphasia consensus criteria. Neurology 82, 1119–1126.

McKhann

, Knopman

, Chertkow

, Hyman

, Jack

Jr , Kawas

, Klunk

, Koroshetz

, Manly

, Mayeux

, Mohs

, Morris

, Rossor

, Scheltens

, Carrillo

, Thies

, Weintraub

, Phelps

(2011) The diagnosis of dementia due to Alzheimer’s disease: Recommendations from the National Institute on Aging-Alzheimer’s Association workgroups on diagnostic guidelines for Alzheimer’s disease. Alzheimers Dement 7, 263–269.

Kapaki

, Paraskevas

, Papageorgiou

, Bonakis

, Kalfakis

, Zalonis

, Vassilopoulos

(2008) Diagnostic value of CSF biomarker profile in frontotemporal lobar degeneration. Alzheimer Dis Assoc Disord 22, 47–53.

Borroni

, Benussi

, Archetti

, Galimberti

, Parnetti

, Nacmias

, Sorbi

, Scarpini

, Padovani

(2015) CSF p-tau181/tau ratio as biomarker for TDP pathology in frontotemporal dementia. Amyotroph Lateral Scler Frontotemporal Degener 16, 86–91.

Goodglass

, Kaplan

, Barresi

(2000) Boston Diagnostic Aphasia Examination –.3rd Edition (BDAE-3), Pearson, San Antonio.

10.

del Campo

, Mollenhauer

, Bertolotto

, Engelborghs

, Hampel

, Simonsen

, Kapaki

, Kruse

, Le Bastard

, Lehmann

, Molinuevo

, Parnetti

, Perret-Liaudet

, Sáez-Valero

, Saka

, Urbani

, Vanmechelen

, Verbeek

, Visser

, Teunissen

(2012) Recommendations to standardize preanalytical confounding factors in Alzheimer’s and Parkinson’s disease cerebrospinal fluid biomarkers: An update. Biomark Med 6, 419–430.

11.

Molina

, Probst

, Villanueva

, Jiménez-Jiménez

, Madero

, Torres

, Bermejo

(1998) Primary progressive aphasia with glial cytoplasmic inclusions. Eur Neurol 40, 71–77.

12.

Mandell

, Alexander

, Carpenter

(1989) Creutzfeldt-Jakob disease presenting as isolated aphasia. Neurology 39, 55–58.

13.

Gil-Navarro

, Lladó

, Rami

, Castellví

, Bosch

, Bargalló

, Lomeña

, Reñé

, Montagut

, Antonell

, Molinuevo

, Sánchez-Valle

(2013) Neuroimaging and biochemical markers in the three variants of primary progressive aphasia. Dement Geriatr Cogn Disord 35, 106–117.

14.

Santangelo

, Coppi

, Ferrari

, Bernasconi

, Pinto

, Passerini

, Comi

, Magnani

(2015) Cerebrospinal fluid biomarkers can play a pivotal role in the diagnostic work up of primary progressive aphasia. . J Alzheimers Dis 43, 1429–1440.

15.

Josephs

, Duffy

, Strand

, Whitwell

, Layton

, Parisi

, Hauser

, Witte

, Boeve

, Knopman

, Dickson

, Jack

Jr , Petersen

(2006) Clinicopathological and imaging correlates of progressive aphasia and apraxia of speech. Brain 129, 1385–1398.

16.

Snowden

, Neary

, Mann

(2007) Frontotemporal lobar degeneration: Clinical and pathological relationships. Acta Neuropathol 114, 31–38.

17.

Kertesz

, McMonagle

, Blair

, Davidson

, Munoz

(2005) The evolution and pathology of frontotemporal dementia. Brain 128, 1996–2005.

18.

Harris

, Gall

, Thompson

, Richardson

, Neary

, du Plessis

, Pal

, Mann

, Snowden

, Jones

(2013) Classification and pathology of primary progressive aphasia. Neurology 81, 1832–1839.

19.

Bonner

, Ash

, Grossman

(2010) The new classification of primary progressive aphasia into semantic, logopenic, or nonfluent/agrammatic variants. Curr Neurol Neurosci Rep 10, 484–490.