Effect of APOE4 genotype and Alzheimer’s-related cerebrospinal fluid markers on cognitive decline in non-demented elderly

Contents

Abstract

Introduction

Materials and Methods

Results

Discussion

Conclusion

References

Appendix

Abstract

Keywords: Alzheimer's disease, APOE4, cerebrospinal fluid, mnemonic discrimination

Background. Alzheimer's disease (AD) affects millions worldwide, primarily in the elderly population. To date, virtually all drug therapies have failed to effectively treat AD. Subjects carrying the APOE4 gene are known to be at up to 12-times increased risk for AD. The underlying pathological changes in the form of ß-amyloid plaques (Aß) and tau tangles are detectable two decades before symptoms of memory decline arise and can be measured in cerebrospinal fluid (CSF). Early detection of AD in cognitively normal elderly would open a window of opportunity for treatment and preventative measures. This study aims to determine whether APOE4 genotype and AD-related pathological load in the form of CSF-derived Aß and tau can be used for prediction of mnemonic discrimination performance in non-demented elderly.

Methods. A total of 37 non-demented, healthy elderly were tested for APOE4 genotype. A subset of 16 female subjects underwent lumbar punctures to acquire CSF measures for Aß38, Aß40 and Aß42. Eight subjects were additionally tested for total tau levels, and 11 subjects underwent hyperphosphorylated tau measures. Cognition was measured through a behavioural mnemonic discrimination task for object, spatial and temporal domains.

Results. Age had no effect on CSF Aß and tau measures. APOE4 genotype, CSF Aß42 and p-tau status did not affect all three mnemonic discrimination domains. In a subset of females, performance on the object discrimination task decreased with age and increased with continuous CSF Aß38 levels. Conclusion. Neither APOE4 genotype nor CSF Aß and tau levels could explain differences in mnemonic discrimination ability, likely due to complex underlying interactions among several factors other than genetic risk or pathological load present in preclinical AD. Effects on object discrimination are likely caused by selective disruption of the perirhinal cortex (PrC) - lateral entorhinal cortex (LEC) object processing pathway in earlier stages of pathological spread. In future, multivariate studies are needed to investigate changes of CSF longitudinally, which could potentially lead to the discovery of AD biomarkers.

Introduction

The global prevalence of Alzheimer's disease (AD) and other dementias was 43.8 million in 2016 and is likely to exceed 152 million by 2050 (1-3). It affects 40% of individuals above 85 and is characterised by episodic memory decline, confusion, as well as other cognitive deficits in later stages of the disease (4). This global epidemic is estimated to cost over US$1 trillion, creating a serious burden for healthcare systems worldwide (3).

At an earlier stage, Mild Cognitive Impairment (MCI) describes abnormal forms of cognitive decline that have not reached the severity of dementia. MCI affects approximately 16-20 percent of elderly aged 65 or above and 24-32% of individuals with MCI progress to dementia (5, 6). In contrast to MCI, the diagnosis of dementia requires low performance of one or more cognitive domains with significant interference in everyday activities (7). AD is more prevalent in women, who make up two-thirds of all cases in the US (3). Further, African Americans are twice as likely and Hispanics one-and-a-half times as likely to be diagnosed with AD than Caucasians. While biological differences play a partial role, e.g. through a stronger AD-association with the ABCA7 risk gene in African Americans, socioeconomic characteristics such as education level are more likely to drive this disproportion. Further, African Americans have a higher prevalence of diabetes, obesity and hypertension, all of which are associated with AD (3, 8, 9).

With an ageing population, effective prevention and treatment are needed to cope with this burden. Research on drugs for AD treatment has been extremely unsuccessful with a failure rate of 99.6% and only one approved agent since 2004 (memantine) (10). These drugs trials involved acetylcholinesterase inhibitors, y/ß-secretase inhibitors and immunotherapeutics. The general failure of pharmaceuticals may have been due to late intervention - early introduction of pharmaceuticals may yield better results, given thorough investigation of potential side effects. Recently, off-label usage of drugs with indications other than AD has gained momentum with increased funding for drug repurposing studies. Trials of the diabetes medication Liraglutide are ongoing, and preclinical trials suggest possible neuroprotective effects through normalization of cerebral glucose uptake (11). Dysfunctions in insulin metabolism are thought to contribute to AD, causing a 73% increased risk of dementia in patients with type 2 diabetes mellitus (12). Comorbidity with depression is common in AD and often treated with selective serotonin reuptake inhibitors (SSRIs) which, despite controversial efficacy for depression symptoms, may delay disease progression ofAD (13).

The current lack of effective treatment emphasises the urge for early detection and prevention of the disease and the exploration of new biomarkers to predict the course of cognitive decline as early as possible. Early detection of AD-risk would open a window of opportunity for preventative measures in the preclinical stage and could further predict whether patients with MCI revert to normal cognition or develop dementia. Age alone fails to accurately predict this decline as there is high variability across age strata (subsets) with differential neuropathological and genomic profiles in the population.

As a major hallmark of the disease, the peptide ß-amyloid (Aß) is increased in AD patients and thought to contribute to synaptic loss and neurodegeneration (14, 15). The notion of Aß oligomer aggregation being the initial trigger of a cascade, involving the formation of neurofibrillary tau tangles (NFTs) leading to neuronal loss and inflammation (amyloid hypothesis) has been recently challenged. Synaptic loss can occur without Aß deposition and conversely be spared in Aß positive subjects, suggesting a more complex network involved in AD-related pathology with e.g. Aß accumulation being the downstream result of earlier events (16, 17).

The entorhinal cortex, more specifically the lateral entorhinal cortex (LEC) which connects the hippocampus to cortical areas of the brain, has been identified as the origin of amyloid deposition in preclinical AD (18). Epigenetic dysrégulation in the brainstem may precede this accumulation, with more evidence pointing to NFT formation in the earliest pre-symptomatic stage (19). The brainstem, particularly the dorsal raphe nucleus is further involved in neuropsychological manifestations of AD through its serotonergic projections (20).

More recently, the involvement of cholesterol as a catalyst for Aß deposition has been demonstrated (21). Cholesterol in the brain is known to be regulated by apolipoprotein E, which plays a key role in AD. Heterozygous carriers of the apolipoprotein E-e4 (APOE4) genotype have a 2-3-times higher risk for AD while homozygous carriers have a 12-times higher risk for AD (22).

To date, virtually all anti-Aß treatments have failed clinically (10, 16). Findings on potential effects of SSRIs on Aß levels are contradictory. While some studies found common SSRIs to prevent Aß formation, others suggested a nearly 2-fold increase in dementia risk (23-25). Trials on the effectiveness ofearly pharmacological intervention in asymptomatic Aß-positive individuals such as the A4 solanezumab study are ongoing (26).

The deposition of Aß is slow and can occur over decades before individuals reach the threshold of Aß positivity and the onset of hippocampal atrophy and memory impairment (27, 28). These changes in cerebrospinal fluid (CSF) and blood are measurable in subjects with asymptomatic, preclinical AD (29­31). However, it is important to note that while Aß positivity is a precondition for structural decline, it is also found in at least 20% of cognitively healthy elderly along with other neuropathologies (32, 33). This underscores the need for studies investigating early changes in non-demented elderly to identify shared properties among those with AD-related pathological load and absence of cognitive dysfunction.

Previous studies found a dose-dependent association between Aß burden and poor processing speed, working memory and reasoning, but not simple episodic memory in a full-lifespan sample of healthy individuals (34). Larger trials found a significant relation between Aß burden and decreased episodic memory and visuospatial performance in women and associative memory overall while other literature suggests no significant relationship between amyloid positivity and cognitive memory performance (32, 35-37). Nevertheless, some of these contradicting findings were based on limited function tests and lacked an adequate sample size.

Amyloid deposition does not necessarily precede NFT formation as tau positivity can occur in Aß- negative subjects, further challenging the notion of a sequential progression from Aß accumulation to NFT formation (38, 39). Elevated levels of tau have been consistently found in AD patients of which hyperphosphorylated tau (p-tau) levels show high specificity for AD (40, 41). An increase of total tau (t-tau) is found in numerous CNS disorders with neuronal and axonal degeneration such as acute stroke and is therefore unsuitable to discriminate AD from other disorders (41). On the contrary, t-tau levels in CSF can discriminate stable MCI patients from those that progressed to AD with 90% sensitivity and 100% specificity (42, 43).

The precise relationship between AD-related pathological status, genetic risk factors, cognitive decline and neurodegeneration remains unclear. Aß and tau status in non-demented elderly may predetermine their cognitive decline trajectory (34). Other factors such as genetics, sex-differences and overall cognitive reserve are known to modify this relationship, calling for a multivariate approach (44).

Longitudinal samples of CSF biomarkers are associated with later amyloid positivity in preclinical early middle-aged individuals which include the main constituent of plaques Aß42, t-tau as a measure of neuronal injury/death and p-tau as a measure for NFTs (30, 41). While CSF levels of Aß42 levels are expected to decline as AD-related pathologies develop, t-tau and p-tau levels are expected to increase. To date, CSF is regarded as a comparatively accurate measure for Aß and tau levels as it is in direct contact with the interstitial fluid and detects abnormal changes in Aß42 earlier than cortical Pittsburgh Compound B (PiB) positron emission tomography (PET) imaging (29-31).

Episodic memory loss is known to be a major hallmark of AD, with numerous studies on animals and humans testing episodic memory in the form of mnemonic discrimination in young and aged subjects. These studies demonstrated lower discrimination performance or 'memory rigidity' in object, spatial and temporal domains with higher age (45, 46). Beyond age, the relationship between episodic memory performance and pathological load remains unclear. The Mnemonic Similarity Test (MST) can assess hippocampal function on object, spatial and temporal discrimination. The MST (formerly known as Behavioural Pattern Separation Test, BPS) has been shown to have a high test-retest reliability and little practice effects. As a widely used measure for mnemonic discrimination performance and hippocampal memory, the MST provides a suitable outcome measure for this trial (46).

In this study, initial visit data from a subset of subjects enrolled in an ongoing longitudinal study on AD-biomarkers (PI. Michael Yassa) is analysed.

We aim to determine whether APOE4 genotype and AD-related pathological load in the form of CSF- derived Aß and tau can be used for prediction of mnemonic discrimination performance in cognitively healthy, non-demented elderly.

Materials and Methods

Participants

Participants were recruited as part of an ongoing study on biomarker exploration at the Yassa Lab for Translational Neuroscience at the University of California, Irvine (UCI). Recruitment involved phone calls, advertisements, flyers, brochures and other media. This analytic, cross-sectional study analysed initial visit data of a subset of 37 older adults (29 female, range 64-81 years, mean, M = 73.71 years, standard deviation, SD = 4.2) who were free of dementia at the time of enrolment, and screened against major neurological and medical disorders as well as magnetic resonance imaging (MRI) and PET contraindications (Table 1). Subjects with significant co-morbid neurological diseases (e.g. Parkinson's disease, multiple sclerosis, brain cyst, brain tumours), major health conditions (e.g. uncontrolled diabetes mellitus, uncontrolled hypertension) or psychiatric disorders (e.g. schizophrenia, bipolar disorder, anxiety disorder, major depression) were excluded from this study. Participants were tested on the Beck Depression Inventory (BDI) to identify depression co-morbidity and scored 3.54 on average, SD = 2.78 (n = 24), indicating no or minimal depression (Table 1) (47). Participants were fluent in English and had adequate visual and auditory acuity for neuropsychological and computerised testing. Most subjects were female (n=29), Caucasian (n=24) and highly educated (M = 16 education years, SD = 2.58, equivalent to a BA degree), limiting the generalizability of this cohort (Table 1). All subjects provided a written statement of informed consent prior to enrolment and were debriefed in accordance with guidelines set by the UCI Institutional Review Board (IRB).

Table 1. Demographics and neuropsychological tests

Abbildung in dieser Leseprobe nicht enthalten

Neuropsychological testing

Normal cognition, including dementia status and severity, were tested using the Clinical Dementia Rating (CDR) and Mini-Mental State Exam (MMSE), which are well-established measures for capturing clinical and cognitive decline across domains.

The CDR combines measures from over six domains, namely memory, orientation, judgement and problem solving, community affairs, home and hobbies, as well as personal care into one global score. Structured, standardised interviews gathered information on all domains from a collateral source (e.g. spouse) and the subject itself (48, 49). Subjects with a global CDR score of > 0 indicating questionable, mild, moderate or severe dementia were excluded.

The MMSE consists of a short, 30-point questionnaire primarily used to assess overall cognitive function. Subjects were tested on their orientation, verbal registration, attention and calculation, verbal recall, language and praxis. All participants must have obtained an MMSE score of > 27 (out of 30), which yields the optimal balance for both sensitivity and specificity (50).

Further neuropsychological testing involved the Alzheimer's Disease Cooperative Study (ADCS) Uniform Dataset (UDS) Neuropsychological Battery V3. Participants scoring 1.5 standard deviations (SD) or more outside age-adjusted norms on tests of the battery, such as the Rey Auditory Verbal Learning Memory Test (RAVLT) were excluded.

The RAVLT is a widely used tool to assess performance on attention and short-term verbal memory. During the test, subjects listened to a list of words (List A) and were asked to recall them immediately after. The listening and recall part were repeated over five times, testing for learning. The subjects then listened to a new, interfering list of words (List B) and were asked to recall the words from list A again immediately after, which was measured through the RAVLT immediate recall, A6-score. After a 20-minute delay, subjects were asked to recall list B, which was measured through the RAVLT delayed recall, A7-score. Finally, a mixed list of words from both List A and B was read to the subjects, who then had to recognise words they have heard before. This was measured through the RAVLT recognition, A8-score. On average, subjects performed above their age and education years stratified norm (means of4.3, 7.31 and 13.6 for trials A6, A7 and A8, respectively) as discussed by Correia et al. (2013), further warranting cognitive normality (Table 1) (51). Neuropsychological testing and scoring were conducted by trained neuropsychology testing technicians with established inter-rater reliability.

APOE4 genotyping

Genotyping was performed within the UCI Alzheimer's Disease Research Center (ADRC) to identify APOE genotype (e4 carriers vs noncarriers) in all test subjects. DNA was extracted from salivary samples, with each sample containing approximately 1-3 ml of whole saliva obtained by passive drool. Saliva samples were collected using DNA Genotek Oragene Discover OGR-500 kits with a median DNA yield of 110 pg.

To perform genotyping via polymerase chain reaction (PCR), an Applied Biosystems MicroAmp optical 384-well reaction plate, TaqMan primers, PCR MasterMix were used in conjunction with an Applied Biosystems Viia 7 qPCR machine.

CSF collection and analyses

Routine lumbar punctures (spinal tap) were performed by experienced clinicians within the UCI ADRC to provide samples of CSF in a subsample of 16 female older adults (range 71-81 years, M = 75.40 years, SD = 3.26). Lumbar punctures were completed under local anaesthesia and followed the Alzheimer's Disease Neuroimaging Initiative (ADNI) recommendations to ensure reproducible cross­laboratory performance standardisation. All subjects underwent an overnight fast of at least 6 hours prior to the procedure, which was scheduled in the morning to avoid time-of-day effect. To prevent post procedure leakage headaches, a small gauge needle was used to obtain 15-20 ml of CSF. Vital signs were taken before and after the procedure. Samples were stored in 0.5 ml aliquots and frozen at -800 °C. Using the Meso Scale Discovery Assay kit, CSF was sampled for Aß38, Aß40, Aß42, t-tau and p-tau levels. Duplicate measurements were taken and averaged through a calculate concentration mean (ccm). Positivity for Aß and tau was defined as Aß42 < 459 pg/ml, t-tau > 339 pg/ml and p-tau > 67 pg/ml, respectively, as discussed by Vos et al. (2013). These cut-offs were determined to best differentiate subjects with a global CDR score of 0 from those with a CDR score of 0.5 (52, 53). Subjects were contacted via phone call 24-28 hours after the procedure to confirm well-being and collect information about possible adverse effects. The mean delay between CSF collection and neuropsychological testing was 264 days or 8.6 months.

Cognitive performance

Participants were given the mnemonic similarity task (MST) to assess hippocampal memory function, as described by Stark, Yassa et al. (2013). This behavioural task was designed to test discrimination performance on object, spatial, and temporal domains (46).

Each test consisted of two phases: an incidental encoding phase and a subsequent testing phase where discrimination abilities were assessed. For all three domains, alternating images of everyday items were presented on a computer screen for 2 seconds at a time (500 ms Interstimulus interval), and participants were asked to respond to each image using two buttons (Figure 1). Participants were centred in front of the screen and kept their hands on the response buttons to reduce reaction time variations. Visual acuity and comprehension of task instructions have been ensured for all participants.

Abbildung in dieser Leseprobe nicht enthalten

Figure 1. Task design of MST-Object(A), Spatial (B) and Temporal (C) task. Abbreviations: MST, mnemonic similarity task; ISI, Interstimulus interval

During the encoding phase of the object task (MST-O), 120 items were shown one at a time, and subjects were asked whether the object belonged indoors or outdoors (concurrent indoor/outdoor judgement). For the subsequent testing phase, 160 items were presented to the subject, of which 40 were exact repetitions of the images shown before (targets), 80 were similar (lures), and 40 were completely new (foils). Lures varied in their similarity over two levels (high lures being more similar, low lures being more dissimilar) to manipulate object interference with the most similar lures having the highest interference. This time, subjects were asked to discriminate between items they saw previously (old) and similar or novel items (new). Falsely identifying lures as identical, old items (false alarm) is most likely driven by pattern completion processes (i.e. overgeneralization) whereas correctly discriminating lures as new images (correct rejection) is likely driven by pattern separation (Figure 2) (46).

During the encoding phase of the spatial task (MST-S), 160 items were shown, one at a time, in alternating positions (31 possible locations) on the screen, and subjects were asked for an indoor/outdoor judgement. For the testing phase, the same 160 items were presented on either the same or a different location than in the previous phase. The distance to the old location was varied over four levels (same location, small change, large change, corner-to-corner switch) with 40 images in each category to manipulate spatial interference. Subjects were then asked whether the item has remained in its old location or moved to a new location, testing their spatial judgement. During the encoding phase of the temporal task (MST-T), 32 items were shown one at a time, and subjects were again asked for an indoor/outdoor judgement. During the following testing phase, 16 pairs of two previously shown items were shown, side-by-side and subjects were asked which of the two (left vs right) was presented earlier than the other during the previous phase, testing their order judgement. The number of items shown in between (intervening images) was varied over four levels (adjacent, eight, sixteen, primacy/recency) to manipulate temporal interference. The temporal encoding- and testing phases were repeated over ten times.

Image type

Abbildung in dieser Leseprobe nicht enthalten

Figure 2. Behavioural classification chart for MST-0 and MST-S task. Image type describes if a target/repeat (MST-O: same image; MST-S: same location) or lure/foil (MST-O: similar/new image; MST- S: new location) was shown to the subject. Subject response indicates which button the participant pressed (MST-O: "Old" vs "New Item"; MST-S: "Old" vs "New Location").

As a score of mnemonic discrimination ability, a Lure Discrimination Index (LDI) was computed for each subject in the object and spatial task: p ("New" I Lure) - p ("New" I Target), i.e. the percentage of correct rejections (CR) minus the percentage of target misses (Figure 2). The LDI yields the advantage of correcting for response bias, e.g. for subjects who solely press one of two buttons. If subjects exclusively respond with "New" throughout the entire task, they will achieve all correct rejections for lures/foils but also get all misses for targets/repeats. Through deduction of the misses from the correct rejections, the LDI corrects for this bias (54). Subjects with a high LDI score were able to correctly identify similar (lure) and novel (foil) images as "new" (correct rejection) and less often falsely identify target repeats as "new" (miss). Thus, a high LDI score corresponds to a high discrimination ability, which in turn is associated with the neural mechanisms of pattern separation (55). Apart from a higher number of misses, lower LDI scores may reflect the false identification (i.e. overgeneralization) of similar lures and novel foils as "old" (false alarm) - which is associated with a bias towards pattern completion.

The testing phase of the temporal task had different response options, namely which stimuli were presented first instead of whether the stimulus was old or new, making the temporal task unsuitable for the LDI score. Instead, the score will be presented as the percentage of correct responses, P(Correct).

When available, alternate parallel versions ofthe tests were used to avoid practice effects. The order of tasks administered (object, spatial, temporal) was randomised and first/last items presented were corrected for to avoid primacy/recency (serial-position) effects. Breaks were provided to the subject when needed, and all tests were scored according to standardised procedures. The MST was run on PsychoPy2 1.85.

Data analysis

Output of the MTS data was organised, and scores were calculated using custom-scripted Python 3.7.3 and RStudio 1.2. The performance on each MST task was analysed through unpaired t-tests comparing test scores on the binary indicator of CSF Aß and tau status. Further, a two-way ANOVA was performed to test for interaction in an MST-0 subsample. Linear regressing was used to test for effects of CSF Aß and tau level on MST performance. Statistics were run, and graphs were computed using GraphPad Prism 8. Post hoc power analyses were based on an a error probability of 0.05 and were run using G Power 3.1.9.

Results

Age did not differ between APOE4 genotypes and did not predict levels ofCSFAßand tau

To compare mean ages between groups of subjects based on APOE4 genotype, unpaired t-tests were performed. Of the total study population of 37 subjects, 15 were tested positive for at least one allele of e4 (M = 73.33 years, SD = 4.29) whereas 22 were e4-negative (M = 73.97 years, SD = 4.22) with no significant age difference between the groups, t (35) = 0.45, p = 0.65.

Of the subsample of 16 females that underwent CSF analysis, seven were tested positive for at least one allele of e4 (M = 74.16 years, SD = 2.85) whereas nine were e4-negative (M = 76.37 years, SD = 3.38) with no significant age difference between the groups, t (14) = 1.39, p = 0.19.

Those subjects were tested for CSF levels of Aß peptides with chain lengths of 38 (M = 2916.09 pg/ml, SD = 895.91), 40 (M = 6014.57 pg/ml, SD = 1660.05) and 42 (M = 494.30 pg/ml, SD = 249.67) amino acids. Of those, eight subjects were additionally sampled for total tau (M = 655.90 pg/ml, SD = 359.32) and 11 for hyperphosphorylated tau (M = 56.51 pg/ml, SD = 17.52).

Using the cut-off values described earlier, eight individuals were classified as Aß42-positive, six as t- tau-positive and four as p-tau-positive (Table 2). CSF sample sizes were compromised due to a lack of available data in the early stage of the parent study and exclusions due to delay times of > 365 days between genotyping, CSF sampling and neuropsychological testing.

To test the effect of age on CSF measures, linear regressions were performed.

Age did not significantly predict levels of Aß38 (ß = 70.75, R[2] = 0.07, F (1,14) = 1.06, p = 0.32), Aß40 (ß = 127.2, R[2] = 0.12, F (1,14) = 1.99, p = 0.18), Aß42 (ß = 19.98, R[2] = 0.05, F (1,14) = 0.67, p = 0.43), t-tau (ß = 50.04, R[2] = 0.01, F (1,6) = 0.08, p = 0.79) and p-tau (ß = 1.70, R[2] = 0.16, F (1,9) = 1.76, p = 0.22).

[...]