jpad journal

AND option

OR option

COMPARATIVE ANALYSIS OF DIFFERENT DEFINITIONS OF AMYLOID-Β POSITIVITY TO DETECT EARLY DOWNSTREAM PATHOPHYSIOLOGICAL ALTERATIONS IN PRECLINICAL ALZHEIMER

 

M. Milà-Alomà1,2,3,4, G.Salvadó1,2, M. Shekari1,2,3, O. Grau-Rivera1,2,4,5, A. Sala-Vila1,2, G. Sánchez-Benavides1,2,4, E.M. Arenaza-Urquijo1,2,4, J.M. González-de-Echávarri1,2, M. Simon6, G. Kollmorgen7, H. Zetterberg8,9,10,11, K. Blennow8,9, J.D. Gispert1,2,3,12, M. Suárez-Calvet1,2,4,5,
J.L. Molinuevo1,2,3,4 for the ALFA study†

 

1. Barcelonaβeta Brain Research Center (BBRC), Pasqual Maragall Foundation. Barcelona, Spain; 2. IMIM (Hospital del Mar Medical Research Institute), Barcelona, Spain; 3. Universitat Pompeu Fabra, Barcelona, Spain; 4. Centro de Investigación Biomédica en Red de Fragilidad y Envejecimiento Saludable (CIBERFES), Madrid, Spain; 5. Servei de Neurologia, Hospital del Mar, Barcelona, Spain; 6. Roche Diagnostics International Ltd, Rotkreuz, Switzerland; 7. Roche Diagnostics GmbH, Penzberg, Germany; 8. Department of Psychiatry and Neurochemistry, Institute of Neuroscience and Physiology, Sahlgrenska Academy, University of Gothenburg, Sweden;
9. Clinical Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, Sweden; 10. Department of Neurodegenerative Disease, UCL Institute of Neurology, Queen Square, London, United Kingdom; 11. UK Dementia Research Institute at UCL, London, United Kingdom; 12. Centro de Investigación Biomédica en Red Bioingeniería, Biomateriales y Nanomedicina, Madrid, Spain; †The complete list of collaborators of the ALFA Study can be found in the acknowledgements section.

Corresponding Authors: José Luis Molinuevo, Alzheimer Prevention Program – Barcelonaβeta Brain Research Center, Wellington 30, 08005, Barcelona, Spain, +34933160990, E-mail: jlmolinuevo@barcelonabeta.org; Marc Suárez-Calvet, Alzheimer Prevention Program – Barcelonaβeta Brain Research Center, Wellington 30, 08005, Barcelona, Spain
+34933160990, E-mail: msuarez@barcelonabeta.org

J Prev Alz Dis 2021;1(8):68-77
Published online September 29, 2020, http://dx.doi.org/10.14283/jpad.2020.51

 


Abstract

Amyloid-β (Aβ) positivity is defined using different biomarkers and different criteria. Criteria used in symptomatic patients may conceal meaningful early Aβ pathology in preclinical Alzheimer. Therefore, the description of sensitive cutoffs to study the pathophysiological changes in early stages of the Alzheimer’s continuum is critical. Here, we compare different Aβ classification approaches and we show their performance in detecting pathophysiological changes downstream Aβ pathology.
We studied 368 cognitively unimpaired individuals of the ALFA+ study, many of whom in the preclinical stage of the Alzheimer’s continuum. Participants underwent Aβ PET and CSF biomarkers assessment. We classified participants as Aβ -positive using five approaches: (1) CSF Aβ42 < 1098 pg/ml; (2) CSF Aβ42/40 < 0.071; (3) Aβ PET Centiloid > 12; (4) Aβ PET Centiloid > 30 or (5) Aβ PET Positive visual read. We assessed the correlations between Aβ biomarkers and compared the prevalence of Aβ positivity. We determined which approach significantly detected associations between Aβ pathology and tau/neurodegeneration CSF biomarkers.
We found that CSF-based approaches result in a higher Aβ-positive prevalence than PET-based ones. There was a higher number of discordant participants classified as CSF
Aβ-positive but PET Aβ-negative than CSF Aβ-negative but PET Aβ-positive. The CSF Aβ 42/40 approach allowed optimal detection of significant associations with CSF p-tau and t-tau in the Aβ-positive group.
Altogether, we highlight the need for sensitive Aβ -classifications to study the preclinical Alzheimer’s continuum. Approaches that define Aβ positivity based on optimal discrimination of symptomatic Alzheimer’s disease patients may be suboptimal for the detection of early pathophysiological alterations in preclinical Alzheimer.

Key words: Alzheimer’s disease, biomarkers, cerebrospinal fluid, positron emission tomography, preclinical Alzheimer.


 

Introduction

Alzheimer’s disease (AD) is preceded by a long preclinical phase, which is characterized by the emergence of pathological brain changes in the absence of evident clinical symptoms. Cognitive symptoms may arise decades after the first pathological brain changes and can eventually progress to dementia. AD is therefore understood as a continuum (the Alzheimer’s continuum) entailing a preclinical (i.e., asymptomatic) and a clinical (i.e., symptomatic) phase (1, 2). Following the latest NIA-AA guidelines, the term ‘Alzheimer’s disease’ should be applied when there is biomarker evidence of both amyloid-β (Aβ) and tau pathologies (2, 3).
A key aspect in preclinical Alzheimer’s studies is the biomarker-based definition of Aβ pathology. Currently established Aβ pathology biomarkers can be divided in two main categories: fluid [mainly cerebrospinal fluid (CSF) and, more recently, also blood] and neuroimaging (Aβ PET) biomarkers (4, 5). Although all these biomarkers may be used on a continuous scale, cutoffs are needed to dichotomize biomarker values for diagnosis or research individual categorization purposes. Different approaches may be used to establish biomarker cutoffs. These mainly include: (1) comparison against a gold standard (neuropathology, other validated biomarkers reflecting pathology), (2) optimal discrimination between a control and a pathological group (case-control studies), (3) prediction of disease progression, or (4) description of abnormality based on population percentiles in a ‘normal’ or reference population (e.g., 95% of the unaffected population have lower/higher values) (2, 6, 7).
Regarding CSF Aβ biomarkers, a decrease in CSF Aβ42 is the widely known pattern in AD, which also correlates with increased Aβ PET uptake (8, 9). In addition, CSF Aβ42 cutoffs have been validated against Aβ PET as gold standard (10, 11). Nevertheless, the use of CSF Aβ42 may be limited by interindividual differences in Aβ42 production, and by the impact of pre-analytical and assay-related factors. To overcome some of these limitations, normalization of Aβ42 with Aβ40, by using the Aβ42/40 ratio, was proposed (12). Studies have shown that the CSF Aβ42/40 ratio is a better predictor of abnormal Aβ PET and that the ratio allows a more accurate identification of AD patients than CSF Aβ42 (13, 14).
In the clinical setting, Aβ PET imaging is used to detect Aβ pathology in patients with cognitive impairment suspected to be AD-related. The definition of a positive or negative scan is done through visual read (VR) by trained specialists, who categorize images according to well-established criteria. The VR method has been validated against neuropathology and has shown high concordance with quantitative methods (15–17). Aβ PET images can be quantified using standardized uptake value ratios (SUVR), as well as with the “Centiloid” (CL) scale. The CL scale is a method for standardizing measurements to a uniform scale, thus facilitating comparisons across radiotracers and also across studies (18), and has a robust association with pathology (19,20).
It is important to note that, although binary cutoffs allow us to stratify patients or research participants according to their biomarker profile, there is a continuum of Aβ pathology changes, that is a dynamic transition from Aβ-negative to Aβ-positive states. Therefore, binarization may leave emerging Aβ pathology undetected. This is especially important in the study of preclinical Alzheimer, in which the use of cutoffs that have been defined for a clinical use possibly conceals meaningful amounts of early Aβ pathology. Thus, the description of cutoffs that are sensitive enough to detect subtle pathological changes in the very early stage of the Alzheimer’s continuum is of great importance (21, 22).
The aim of this study is to describe and compare different Aβ classification approaches, derived through both CSF and PET measures, and to show their performance in the detection of other early pathophysiological changes in the Alzheimer’s continuum. We hypothesize that an optimal Aβ classification approach at this early stage would set the point where Aβ-downstream pathophysiological changes start.

 

Methods

Participants

The ALFA+ cohort is a nested longitudinal study of the ALFA (for ALzheimer and FAmilies) study, which aims at characterizing the preclinical stage of the Alzheimer’s continuum (23). The ALFA study includes 2,743 cognitively unimpaired individuals, aged between 45 and 75 years old, which are enriched for family history of AD and APOE-ε4 carriership. In the ALFA+ cohort, participants undergo a more detailed characterization, which entails acquisition of CSF samples for biomarker levels determination, neuroimaging biomarkers (MRI and FDG and Aβ PET), APOE genotyping and cognitive assessments. Among the ALFA+ participants, 381 had available CSF biomarker measurements. Since we aimed at studying participants in the Alzheimer’s continuum, we excluded 13 participants that had a CSF biomarker of suspected non-Alzheimer’s pathology (2), as defined by CSF Aβ42/40 ≥ 0.071 and CSF p-tau >24 pg/ml. None of them was Aβ-positive using Aβ PET in any of the approaches defined below. Thus, 368 participants were finally included in the present study. Of those, a subset of 303 and 301 participants also had CL quantification or VR assessment of Aβ PET imaging, respectively. For all participants, the time difference between CSF collection and Aβ PET assessments was less than one year.

CSF biomarkers assessment

CSF samples were obtained by lumbar puncture and collected, processed and stored following standard procedures (24). Total tau (t-tau) and phosphorylated tau (p-tau) measurements were performed using the electrochemiluminiscence Elecsys® Total-Tau CSF and Phospho-Tau(181P) CSF immunoassays, respectively. CSF Aβ42 and Aβ40 were measured with the prototype NeuroToolKit (Roche Diagnostics International Ltd.). All assays were performed on a fully automated cobas e 601 or e 401 instrument (Roche Diagnostics International Ltd.) at the Clinical Neurochemistry Laboratory, Sahlgrenska University Hospital, Mölndal, Sweden.

PET imaging acquisition and processing

Imaging acquisition and preprocessing protocols have been described previously (20). In brief, a T1-weighted MRI and an [18F] Flutemetamol PET scan was acquired in all participants. The T1-weighted 3D-TFE sequence was acquired in a Philips 3 T Ingenia CX scanner. PET imaging was conducted in a Siemens Biograph mCT, following a cranial CT scan for attenuation correction. Participants were injected with 185 MBq (range 166.5–203.5 Mbq) of [18F] Flutemetamol, and 4 frames of 5 min each were acquired 90 min post-injection.
PET processing was performed following the standard CL pipeline (18) using SPM12. In brief, PET images are first coregistered to their respective T1-weighted images and, afterwards, moved to MNI space using the normalization transformation derived from the segmentation of the T1-weighted image. PET images were intensity normalized using whole cerebellum as reference region. CL values were calculated from the mean values of the standard CL target region (http://www.gaain.org/centiloid-project) using the transformation previously calibrated (20).
A nuclear medicine physician visually rated the scans as Aβ-positive or Aβ-negative using standard clinical criteria as specified in the Summary of Product Characteristics of the tracer (https://www.ema.europa.eu/en/documents/product-information/vizamyl-epar-product-information_en.pdf).

Aβ pathology biomarkers and cutoffs

We classified the participants as Aβ-positive or Aβ-negative using five different Aβ biomarker/cutoff combinations (Aβ classification approaches; Table 1). Each of the approaches were determined as follows. In approach 1, participants with a CSF Aβ42 < 1098 pg/ml were considered as Aβ-positive. This cutoff had been previously established in concordance with Aβ PET in a similar cohort and had area under the ROC curve (AUC) of 0.85 to discriminate between Aβ PET-positive or negative (10). In approach 2, we used CSF Aβ42/40 as a biomarker of Aβ pathology and we determined a cutoff using a Gaussian mixture model (GMM). The cutoff was defined as the mean minus 2 standard deviations (SD) of the non-pathologic Gaussian distribution. As the fitting of the GMM can vary depending on initial conditions, the process was repeated 1000 times with random initial conditions. The final cutoff for CSF Aβ42/40, derived as the median value of the 1000 repetitions, was 0.071. In both approaches 3 and 4, we used the Aβ PET CL scale as biomarker of Aβ pathology, but with different cutoffs. In approach 3, we defined Aβ positivity when CL > 12; this cutoff has been previously proposed to detect subtle Aβ pathology in a comparison against CSF Aβ42 (20). In approach 4, we defined Aβ positivity when CL > 30, a cutoff that falls within the optimal range of agreement with VR method in clinical populations (24 – 35 CL), and has been proposed to detect established Aβ pathology (19,20,25–27). Finally, in approach 5, we categorized participants as Aβ-positive or Aβ-negative using the Aβ PET VR, the approach most commonly used in the clinical setting.

Table 1. Summary of Aβ biomarkers/cutoffs combinations used in each Aβ classification approach

Abbreviations: Aβ, amyloid-β; CSF, cerebrospinal fluid; CL: Centiloids.

Statistical analysis

CSF biomarker distributions were tested for normality with the Kolmogorov-Smirnov test and histograms visual inspection. CSF Aβ42, p-tau and t-tau did not follow a normal distribution and were thus log10-transformed. In contrast, CSF Aβ42/40 ratio followed a normal distribution and was not transformed. CSF biomarker extreme values were identified following Tukey criterion (i.e., values falling at over three times the interquartile range below the first quartile or above the third quartile) and excluded from the analyses.
Differences in demographic characteristics and CSF biomarker values as a function of Aβ classification schemes were assessed by means of chi-squared or one-way analysis of variance (ANOVA) tests, as appropriate. Correlations between Aβ biomarkers were tested using partial Spearman rank correlation, corrected by age. We used a chi-squared analysis in order to test significant differences in Aβ-positive prevalence using different Aβ classification approaches.
We tested the association between tau and neurodegeneration CSF biomarkers (i.e., CSF p-tau, t-tau) and the Aβ biomarker (i.e., CSF Aβ42, Aβ42/40 or CL scale) in a linear model, adjusting by the effect of age and sex. These analyses were performed separately in Aβ-positive or Aβ-negative groups as defined by each approach. Moreover, in order to determine which of the approaches was more sensitive to detect early pathophysiological changes, we conducted the same analyses in the whole cohort but additionally introducing in the model the interaction term ‘Aβ biomarker x Aβ group’. A significant interaction term denotes that the association of the Aβ biomarker with CSF p-tau and t-tau differs between the Aβ-negative and Aβ-positive groups, as defined by that approach. These analyses were performed in each of the five Aβ classification approaches.
All statistical models were adjusted for the effects of age and sex. For all analysis, we applied a false discovery rate (FDR) multiple comparison correction following the Benjamini-Hochberg procedure (28). All tests were 2-tailed, with a significance level of α = 0.05.

 

Results

Participants’ demographics

Participants’ characteristics are depicted in Table 2. We compared the demographical characteristics of the Aβ-positive and Aβ-negative groups defined by the different Aβ classification approaches. We observed that Aβ-positive participants were significantly older using PET-based versus CSF-based approaches. There were not significant differences in years of education, frequency of the APOE-ε4 allele or sex between the different Aβ-classification approaches.
Most of the AD biomarkers significantly differed between the five Aβ classification approaches within the Aβ positive or Aβ-negative group (Table 2). PET-based Aβ classification approaches (namely, approach 3, 4 and 5) had a more pathologic biomarker profile (that is, lower CSF Aβ42 and CSF Aβ42/40 ratio, but higher CL scale and CSF p-tau and t-tau), than the CSF-based Aβ classification approaches (namely, approach 1 and 2). These results suggest that CSF-based Aβ biomarkers are capturing Aβ accumulation at an earlier stage of the Alzheimer’s continuum. This idea is supported by the fact that Aβ-negative groups defined by PET-based approaches had higher levels of CSF p-tau and t-tau (Table 2).

Table 2. Participants’ demographic characteristics and biomarker levels by five Aβ classification approaches

Data are expressed as mean (M) and standard deviation (SD) or number of participants (n) and percentage (%), as appropriate. One-way ANOVA followed by Tukey corrected post hoc comparisons was used to compared age and education and Pearson’s Chi-square test to compare sex and APOE-ε4 status between approaches within each Aβ-group. CSF biomarkers and centiloid scale were compared with an ANCOVA adjusted by age and sex followed by Tukey corrected post hoc comparison. The P-values indicated in the last column refer to the approach main effect. All P-values are corrected for multiple comparisons using the FDR method; Abbreviations: Aβ42, amyloid-β 42; p-tau, phosphorylated tau; t-tau, total tau; * P<0.05 vs approach 1; † P<0.001 vs approach 1; ‡ P<0.01 vs approach 2; § P<0.01 vs approach 1; ||P<0.0001 vs approach 1; { P<0.0001 vs approach 2; # P<0.001 vs approach 2; ** P<0.05 vs approach 2; ‡‡ P< 0.05 vs approach 4

 

Correlations between Aβ biomarkers

We tested the correlations between CSF and PET Aβ biomarkers, corrected by age. There was a moderate correlation between CSF Aβ42 and CL values (Spearman’s rho = -0.41; P<0.0001; Figure 1A and B) and CSF Aβ42/40 and CL values (Spearman’s rho = -0.54; P<0.0001; Figure 1C and D). There was also a high correlation between CSF Aβ42 and CSF Aβ42/40 ratio (Spearman’s rho = -0.75; P<0.0001; Figure 1E). Prevalence of Aβ positivity using different Aβ classification approaches We next determined the prevalence of Aβ-positive and Aβ-negative participants according to each Aβ classification approach and, remarkably, we found a wide range of Aβ-positive prevalences ranging from 8.25 to 44.3% (Figure 2; Table 3). Approach 1 (i.e., CSF Aβ42) resulted in the highest prevalence of Aβ-positive participants (N = 163, 44.3%), followed by approach 2 (CSF Aβ42/40 ratio; Aβ-positive: N = 131, 35.6%), approach 3 (CL > 12; Aβ-positive: N = 49, 16.2%), approach 5 (VR; Aβ-positive: N = 40, 13.3%) and, finally, approach 4 (CL > 30; Aβ-positive: N = 25, 8.25%) (Figure 2). The prevalence of Aβ-positive participants in CSF-based classification approaches (approach 1 and 2) was significantly higher than that in PET-based classification approaches (approach 3, 4 and 5, P < 0.0001; Figure 2). Notably, approach 3 (CL > 12) showed a statistically significant higher prevalence of Aβ-positive participants compared to approach 4 (CL > 30, P = 0.043, Figure 2).

Figure 1. Correlation between Aβ classification approaches

Scatter plots representing the associations between each Aβ biomarker. Each point depicts the value of the Aβ biomarker of an individual. Green areas depict discordant subjects and percentage of them is shown in red. The red dashed lines indicate the Aβ biomarker cutoffs used in each Aβ classification approach: CSF Aβ42 = 1098 pg/ml, CSF Aβ42/40 = 0.071, CL = 12 and CL = 30. A, B. Correlation and comparison between CSF Aβ42 and CL; cutoff CL = 12 (A) and CL = 30 (B). C, D. Correlation and comparison between CSF Aβ42/40 and CL; cutoff CL = 12 (C) and CL = 30 (D). E. Correlation and comparison between CSF Aβ42 and CSF Aβ42/40. F. Comparison between CSF Aβ42 and VR. G. Comparison between CSF Aβ42/40 and VR. H, I. Comparison between VR and CL; cutoff CL = 12 (H) and CL = 30 (I). Abbreviations: Aβ, amyloid β; CL, Centiloids; CSF, Cerebrospinal fluid; PET, Positron Emission Tomography

 

Figure 2. Prevalence of Aβ-negative and Aβ-positive groups in each Aβ classification approach

The graph depicts the prevalence of the Aβ-positive (Aβ+; color) and Aβ-negative (Aβ-; grey) individuals in each Aβ classification approach. We conducted a pair wise chi squared comparison between all the approaches and we found significant statistical differences between them (P < 0.0001). Bonferroni-corrected pair-wise post hoc comparisons: * P<0.0001 vs approach 1 and approach 2¸† P<0.05 vs approach 3. Abbreviations: Aβ, amyloid-β; CL, Centiloids; CSF, Cerebrospinal fluid; PET, Positron Emission Tomography; VR, Visual Reads.

 

We assessed the differences between the CSF- and PET-defined Aβ-positive participants (i.e., CSF/PET-discordant participants) within each Aβ classification approach (shown in Table 3 and Figure 1). The participants that were classified as CSF Aβ-positive but PET Aβ-negative range from 19.8 to 37.0% (Table 3; green areas in Figure 1). The highest discordance occurred between Aβ classification approach 4 (CL < 30) and approach 1 (CSF Aβ42; 37.0% discordance; Figure 1B), and the lowest occurred between approach 3 (CL < 12) and approach 2 (CSF Aβ42/40; 19.8% discordance, Figure 1C). By contrast, fewer participants showed discordance in the other direction, i.e., a CSF Aβ-negative but a PET Aβ-positive biomarker (range: 0.33% to 2.99%; Figure 1; Table 3). Of note, the proportion of discordant participants was higher for CSF Aβ42 (approach 1) than for CSF Aβ42/40 ratio (approach 2) in all comparisons. Association of Aβ biomarkers with tau pathology and neurodegeneration CSF biomarkers We next assessed whether changes in tau pathology and neurodegeneration CSF biomarkers as a function of Aβ pathology differed before and after the computed Aβ cutoffs. We hypothesized that an optimal Aβ biomarker cutoff would discriminate between an Aβ-negative state (where Aβ pathology is not associated with downstream tau pathology and neurodegeneration changes) and an Aβ-positive state (where those downstream changes arise). For these purposes, we computed the association between each Aβ biomarker and CSF p-tau and t-tau in Aβ-negative and Aβ-positive groups defined by the different Aβ classification approaches. We also computed the interaction term ‘Aβ biomarker x Aβ group’. The results of the analyses are summarised in Table 4 and Figure 3. Remarkably, robust statistically significant interactions occurred using approach 2. In other words, the cutoff derived from the CSF Aβ42/40 ratio significantly discriminated between two states: (1) an Aβ-negative state where tau and neurodegeneration CSF biomarkers were not associated with CSF Aβ42/40 levels and (2) an Aβ-positive state where CSF p-tau and t-tau significantly increase as a function of higher Aβ pathology (i.e., decreased CSF Aβ42/40 ratio). In approach 1 (i.e., CSF Aβ42), we also observed statistical significance in the interaction term ‘Aβ biomarker x Aβ group’ for CSF p-tau and t-tau. Nevertheless, both CSF p-tau and t-tau were positively associated with CSF Aβ42 in the Aβ-negative group. None of the Aβ-positive and Aβ-negative groups defined with Aβ PET-based approaches differed in their association between CL scale and CSF p-tau and t-tau. However, there was a tendency to statistical significance in the interaction term ‘Aβ biomarker x Aβ group’ for CSF p-tau in approach 3 (i.e., CL > 12). Moreover, there was a significant association between CSF p-tau and t-tau with CL scale in the Aβ-positive group using approach 3 and approach 5 (i.e., VR).

Table 3. Contingency table with the percentage of agreement and discordance between the different Aβ classification approaches

Data are expressed as percentage (%) of participants in each category. Bold letters indicate discordance. We performed a chi-square test of independence to examine the relation between Aβ positivity and the Aβ classification approach used and we observed that the percentage of Aβ positivity was not equally distributed between Aβ-classification approaches X2 (4, N = 301) = 176, P < 0.0001. A Bonferroni corrected pair-wise post hoc comparisons showed significant differences between the approach 1 and approach 3, 4 and 5 (P < 0.0001)¸ between the approach 2 and approach 3, 4 and 5 (P < 0.0001), and between approach 3 and 4 (P < 0.05). Abbreviations: Aβ, amyloid-β; Aβ+, amyloid-β positive; Aβ-, amyloid-β negative; CL, Centiloids; VR, Visual Read.

Figure 3. Association of each Aβ biomarker with CSF p-tau and t-tau by Aβ classification approach

Scatter plots representing the associations of CSF p-tau and t-tau with each Aβ biomarker in Aβ-positive and Aβ-negative groups. Each point depicts the value of the CSF biomarker of an individual and the solid lines indicate the regression line for each group. The P-values shown for the ‘Aβ biomarker x Aβ group’ interaction were computed using a linear model adjusting for age and sex. All P-values are corrected for multiple comparisons using the FDR method. The dashed lines indicate the Aβ biomarker cutoffs used in each Aβ classification approach: CSF Aβ42 = 1098 pg/ml, CSF Aβ42/40 = 0.071, CL = 12 and CL = 30; Abbreviations: Aβ, amyloid β; CL, Centiloids; CSF, Cerebrospinal fluid; PET, Positron Emission Tomography; p-tau, phosphorylated tau; t-tau, total tau; VR, Visual Reads.

 

Table 4. CSF p-tau and t-tau associations with Aβ biomarkers by Aβ classification approach

CSF p-tau or t-tau were assessed by a linear model with the Aβ biomarker (CSF Aβ42, CSF Aβ42/40 or CL) as main effect and age and sex as covariates. This analysis was performed separately in Aβ-negative or Aβ-positive groups as defined by each approach. Moreover, we added the interaction term ‘Aβ biomarker x Aβ group’ in order to test statistical differences in the regression slopes between Aβ-negative and –positive groups. The standardized regression coefficients (β) and standard errors (SE) are depicted. P-values are corrected for multiple comparisons using FDR method. *Significant values; Abbreviations: Aβ42, amyloid-β 42; CL, Centiloids; p-tau, phosphorylated tau; t-tau, total tau; VR, Visual Read.

 

Discussion

In this study, we compared five different approaches to classify cognitively unimpaired individuals as Aβ-positive or Aβ-negative, and we found considerable differences between them. Specifically, our results show that: (1) Aβ-positive prevalence vary from 8 to 44% depending on the approach used; (2) Aβ CSF-based approaches result in the highest prevalence of Aβ-positive subjects compared with PET-based approaches; (3) Discordant classification appears more frequently as a CSF Aβ-positive but PET Aβ-negative (19.8 – 37.0%), while only a few individuals were CSF Aβ-negative but PET Aβ-positive (0.33 – 2.99%); (4) In Aβ PET, the use of a lower CL cutoff (12 CL instead of 30 CL) increases the detection of Aβ-positive individuals based on CSF levels; (5) among all Aβ classification approaches assessed, the CSF Aβ42/40 ratio cutoff shows the greater concordance with Aβ PET imaging (using the 12 CL cutoff), and the most significant difference of regression slopes between the Aβ-positive and Aβ-negative groups in the association with Aβ downstream pathology (e.g., tau pathology and neurodegeneration).
The ATN classification is widely used in AD research to classify patients or study participants based on biomarker evidence of Aβ pathology (A), tau pathology (T) or neurodegeneration/neuronal injury (N) (7). Each biomarker group can be defined with different biomarkers. For Aβ pathology (A), the most accepted biomarkers are CSF Aβ42, the CSF Aβ42/40 ratio and Aβ PET.
In regard to CSF Aβ biomarkers, CSF Aβ42 has been the most widely used, and cutoffs have been established against Aβ PET (10,11). However, increasing evidence favors the use of CSF Aβ42/40 ratio to normalize both preanalytical and interindividual differences, which provides a higher reproducibility and specificity for AD pathology (13,14). We show that the CSF Aβ42 approach results in a higher number of Aβ-positive participants than any other approach. However, CSF Aβ42 approach also results in a higher frequency of discrepancies with the PET-based approaches compared to the CSF Aβ42/40 ratio approach. This observation is in line with the idea that the CSF Aβ42/40 ratio predicts more accurately Aβ PET than CSF Aβ42. Furthermore, a high concordance between Aβ PET VR and ratios of CSF biomarkers have been shown to best distinguish Aβ-positive from negative individuals in both cognitively unimpaired and impaired populations. The CSF p-tau/Aβ42 ratio has a high diagnostic clinical utility as its cutoff has been derived using Aβ PET VR concordance (10). However, we have not used this ratio in this setting because it represents a mixed pathology of Aβ and tau, and tau biomarkers become abnormal later in the disease stage. For CSF Aβ42, we used the cutoff of 1098 pg/ml, that was previously defined in a study where CSF Aβ42 was measured by a Roche Elecsys assay and a ROC curve was built in comparison against Aβ PET. This cutoff is also very similar to that obtained in the BioFINDER study (1100 pg/ml) which is anchored to a positive Aβ PET VR (11).
For CSF Aβ42/40, we rendered a cutoff with a different approach. Instead of comparing the CSF values with Aβ PET as the gold standard, we applied a data-driven approach, namely a Gaussian Mixture Modelling (GMM). This method allows to describe a normal range of CSF Aβ42/40 ratios in the non-pathological group and derive a cutoff defined as the mean minus 2 standard deviations of this Gaussian distribution. By definition, under this criterion, 97.7% of Aβ-negative subjects are expected to display high CSF Aβ42/40 values. In other words, this criterion is highly specific by design. We hypothesised that this method would be more sensitive to detect incipient Aβ pathology in cognitively unimpaired individuals than criteria based on discriminating asymptomatic from symptomatic AD populations. In fact, GMM has already been used to derive cohort-specific CSF Aβ cutoffs, showing a robust ability to discriminate normal and pathologic distributions of Aβ42 (29–31). Our results confirm the suitability of both using the CSF Aβ42/40 ratio and deriving its cutoff with a GMM method. In fact, approach 2 shows the lowest number of discordant values compared with any other approach. According with our hypothesis, we observed that this approach provides optimal discrimination between Aβ-negative participants, showing no associations with Aβ-downstream pathologies (as measured by CSF p-tau and t-tau), and Aβ-positive participants, which display significant association with them. This is shown by the fact that this approach results in a significant difference of regression slopes between the Aβ-positive and Aβ-negative groups in the association between CSF p-tau or t-tau as a function of Aβ pathology. In fact, visual inspection of Figure 4 shows that the CSF Aβ42/40 cutoff we have computed by GMM falls close to the intersection of the regression lines in the Aβ-positive and Aβ-negative groups. Remarkably, this result is not seen when using the CSF Aβ42 approach. In this approach, we found unexpected significant positive correlations between CSF Aβ42 and both CSF p-tau and t-tau in the Aβ-negative group. Without the Aβ40 normalization, CSF Aβ42 may unspecifically correlate with other proteins because of the abovementioned limitations in terms of interindividual differences (high and low Aβ producers) and preanalytical variability.
In relation to Aβ PET biomarkers, Aβ PET classification can be performed through VR, widely used in the clinical practice, or using a quantitative approach such as the CL scale. This scale is expressed as universal units and allows the comparison among different tracers and studies (18). Different cutoffs to categorize participants according to CL might be used, depending on the sample and on the study objectives. Optimal cutoffs for CL against VR typically fall within the range of 24 – 35 CL in clinical populations (25–27). However, these VR-based CL cutoffs might not be optimal to study the preclinical stage of the Alzheimer’s continuum, when the goal is to detect subtle Aβ accumulation. Therefore, here we compared three different PET-based approaches: CL > 12, which was established as a cutoff with an optimal agreement against the standard cutoff of CSF Aβ42 < 1098 pg/ml (20) and is expected to detect subtle early Aβ changes; CL > 30, a cutoff falling within the range proposed to robustly detect established Aβ pathology (19,20,25–27); and VR, expected to represent a cutoff closer to the clinical practice. As expected, a cutoff of CL > 12 showed a higher prevalence of Aβ-positive individuals than the other two PET-based approaches. The slopes of CSF p-tau or t-tau as a function of CL did not significantly differ between the Aβ-positive and Aβ-negative groups defined by the CL12 approach, although tendency to this direction is observed, especially for CSF p-tau. However, the CL12 approach allowed the detection of significant associations of CSF p-tau and t-tau in the Aβ-positive group, an association that is also observed with the VR approach, but not with the CL30 one. These results suggest that the CL12 approach may be detecting Aβ deposition that is already associated with increasing tau pathology and neurodegeneration biomarkers. On the contrary, classifying Aβ groups using CL > 30 as cutoff, resulted in the lowest prevalence of Aβ-positive individuals and did not allow the discrimination of changes in Aβ-downstream CSF biomarkers changes between the two Aβ-groups. This suggests that 30 CL is a late cutoff to apply in preclinical Alzheimer’s studies, especially in cohort such as ALFA+, comprised by individuals in the very earliest phase of the Alzheimer’s continuum. Aβ-positive subjects according to the CL30 criterion are probably in a more advanced stage of Aβ pathology. We did not find a significant association between either CSF t-tau or p-tau with CL in the Aβ-positive group defined by the CL30 approach. This might be explained by the reduced sample size in that Aβ-positive group (n = 25), but also because Aβ-pathology might start to plateau in this late stage of preclinical Alzheimer. Conversely, a positive association is observed in the Aβ-negative group, supporting the idea that CL30 is a late cutoff and that Aβ-downstream processes have arisen before this cutoff. Interestingly, the VR approach rendered an Aβ-positive prevalence closer to the CL12 approach than to the CL30 approach. Of note, the raters in our study are highly trained to detect early Aβ deposition, but may differ in other studies or in the clinical setting. Similarly to the CL30 approach, the VR approach does not discriminate between the associations of the CL scale with CSF t-tau or p-tau in the Aβ-positive and Aβ-negative groups, although in this case positive associations are found in the Aβ-positive group.
Finally, we found obvious differences between the CSF- and PET-based Aβ classification approaches. Our results show only a moderate correlation between Aβ CSF and Aβ PET, which is not surprising as they may reflect different aspects of Aβ pathology, i.e., soluble Aβ and fibrillar Aβ aggregates, respectively (1). Noteworthy, the proportion of discordant values that were CSF Aβ-positive but PET Aβ-negative was considerably higher than those that were CSF Aβ-negative but PET Aβ-positive. This is consistent with the idea that Aβ CSF biomarkers detect earlier stages of Aβ pathology than Aβ PET, that is Aβ dysregulation that precedes Aβ plaques formation (32). Furthermore, this earlier dysregulation seems to be already associated to changes in Aβ-downstream events (i.e., tau pathology and neurodegeneration), thus proving the importance of studying this very early stage.
Our study has some limitations: (1) it is a cross sectional study and we could not track the progression of Aβ biomarkers, which would be particularly relevant in the Aβ-discordant biomarkers individuals; (2) it is a comparative analysis between different approaches, lacking a gold standard, so conclusions on sensitivity and specificity for each approach cannot be drawn. However, since comparisons among Aβ positivity criteria are made in the same dataset, conclusions on the relative sensitivity to detect associations with downstream pathophysiological events are valid; (3) results might not be generalizable to other cohorts because ALFA+ is a very specific cohort aimed at studying the preclinical stage of the Alzheimer’s continuum.
Overall, in this study we show that, in the preclinical stage of the Alzheimer’s continuum, there are important differences in the prevalence and the characteristics of Aβ-positive and Aβ-negative study participants depending on the Aβ classification approach used for the dichotomization. This emphasizes the importance of considering the sample and the study aims when deriving cutoffs for participant dichotomization. This is particularly important when studying the preclinical Alzheimer’s ccontinuum stage, where the approaches usually used in symptomatic AD patients may be not suitable to detect subtle changes in Aβ pathology.

Acknowledgements: This publication is part of the ALFA study (ALzheimer and FAmilies). The authors would like to express their most sincere gratitude to the ALFA project participants and relatives without whom this research would have not been possible.

Collaborators of the ALFA study are: Annabella Beteta, Raffaele Cacciaglia, Alba Cañas, Carme Deulofeu, Irene Cumplido, Natalia Vilor Tejedor, Ruth Dominguez, Maria Emilio, Carles Falcon, Karine Fauria, Sherezade Fuentes, Laura Hernandez, Gema Huesa, Jordi Huguet, Paula Marne, Tania Menchón, Carolina Minguillon, Grégory Operto, Albina Polo, Sandra Pradas, Anna Soteras, and Marc Vilanova. The authors thank Roche Diagnostics International Ltd. for providing the kits to measure CSF biomarkers and GE Healthcare for kindly providing 18F-flutemetamol doses of ALFA+ study participants. ELECSYS, COBAS, and COBAS E are registered trademarks of Roche. The Elecsys β-Amyloid (1-42) CSF immunoassay in use is not a commercially available IVD assay. It is an assay that is currently under development and for investigational use only. The measuring range of the assay is 200 (lower technical limit) – 1700 pg/mL (upper technical limit). The performance of the assay beyond the upper technical limit has not been formally established. Therefore, use of values above the upper technical limit, which are provided based on an extrapolation of the calibration curve, is restricted to exploratory research purposes and is excluded for clinical decision making or for the derivation of medical decision points.

Funding: The project leading to these results has received funding from “la Caixa” Foundation (ID 100010434), under agreement LCF/PR/GN17/50300004 and the Alzheimer’s Association and an international anonymous charity foundation through the TriBEKa Imaging Platform project (TriBEKa-17-519007). Additional support has been received from the Universities and Research Secretariat, Ministry of Business and Knowledge of the Catalan Government under the grant no. 2017-SGR-892. OGR is supported by the Spanish Ministry of Science, Innovation and Universities (FJCI-2017-33437). ASV is the recipient of an Instituto de Salud Carlos III Miguel Servet II fellowship (CP II 17/00029). EMAU is supported by the Spanish Ministry of Science, Innovation and Universities – Spanish State Research Agency (RYC2018-026053-I). JDG is supported by the Spanish Ministry of Science and Innovation (RYC-2013-13054). MSC received funding from the European Union’s Horizon 2020 Research and Innovation Program under the Marie Sklodowska-Curie action grant agreement No 752310, and currently receives funding from Instituto de Salud Carlos III (PI19/00155) and from the Spanish Ministry of Science, Innovation and Universities (Juan de la Cierva Incorporación grant IJC2018-037478-I). KB is supported by the Swedish Research Council (#2017-00915), the Alzheimer Drug Discovery Foundation (ADDF), USA (#RDAPB-201809-2016615), the Swedish Alzheimer Foundation (#AF-742881), Hjärnfonden, Sweden (#FO2017-0243), the Swedish state under the agreement between the Swedish government and the County Councils, the ALF-agreement (#ALFGBG-715986), and European Union Joint Program for Neurodegenerative Disorders (JPND2019-466-236). HZ is a Wallenberg Scholar supported by grants from the Swedish Research Council (#2018-02532), the European Research Council (#681712), the Swedish state under the agreement between the Swedish government and the County Councils, the ALF-agreement (#ALFGBG-720931), the Alzheimer Drug Discovery Foundation (ADDF), USA (#201809-2016862), and the UK Dementia Research Institute at UCL.

Conflicts of Interest: JLM has served/serves as a consultant or at advisory boards for the following for-profit companies, or has given lectures in symposia sponsored by the following for-profit companies: Roche Diagnostics, Genentech, Novartis, Lundbeck, Oryzon, Biogen, Lilly, Janssen, Green Valley, MSD, Eisai, Alector, BioCross, GE Healthcare, ProMIS Neurosciences. JDG has given lectures in symposia sponsored by the following for-profit companies: General Electric, Philips and Biogen. KB has served as a consultant or at advisory boards for Abcam, Axon, Biogen, Lilly, MagQu, Novartis and Roche Diagnostics, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB (BBS), which is a part of the GU Ventures Incubator Program. HZ has served at scientific advisory boards for Denali, Roche Diagnostics, Wave, Samumed and CogRx, has given lectures in symposia sponsored by Fujirebio, Alzecure and Biogen, and is a co-founder of Brain Biomarker Solutions in Gothenburg AB, a GU Ventures-based platform company at the University of Gothenburg. GK is a full-time employee of Roche Diagnostics GmbH. MS is a full-time employee of Roche Diagnostics International Ltd. The remaining authors declare that they have no conflict of interest.

Ethical Standards: The ALFA+ study (ALFA-FPM-0311) was approved by the Independent Ethics Committee “Parc de Salut Mar,” Barcelona, and registered at Clinicaltrials.gov (Identifier: NCT02485730). All participants signed the study’s informed consent form that had also been approved by the Independent Ethics Committee “Parc de Salut Mar,” Barcelona.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

References

1. Dubois B, Hampel H, Feldman HH, Scheltens P, Aisen P, Andrieu S, et al. Preclinical Alzheimer’s disease: Definition, natural history, and diagnostic criteria. Alzheimer’s Dement. 2016 Mar;12(3):292–323.
2. Jack CR, Bennett DA, Blennow K, Carrillo MC, Dunn B, Haeberlein SB, et al. NIA-AA Research Framework: Toward a biological definition of Alzheimer’s disease. Alzheimer’s Dement. 2018;14(4):535–62.
3. Burnham SC, Coloma PM, Li Q-X, Collins S, Savage G, Laws S, et al. Application of the NIA-AA Research Framework: Towards a Biological Definition of Alzheimer’s Disease Using Cerebrospinal Fluid Biomarkers in the AIBL Study. J Prev Alzheimer’s Dis. 2019;6(4):248–55.
4. Milà-Alomà M, Suárez-Calvet M, Molinuevo JL. Latest advances in cerebrospinal fluid and blood biomarkers of Alzheimer’s disease. Ther Adv Neurol Disord. 2019 Jan 18;12:175628641988881.
5. Cavedo E, Lista S, Khachaturian Z, Aisen P, Amouyel P, Herholz K, et al. The Road Ahead to Cure Alzheimer’s Disease: Development of Biological Markers and Neuroimaging Methods for Prevention Trials Across all Stages and Target Populations. J Prev Alzheimer’s Dis. 2014 Dec;1(3):181–202.
6. Jack CR, Wiste HJ, Weigand SD, Therneau TM, Lowe VJ, Knopman DS, et al. Defining imaging biomarker cut points for brain aging and Alzheimer’s disease. Alzheimer’s Dement. 2017 Mar 1;13(3):205–16.
7. Jack CR, Bennett DA, Blennow K, Carrillo MC, Feldman HH, Frisoni GB, et al. A / T / N : An unbiased descriptive classification scheme for Alzheimer disease biomarkers. Neurology. 2016 Aug;0(5):1–10.
8. Palmqvist S, Zetterberg H, Blennow K, Vestberg S, Andreasson U, Brooks DJ, et al. Accuracy of Brain Amyloid Detection in Clinical Practice Using Cerebrospinal Fluid β-Amyloid 42. JAMA Neurol. 2014 Oct 1;71(10):1282.
9. Landau SM, Lu M, Joshi AD, Pontecorvo M, Mintun MA, Trojanowski JQ, et al. Comparing positron emission tomography imaging and cerebrospinal fluid measurements of β-amyloid. Ann Neurol. 2013 Dec;74(6):826–36.
10. Schindler SE, Gray JD, Gordon BA, Xiong C, Batrla-Utermann R, Quan M, et al. Cerebrospinal fluid biomarkers measured by Elecsys assays compared to amyloid imaging. Alzheimer’s Dement. 2018 Nov;14(11):1460–9.
11. Hansson O, Seibyl J, Stomrud E, Zetterberg H, Trojanowski JQ, Bittner T, et al. CSF biomarkers of Alzheimer’s disease concord with amyloid-β PET and predict clinical progression: A study of fully automated immunoassays in BioFINDER and ADNI cohorts. Alzheimer’s Dement. 2018 Nov;14(11):1470–81.
12. Vanderstichele HMJ, Janelidze S, Demeyer L, Coart E, Stoops E, Herbst V, et al. Optimized Standard Operating Procedures for the Analysis of Cerebrospinal Fluid Aβ42 and the Ratios of Aβ Isoforms Using Low Protein Binding Tubes. Galimberti D, editor. J Alzheimer’s Dis. 2016 Aug 3;53(3):1121–32.
13. Janelidze S, Zetterberg H, Mattsson N, Palmqvist S, Vanderstichele H, Lindberg O, et al. CSF A β 42/A β 40 and A β 42/A β 38 ratios: better diagnostic markers of Alzheimer disease. Ann Clin Transl Neurol. 2016 Mar;3(3):154–65.
14. Hansson O, Lehmann S, Otto M, Zetterberg H, Lewczuk P. Advantages and disadvantages of the use of the CSF Amyloid β (Aβ) 42/40 ratio in the diagnosis of Alzheimer’s Disease. Alzheimers Res Ther. 2019 Dec 22;11(1):34.
15. Thurfjell L, Lilja J, Lundqvist R, Buckley C, Smith A, Vandenberghe R, et al. Automated quantification of 18F-flutemetamol PET activity for categorizing scans as negative or positive for brain amyloid: Concordance with visual image reads. J Nucl Med. 2014 Oct 1;55(10):1623–8.
16. Cohen AD, Mowrey W, Weissfeld LA, Aizenstein HJ, McDade E, Mountz JM, et al. Classification of amyloid-positivity in controls: Comparison of visual read and quantitative approaches. Neuroimage. 2013 May 1;71:207–15.
17. Schreiber S, Landau SM, Fero A, Schreiber F, Jagust WJ. Comparison of visual and quantitative florbetapir F 18 positron emission tomography analysis in predicting mild cognitive impairment outcomes. JAMA Neurol. 2015 Oct 1;72(10):1183–90.
18. Klunk WE, Koeppe RA, Price JC, Benzinger TL, Devous MD, Jagust WJ, et al. The Centiloid project: Standardizing quantitative amyloid plaque estimation by PET. Alzheimer’s Dement. 2015 Jan;11(1):1-15.e4.
19. La Joie R, Ayakta N, Seeley WW, Borys E, Boxer AL, DeCarli C, et al. Multisite study of the relationships between antemortem [ 11 C]PIB-PET Centiloid values and postmortem measures of Alzheimer’s disease neuropathology. Alzheimer’s Dement. 2019 Feb 1;15(2):205–16.
20. Salvadó G, Molinuevo JL, Brugulat-Serrat A, Falcon C, Grau-Rivera O, Suárez-Calvet M, et al. Centiloid cut-off values for optimal agreement between PET and CSF core AD biomarkers. Alzheimers Res Ther. 2019 Dec 21;11(1):27.
21. Aisen PS, Cummings J, Doody R, Kramer L, Salloway S, Selkoe DJ, et al. The Future of Anti-Amyloid Trials. J Prev Alzheimer’s Dis. 2020 Mar 1;7(3):146–51.
22. Blennow K, Zetterberg H. Amyloid and Tau Biomarkers in CSF. J Prev Alzheimer’s Dis. 2015;2(1):46–50.
23. Molinuevo JL, Gramunt N, Gispert JD, Fauria K, Esteller M, Minguillon C, et al. The ALFA project: A research platform to identify early pathophysiological features of Alzheimer’s disease. Alzheimer’s Dement (New York, N Y). 2016 Jun;2(2):82–92.
24. Teunissen CE, Tumani H, Engelborghs S, Mollenhauer B. Biobanking of CSF: International standardization to optimize biomarker development. Clin Biochem. 2014 Mar;47(4–5):288–92.
25. Amadoru S, Doré V, McLean CA, Hinton F, Shepherd CE, Halliday GM, et al. Comparison of amyloid PET measured in Centiloid units with neuropathological findings in Alzheimer’s disease. Alzheimers Res Ther. 2020 Dec 4;12(1):22.
26. Leuzy A, Chiotis K, Hasselbalch SG, Rinne JO, de Mendonça A, Otto M, et al. Pittsburgh compound B imaging and cerebrospinal fluid amyloid-β in a multicentre European memory clinic study. Brain. 2016 Sep;139(9):2540–53.
27. Rowe CC, Doré V, Jones G, Baxendale D, Mulligan RS, Bullich S, et al. 18F-Florbetaben PET beta-amyloid binding expressed in Centiloids. Eur J Nucl Med Mol Imaging. 2017 Nov 1;44(12):2053–9.
28. Benjamini, Yoav ; Hochberg Y. Controlling the False Discovery Rate – a Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society Series B-Methodological 1995.pdf. J R Stat Soc Ser B. 1995;57(1):289–300.
29. Bertens D, Tijms BM, Scheltens P, Teunissen CE, Visser PJ. Unbiased estimates of cerebrospinal fluid β-amyloid 1-42 cutoffs in a large memory clinic population. Alzheimer’s Res Ther. 2017 Feb 14;9(1):8.
30. Willemse EAJ, van Maurik IS, Tijms BM, Bouwman FH, Franke A, Hubeek I, et al. Diagnostic performance of Elecsys immunoassays for cerebrospinal fluid Alzheimer’s disease biomarkers in a nonacademic, multicenter memory clinic cohort: The ABIDE project. Alzheimer’s Dement Diagnosis, Assess Dis Monit. 2018 Jan 1;10:563–72.
31. De Meyer G, Shapiro F, Vanderstichele H, Vanmechelen E, Engelborghs S, De Deyn PP, et al. Diagnosis-independent Alzheimer disease biomarker signature in cognitively normal elderly people. Arch Neurol. 2010 Aug;67(8):949–56.
32. Palmqvist S, Mattsson N, Hansson O. Cerebrospinal fluid analysis detects cerebral amyloid-β accumulation earlier than positron emission tomography. Brain. 2016 Apr;139(4):1226–36.

CROSS-SECTIONAL CHARACTERIZATION OF ALBUMIN GLYCATION STATE IN CEREBROSPINAL FLUID AND PLASMA FROM ALZHEIMER’S DISEASE PATIENTS

 

M. Costa1, A. Mestre1, R. Horrillo1, A. M. Ortiz1, A. Pérez1, A. Ruiz2, M. Boada2,3, S. Grancha1

 

1. Grifols Bioscience Research Group, Grifols, Barcelona, Spain; 2. Research Center and Memory Clinic, Fundació ACE, Institut Català de Neurociències Aplicades – Universitat Internacional de Catalunya, Barcelona, Spain; 3. Department of Neurology, Hospital General Universitari Vall d’Hebron, Barcelona, Spain

Corresponding Author: Dr. Montserrat Costa, Grifols, Research & Development Area, Carrer Can Guasch, 2, 08150 Parets del Vallès, Barcelona, Spain, Tel: +34 935 710 853; Fax: +34 935 710 381, E-mail: montse.costa@grifols.com

J Prev Alz Dis
Published online December 28, 2018, http://dx.doi.org/10.14283/jpad.2018.48

 


Abstract

We determined albumin post-translational modifications (PTMs) by mass spectrometry (MS) in plasma and cerebrospinal fluid (CSF) from 31 Alzheimer’s disease (AD) patients (with 27 samples of paired plasma-CSF from the same patients). Results were cross-sectionally compared with healthy controls. For percentage of relative intensity of glycated isoforms, plasma albumin was globally more glycated in AD patients than in healthy controls (P<0.01). MS results in plasma were confirmed by a quantitative enzymatic assay (Lucica GA-L) for albumin early-glycation detection. In CSF there were no global glycation differences detected by MS, although a different pattern of glycated isoforms was observed. Oxidized+glycated and cysteinylated+glycated isoforms were increased in both plasma and CSF of AD patients in comparison with healthy controls (P<0.001). Furthermore, AD patients showed higher glycation in plasma than in CSF (P<0.01). Our data support the role of glycation and oxidative stress in AD.

Key words: Albumin, glycation status, cerebrospinal fluid, plasma; Alzheimer’s disease.


 

 

Introduction

Albumin is the most abundant protein in the human body. In addition to its well-known role as a plasma expander, albumin has a number of relevant physiologic functions such as being the main extracellular antioxidant and transporter, and exerting metal-binding and scavenger activities (1). In albumin, glycation-induced conformational changes can have a deleterious effect in both its binding capacity (2-5) and antioxidant capacity (6-8). Elevated levels of glycated albumin are also associated with aging (9) and related conditions such as retinopathy, nephropathy, neuropathy, cardiovascular diseases and Alzheimer’s disease (AD) (4).
Serum albumin is thought to inhibit neuritic plaque formation tissue through regulating β-amyloid protein (Aβ) fibril growth in the brain (10), with glycated albumin being less effective than native albumin in preventing Aβ aggregation (11). AD patients have been shown to have elevated levels of glycated proteins in serum (12, 13), plasma (14) and cerebrospinal fluid (CSF) (15, 16). More specifically, glycated albumin has been found increased in AD plasma and serum (11, 16), but no consistent results have been obtained in AD CSF (11, 16). Although a paired comparison of protein glycation in plasma and CSF from the same AD patient has already been assessed (17), the same approach for albumin glycation has not been studied. In addition, to the best of our knowledge, a quantitative approach to specifically assess early-glycated albumin in AD plasma has not been addressed, despite the fact that the early glycation process has been reported as playing a role in AD (16).
In a previous paper we reported that albumin of AD patients was significantly more oxidized than in healthy subjects, with this effect more marked in CSF than in plasma (18). These data supported the involvement of oxidative stress in AD, thus pointing out the need for further research to better understand the role of albumin in AD. The aim of this cross-sectional study was to extend the same approach to albumin glycation in plasma and CSF from AD patients, which was also characterized in comparison with age-matched healthy subjects.

 

Methods

Experimental design

Determination of albumin post-translational modifications (PTMs) for glycation in terms of their identification and relative quantification was performed by means of mass spectrometry (MS) technique. Albumin glycation levels were assessed in plasma (n=28) and CSF (n=30) samples from 31 patients with mild-moderate AD (there were 27  plasma and CSF samples drawn from the same individuals). Results were compared with albumin glycation levels assessed in plasma (n=20) and CSF (n=10) samples from healthy −not diagnosed with AD− age-matched subject controls. Additionally, quantification of early-glycated albumin in plasma (n=34 for AD patients; n=31 for controls) was measured by quantitative enzymatic assay (Lucica GA-L). Complementarily, the paired plasma versus CSF comparison of albumin glycation was performed in the 27 samples available from the same AD patients.

Sampling

Plasma and CSF samples taken from mild-moderate AD patients (53 to 78 years of age) meeting NINCDS-ADRDA criteria for probable AD, and Mini-Mental State Examination (MMSE) scores between ≥18 and ≤26 were baseline specimens provided by the centers participating in the clinical trial EudraCT #2007-000414-36. Detailed patient population demographics, clinical profiles and inclusion/exclusion criteria are available in previous publications (19, 20).
Samples from control patients (48 to 81 years of age for plasma and 65 to 72 years of age for CSF) were kindly supplied by ACE Foundation, Catalan Institute of Applied Neurosciences (Barcelona, Spain) from their sample repository and purchased from Sera Laboratories International Ltd (West Sussex, UK), respectively.
All donors gave their informed consent in compliance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) for studies involving humans.
Blood sampling (plasma) and lumbar puncture (CSF) were carried out following the standard techniques of each center (19). Albumin concentration in plasma and CSF from both AD patients and controls was similar (18).

Liquid chromatography-Mass spectrometry

PTMs of albumin were assessed on the intact mass protein (top-down) by using ultra high-performance liquid chromatography coupled to electrospray ionization mass spectrometer (LC-ESI-qTOF-MS) (21).
For plasma and CSF, albumin-enriched samples were obtained, separated by liquid chromatography and subsequently analyzed by MS on a Quadrupole Time of Flight analyzer with electrospray ionization source (ESI) (Agilent qTOF 6550 Jet Stream, Agilent Technologies, Santa Clara CA, USA). All procedures and mass spectrometry data analysis have been described in a previous publication (18). Albumin controls were processed in parallel to albumin-enriched samples to ensure sample stability during analysis.
Isoform relative intensity (%) was calculated by dividing the isoform absolute intensity by the sum of all isoform absolute intensities, expressed as a percentage. For every comparison, a filter of frequency was applied to consider only those isoforms that were present in all analyzed samples of at least one of the compared groups.
The identification of albumin PTMs was performed as previously described (22-24) by analyzing 60,000-75,000 Da mass range and comparing the theoretical mass with the experimental value, allowing a mass drift of 10 Da (Agilent BioConfirm software).

Quantitative Enzymatic assay (Lucica GA-L)

Early-albumin glycation (fructosyl groups bound to albumin) in plasma samples was determined by an in vitro diagnostic assay, Lucica® GA-L (25) (Asahi Kasei Pharma Corp., Tokio, Japan). This kit includes the measurement of glycated albumin by an enzymatic method using an albumin-specific protease and a ketoamine oxidase, and the measurement of total albumin using bromocresol purple reagent. Glycated albumin results were expressed in a percentage relative to the total albumin and corrected to match HPLC results: Glycated albumin (%)=(glycated albumin[g/dL] / total albumin [g/dL])/1.14×100+2.9, following the instructions provided by the kit manufacturer.  CSF analysis was not performed because the method’s sensitivity is insufficient for low albumin concentration samples, such as those typical in CSF.

Statistical analysis

Variables are expressed as the median and interquartile range (IQR). Unpaired and paired Student’s t test and Mann-Whitney tests were used for the comparisons. Benjamini-Hochberg’s correction for multiple comparisons was applied when comparing differences in albumin PTMs’ relative abundance between groups (26).
To examine the magnitude of the differences in glycated albumin, the effect sizes based on standardized differences between mean scores were calculated (Cohen’s delta) (27). Effect sizes of 0.20, 0.50 and 0.80 were considered small, medium and large, respectively (27).
Software used for charting and calculations was Graph-Pad Prism v6 (San Diego CA, USA) and GeneSpring 13.1 (Agilent Technologies, Santa Clara CA, USA).

 

Results

Albumin glycation in plasma and CSF: AD patients versus controls

Seventeen albumin PTMs were identified by intact mass protein analysis. In plasma, seven of these 17 PTMs corresponded to glycated PTMs: glycated (Alb+Glyc); oxidized+glycated (Alb+SO3H+Glyc); cysteinylated+glycated+dehydroalanine (Alb+Cys+Glyc-DHA); cysteinylated+glycated (Alb+Cys+Glyc); two glycation modifications (Alb+2Glyc); cysteinylated+two glycation modifications (Alb+Cys+2Glyc); and three glycation modifications (Alb+3Glyc). In CSF, four of these 17 albumin PTMs were identified as glycated PTMs: Alb+Glyc; Alb+SO3H+Glyc; Alb+Cys+Glyc; and Alb+2Glyc.
MS values for percentage of relative intensity of albumin glycation in plasma showed that AD patients had higher amount of glycated albumin isoforms than controls when considering the sum of the seven glycated PTMs (P<0.01, Cohen’s delta: 1.0) (Figure 1a). Separately analyzed, four out of seven glycated PTMs showed a statistically significant increase in AD versus controls: oxidized (Alb+SO3H+Glyc [P<0.001; Cohen’s delta: 1.2]), and cysteinylated (Alb+Cys+Glyc-DHA [P<0.0001; Cohen’s delta: 1.0], Alb+Cys+Glyc [P<0.0001; Cohen’s delta: 1.7], and Alb+Cys+2Glyc [P<0.001; Cohen’s delta: 1.0]) (Figure 1b).

Figure 1. Albumin glycation in plasma as isoform relative intensity (%) identified in Alzheimer’s disease (AD) patients and healthy controls. A) Sum of all 7 glycated post-translational modifications (PTMs). B) The 7 glycated PTMs separately (glycated [Alb+Glyc]; oxidized+glycated [Alb+SO3H+Glyc]; cysteinylated+glycated+dehydroalanine [Alb+Cys+Glyc-DHA]; cysteinylated+glycated [Alb+Cys+Glyc]; two glycation modifications [Alb+2Glyc]; cysteinylated+two glycation modifications [Alb+Cys+2Glyc]; three glycation modifications [Alb+3Glyc])

Figure 1. Albumin glycation in plasma as isoform relative intensity (%) identified in Alzheimer’s disease (AD) patients and healthy controls. A) Sum of all 7 glycated post-translational modifications (PTMs). B) The 7 glycated PTMs separately (glycated [Alb+Glyc]; oxidized+glycated [Alb+SO3H+Glyc]; cysteinylated+glycated+dehydroalanine [Alb+Cys+Glyc-DHA]; cysteinylated+glycated [Alb+Cys+Glyc]; two glycation modifications [Alb+2Glyc]; cysteinylated+two glycation modifications [Alb+Cys+2Glyc]; three glycation modifications [Alb+3Glyc])

Data are shown as median and IQR (**, p<0.01; ***, p<0.001; ****, p<0.0001; AD versus control).

 

MS results in plasma were confirmed by the quantitative enzymatic assay (Lucica GA-L) for albumin early-glycation detection. The percentage of glycated plasma albumin in AD patients was 12.9% (12.3%-14.9%) versus 11.6% (10.5%-13.4%) in controls (median–IQR; P<0.001; Cohen’s delta: 0.7).
MS values of albumin glycation in CSF showed no differences in percentage of relative intensity when comparing the sum of the four glycated PTMs in AD patients versus controls (Figure 2a). However, if separately analyzed, three out of four glycated PTMs showed statistically significant differences in AD versus controls: a decrease in Alb+Glyc isoform (P<0.0001; Cohen’s delta: 1.7) and an increase in the oxidized and cysteinylated forms (Alb+SO3H+Glyc [P<0.0001: Cohen’s delta: 3.0] and Alb+Cys+Glyc [P<0.0001: Cohen’s delta: 1.8]) (Figure 2b).

Figure 2. Albumin glycation in cerebrospinal fluid (CSF) as isoform relative intensity (%) identified in Alzheimer’s disease (AD) patients and healthy controls. A) Sum of 4 glycated PTMs. B) The 4 glycated PTMs separately (glycated [Alb+Glyc]; oxidized+glycated [Alb+SO3H+Glyc]; cysteinylated+glycated [Alb+Cys+Glyc]; two glycation modifications [Alb+2Glyc])

Figure 2. Albumin glycation in cerebrospinal fluid (CSF) as isoform relative intensity (%) identified in Alzheimer’s disease (AD) patients and healthy controls. A) Sum of 4 glycated PTMs. B) The 4 glycated PTMs separately (glycated [Alb+Glyc]; oxidized+glycated [Alb+SO3H+Glyc]; cysteinylated+glycated [Alb+Cys+Glyc]; two glycation modifications [Alb+2Glyc])

Data are shown as median and IQR (****, p<0.0001; AD versus control).

 

Albumin glycation in AD patients: plasma versus CSF

When comparing albumin glycated PTMs in the subset of 27 AD samples consisting of paired plasma and CSF specimens both obtained from the same patient, plasma albumin was found significantly more glycated than in their paired CSF samples (Table 1). Moreover, the specific analysis showed that six out of seven glycated PTMs of albumin were statistically significantly different between compartments. The most relevant ones were Alb+Glyc and Alb+Cys+Glyc in plasma, and Alb+Glyc and Alb+SO3H+Glyc in CSF. Details are shown in Table 1.

Table 1. Albumin glycation in paired samples of plasma and cerebrospinal fluid (CSF) of Alzheimer’s disease (AD) patients (n=27). Compared relative intensity (%) of 7 individual glycated post-translational modifications (PTMs) (glycated [Alb+Glyc]; oxidized+glycated [Alb+SO3H+Glyc]; cysteinylated+glycated+dehydroalanine [Alb+Cys+Glyc-DHA]; cysteinylated+glycated [Alb+Cys+Glyc]; two glycation modifications [Alb+2Glyc]; cysteinylated+two glycation modifications [Alb+Cys+2Glyc]; three glycation modifications [Alb+3Glyc]); and the sum of all PTMs are shown (median and IQR)

Table 1. Albumin glycation in paired samples of plasma and cerebrospinal fluid (CSF) of Alzheimer’s disease (AD) patients (n=27). Compared relative intensity (%) of 7 individual glycated post-translational modifications (PTMs) (glycated [Alb+Glyc]; oxidized+glycated [Alb+SO3H+Glyc]; cysteinylated+glycated+dehydroalanine [Alb+Cys+Glyc-DHA]; cysteinylated+glycated [Alb+Cys+Glyc]; two glycation modifications [Alb+2Glyc]; cysteinylated+two glycation modifications [Alb+Cys+2Glyc]; three glycation modifications [Alb+3Glyc]); and the sum of all PTMs are shown (median and IQR)

Discussion

In this study we observed that plasma albumin was overall more glycated in AD patients than in healthy age-matched controls, when analyzed as the percentage of relative intensity of albumin glycation isoforms assessed by MS. Conversely, in CSF albumin, there were no global glycation differences, although a different pattern of glycation isoforms was observed. Moreover, AD patients had their albumin significantly more glycated in plasma than in CSF.
In plasma samples from AD patients, when considering all the glycated albumin PTMs separately, statistical significance versus controls was shown with the oxidized+glycated albumin forms and especially the cysteinylated+glycated albumin forms, rather than with the purely glycated albumin. Since cysteinylated forms are considered to be reversibly oxidized albumin modifications (1), these findings would support the higher content of reversibly oxidized albumin (HNA1) detected by HPLC-FLD analysis in AD plasma samples compared to control plasma (18). Interestingly, cysteinylated albumin has been described as a sensitive plasma marker in oxidative stress-related chronic diseases (28).
In CSF, overall albumin glycation levels were similar between AD patients and controls but, interestingly, the glycation pattern was different. AD patients showed lower content of glycated albumin but increased content of both the oxidized+glycated and cysteinylated+glycated albumin forms. Again, since cysteinylated forms are considered to be reversibly oxidized albumin modifications while SO3H forms are considered to be irreversibly oxidized albumin modifications (2), these findings are in accordance to a higher content in HNA1 and irreversibly oxidized albumin (HNA2) observed by HPLC-FLD analysis in AD CSF (18).
Furthermore, MS results obtained in plasma from AD patients and healthy controls were confirmed by a quantitative enzymatic assay for early glycation detection of albumin (Lucica GA-L).
Albumin glycation in AD patients has been previously addressed by a few other authors. Increased levels of early glycation products in all major proteins of CSF of AD patients, including albumin, apolipoprotein E and transthyretin, have been reported (16). Furthermore, targeted assessment of advanced glycation end products in albumin has been addressed in plasma (11) and serum (13), showing higher levels than in a control population, while presenting no differences in CSF (11). Our findings in CSF albumin glycation are in accordance with those obtained by Ramos-Fernández et al. (11). Differences in the AD study population, as well as in the experimental approach, might explain the lack of consensus with all of the investigations.
Based on our literature search, albumin glycation in paired plasma and CSF samples taken from the same patient has not been previously reported. Our results showed higher overall albumin glycation in plasma compared to CSF, with both compartments being different in terms of PTMs relative quantification. Besides the glycated albumin form, in plasma the reversibly oxidized cysteinylated+glycated albumin was relevant, while in CSF the irreversibly oxidized SO3H+glycated isoform gained predominance. These results suggest a restriction of albumin communication in both compartments, which would present different environments in terms of glycation and oxidation.
To summarize, accepting the limitations of this study, such as the sample size and the use of MS as a semi-quantitative method to analyze albumin glycation, the results presented here suggest that in both plasma and CSF samples taken from mild to moderate AD patients there is a marked increase in the cysteinylated+glycated and oxidated+glycated albumin forms, compared to healthy age-matched controls. Furthermore, AD patients showed higher glycation in plasma than in CSF. These data support the role of glycation and oxidative stress in AD and deserve further investigation.

 

Acknowledgements: We thank Pol Herrero (Proteomics facility of the Centre for Omic Sciences [COS] Joint Unit of the Universitat Rovira i Virgili-Eurecat) for his contribution to the mass spectrometry analysis. Cristina Aparicio, Francisca Doncel, Santiago Garcia, Carlota Gelabert, Aida Raventós, Ana María Siles, Jordi Vidal, and Eva Vior (Grifols) are acknowledged for their expert technical assistance. Jordi Bozzo PhD, CMPP (Grifols) is acknowledged for medical writing and editorial support in the preparation of this manuscript.

Conflict of interest: MC, AM, RH, AMO, AP and SG are employees of Grifols, a manufacturer of plasma derivatives. MB has consulted for Araclon, Avid, Grifols, Lilly, Nutricia, Roche and Servier. She received fees for lectures, and/or reimbursement of expenses for congresses attendance, and/or funds for research from Araclon, Grifols, Nutricia, Roche and Servier. She has not received personal compensations from these organizations. AR has consulted for Grifols. He is a member of the scientific advisory board of Landsteiner Genmed. He received funds for research from Araclon, Grifols, Nutricia, Roche and Servier. He received reimbursement of expenses for congresses attendance from Araclon, Grifols, and Landsteiner Genmed. He has not received personal compensations from these organizations..

Ethical standards: All donors gave their informed consent in compliance with the Code of Ethics of the World Medical Association (Declaration of Helsinki) for studies involving humans.

Open Access: This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits use, duplication, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license and indicate if changes were made.

 

References

1.    Colombo G, Clerici M, Giustarini D, Rossi R, Milzani A, Dalle-Donne I.  Redox albuminomics: oxidized albumin in human diseases. Antioxid Redox Signal 2012;17, 1515-1527.
2.    Anguizola J, Matsuda R, Barnaby OS, Hoy KS, Wa C, DeBolt E, Koke M, Hage DS. Review: Glycation of human serum albumin. Clinica Chimica Acta 2013;425, 64-76.
3.    Dozio E, Di Gaetano N, Findeisen P, Corsi Romanelli MM. Glycated albumin: from biochemistry and laboratory medicine to clinical practice. Endocrine 2016;55, 682-690.
4.    Rondeau P, Bourdon E.  The glycation of albumin: Structural and functional impacts. Biochimie 2011;93, 645-658.
5.    Vetter SW. in Advances in Clinical Chemistry Elsevier, 2015; pp. 205-275.
6.    Baraka-Vidot J, Guerin-Dubourg A, Dubois F, Payet B, Bourdon E, Rondeau P. New insights into deleterious impacts of in vivo glycation on albumin antioxidant activities. Biochim Biophys Acta 2013;1830, 3532-3541.
7.    Khan MS, Tabrez S, Rabbani N, Shah A. Oxidative Stress Mediated Cytotoxicity of Glycated Albumin: Comparative Analysis of Glycation by Glucose Metabolites. J Fluorescence 2015;25, 1721-1726.
8.    Rondeau P, Singh NR, Caillens H, Tallet F, Bourdon E. Oxidative stresses induced by glycoxidized human or bovine serum albumin on human monocytes. Free Radic Biol Med 2008;45, 799-812.
9.    Arasteh A, Farahi S, Habibi-Rezaei M, Moosavi-Movahedi A. Glycated albumin: an overview of the In Vitro models of an In Vivo potential disease marker. J Diabetes Metab Disord 2014;13, 49.
10.    Stanyon HF, Viles JH. Human Serum Albumin Can Regulate Amyloid-β Peptide Fiber Growth in the Brain Interstitium. J Biol Chem 2012;287, 28163-28168.
11.    Ramos-Fernández E, Tajes M, Palomer E, Ill-Raga G, Bosch-Morató M, Guivernau B, Román-Dégano I, Eraso-Pichot A, Alcolea D, Fortea J, Nuñez L, Paez A, Alameda F, Fernández-Busquets X, Lleó A, Elosúa R, Boada M, Valverde MA, Muñoz FJ. Posttranslational Nitro-Glycative Modifications of Albumin in Alzheimer’s Disease: Implications in Cytotoxicity and Amyloid-β Peptide Aggregation. J Alzheimers Dis 2014;40, 643-657.
12.    Lin CY, Sheu JJ, Tsai IS, Wang ST, Yang LY, Hsu IU, Chang HW, Lee HM, Kao SH, Lee CK, Chen CH, Lin YF. Elevated IgM against Nepsilon-(Carboxyethyl)lysine-modified Apolipoprotein A1 peptide 141-147 in Taiwanese with Alzheimer’s disease. Clin Biochem 2018;56, 75-82.
13.    Meli M, Perier C, Ferron C, Parssegny F, Denis C, Gonthier R, Laurent B, Reynaud E, Frey J, Chamson A. Serum pentosidine as an indicator of Alzheimer’s disease. J Alzheimers Dis 2002;4, 93-96.
14.    Riviere S, Birlouez-Aragon I, Vellas B. Plasma protein glycation in Alzheimer’s disease. Glycoconj J 1998;15, 1039-1042.
15.    Ahmed N, Ahmed U, Thornalley PJ, Hager K, Fleischer G, Munch G. Protein glycation, oxidation and nitration adduct residues and free adducts of cerebrospinal fluid in Alzheimer’s disease and link to cognitive impairment. J Neurochem 2005;92, 255-263.
16.    Shuvaev VV, Laffont I, Serot J-M, Fujii J, Taniguchi N, Siest G. Increased protein glycation in cerebrospinal fluid of Alzheimer’s disease. Neurobiol Aging 2001;22, 397-402.
17.    Bar KJ, Franke S, Wenda B, Muller S, Kientsch-Engel R, Stein G, Sauer H. Pentosidine and N(epsilon)-(carboxymethyl)-lysine in Alzheimer’s disease and vascular dementia. Neurobiol Aging 2003;24, 333-338.
18.    Costa M, Horrillo R, Ortiz AM, Perez A, Mestre A, Ruiz A, Boada M, Grancha S. Increased Albumin Oxidation in Cerebrospinal Fluid and Plasma from Alzheimer’s Disease Patients. J Alzheimers Dis 2018;63, 1395-1404.
19.    Boada M, Anaya F, Ortiz P, Olazaran J, Shua-Haim JR, Obisesan TO, Hernandez I, Munoz J, Buendia M, Alegret M, Lafuente A, Tarraga L, Nunez L, Torres M, Grifols JR, Ferrer I, Lopez OL, Paez A. Efficacy and Safety of Plasma Exchange with 5% Albumin to Modify Cerebrospinal Fluid and Plasma Amyloid-beta Concentrations and Cognition Outcomes in Alzheimer’s Disease Patients: A Multicenter, Randomized, Controlled Clinical Trial. J Alzheimers Dis 2017;56, 129-143.
20.    Cuberas-Borrós G, Roca I, Boada M, Tárraga L, Hernández I, Buendia M, Rubio L, Torres G, Bittini A, Guzmán de Villoria J, Pujadas F, Torres M, Núñez L, Castell J, Páez A. Longitudinal neuroimaging analysis in mild-moderate Alzheimer’s disease patients treated with plasma exchange with 5% human albumin. J Alzheimers Dis 2018;61, 321-332.
21.    Kawakami A, Kubota K, Yamada N, Tagami U, Takehana K, Sonaka I, Suzuki E, Hirayama K. Identification and characterization of oxidized human serum albumin. A slight structural change impairs its ligand-binding and antioxidant functions. FEBS J 2006;273, 3346-3357.
22.    Naldi M, Giannone FA, Baldassarre M, Domenicali M, Caraceni P, Bernardi M, Bertucci C. A fast and validated mass spectrometry method for the evaluation of human serum albumin structural modifications in the clinical field. Eur J Mass Spectrom 2013;19, 491-496.
23.    Domenicali M, Baldassarre M, Giannone FA, Naldi M, Mastroroberto M, Biselli M, Laggetta M, Patrono D, Bertucci C, Bernardi M, Caraceni P. Posttranscriptional changes of serum albumin: clinical and prognostic significance in hospitalized patients with cirrhosis. Hepatology 2014;60, 1851-1860.
24.    Bar-Or R, Rael LT, Bar-Or D. Dehydroalanine derived from cysteine is a common post-translational modification in human serum albumin. Rapid Commun Mass Spectrom 2008;22, 711-716.
25.    Kohzuma T, Yamamoto T, Uematsu Y, Shihabi ZK, Freedman BI. Basic performance of an enzymatic method for glycated albumin and reference range determination. J Diabetes Sci Technol 2011;5, 1455-1462.
26.    Benjamini Y, Heller R. Screening for partial conjunction hypotheses. Biometrics 2008;64, 1215-1222.
27.    Cohen J. Statistical Power Analysis for the Behavioral Sciences, Hillsdale, N.J. : Lawrence Erlbaum Associates, 1988.
28.    Nagumo K, Tanaka M, Chuang VT, Setoyama H, Watanabe H, Yamada N, Kubota K, Tanaka M, Matsushita K, Yoshida A, Jinnouchi H, Anraku M, Kadowaki D, Ishima Y, Sasaki Y, Otagiri M, Maruyama T. Cys34-cysteinylated human serum albumin is a sensitive plasma marker in oxidative stress-related chronic diseases. PLoS One 2014;9, e85216.

AMYLOID AND TAU BIOMARKERS IN CSF

 

K. Blennow1, H. Zetterberg1,2

1. Clinical Neurochemistry Laboratory, Institute of Neuroscience and Physiology, the Mölndal Campus at the Sahlgrenska Academy at University of Gothenburg, SE-43180 Mölndal, Sweden; 2. UCL Institute of Neurology, Queen Square, London, UK

Corresponding Author: Kaj Blennow, MD, Ph.D. Clinical Neurochemistry Laboratory, Institute of Neuroscience and Physiology, The Sahlgrenska Academy at University of Gothenburg, Mölndal Campus, Sahlgrenska University Hospital, SE-431 80 Mölndal, Sweden, Tel:  + 46 31 3431791, Fax: + 43 31 3432426, E-mail: kaj.blennow@neuro.gu.se

 

J Prev Alz Dis 2015;2(1):46-50
Published online Januay 20, 2015, http://dx.doi.org/10.14283/jpad.2015.41


Abstract

The number of failed Alzheimer’s disease (AD) clinical trials on Aβ-targeting drugs is increasing. The explanation for this is most likely multi-factorial. An optimistic standpoint is that trials have to be on patients in an earlier stage of the disease, before neurodegeneration is too severe, to show efficacy, and probably also of longer duration. Further, there is a general agreement that enrolled patients have to be diagnosed based on combined clinical and biomarker criteria, to avoid noise from the large proportion (20%) of cases that are misdiagnosed if only clinical criteria are used. Last, the poor predictive power of translating an “anti-Aβ” or “anti-plaque” effect from AD transgenic animal models to AD patients also calls for biomarkers to verify target engagement in man, and to show downstream effects of Aβ-targeting drug candidates in AD patients. The focus of this review is on the possible role of cerebrospinal fluid (CSF) biomarkers in AD clinical trials for diagnostics, and thus patient enrichment, and for theragnostics, to provide evidence of target engagement of the drug on Aβ metabolism or aggregation, and of effects on the molecular pathology of the disease.

 

Key words: Alzheimer’s disease, Biomarker, β-amyloid (Aβ), Cerebrospinal fluid, Clinical trial, Diagnosis, Mild cognitive impairment (MCI), Phosphorylated tau, Plasma, tau protein, Theragnostic. 


 

Introduction

Recent failures of several late phase clinical trials on drug candidates targeting β-amyloid (Aβ) is causing concern among Alzheimer’s disease (AD) researchers as well as pharmaceutical companies, and other stakeholders. A positive and realistic standpoint is that we need to initiate treatment before the dementia stage of the disease, before neurodegeneration is too severe and plaque and tangle load to large. As an example, in the solanezumab clinical phase III trials, there was a trend for a treatment effect in patients with mild AD (1). Further, it has been known since long that the accuracy of the clinical diagnosis of AD based on purely clinical grounds is poor. In line with this, it was reported that approximately 25% of clinically diagnosed AD patients in the bapineuzumab phase III clinical trial had negative amyloid PET scans, i.e., were misdiagnosed as having AD (2). It is logical that inclusion of a large percentage of cases with other diseases than AD in an anti-Aβ trial will increase variation and noise and markedly affect the possibility to identify a beneficial clinical effect of a drug candidate..

The failures of the anti-Aβ trials have also caused distress that the amyloid cascade hypothesis will be falsified, i.e., that Aβ aggregation and deposition is not the primary cause, but merely a by-stander of the disease, or that the hypothesis may only be valid for the familial form of AD. An alternative explanation for at least some of the trial failures may be that poor drug candidates have been taken directly into phase II and III clinical trials, based only on promising, but misleading, results from preclinical studies performed in AD transgenic mice (3).

In this paper, we review the position of biomarkers in AD drug development and clinical trials as fundamental tools to enable selection of the optimal clinical cohorts, i.e., for diagnostics, and as theragnostic tools, to examine the pharmacodynamic properties of drug candidates, including identification of target engagement and monitoring downstream effects on neurodegeneration; an application we call theragnostics (4).

CSF Biomarkers in clinical trials

As reviewed elsewhere (5), there is a consensus among the academy, pharmaceutical industry and regulatory authorities that biomarkers may have several uses in clinical trials. Given the diagnostic challenges not only to identify prodromal AD cases in MCI cohorts, but also to make a diagnosis cases with mild AD dementia, an attractive application of CSF biomarkers is as diagnostic tools to enrich the trial cohort with pure AD cases. This procedure will increase the proportion of patients with Alzheimer pathology, thereby increasing the chance to identify a clinical effect of the drug. Regulatory authorities such as the European Medicines Agency (EMA) also recommend the use of CSF for enrichment of clinical trial populations with prodromal AD cases (6).

While pharmacokinetics, i.e., measurement of drug absorption, distribution, metabolism, and excretion is not the theme of this paper, some aspects of pharmacodynamics, intended biochemical drug effects in particular, will be reviewed. Theragnostic biomarker Phase I studies to verify appropriate target engagement in man may be important in for the decision whether to take the drug candidate into large and expensive Phase II or III clinical trials (Table 1). Further, Phase II biomarker studies providing evidence of downstream effects on the neurodegenerative process may in a similar way be important in the decision making on launching large multi-center Phase III trials (Table 1). Last, this type of biomarker-based evidence of an effect of the drug on neurodegeneration may also be important in Phase III registration trials to provide objective evidence of disease modification (5).

Enrichment of true AD cases 

For anti-Aβ trials, identification of cases with biomarker evidence of cortical amyloid pathology is a logical strategy. The reduction in CSF Aβ42 correlates inversely with cerebral plaque load at autopsy (7) and global cortical amyloid ligand binding as measured by positron emission tomography (PET) (8). Importantly, recent large studies have shown an agreement between CSF Aβ42, also measured in clinical routine, and amyloid PET measurements of cortical Aβ load with more than 90% concordance (9), and almost identical diagnostic performance (10). Further, regional assessments of amyloid load on PET scans provide no improvement in diagnostic performance as compared with mean global cortical binding or CSF Aβ42 (10). Thus, provided that harmonized cut-offs for both measurements can be identified, amyloid PET and CSF Aβ42 can be used interchangeably for enrichment in clinical trials.

Except for Aβ42, there is a wide range of other Aβ species in CSF, of which Aβ40 is the most abundant (11). Although there is no major change in CSF Aβ40 in AD, there is a marked decrease in the ratio of CSF Aβ42/Aβ40 in both AD dementia and MCI, which is more pronounced than the reduction in CSF Aβ42 alone (12, 13). The use of the CSF Aβ42/Aβ40 ratio may improve diagnostic performance as compared with Aβ42 by itself (14, 15). The hypothesis behind this is that the ratio compensates for high or low production of total Aβ (all isoforms), while single analysis of Aβ42 may give false positive cases among “low producers” and false negative cases among “high producers”.

It is likely that combination of CSF T-tau and P-tau adds to the diagnostic performance of Aβ42 to identify prodromal AD among MCI cases (16, 17). However, the major contribution of CSF T-tau and P-tau may be in predicting progression (Table 1). While amyloid biomarkers become positive 5-10 years before symptoms in the preclinical phase and before dementia in the MCI stage, addition of CSF T-tau and P-tau improves predictive power of progression during a clinically relevant time period (18, 19, 20). Indeed, the updated International Working Group (IWG) criteria for AD also recommend the algorithm low CSF Aβ42 together with high T-tau and/or P-tau (21). 

Biomarkers to verify target engagement

Testing new anti-Aβ drug candidates in AD transgenic mouse, using a reduction in plaque load as the indication to launch large Phase II trials on AD patients has been mainstream in AD drug development. As reviewed elsewhere (22), the very large number of compounds found to reduce Aβ pathology in these models but not in AD patients makes them poor predictors of treatment success in sporadic AD. Increasing attention is drawn to the low predictivity of transgenic animal models for success in later clinical trials, a problem that by no means is unique to AD drugs development (23). For this reason, it may be wise not to rely only on preclinical findings of target engagement and a plaque-lowering effect in AD transgenic mice for the decision to move into large and expensive phase II or III clinical trials, without any data in man speaking for appropriate target engagement.

Target engagement on Aβ metabolism or clearance may be identified in short-term proof-of-principle studies on a small number of healthy volunteers in phase I, or proof-of-concept studies on AD patients in phase IIa, the design of these depends on the type of anti-Aβ drug. As a reassuring example, solid target engagement data for a BACE1 inhibitor was obtained in a single-dose study on 30 healthy volunteers (24). The CSF biomarker response to Aβ immunotherapy is complex, and differ depending on whether active immunization or passive immunotherapy with anti-Aβ antibodies were used, for review see (25). Analysis of a change in CSF levels of Aβ oligomers have a potential for use as a pharmacodynamic biomarker to identify target engagement in immunotherapy trials, but due to the very low levels of Aβ oligomers in CSF, assays have been hampered by low analytical sensitivity. A recent paper reported on a novel immunoassay based on the ultrasensitive Erenna platform, that was highly selective for Aβ oligomers over monomers, and had a limit of detection far below standard ELISA methods (26). This type of technical developments brings hope that CSF Aβ oligomers levels can be applied in early trials to identify target engagement and monitor treatment effects on Aβ aggregation and plaques (Table 1).

This type of early clinical biomarker study would aid to select drug candidates with a proven effect on Aβ metabolism or clearance also in man, which would be of value in the decision making whether to embark on expensive phase II and III trials. Hopefully, such an approach will improve the success rate of future clinical trials. 

Table 1. Position of cerebrospinal fluid biomarkers in clinical trials – anti-Aβ trials as example

Abbreviations: Aβ, amyloid-β; AD, Alzheimer disease; BACE1, β-site APP cleaving enzyme 1; CSF, cerebrospinal fluid; H-FABP, Heart fatty acid-binding protein; P-tau, phosphorylated tau; sAPP, soluble amyloid precursor protein extracellular domain; T-tau, total tau; VLP-1, visinin-like protein-1.

Biomarkers to identify downstream effects on pathology

The term downstream biomarkers can be applied to biomarkers used to identify and monitor effects downstream of the primary target of the drug (3). For example, evidence from biomarkers reflecting the intensity of the neuronal degeneration (such as T-tau) or phosphorylation state of tau (P-tau) may provide valuable indications that an anti-Aβ drug candidate has effects downstream of Aβ production or aggregation (Table 1). An example was found in the bapineuzumab Phase II and III trials, in which a lowering of P-tau (2) or P-tau and t-tau (27) was found.

Since tau is involved in the pathogenic processes in AD, complementary data on downstream effects on neurodegeneration from other biomarkers may increase the validity of results (Table 1). Such CSF biomarkers may include the neuronal biomarkers heart-type fatty acid-binding protein (HFABP or FABP3) (28, 29) and visinin-like protein 1 (VLP-1 or VILIP-1) (30, 31). Both HFABP and VLP-1 show a clear increase in AD, also in the early stages of the disease, and correlate with CSF tau levels (28-31) and may thus provide valuable information on disease-modifying effects on novel drug candidates.

Synapses are the primary functional unit for neuronal communication, composed on a pre-synaptic unit containing synaptic vesicles filled with neurotransmitters that upon release regulated by a delicate machinery of presynaptic proteins and a post-synaptic unit with receptors and proteins involved in advancing the signal [32, 33]. The degree of synaptic degeneration show the best correlation with severity of cognitive symptoms in AD, for review see (34). Measurement of synaptic proteins in CSF may thus serve as valuable biomarkers for synaptic function and degeneration, which may correlate well with a change in cognitive function due to treatment. Although all of presynaptic vesicle proteins (such as synaptotagmin and rab3a), presynaptic membrane proteins such as SNAP-25, and dendritic proteins such as neurogranin have been shown to be present in human CSF (35, 36), assay development for synaptic proteins has proven difficult due to their low abundance. Nevertheless, a pilot study based on immunoprecipitation and Western blot showed a marked increase in CSF neurogranin in AD (37), a finding that recently has been confirmed using a novel ELISA method (38). Importantly, high CSF neurogranin levels predicted progression to AD dementia in mild cognitive impairment (MCI) patients, and within (amyloid positive) prodromal AD cases, high CSF neurogranin levels correlated with a more rapid deterioration in cognitive symptoms during clinical follow-up [38]. These findings suggest that synaptic biomarkers may be useful in the earlier phases of AD as biomarkers linked to cognition.

Importantly, biomarker-based evidence of both target engagement in man and an effect of the drug candidate on downstream AD-related molecular mechanisms and a general effect on neuronal and synaptic degeneration is likely to be important in Phase III registration trials, to provide objective measures of disease modification (5). CSF biomarkers may also qualify as surrogate markers if a change in CSF biomarkers with treatment, e.g. for synaptic proteins, can be shown to predict clinical outcome. This would be highly useful in future clinical trails on AD disease-modifying drugs, allowing shorter trial duration and a lower number of subjects.

References

1. Doody RS, Thomas RG, Farlow M, Iwatsubo T, Vellas B, Joffe S, et al. Phase 3 trials of solanezumab for mild-to-moderate Alzheimer’s disease. N Engl J Med 2014;370:311-321.

2. Salloway S, Sperling R, Fox NC, Blennow K, Klunk W, Raskind M, et al. Two Phase 3 Trials of Bapineuzumab in Mild-to-Moderate Alzheimer’s Disease. N Engl J Med 2014;370:322-33.

3. Blennow K. Biomarkers in Alzheimer’s disease drug development. Nature Med 2010;16:1218-1222.

4. Blennow K, Hampel H, Weiner M, Zetterberg H. Cerebrospinal fluid and plasma biomarkers in Alzheimer’s disease. Nat Rev Neurol 2010;6:131-144.

5. Hampel H, Frank R, Broich K, Teipel SJ, Katz RG, Hardy J, et al. Biomarkers for Alzheimer’s disease: academic, industry and regulatory perspectives. Nat Rev Drug Discov 2010;9:560-574.

6. Isaac M, Vamvakas S, Abadie E, Jonsson B, Gispen C, Pani L. Qualification opinion of novel methodologies in the predementia stage of Alzheimer’s disease: cerebrospinal fluid related biomarkers for drugs affecting amyloid burden – regulatory considerations by European Medicines Agency focusing in improving benefit/risk in regulatory trials. Eur Neuropsychopharmacol  2011;21:781-788.

7. Strozyk D, Blennow K, White LR, Launer LJ. CSF Aβ42 levels correlate with amyloid-neuropathology in a polpulation-based autopsy study. Neurology 2003;60:652-656.

8. Fagan AM, Mintun MA, Mach RH, Lee SY, Dence CS, Shah AR, et al. Inverse relation between in vivo amyloid imaging load and cerebrospinal fluid Abeta42 in humans. Ann Neurol 2006;59:512-519.

9. Palmqvist S, Zetterberg H, Blennow K, Vestberg S, Andreasson U, Brooks DJ, et al. Accuracy of brain amyloid detection in clinical practice using cerebrospinal fluid β-amylod 42: a cross-validation study against amyloid positron emission tomography. JAMA Neurol 2014, JAMA Neurol 2014;71:1282-1289.

10. Mattsson N, Insel PS, Landau S, Jagust W, Donohue M, Shaw LM, Trojanowski JQ, Zetterberg H, Blennow K, Weiner M, for the Alzheimer’s Disease Neuroimaging Initiative. Diagnostic accuracy of CSF Aβ42 and florbetapir PET for Alzheimer’s disease. Ann Clin Transl Neurol 2014, Ann Clin Transl Neurol 2014;1:534-543.

11. Portelius E, Westman-Brinkmalm A, Zetterberg H, Blennow K. Determination of β-amyloid peptide signatures in cerebrospinal fluid using immunoprecipitation-mass spectrometry. J Proteome Res 2006;5:1010–1016.

12. Mehta PD, Pirttila T, Mehta SP, Sersen EA, Aisen PS, Wisniewski HM. Plasma and cerebrospinal fluid levels of amyloid β proteins 1-40 and 1-42 in Alzheimer disease. Arch Neurol 2000;57:100-105.

13. Hansson O, Zetterberg H, Buchhave P, Andreasson U, Londos E, Minthon L, Blennow K. Prediction of Alzheimer’s disease using the CSF Aβ42/Aβ40 ratio in patients with mild cognitive impairment. Dement Geriatr Cogn Disord 2007;23:316-320.

14. Wiltfang J, Esselmann H, Bibl M, Hüll M, Hampel H, Kessler H, et al. Amyloid beta peptide ratio 42/40 but not A beta 42 correlates with phospho-Tau in patients with low- and high-CSF A beta 40 load. J Neurochem 2007;101:1053-1059.

15. Lewczuk P, Lelental N, Spitzer P, Maler JM, Kornhuber J. Amyloid-β 42/40 Cerebrospinal Fluid Concentration Ratio in the Diagnostics of Alzheimer’s Disease: Validation of Two Novel Assays. J Alzheimers Dis 2014, in press.

16. Hansson O, Zetterberg H, Buchhave P, Londos E, Blennow K, Minthon L.  Association between CSF biomarkers and incipient Alzheimer’s disease in patients with mild cognitive impairment: a follow-up study. Lancet Neurol 2006;5:228-34.

17. Shaw LM, Vanderstichele H, Knapik-Czajka M, Clark CM, Aisen PS, Petersen RC, et al. Cerebrospinal fluid biomarker signature in Alzheimer’s disease neuroimaging initiative subjects. Ann Neurol 2009;65:403-413.

18. Buchhave P, Minthon L, Zetterberg H, Wallin AK, Blennow K, Hansson O. Cerebrospinal fluid levels of β-amyloid 1-42, but not of tau, are fully changed already 5 to 10 years before the onset of Alzheimer´s dementia. Arch Gen Psychiatry 2012;69:98-106.

19. van Rossum IA, Vos SJB, Burns L, Knol DL, Scheltens P, Soininen H, et al. Injury markers predict cognitive decline in subjects with MCI and amyloid pathology. Neurology 2012;79:1809-1816.

20. Roe CM, Fagan AM, Grant EA, Hassenstab J, Moulder KL, Maue Dreyfus D, et al. Amyloid imaging and CSF biomarkers in predicting cognitive impairment up to 7.5 years later. Neurology 2013;80:1784-1791.

21. Dubois B, Feldman HH, Jacova C, Hampel H, Molinuevo JL, Blennow K, et al. Advancing research diagnostic criteria for Alzheimer’s disease: the IWG-2 criteria. Lancet Neurol 2014;13:614–629.

22. Blennow K, de Leon MJ, Zetterberg H. Alzheimer’s disease. Lancet 2006;368:387-403.

23. Prinz F, Schlange T, Asadullah K. Believe it or not: how much can we rely on published data on potential drug targets? Nat Rev Drug Discov 2011;10:712.

24. May PC, Dean RA, Lowe SL, Martenyi F, Sheehan SM, Boggs LN, et al. Robust central reduction of amyloid-β in humans with an orally available, non-peptidic β-secretase inhibitor. J Neurosci 2011;31:16507-16.

25. Blennow K, Hampel H, Zetterberg H. Biomarkers in β-amyloid immunotherapy trials in Alzheimer’s disease. Neuropsychopharmacology 2014;39:189-201.

26. Savage MJ, Kalinina J, Wolfe A, Tugusheva K, Korn R, Cash-Mason T, et al. A sensitive aβ oligomer assay discriminates Alzheimer’s and aged control cerebrospinal fluid. J Neurosci 2014;34:2884-2897.

27. Blennow K, Zetterberg H, Rinne JO, Salloway S, Wei J, Black R, et al. Effect of immunotherapy with bapineuzumab on cerebrospinal fluid biomarker levels in patients with mild to moderate Alzheimer’s disease. Arch Neurol 2012;69:1002-1010.

28. Desikan RS, Thompson WK, Holland D, Bartsch H, Roddey C, Mills IG, Brewer JB, Zetterberg H, Blennow K, Hyman BT, Sperling RA, McEvoy LK, Andreassen OA, Dale AM, for the Alzheimer’s Disease Neuroimaging Initiative. Heart fatty acid binding protein modulates Aβ-associated Alzheimer’s neurodegeneration. Neuromol Med 2013;8:39.

29. Olsson B, Hertze J, Ohlsson M, Nägga K, Höglund K, Basun H, et al. Cerebrospinal Fluid Levels of Heart Fatty Acid Binding Protein are Elevated Prodromally in Alzheimer’s Disease and Vascular Dementia. J Alzheimers Dis 2013;34:673-679.

30. Lee JM, Blennow K, Andreasen N, Laterza O, Modur V, Olander J, et al. The Brain Injury Biomarker VLP-1 Is Increased in the Cerebrospinal Fluid of Alzheimer’s Disease Patients. Clin Chem 2008;54:1617-1623.

31. Mroczko B, Groblewska M, Zboch M, Muszyński P, Zajkowska A, Borawska R, et alEvaluation of Visinin-Like Protein 1 Concentrations in the Cerebrospinal Fluid of Patients with Mild Cognitive Impairment as a Dynamic Biomarker of Alzheimer’s Disease. J Alzheimers Dis 2014, J Alzheimers Dis 2015;43:1031-1037.

32. Díez-Guerra FJ. Neurogranin, a link between calcium/calmodulin and protein kinase C signaling in synaptic plasticity. IUBMB Life 2010;62:597-606.

33. Jahn R1, Fasshauer D. Molecular machines governing exocytosis of synaptic vesicles. Nature 2012;490:201-207.

34. Davidsson P, Blennow K. Neurochemical dissection of synaptic pathology in Alzheimer’s disease. Int Psychogeriatrics 1998;10:11-23.

35. Davidsson P, Jahn R, Bergquist J, Ekman R, Blennow K. Synaptotagmin, a synaptic vesicle protein, is present in human cerebrospinal fluid: a new biochemical marker for synaptic pathology in Alzheimer’s disease? Mol Chem Neuropathol 1996;27:195-210.

36. Davidsson P, Punchades M, Blennow K. Identification of synaptic vesicle, pre- and postsynaptic proteins in human cerebrospinal fluid using liquid phase isoelectric focusing. Electrophoresis 1999;20:431-437.

37. Thorsell A, Bjerke M, Gobom J, Brunhage E, Vanmechelen E, Andreasen N, Hansson O, Minthon L, Zetterberg H, Blennow K. Neurogranin in cerebrospinal fluid as a marker of synaptic degeneration in Alzheimer’s disease. Brain Res 2010;1362:13-22.

38. Kvartsberg, H, Duits FH, Ingelsson M, Andreasen N, Öhrfelt A, Andersson K, et al. Cerebrospinal fluid levels of the synaptic protein neurogranin correlates with cognitive decline in prodromal Alzheimer’s disease. Alzheimers Dement 2014, in press.