key: cord-1015027-40zbm3r0
authors: Wiesman, Alex I.; Murman, Daniel L.; May, Pamela E.; Schantell, Mikki; Losh, Rebecca A.; Johnson, Hallie J.; Willet, Madelyn P.; Eastman, Jacob A.; Christopher‐Hayes, Nicholas J.; Knott, Nichole L.; Houseman, Lisa L.; Wolfson, Sara L.; Losh, Kathryn L.; Johnson, Craig M.; Wilson, Tony W.
title: Spatio‐spectral relationships between pathological neural dynamics and cognitive impairment along the Alzheimer's disease spectrum
date: 2021-05-31
journal: Alzheimers Dement (Amst)
DOI: 10.1002/dad2.12200
sha: 49b5343e594e3c9c75e88c2a2e5dde87b3986bb3
doc_id: 1015027
cord_uid: 40zbm3r0

INTRODUCTION: Numerous studies have described aberrant patterns of rhythmic neural activity in patients along the Alzheimer's disease (AD) spectrum, yet the relationships between these pathological features and cognitive decline are uncertain. METHODS: We acquired magnetoencephalography (MEG) data from 38 amyloid‐PET biomarker‐confirmed patients on the AD spectrum and a comparison group of biomarker‐negative cognitively normal (CN) healthy adults, alongside an extensive neuropsychological battery. RESULTS: By modeling whole‐brain rhythmic neural activity with an extensive neuropsychological profile in patients on the AD spectrum, we show that the spectral and spatial features of deviations from healthy adults in neural population‐level activity inform their relevance to domain‐specific neurocognitive declines. DISCUSSION: Regional oscillatory activity represents a sensitive metric of neuronal pathology in patients on the AD spectrum. By considering not only the spatial, but also the spectral, definitions of cortical neuronal activity, we show that domain‐specific cognitive declines can be better modeled in these individuals.

As scientists researching AD have shifted toward viewing it as a disease without strict clinical stages, objective and continuous measures of direct relevance to cognitive/functional impairments are sorely needed.

Among functional neuroimaging studies of patients on the AD spectrum, spectrally defined patterns of rhythmic population-level neural activity have emerged as a particularly sensitive method for differentiating patients with probable AD from cognitively normal (CN) older adults. [7] [8] [9] In particular, electro-and magneto-encephalography (MEG) studies of patients with AD have consistently shown increased spontaneous neural activity in the slower delta (2 to 4 Hz) and theta (5 to 7 Hz) frequency bands, as well as decreased activity in the faster alpha (8 to 12 Hz) and beta (15 to 30 Hz) bands. [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] Though widely replicated, the relationship between these findings and domain-specific cognitive function has not been extensively investigated. Because of this shortcoming, it is not clear which, if any, of these deviations from healthy rhythmic neural activity covary meaningfully with cognitive decline in patients on the AD spectrum. Identifying the relationships between rhythmic neural activity and cognitive decline in AD is critical, as the current knowledge gap in this area limits our understanding of the clinical implications of these findings. More specifically, knowledge of the direction of these relationships would indicate whether altered rhythmic neuronal activity represents pathological or compensatory processes, and thus could be useful in bolstering diagnosis and informing intervention. Additionally, previous studies of oscillatory aberrations in patients with AD have often been conducted at a coarse spectral and/or spatial resolution, and the vast majority were performed in patient groups who were not biomarker-confirmed.

In this study, we examine MEG data collected from 38 patients on the AD spectrum, all of whom were assessed as being biomarkerpositive using 18 F florbetapir positron emission tomography (PET), and 20 demographically matched CNs (19 biomarker-negative) using methods that provide remarkably high spatial and spectral resolution. By combining these data with an extensive neuropsychological battery, we show that spectrally and spatially defined neural oscillations significantly predict domain-specific neuropsychological function and instrumental activities of daily living in ways that are clinically informative.

Further, we find that the spatial and spectral properties of the relationships with cognitive and functional impairments only partially overlap with commonly reported group differences from CNs in the amplitude of oscillations. This demonstrates that not all neural oscillatory differences in patients on the AD spectrum necessarily reflect pathology that is directly relevant to cognitive and functional impairments, while also providing key new pathways for research aimed at improving patient outcomes.

Forty-four patients with amnestic mild cognitive impairment (aMCI) or mild probable AD, as determined by a fellowship-trained neurologist course of the study, as well as to complete certain questionnaires (e.g., the Functional Activities Questionnaire), and as such informed consent was obtained from each informant as well. To ensure that the interests of all patients were represented appropriately, for patients whose capacity to consent was questionable, informed assent was obtained from the research participant, in addition to informed consent from a legally authorized representative.

After screening and informed consent, participants underwent a battery of neuropsychological tests, with raw scores for each participant being converted to demographically adjusted z-scores based on published normative data. [19] [20] [21] [22] 

Combined PET/computed tomography (CT) data using 18 F-florbetapir (Amyvid, Eli Lilly) and a GE Discovery MI digital scanner were collected following the standard procedures described by the Society of Nuclear Medicine and Molecular Imaging (3D acquisition; single intravenous slow-bolus < 10 mL; dose = 370 MBq; waiting period = 30-50 minutes; acquisition = 10 minutes). 28 Images were attenuation corrected using the CT data, reconstructed in MIMneuro (slice thickness = 2 mm), 29 converted to voxel standardized uptake values (SUV) based on body weight, and normalized into Montreal Neurological Institute space. Each scan was read by a fellowship-trained neuroradiologist blinded to group assignment and assessed as being "amyloidpositive" or "amyloid-negative" using established clinical criteria. 29 At this stage, patients who were amyloid-negative were excluded from the AD spectrum group. Images were then normalized to the crus of the cerebellum (SUIT template) 30 to generate voxel-wise maps of SUV ratios (SUVr), 31 and back-transformed into each patient's native magnetic resonance imaging (MRI) space using their FreeSurfer-processed T1 data. The PET data overlapping with each individual's cortical gray-matter ribbon was then projected onto a tessellated FSAverage template surface using mri_vol2surf (maximum value; projection fraction = 1; steps of 2). 32

Our MEG recording and preprocessing pipeline has been extensively described in previous papers. [33] [34] [35] [36] Eight minutes of seated eyes-closed resting state MEG data were collected from each participant using a 306-sensor Elekta/MEGIN system at 1 kHz (bandwidth 0.1 to 330 Hz).

Continuous head position indicator coils were used to measure slight movements during each recording, and the position of these coils was digitized, along with each participant's fiducials and scalp surface, using a 3D digitizer (Fastrak 3SF0002, Polhemus Navigator Sciences).

Head motion correction and signal space separation with a temporal extension 37 were implemented to reduce noise, and only the data from the 204 gradiometers were used for analysis. Using the digitized head points, each participant's MEG data were co-registered with their own high-resolution structural T1-weighted MRI data (Siemens Prisma 3T; 64-channel head coil; TR: 2.3 seconds; TE: 2.98 ms; flip angle: 9 • ; FOV: 256 mm; slice thickness: 1 mm; voxel size: 1 mm 3 )

using an iterative closest-point rigid-body registration in Brainstorm (September 3, 2020 distribution) 38 and, after visual inspection, these fits were manually corrected. Triangulated cortical surfaces were computed from the T1 MRI data using FreeSurfer recon_all 32 and imported into Brainstorm. Individual cortical surfaces (including the cerebellum) were down-sampled to ≈17,000 vertices for computation of the forward model for use in MEG source imaging.

MEG data were bandpass filtered between 1 and 200 Hz and notch filtered at 60, 120, and 180 Hz, and ocular and cardiac artifacts were identified using an automated identification algorithm, supplemented by visual inspection of their temporal and spatial topography. From these topographies, signal-space projectors (SSPs) were generated and reviewed for each type of artifact, and those accounting for ocular and cardiac components were removed from the gradiometer data.

Artifact-reduced MEG data were then epoched into non-overlapping blocks of 4 seconds and down-sampled to 500 Hz. Epochs still containing major artifacts (e.g., SQUID jumps) were excluded within each participant using the ∪ of standardized thresholds of ± 2.5 median absolute deviations from the median for signal amplitude and gradient. After exclusions, a mean of 98.90 (SD: 8.56) and 96.89 (SD: 7.93) epochs were included for further analysis for the CN and AD spectrum groups, respectively. Importantly, there was no significant difference in the amount of data used between the two groups (P = .377).

Empty-room recordings of ≥ 2 minutes, collected around each individual scanning session, were processed using an identical pipeline to the grouped into canonical frequency bands (delta: 2 to 4 Hz; theta: 5 to 7 Hz; alpha: 8 to 12 Hz; beta: 15 to 29 Hz), and these spectral maps were normalized to the total power across the frequency spectrum.

The norm of the three unconstrained orientations per location and map were then projected onto a common FSAverage template surface (including the cerebellum) for statistical modeling.

Statistical comparisons were performed, accounting for the effects of age, using SPM12. Initial tests using parametric general linear models for interpretation of directional effects. Labels for peak-vertex data were derived from the Desikan-Killiany atlas. 41 Aβ SUVr data were also extracted from the same peak-vertices of the surface-based florbetapir PET images, for use in secondary analyses in the statistical software R. 42 Mediations were tested using a hierarchical regression approach. 43 

To examine the spatially and spectrally specific relationships between 

Replicating an extensive literature, we observed a robust increase in theta-frequency (5 to 7 Hz) activity in patients on the AD spectrum, F I G U R E 1 Dynamic mapping of Alzheimer's pathology (DMAP) study flowchart. After initial screening and recruitment, participants performed an extensive series of neuropsychological tests designed to tap five cognitive domains: attention, memory, verbal function, processing speed, and learning. This testing visit was followed by another, neuroimaging-focused visit, during which participants underwent functional and structural neuroimaging with MEG and MRI. Participants in the patient group returned for a third visit, in which they underwent a quantitative PET/CT scan with florbetapir 18 F, and were excluded from further analysis at this stage if they were biomarker-negative. Previous PET data was available for 19 CN adults. 

To investigate the relevance of these spectrally specific neural deviations to cognitive decline in patients on the AD spectrum, we next These relationships were largely non-overlapping in regard to their spatial, spectral, and cognitive definitions (Figure 3 ). In the delta band, pathologically higher levels of neural activity in the right temporal pole (r peak [35] = -.54, P < .001) and left superior temporal cortex (STC; r peak [35] = -.46, P = .004) predicted declines in processing speed.

In contrast, greater right STC (r peak [35] = .45, P = .005) and supramarginal gyrus (SMG; r peak [35] = .45, P = .005) deviations from CNs in the theta band predicted better memory performance, indicating that these commonly reported differences are likely compensatory in nature. Pathological beta-frequency deviations from CNs in the right (r peak [35] = .45, P = .005) and left (r peak [35] ) = .36, P = .027) IPC predicted declines in attention function, as well as reduced processing speed in the left MTC (r peak [35] = .37, P = .023). Notably, despite robust differences in alpha-frequency amplitude between patients on the AD spectrum and CNs, replicating many past studies, no significant relationships were found between cognitive function and alpha-frequency cortical activity.

Spatio-spectral neurocognitive relationships are not mediated by regional Aβ accumulation To examine whether regional Aβ uptake was responsible for any of these effects, we extracted SUVrs from each patient's florbetapir 18 F PET images for each of the vertices that exhibited a significant F I G U R E 2 Spatio-spectral group differences in cortical neural oscillatory amplitude. Surface maps to the left indicate significant statistical differences in oscillatory amplitude between patients (Alzheimer's disease spectrum [ADS]) and cognitively normal older controls (CN), beyond the effects of age, and corrected for multiple comparisons using a stringent threshold-free cluster enhancement approach (p FWE = .05). Theta maps are shown at the top, with alpha maps in the middle and beta maps at the bottom. Plots to the right of each map indicate the direction and nature of these effects at the vertices where they were most pronounced. Box plots represent conditional means, first and third quartiles, and minima and maxima, and violin plots show the probability density. L-IPC, left inferior parietal cortex; L-MTC, left middle temporal cortex relationship between neuronal oscillatory activity and cognitive impairment in patients on the AD spectrum. We found no evidence for a mediation of any of these spatio-spectral neurocognitive relationships by Aβ uptake. In addition, Bayesian analysis of these models revealed evidence that Aβ uptake provides no additional predictive information regarding cognitive status for any relationship in the theta (all BF 01 > 2.30) or delta (all BF 01 > 2.40) band, nor for the right IPCattention relationship in the beta band (BF 01 = 2.30). In contrast, Aβ did not exhibit robust evidence for or against the null hypothesis of no relationship to attention in the left IPC (BF 01 = 1.02), and predicted processing speed in the left MTC, above and beyond the effects of betafrequency neural activity in the same region (r [34] = -.34, P = .044; BF 10 = 2.07). However, this relationship should be interpreted cautiously, as it did not survive correction for multiple comparisons.

Finally, to establish which, if any, of these spectro-spatial neuronal maps related to patient functional independence, we regressed each spectrally specific map on IADLs (measured with the FAQ) in the patient group. No significant relationships between neuronal activity and functional independence were found in the delta, alpha, or beta bands; however, a robust relationship was observed in the theta band in right lateral occipital cortex (LOC; r peak [35] = -.46, P = .004; Figure 4 ). This effect again suggested compensation, such that stronger theta activity in this region predicted increased functional independence (i.e., lower FAQ scores).

Decades of research have found pathological population-level neuronal activity in patients on the AD spectrum, and this study provides key new knowledge in this established area by comprehensively associating these neural changes with domain-specific cognitive declines.

By doing so, we find support for the traditional conceptualization of delta and beta frequency deviations from healthy aging as signaling pathological changes, and show these changes are regionally specific and linked to cognitive decline in specific domains. Conversely, we find no such cognitive correlate of the commonly found decreases in alpha oscillatory amplitude. Unexpectedly, we also find that the oftenreported perturbations in theta amplitude in patients on the AD spectrum appear to predict better cognition and functional outcomes.

Spatio-spectral neural oscillations predict cognitive decline along the Alzheimer's disease (AD) spectrum. Surface maps to the left indicate significant statistical outputs of whole-brain models relating oscillatory amplitude and domain-specific cognitive function in patients on the AD spectrum, beyond the effects of age, and corrected for multiple comparisons using a stringent threshold-free cluster enhancement approach (p FWE = .05). Plots to the right of each map indicate the direction and nature of these effects at the vertices where they were most pronounced, with lines-of-best-fit and corresponding confidence intervals overlaid. L-IPC, left inferior parietal cortex; L-MTC, left middle temporal cortex; R-STC, right superior temporal cortex; R-TP, right temporal pole

We first replicated previous findings of increased low-frequency neural activity and decreased high-frequency activity in patients on the AD spectrum. Importantly, by leveraging a powerful source imaging approach, our findings provide enhanced spatial resolution compared to otherwise comparable investigations conducted previously.

This increased sensitivity revealed that commonly observed decreases To understand the nature of such relationships between oscillatory neural deviations and cognitive impairments, we regressed spectrally specific maps of cortical neural activity on five cognitive domains, measured with a thorough neuropsychological battery. As expected, many of these neural deviations from healthy aging predicted declines in cognitive performance, indicating pathology. Increased delta amplitude across a bilateral network of inferior parietal, anterior and posterior temporal, and dorsolateral prefrontal regions predicted worse processing speed, which aligns well with previous research indicating the importance of delta rhythms for temporal expectation and reaction F I G U R E 4 Spatio-spectral neural oscillations predict functional independence along the Alzheimer's disease spectrum. The surface maps to the left indicate the significant statistical output of a whole-brain model relating oscillatory amplitude and functional independence (instrumental activities of daily living) in patients on the AD spectrum, beyond the effects of age, and corrected for multiple comparisons using a stringent threshold-free cluster enhancement approach (p FWE = .05). The plot to the right indicates the direction and nature of this effect at the vertex where it was most pronounced, with the line-of-best-fit and corresponding confidence interval overlaid. FAQ, Functional Activities Questionnaire; R-LOC, right lateral occipital cortex time. 44 Decreased beta-frequency amplitude similarly predicted declines in processing speed. However, beta's relationship to processing speed was conspicuously spatially non-overlapping with that of delta, and instead included a left-lateralized network of middle temporal, supramarginal, and somato-motor cortices. Emerging work has suggested an overarching role for human beta oscillations as a "top-down,"

bi-directional regulator of cognitive flexibility and prediction. 45, 46 Thus, these complementary relationships between delta and beta oscillatory amplitude and processing speed point to a multi-spectral, distributed pathology affecting both "early" temporal readiness and "late" temporal prediction signals in patients on the AD spectrum. Supporting this conceptualization further, pathologically low levels of beta, but not delta, oscillations also predicted declines in attention across a bilateral network of inferior parietal, middle temporal, and prefrontal regions.

In contrast to the relationships observed in the delta and beta frequencies, deviations from healthy aging in theta oscillatory amplitude predicted better cognitive function. Specifically, increases in theta amplitude in right superior temporal and supramarginal cortices were related to better memory function, while similar increases in right parieto-occipital cortices signaled greater functional independence.

This was unexpected, and in fact, many previous studies have combined theta and delta frequency information into one "low-frequency" metric for simplicity. This compensatory theta effect suggests that theta and delta frequency deviations from healthy aging in patients on the AD spectrum are functionally distinct, both in terms of their relationships to differing cognitive domains, as well as in the ultimate direction of such relationships. The spatial definitions of the theta-frequency effects are also of interest, as better memory performance was predicted by neuronal deviations in the superior temporal cortex, while better daily functioning (i.e., IADLs) was predicted by similar deviations in parieto-occipital cortices. This signals that, while compensatory theta deviations in key association cortices may robustly predict alterations in memory function, those deviations in "lower-order" regions typically associated with visuospatial and visual attention function appear to be more directly relevant to the challenges to independence that these patients face.

Absent from these neuro-cognitive relationships was alpha activity, which was conspicuous as oscillations in this range are likely the most consistently reported to differ in the AD spectrum relative to healthy aging. Although this null finding is certainly far from conclusive, it was not due to power or a lack of sensitivity in the alpha band, as we observed robust amplitude differences in the alpha range between CNs and patients. It remains possible that these deviations from healthy aging in the alpha band do not represent functional insults until later in the course of the disease, or alternatively could represent other psychiatric or situational comorbidities not examined here. Also absent were any relationships between oscillatory amplitude and regional Aβ accumulation, but this was less surprising. A growing literature has shown that regional tau, and not Aβ, pathology better predicts cognitive and functional outcomes in clinical AD. [47] [48] [49] [50] [51] Accordingly, while these data allowed us to confirm that all patients in our AD spectrum group were Aβ-positive, they appear to yield little information regarding functional neural or cognitive declines, at least at the stage of the disease that we study herein.

First and foremost, we would like to acknowledge the efforts of our research participants. Without their selflessness and kind demeanor, none of this work would have been possible. We would also like to thank the research and clinical staff who sustained patient recruitment and data collection for this study, Dr. Clifford Jack for helpful input regarding normalization of our PET data, as well as our funding sources for their support. Finally, we would like to extend our 

Cascading network failure across the Alzheimer's disease spectrum

On the path to 2025: understanding the Alzheimer's disease continuum

The neuropsychology of preclinical Alzheimer's disease and mild cognitive impairment

Measuring information processing speed in mild cognitive impairment: clinical versus research dichotomy

Prediction of conversion from mild cognitive impairment to Alzheimer's disease dementia based upon biomarkers and neuropsychological test performance

Recent publications from the Alzheimer's Disease Neuroimaging Initiative: reviewing progress toward improved AD clinical trials

A comprehensive review of magnetoencephalography (MEG) studies for brain functionality in healthy aging and Alzheimer's disease (AD)

Neural oscillations and brain stimulation in Alzheimer's disease

Alzheimer's disease: the state of the art in resting-state magnetoencephalography

Altered generation of spontaneous oscillations in Alzheimer's disease. Neuroimage

Discrimination of Alzheimer's disease and mild cognitive impairment by equivalent EEG sources: a cross-sectional and longitudinal study

Quantitative analysis of occipital EEG in different stages of Alzheimer's disease. Electroencephalogr

Quantitative EEG analysis in early onset Alzheimer's disease: a controlled study

Magnetoencephalographic analysis of cortical activity in Alzheimer's disease: a pilot study

Focal temporoparietal slow activity in Alzheimer's disease revealed by magnetoencephalography

Altered temporal correlations in parietal alpha and prefrontal theta oscillations in early-stage Alzheimer disease

MEG spectral profile in Alzheimer's disease and mild cognitive impairment

Resting-state oscillatory brain dynamics in Alzheimer disease

Advanced Clinical Solutions For The WAIS-IV And WMS-IV

Revised Comprehensive Norms for An Expanded Halstead-Reitan Battery: Demographically Adjusted Neuropsychological Norms for African American and Caucasian Adults

Hopkins Verbal Learning Test-Revised: normative data and analysis of inter-form and test-retest reliability

Wechsler Adult Intelligence Scale-Fourth Edition (WAIS-IV)

The Psychological Corporation

Hopkins Verbal Learning Test-Revised: Professional Manual

Measurement of functional activities in older adults in the community

MoCA: a brief screening tool for mild cognitive impairment

Mini-mental state": a practical method for grading the cognitive state of patients for the clinician

SNMMI procedure standard/EANM practice guideline for amyloid PET imaging of the brain 1.0

Performance characteristics of amyloid PET with florbetapir F 18 in patients with Alzheimer's disease and cognitively normal subjects

A spatially unbiased atlas template of the human cerebellum

Defining imaging biomarker cut points for brain aging and Alzheimer's disease

Aberrant Occipital Dynamics Differentiate HIV-infected Patients With and Without Cognitive Impairment. Brain

Age-related visual dynamics in HIV-infected adults with cognitive impairment

Neural dynamics of selective attention deficits in HIV-associated neurocognitive disorder

Quiet connections: reduced frontotemporal connectivity in nondemented Parkinson's disease during working memory encoding

Spatiotemporal signal space separation method for rejecting nearby interference in MEG measurements

Brainstorm: a user-friendly application for MEG/EEG analysis

Threshold-free cluster enhancement: addressing problems of smoothing, threshold dependence and localisation in cluster inference

ggplot2: elegant graphics for data analysis

An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest

R: A Language and Environment for Statistical Computing

The moderator-mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations

Phase entrainment of human delta oscillations can mediate the effects of expectation on reaction speed

Beta-band oscillations-signalling the status quo

Magnetoencephalography for brain electrophysiology and imaging

Relationships between flortaucipir PET tau binding and amyloid burden, clinical diagnosis, age and cognition

Tau PET patterns mirror clinical and neuroanatomical variability in Alzheimer's disease

Associations between tau, Aβ, and cortical thickness with cognition in Alzheimer disease

Amyloid-β-independent regulators of tau pathology in Alzheimer disease

Predicting diagnosis and cognition with 18F-AV-1451 tau PET and structural MRI in Alzheimer's disease

SUPPORTING INFORMATION Additional supporting information may be found online in the Supporting Information section at the end of the article. How to cite this article: Wiesman AI, Murman DL, May PE, et al. Spatio-spectral relationships between pathological neural dynamics and cognitive impairment along the

Alzheimer's disease spectrum

The authors declare no competing conflicts of interest, financial or otherwise.