The CSF p-tau/β-amyloid 42 ratio correlates with brain structure and fibrillary β-amyloid deposition in cognitively unimpaired individuals at the earliest stages of pre-clinical Alzheimer's disease

Abstract

CSF concentrations of β-amyloid 42 (Aβ42) and phosphorylated tau (p-tau) are well-established biomarkers of Alzheimer's disease and have been studied in relation to several neuropathological features both in patients and in cognitively unimpaired individuals. The CSF p-tau/Aβ42 ratio, a biomarker combining information from both pathophysiological processes, has emerged as a promising tool for monitoring disease progression, even at pre-clinical stages. Here, we studied the association between the CSF p-tau/Aβ42 ratio with downstream markers of pre-clinical Alzheimer's disease progression including brain structure, glucose metabolism, fibrillary Aβ deposition and cognitive performance in 234 cognitively unimpaired individuals, who underwent cognitive testing, a lumbar puncture, MRI, 18F-fluorodeoxyglucose and 18F-flutemetamol PET scanning. We evaluated both main effects and interactions with Alzheimer's disease risk factors, such as older age, female sex and the apoliporoptein E (APOE)-ɛ4 allele, in a priori defined regions of interest and further examined the associations on the whole-brain using voxel-wise regressions. In addition, as the association between CSF Alzheimer's disease biomarkers and brain structure and function may be non-linear, we tested the interaction between the CSF p-tau/Aβ42 ratio and stages of pre-clinical Alzheimer's disease defined using the amyloid (A) and tau (T) classification. We found significantly positive associations between CSF p-tau/Aβ42 and both cortical Aβ deposition and regional grey matter volume while no effect was observed for brain metabolism. A significant interaction with age indicated that, for the same level of CSF p-tau/Aβ42, older individuals displayed both increased Aβ deposition and lower grey matter volume, in widespread cortical areas. In addition, we found that women compared with men had a greater Aβ fibrillary accumulation in midline cortical areas and inferior temporal regions, for the same level of the CSF biomarker. The impact of CSF p-tau/Aβ42 on grey matter volume was modulated by AT stages, with A+T+ individuals displaying significantly less positive associations in areas of early atrophy in the Alzheimer's continuum. Finally, we found that sex and APOE-ɛ4 modulated the association between the CSF biomarker and episodic memory as well as abstract reasoning, respectively. Our data indicate that the CSF p-tau/Aβ42 ratio is strongly associated with multiple downstream neuropathological events in cognitively unimpaired individuals and may thus serve as a potent biomarker to investigate the earliest changes in pre-clinical Alzheimer's disease. Given that its impact on both Aβ deposition and grey matter volume is modulated by specific risk factors, our results highlight the need to take into account such predisposing variables in both clinical practice and prevention trials.

Keywords: Alzheimer’s disease; CSF biomarkers; amyloid deposition; brain structure; p-tau/Aβ42 ratio.

Graphical Abstract

Figure 1

CSF p-tau/Aβ42 and Aβ deposition. Main effects of CSF p-tau/Aβ42 on fibrillar Aβ deposition. (A) Surface representation of the four composite ROIs of cortical Aβ deposition as in the study by Collij et al. (B) Scatterplots showing the significant linear positive associations between CSF p-tau/Aβ42 and Aβ SUVR in a priori defined ROIs. Each data point represents an individual subject (n = 221). Statistical analyses were performed using linear regression, with the following F-values for each stage: F = 144.51 for Aβ PET Stage 1, F = 145.84 for Aβ PET Stage 2, F = 104.54for Aβ PET Stage 3 and F = 96.76 for Aβ PET Stage 4. P-values were all significant at pFDR < 0.05. (C) Surface rendering of the statistical probability maps indicating higher fibrillar Aβ deposition, indicated by a higher SUVR), as a function of CSF p-tau/Aβ42 continuous levels. Statistical analyses were performed using linear regression, and the projected P-values correspond to their respective F-test. LL, left lateral; LM, left medial; RL, right lateral; RM, right medial.

Figure 2

CSF p-tau/Aβ42 and GMV. Main effects of CSF p-tau/Aβ42 on GMV. (A) Surface representation of the three composite ROIs as defined in Braak and Braak. (B) Scatterplots showing the significant linear positive associations between CSF p-tau/Aβ42 and GMV in a priori defined ROIs. Each data point represents an individual subject (n = 234). Statistical analyses were performed using linear regression, with the following F-values for each stage: F = 0.28 for Braak I and II, F = 4.31 for Braak III and IV and F = 2.61 for Braak V and VI. P-value at Braak III and IV was significant at uncorrected P < 0.05. (C) Surface rendering of the statistical probability maps indicating a greater GMV as a function of CSF p-tau/Aβ42 continuous levels. Statistical analyses were performed using linear regression, and the projected P-values correspond to their respective F-test. LL, left lateral; LM, left medial; RL, right lateral; RM, right medial.

Figure 3

CSF p-tau/Aβ42 interactions with age and sex on Aβ deposition. Older age and female sex are associated with a higher Aβ deposition for a given CSF p-tau/Aβ42 level. (A) Group scatterplots showing the significant interactions between continuous CSF p-tau/Aβ42 and age on Aβ SUVR in the a priori defined ROIs. For visualization purposes, age was broken down into three groups using tercile ranking. In each scatterplot, parameter estimates and P-values are provided for each age subgroup. . Each data point represents an individual subject (n = 221). Statistical analyses were performed using linear regression, with the following F-values for each stage: F = 13.98 for Aβ PET Stage 1, F = 18.32 for Aβ PET Stage 2, F = 18.14 for Aβ PET Stage 3 and F = 19.86, for Aβ PET Stage 4. P-values were all significant at pFDR < 0.05. (B) Surface rendering of the statistical probability maps showing the significant interaction between CSF p-tau/Aβ42 and age, on the whole-brain level. Statistical analyses were performed using linear regression, and the projected P-values correspond to their respective F-test. (C) Group scatterplots showing the significant interactions between continuous CSF p-tau/Aβ42 and sex in the a priori defined ROIs. In each scatterplot, parameter estimates and P-values are provided for males and females. Each data point represents an individual subject (n = 221). Statistical analyses were performed using linear regression, with the following F-values for each stage: F = 6.25 for Aβ PET Stage 1, F = 10.55 for Aβ PET Stage 2, F = 9.69 for Aβ PET Stage 3 and F = 11.54 for Aβ PET Stage 4. P-values were all significant at pFDR < 0.05. (D) Surface rendering of the statistical probability maps showing the significant interaction between CSF p-tau/Aβ42 and sex, on the whole-brain level. Statistical analyses were performed using linear regression, and the projected P-values correspond to their respective F-test. LL, left lateral; LM, left medial; RL, right lateral; RM, right medial.

Figure 4

CSF p-tau/Aβ42 interactions with age on GMV. Younger age was associated with a greater GMV for a given CSF p-tau/Aβ42 level. (A) Group scatterplots showing the significant interactions between continuous CSF p-tau/Aβ42 and age on GMV, in the a priori defined ROIs. For visualization purposes, age was broken down into three groups using tercile ranking. In each scatterplot, parameter estimates and P-values are provided for each age subgroup. Each data point represents an individual subject (n = 234). Statistical analyses were performed using linear regression, with an F-value = 4.21 for the Braak V and VI stage, which was significant on a nominal level, P < 0.05. (B) Surface rendering of the statistical probability maps showing the significant interaction between CSF p-tau/Aβ42 and age on GMV, on the whole-brain level. Statistical analyses were performed using linear regression, and the projected P-values correspond to their respective F-test.

Figure 5

CSF p-tau/Aβ42 interaction with Aβ status on GMV. The impact of CSF p-tau/Aβ42 ratio on GMV was significantly modulated by Aβ status. (A) Group scatterplots showing the linear association between CSF p-tau/Aβ42 ratio and GMV in three Braak composite ROIs. For visualization purposes, linear slopes are broken down by the three AT stage groups. In each scatterplot, parameter estimates and P-values are provided for each age subgroup. Each data point represents an individual subject (in = 234). Statistical analyses were performed using linear regression, which yielded with non-significant F-values. By contrast, the interaction with binary category Aβ status was significant in Braak III and IV and Braak V and VI for uncorrected P < 0.05. (B) Volume rendering of GMV clusters of paired comparisons between AT pathophysiology grouping factors.

Figure 6

CSF p-tau/Aβ42 interactions with APOE-ɛ4 on cognitive performance. The association between CSF p-tau/Aβ42 and cognitive performance was modulated by both sex and APOE-ɛ4. Group scatterplots showing that, for the same level of CSF p-tau/Aβ42, women compared with men had a poorer cognitive performance in immediate and delayed paired recall, while APOE-ɛ4 carriers compared with non-carriers had a poorer performance in AR. Each data point represents an individual subject (n = 234). Statistical analyses were performed using linear regression, with F-values being F = 4.98 for immediate recall, F = 4.42 for delayed recall and F = 5.32 for AR. All P-values were significant at uncorrected P < 0.05.