Historically, a clinical diagnosis of Alzheimer’s disease could only be confirmed upon the discovery of β-amyloid plaques and tau tangles post-mortem. Today, biomarkers (imaging and biofluid measurements of β-amyloid, tau, and neurodegeneration) are being used to define the disease in vivo. Chapter 1 of this dissertation examines the long-term performance of baseline cognitive, neuroimaging, and cerebrospinal fluid biomarker-assisted prognoses in patients with mild cognitive impairment. Concordant atrophy, memory impairment, and abnormal β-amyloid and tau was associated with the highest risk for conversion to dementia, while individuals with concordant negative risk remained stable for up to 11 years. These results suggest that baseline biomarker-assisted predictions of decline to dementia are stable over the long term, and that combinations of complementary biomarkers can improve the accuracy of these predictions. Though individuals with mild cognitive impairment are still in the prodromal stages of clinical Alzheimer’s disease, converging evidence suggests the pathologic changes underlying Alzheimer’s disease begin up to 15 years before cognitive impairment. Chapter 2 of this dissertation tests whether the combination of the polygenic and age-specific risk for Alzheimer’s disease can predict elevated β-amyloid in clinically unimpaired individuals in order to enroll these individuals into preclinical Alzheimer’s disease trials. Alzheimer’s disease incidence rates and a polygenic hazard score were used to create a gene- and age-defined ADAge. The ADAge-enrichment screening method identified clinically unimpaired individuals with elevated β-amyloid and lowered screening costs in real-world preclinical Alzheimer’s disease trial data. These results demonstrate the utility of ADAge enrichment as a more efficient and cost-effective means to enroll clinically normal individuals with elevated β-amyloid in clinical trials. Clinically, distinctions between Alzheimer’s disease and related dementias are suboptimal and complicated by shared genetic risk factors and frequent co-pathology. Chapter 3 of this dissertation analyzes the utility of polygenic risk assessments for Alzheimer’s disease, dementia with Lewy bodies, and Parkinson’s disease to differentiate between individuals with distinct underlying pathologies. Polygenic scores were specifically associated with either dementia with Lewy bodies or Alzheimer’s disease pathology, indicating that an assessment of genetic risk may be useful to clinically distinguish between Alzheimer’s disease and dementia with Lewy bodies.

The medial temporal lobe supports integrating the “what,” “where,” and “when” of an experience into a unified memory. However, it remains unclear how representations of these contextual features are neurally encoded and distributed across medial temporal lobe subregions.

This study conducted functional magnetic resonance imaging of the medial temporal lobe, while participants retrieved pair, spatial, and temporal source memories. Multivoxel classifiers were trained to distinguish between retrieval conditions before and after correction for mean signal and response times, to more thoroughly characterize the multivoxel signal associated with memory context.

Activity in perirhinal and parahippocampal cortex dissociated between memory for associated items and memory for their spatiotemporal context, and hippocampal activity was linked to memory for spatial context. However, perirhinal and hippocampal classifiers were, respectively, driven by effects of mean signal amplitude and task difficulty, whereas the parahippocampal classifier survived correction for these effects.

These findings demonstrate dissociable coding mechanisms for episodic memory context across the medial temporal lobe, and further highlight a critical distinction between multivoxel representations driven by spatially distributed activity patterns, and those driven by the regional signal.

Alzheimer’s disease (AD) is the single most common cause of dementia, but AD alone accounts for less than half of all cases. A variety of other brain pathologies can cause cognitive impairment either alone or in combination with AD. A clinical diagnosis often fails to capture this complexity, so evaluation of brain pathology at autopsy remains the gold standard for establishing the underlying causes of dementia. As disease-modifying therapeutics targeting the underlying mechanisms of neurodegenerative diseases are developed, there is a pressing need for clinicopathologic studies to identify distinct clinical presentations that rise from these various pathologies in order to simplify differential diagnosis, improve prognosis of future decline, and help better target interventions.Here, I present a series of clinicopathologic studies that characterize and compare the dementia syndromes associated with neuropathologically-verified AD, Hippocampal Sclerosis, Lewy body disease, and their interactions. The first study demonstrates that Hippocampal Sclerosis, alone or in combination with AD, produces a dementia syndrome that is virtually indistinguishable from AD. The generally slower longitudinal trajectory of cognitive decline in patients with Hippocampal Sclerosis, however, may help clinically distinguish the disorder from AD. The second study revealed double-dissociations in patterns of cognitive deficits and longitudinal declines between Dementia with Lewy Bodies (DLB) and Parkinson’s Disease Dementia (PDD) that likely reflect subtle differences in pathology. These results suggest that pooling DLB and PDD patients in clinical trials targeting Lewy body pathology may reduce the power to see an effect of treatment unless the appropriate cognitive domain for each is targeted by the trial outcome measures. The final two studies demonstrate considerable variability in clinical and cognitive presentation across age of onset within those with severe AD at autopsy, and show that this variability is (at least partly) mediated by the distribution of neurofibrillary tangle (NFT) pathology. Those with younger onset AD have disproportionately greater neocortical NFT pathology relative to their degree of hippocampal NFT pathology. These findings help explain the paradox that those patients with younger onset of symptoms tend to have higher likelihood of atypical clinical presentations of AD, even though they tend to have less concomitant non-AD neuropathology.

The postmitotic nature of neurons allows for the proliferation and accumulation of genomic changes in the brain across an individual’s lifetime. This genomic mosaicism can take many forms, including somatic gene recombination.Chapter 1 is an introduction of the forms of somatic genomic mosaicism in the brain, especially as they relate to the aging and diseased brain. Chapter 2 examines the presence and prevalence of endogenous reverse transcriptase (RT) activity across the human brain. Examination of the prefrontal cortex, basal ganglia, hippocampus and cerebellum revealed widespread RT activity across the human brain. Inhibition of endogenous brain RT activity with nucleoside and non-nucleoside RT inhibitors (NRTIs and NNRTIs) showed highly specific inhibitory profiles, including differences between regions. Endogenous RT activity in the mouse brain exhibited disparate inhibitory profiles, pointing to species specificity in endogenous RT source, activity, and susceptibilities. Chapter 3 examines the diversity of the neural L1 (LINE-1/Long-Interspersed- Element-1) transcriptome. L1 retrotransposons bicistronically express open reading frame 2 (ORF2), which contains RT and endonuclease (EN) domains, along with the RNA-binding protein ORF1. We utilize spatial transcriptomics and targeted long-read (PacBio) sequencing to show discordance in expression of ORF1 and ORF2 and wide variation in sequence diversity and coding potential of L1 transcripts. Bicistronic L1 constituted <0.01% of L1 transcripts, while over 80% of the neural L1 transcriptome is non-coding. Greater than 600 distinct, monocistronic ORF2 variants were identified, with functional variant analyses revealing a range of RT and EN activities in both full-length and truncated ORF2 variants. Thus, monocistronic ORF2 variants predominate, contributing diverse RT and/or EN functions independent of full-length L1 retrotransposition, thus promoting RT-mediated genome instability within the human brain.