Although our understanding of preclinical and early Alzheimer’s disease pathology has significantly improved with the development of cerebrospinal fluid (CSF) biomarkers (such as beta-amyloid 42 [Aβ42], tau, phosphorylated tau 181 [ptau181], and visinin-like protein-1 [VILIP-1]), we severely lack robust noninvasive neuroimaging methods to detect and quantify the alterations in preclinical and early AD. Diffusion tensor imaging (DTI) has been used to detect white matter (WM) microstructure changes, but it failed to differentiate between axon/myelin damage and neuroinflammation. To address these limitations, we developed and validated a novel magnetic resonance imaging technique – diffusion basis spectrum imaging (DBSI) – that can specifically image neuronal injury in neurodegeneration diseases 1, 2. This study aims to examine whether DTI- or DBSI-derived indices correlated with CSF levels of each of the neuronal injury markers (tau, ptau181, and VILIP-1) in preclinical and early AD.

Participants: 175 cognitively normal participants and 82 very mild dementia (CDR0.5)3 participants were selected from a larger population enrolled at Knight Alzheimer’s Disease Research Center (Knight ADRC) of the Washington University School of Medicine (St Louis, MO, USA), in longitudinal studies of memory and aging.

CSF Collection and Analysis: CSF (20-30 mL) was collected by lumbar puncture as previously described 4. Samples were analyzed by ELISA (Innotest; Innogenetics, Ghent, Belgium) after one freeze-thaw for beta amyloid 42 (Aβ42), total tau, and tau phosphorylated at threonine-181 (ptau181) and by ELISA (Quidel, San Diego, CA) after two freeze-thaw cycles for VILIP-1.

NIA-AA preclinical AD stages: Cognitively normal participants were divided into preclinical stage 0 and stage 1 groups according to CSF measures and research criteria proposed by the National Institute on Aging and the Alzheimer’s Association (NIA-AA)5. Participants were classified as preclinical stage 0 (all biomarkers negative [Aβ42 > 500 pg/ml, tau <339 pg/ml, and ptau181 < 80 pg/ml) and preclinical stage 1 (Aβ42 < 500 pg/ml and either tau > 339 pg/ml or ptau 181 > 80 pg/ml).

Imaging Acquisition and Processing: Diffusion MRI images were collected (2x2x2mm voxels, TR=14500ms, TE=102ms, multi-bvalue scheme, 23 directions and bmax = 1400s/mm2). Data were collected in two 6 minute runs on 3T Trio scanner (Siemens, Erlangen, Germany). DTI and DBSI derived mean diffusivity, fractional anisotropy (FA), radial diffusivity and axial diffusivity were computed for all datasets. The whole brain voxel-wise DTI-indices was analyzed using Tract Based Spatial Statistics (TBSS) (available in FSL).

Aβ42, an indicator of amyloid burden, was significantly higher in the preclinical stage 0 group than in the stage 1 and CDR 0.5 groups (Fig. 1A); having a high level of Aβ42 is categorized as being Aβ42-negative. We found no statistical difference in level of Aβ42 between the stage 1 and CDR 0.5 groups (Fig. 1A). There were no significant differences in CSF tau, ptau181, or VILIP-1 between the stage 0 and stage 1 groups (Fig. 1B-D), suggesting that the preclinical stage 1 patients had no neuronal injuries. In contrast, CSF tau, ptau181, and VILIP-1 were significantly higher in the CDR 0.5 group than in the stage 0 and stage 1 groups (Fig. 1B-D), indicating neuronal injury in the CDR 0.5 participants. The CSF ptau181 correlated with only one DTI measure, fractional anisotropy, in 25 WM regions (Fig. 2A). In contrast, the CSF levels of neuronal injury markers correlated with both DBSI-derived fractional anisotropy and radial diffusivity in multiple WM regions. For example, levels of tau negatively correlated with DBSI-derived fractional anisotropy in 14 WM regions (Fig. 2B), and levels of ptau181 positively correlated with DBSI-derived radial diffusivity in 31 WM regions (Fig. 2C). In addition, we observed a significant correlation between levels of VILIP-1 and DBSI-derived fractional anisotropy and radial diffusivity (Fig. 2B) in 36 and 28 WM regions, respectively.CSF measures of tau protein and phosphorylated tau are hypothesized to reflect neurodegeneration and may be the biomarkers that are most relevant to microstructural changes in early AD. VILIP-1, a neuronal calcium-sensor protein, is a marker of neuronal injury, and CSF measures of VILIP-1 are useful for diagnosis and prognosis in early AD. We observed elevated CSF tau, ptau181, and VILIP-1 in the CDR 0.5 participants in this study, indicating that these participants had neuronal injury. DBSI-derived fractional anisotropy and radial diffusivity correlated significantly with CSF tau, ptau181, and VILIP-1, whereas only DTI-derived fractional anisotropy correlated with ptau181 (Fig.2). Unfortunately, DTI-derived fractional anisotropy is sensitive to the mixed contribution from both WM damages and neuroinflammation. By separating the confounding effect from neuroinflammation, DBSI metrics reflect WM abnormalities accurately and specifically.