Association of Longitudinal Changes in Cerebrospinal Fluid Total Tau and Phosphorylated Tau 181 and Brain Atrophy With Disease Progression in Patients With Alzheimer Disease

Key Points Question How do different proposed measurements of tau brain pathologic abnormalities (ie, levels of phosphorylated tau 181 in the cerebrospinal fluid [CSF]) and neurodegeneration (ie, total tau levels in the CSF and brain atrophy) change over the course of Alzheimer disease? Findings This cohort study of 465 participants with dominantly inherited Alzheimer disease found that the rates of change for CSF levels of total tau and phosphorylated tau 181 had a different pattern across the course of Alzheimer disease. The association between the rates of change of CSF levels of total tau and phosphoryated tau 181 and brain atrophy varied by disease stage. Meaning These results may provide a better understanding of the dynamics of Alzheimer disease and have important implications as trials targeting tau brain pathologic abnormalities move forward.


Introduction
The neuropathologic hallmarks of Alzheimer disease (AD) are the presence of neuritic amyloid plaques, primarily of the amyloid-β peptide, and the intraneuronal accumulation of neurofibrillary tangles (NFTs) and neuropil threads composed of hyperphosphorylated, aggregated tau protein. 1,2 Our understanding of AD has evolved substantially over the past 2 decades, with neuroimaging and fluid biomarkers allowing for early detection of AD-related pathologic abnormalities. [3][4][5] In fewer than 1% of patients, AD is caused by autosomal dominant mutations in either the presenilin 1, presenilin 2, or amyloid precursor protein genes. Dominantly inherited AD (DIAD) is considered clinically similar to sporadic AD except for a younger age at onset (AAO). 6 Mutation carriers (MCs) have a somewhat predictable age at AD symptom onset. 7 Biomarker studies 6,8 have contributed to hypothesized trajectories of fluid and imaging biomarker changes that occur over the course of the disease, from the preclinical phase to the end stages characterized by advanced dementia. However, models of biomarker change in AD have been based mostly on cross-sectional data. 6,9,10 More recently, many models have classified biomarkers according to their proposed association with the biological underpinnings of the disease; the recent amyloid/tau/neurodegeneration (A/T/N) framework 11 was developed to provide a more biological rationale to the classification of the disease. Phosphorylated tau 181 (pTau181) in the cerebrospinal fluid (CSF) has been suggested to represent NFT pathologic abnormalities, whereas total tau (tTau) in the CSF is thought to be a marker of neurodegeneration that is passively released with cell death or injury. 8,12 Given these putative mechanisms, it might be expected that levels of both tTau and pTau181 would continue to become more abnormal with disease progression, as NFT pathologic abnormalities increase and neurodegeneration accelerates. However, recent longitudinal studies [13][14][15][16][17] from the Dominantly Inherited Alzheimer Network (DIAN) and Alzheimer's Disease Neuroimaging Initiative (ADNI) cohorts have challenged the linear model from previous cross-sectional studies, 3,9,18 which have consistently found higher CSF levels of tTau and pTau181 as the disease progresses.
These findings highlight a need for accurate determination of the evolution of longitudinal changes in CSF levels of tTau and pTau181 and their association with disease progression. In the present study, we assessed the longitudinal pattern of changes in CSF levels of tTau and pTau181 and their association with brain atrophy as measured by magnetic resonance imaging (MRI). We hypothesized that if CSF tTau and pTau181 were passively released with neurodegeneration, they should be associated with MRI measures of neurodegeneration (eg, rate of atrophy). To evaluate this hypothesis, we used a well-characterized cohort with DIAD from the DIAN study. 19

Methods
All participants were recruited as part of the DIAN study. 20 Participants provided written informed consent or assent with proxy consent. The institutional review boards for each of the participating DIAN sites approved all aspects of the study. This study follows the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

Clinical and Neuropsychological Assessments
Participants underwent extensive clinical evaluation, which included family history of AD, personal medical history, and physical and neurological examination. Clinical dementia status was determined with the CDR in accordance with standard protocols and criteria. 22,24 Clinicians performing the assessments were blinded to mutation status of participants. Full details of participating sites, enrollment, and assessments in DIAN have been published elsewhere. 20

MRI Scanning
Participants underwent volumetric T1-weighted MRI, using the magnetization-prepared rapid acquisition with gradient-echo sequence defined in the ADNI second phase. 25 Sites used a 3-T scanner and were required to pass regular quality control assessments. Volumetric segmentation and cortical surface reconstruction were performed using FreeSurfer image analyzing software version 5.3 (Harvard Medical School), 26 and subcortical volumes were corrected for intracranial volume using a regression approach. 27 Cortical thickness and volume measures were averaged across hemispheres because there were no a priori laterality predictions. Because the focus of our study was to assess the association between longitudinal CSF tTau and pTau181 levels with atrophy rate, a limited number of regions of interest (ROIs) reflecting brain atrophy patterns across AD stages were included. We followed an ROIs approach to include posterior areas with earlier tau deposition (eg, hippocampus,

Statistical Analysis
Data analysis was performed in June 2019. Baseline characteristics of the participants are summarized as mean (SD) for continuous variables and number (column percentage) for categorical variables. P values for comparing the difference between MCs and noncarriers (NCs) were obtained using generalized linear mixed-effects models with random intercepts for family clusters to take into account the associations between participants within the same family; the tests were 2-sided. The and MCs without symptoms (CDR score, 0) separately, using bivariate linear mixed-effects models with random intercepts for family clusters. 35 MCs), the effect size that could be detected with 80% power was 0.27.

Longitudinal Change in CSF tTau and pTau181 Levels
Previous analyses 14,17 of this cohort have found that CSF tTau and pTau181 levels were increased in MCs 15 years before expected symptom onset (EYO = −15). We also examined the longitudinal change in CSF tTau and pTau181 levels in terms of annual rate of change across the EYO (Figure 1 and   years before symptoms appeared (EYO −5) and followed a similar trend through the disease (Table 2 and Figure 1). For all ROIs included in the analysis, the rate of change followed similar trends.

JAMA Network Open | Neurology
To visually compare the differences in the trajectories of the rates of change of CSF tTau and pTau181 levels and structural MRI measures, locally estimated scatterplot smoothing curves were constructed for the standardized rate of change as a function of baseline EYO (Figure 2). Notably, after EYO 0, the rate of change became more negative (structures atrophied more quickly), compared with CSF tTau levels, which continued to have a stable positive rate of change.

Association Between CSF and MRI Measures of Neurodegeneration
We found differences in nearly all posterior and limbic or paralimibic regions for associations in rates of change with tTau and pTau181 according to stage of disease when comparing MCs without symptoms vs those with symptoms (eTable 2 and eTable 3 in the Supplement). Figure 3 shows patterns of correlation coefficients between the rates of change of CSF tTau and pTau181 and brain structure stratified by the absence or presence of symptoms (CDR score of 0 vs CDR score >0) and by posterior (early atrophy) and anterior (later atrophy) cortical or subcortical regions. In individuals without symptoms (CDR score of 0), the rates of change of CSF tTau and pTau181 were inversely  correlated with most brain structure measures, where higher CSF tTau levels were associated with smaller cortical thickness. Generally, there were higher correlation coefficients for CSF tTau in the asymptomatic phase and mostly for the posterior neocortical and allocortical (limbic) regions.
However, after symptom onset, the correlation between CSF tTau and pTau181 and brain structures changed. The intensity of neuronal damage as measured by brain atrophy continued at an increasing rate, whereas the rate of change of CSF tTau levels remained at a somewhat constant rate and the rate of change of CSF pTau181 actually switched from positive to negative. These results suggest that CSF tTau and structural brain measures have distinct patterns later in the disease course.

Discussion
The use of biomarkers has become an essential component of AD research 11,37 and therapeutic trials.
The new framework from the National Institute on Aging and Alzheimer Association 11 capitalizes on the use of biomarkers for early identification of AD, which has substantial implications for early treatment and trial enrollment. However, little work has been done regarding comparisons of longitudinal biomarker trajectories that are currently proposed to represent similar aspects of disease (eg, CSF tTau and MRI are both proposed as markers of neurodegeneration). 38 In this study, on the basis of longitudinal data from the DIAN study, 20 we evaluated the trajectories of CSF tTau and pTau181 over the course of AD. Notably, we included a bigger sample and used newer CSF tTau and pTau181 values generated with a fully automated, high-performance electroluminescence immunoassay. We compared the trajectories of CSF tTau and pTau181 with the atrophy of brain structures as measured by MRI. First, consistent with previous reports, 14,39 our study found mean concentrations of CSF tTau and pTau181 to be higher in MCs from the early stage  of AD, supporting the use of CSF tau as a marker of AD risk and progression. Second, we found that the positive rate of change of CSF tTau remained constant after EYO −10, whereas CSF pTau181 had a positive rate of change early in the disease course, which then reversed and became negative later in the disease. This indicates that our previous results 13,14 were not likely to be an artifact of the measurement used because we used a different method for this study. Third, the associations identified between the rates of change of CSF tTau and pTau181 with brain atrophy do not support the assumption that CSF tTau changes follow a pattern similar to that of structural brain changes. Our findings indicate that neither CSF tTau nor pTau181 has the same pattern of change as brain measures and should be considered as associated but distinct biomarkers in AD. In addition, these findings indicate that, within the current biomarker classification, tTau is an important marker of AD but may not be the ideal marker of neurodegeneration.

JAMA Network Open | Neurology
Recent studies 40,41 have suggested that in the presence of amyloid pathologic abnormalities, more CSF tTau and pTau181 is released. Moreover, increased tau in CSF seems to be dependent of amyloid deposition and occurs in the absence of tau brain pathologic abnormalities. 42 Early disease stages may also be characterized by higher cellular stress 43,44 and inflammation, with higher levels of tTau and pTau181 in CSF representing a response. 45 However, if CSF tTau directly reflected neurodegeneration, it would be expected that the rate of change of CSF tTau would increase in concert with brain atrophy during the period of maximal rate of atrophy (EYO >0). It is possible that during disease progression and neurodegeneration, the loss of neuronal cells results in less neuronal substrate to produce tau. Although this might account for some of the slowing in longitudinal changes in CSF tTau that we observed, it is unlikely that the degree of neurodegeneration is sufficient to fully explain our findings. Early elevations may also be associated with acute neuronal membrane damage, whereas apparent later reductions reflect the death of a smaller number of neurons that remain. Acute neuronal injury may be associated with a stronger inflammatory response at early stages of the disease. 46 The present findings challenge some previous assumptions about AD progression and its association with both pTau181 and tTau. Contrary to the idea that tTau and pTau181 levels continue to increase with greater neurodegeneration and the spread of NFT, 47 we found evidence of a decrease in the rate of change, arguing against the use of these measures as a reflection of a passive release from neuronal death and NFT. These data support the relevance of CSF tTau and pTau181 as markers of amyloid deposition and accompanying changes (eg, inflammation and neuronal membrane damage) 11 ; however, the complex rates of change identified here and in our previous work suggest that using them as measures of therapeutic response requires further investigation, because levels vary as a function of where an individual is in the neuropathological cascade. The apparent disconnect between CSF tTau and MRI measures may reflect the fact that they are measuring different stages of the neurodegenerative process, with CSF tTau accounting for the active phase of neuron injury and damage, and MRI measuring the subsequent structural sequelae of the active death process.
Our findings are consistent with previous studies 15 This study has potential implications for AD trials using tau-based therapies and other putative disease-modifying therapies. First, during trial design, the active group and the placebo group will have to be randomized by disease stage (according to disease severity measured using biomarkers and the severity of cognitive impairment), because minor differences in the neurodegeneration cascade stage might translate into major differences in biomarker trajectories, and as a result, might be misinterpreted as a treatment effect. Second, one must consider how to interpret changes in biomarkers during clinical trials readout; in other words, one must consider how a successful treatment would be expected to affect CSF tau levels or rate of change. This study, along with recent work assessing neurofilament light chains, 51,52 suggests that neurofilament light chains may be an advantageous marker of neurodegeneration in therapeutic trials of AD. 52,53 Limitations

JAMA Network Open | Neurology
The present study is not without limitations. First, our results are dependent on the accuracy of the DIAN EYO and, like any measure, it is subject to error (ie, symptom onset could be a few years earlier or later than expected [randomly]). Second, we recognize that DIAD and late-onset AD are similar but not identical. Sporadic AD disease occurs later in life, and individuals often exhibit a wide range of comorbidities, including vascular disease and the presence of transactive response DNA-binding protein of 43 kDa. [54][55][56][57] These differences may be associated with biomarker profiles or natural progression, so the total generalizability of the present findings to late-onset AD will have to be confirmed in future cohorts. Nevertheless, the results highlight the importance of gathering longitudinal data to refine current biomarker models.

Conclusions
The results of this study support the A/T/N framework, whereby increases in CSF levels of tTau and pTau181 are robust diagnostic markers of pathologic abnormalities and neuronal injury in early Alzheimer disease. However, our data suggest that levels of CSF tTau and pTau181 diverge from atrophy-based measures of neurodegeneration later in the disease. This is an important result to consider in clinical trials targeting tau brain pathologic abnormalities. Future studies should explore the longitudinal rate of change of CSF pTau181 levels and positron emission tomography of tau as markers of tau brain pathologic abnormalities.

ARTICLE INFORMATION
Accepted for Publication: October 21, 2019.