Asymmetric amyloid deposition in preclinical Alzheimer’s disease: A PET study

Abstract

Introduction

The typical spatial pattern of amyloid-β (Aβ) in diagnosed Alzheimer’s disease (AD) is that of a symmetrical hemispheric distribution. However, Aβ may be asymmetrically distributed in early stages of AD. Aβ distribution on PET has previously been explored in MCI and AD, but it has yet to be directly investigated in preclinical AD (pAD). We examined how Aβ was distributed in individuals with pAD and MCI using ¹¹C-Pittsburgh Compound B (PiB) PET.

Methods

In this PET study, 79 subjects were retrospectively enrolled, including 34 controls, 24 pAD, and 21 MCI. All subjects underwent APOE genotyping, ¹¹C-PiB PET, MRI, and cognitive testing. We explored differences in Aβ load, Aβ lateralisation, and Aβ distribution, as well as associations between Aβ distribution and cognition.

Results

The Aβ asymmetry index (AI) differed between groups, with pAD having the highest Aβ AI as compared to both controls and MCI. There was no clear Aβ lateralisation in pAD, but there was a non-significant trend towards Aβ being more left-lateralised in MCI. There were no correlations between the cognitive scores and Aβ AI or Aβ lateralisation in pAD or MCI.

Conclusion

The distribution of Aβ is most asymmetrical in pAD, as Aβ first starts accumulating, and it then becomes less asymmetrical in MCI, when Aβ has spread further, suggesting that more pronounced asymmetrical Aβ distribution may be a distinguishing factor in pAD. Longitudinal studies examining the distribution of Aβ across the AD continuum are needed.

1. Introduction

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder typically affecting individuals above the age of 65. It is associated with worsening cognitive impairment and the most pervasive symptom is memory loss. AD is characterised by a brain pathology consisting of extracellular deposits of amyloid-β (Aβ) plaques [2], intracellular neurofibrillary tau tangles [1], and cortical inflammation via activation of multiple inflammatory pathways mediated through activated microglia [11]. While the exact relationship between these pathologies has yet to be determined, Aβ aggregation is generally believed to be the first pathological event to occur in the cortex [15], ahead of tau aggregation and possibly also microglial activation. Accumulating evidence suggests that AD can be understood as a continuum [31], with the disease beginning in brain areas up to two decades before the onset of clinical symptoms. This disease stage, where the individual has incipient AD pathology but no overt symptoms, has been classified as the preclinical disease stage [28].

The definition of preclinical AD (pAD) is not fully determined, but individuals are generally regarded as having pAD if they are overtly cognitively normal but are accruing Aβ plaques. The most significant genetic modifier of late-onset AD is the Apoliprotein E ε4 (APOE ε4) allele [24]. The human APOE gene is located on chromosome 19 and has two frequent single nucleotide polymorphisms (SNPs) on exon 4, rs429358 and rs7412, which results in four different alleles ε1, ε2, ε3, and ε4, of which the last three are the most common in humans. Carrying one APOE ε4 allele increases the risk of developing AD by 2–4 times. Specifically, APOE ε4 carriage has been shown to influence both the deposition and clearance of Aβ in both neurons and vessels in the brain [26].

Aβ plaques can be identified in vivo with positron emission tomography (PET) using radiotracers that bind to beta sheeted Aβ fibrils. The tracer, ¹¹C-Pittsburgh Compound B (¹¹C-PiB), has a high affinity for Aβ plaques [18], which makes it particularly good at detecting even early signs of Aβ deposition in the brain.

The typical spatial distribution of Aβ fibrils seen with PET in symptomatic AD is that of equal hemispheric distribution, with plaques affecting cortices in a symmetrical pattern, originating in the precuneus and the posterior cingulate cortices and spreading forward into inferior temporal and frontal cortices [8], [20], [30]. There is increasing evidence, however, that this symmetrical pattern of Aβ [6], [25], [34] is not necessarily present in early stages of AD. A recent PET study compared the spatial distribution of Aβ in healthy controls, patients with mild cognitive impairment due to AD (AD-MCI), and AD patients, and they found that the AD-MCI patients had a more asymmetrical Aβ distribution than the AD patients [34]. This would suggest that Aβ is even more asymmetrically distributed in the earliest stages of the disease, and then becomes increasingly symmetrical as the disease progresses. Another PET study found that, when lateralised Aβ is seen in AD-MCI and AD, it is associated with lateralised cognitive symptoms [6]. It has been reported that AD pathology is generally more left-lateralised [32]. This could either suggest an important distinction between right- and left-lateralised pathology along the disease continuum or a bias in symptoms towards word finding rather than perceptual difficulties at patient presentation.

So far, the degree of asymmetry in the distribution of Aβ has, however, not been reported in pAD, but only in AD-MCI and, even more so, in AD, including both typical [34]and atypical [6]AD variants. In this study, we compared the distribution of Aβ in pAD to that in AD-MCI in order to examine how Aβ develops in the earliest disease stages, how it is distributed, whether it is lateralised and, if so, whether either the distribution or the lateralisation is associated with specific cognitive symptoms. Since Aβ distribution has already been examined in the late stages of AD, we chose in this study to focus only on the early AD stages, namely pAD and AD-MCI.

2. Methods

2.1. Subjects

Findings from a total of 79 subjects that had been recruited for our previous MCI and current pAD PiB PET study were used for the analysis. The AD-MCI subjects were recruited from local clinics as well as by advertisement for a study reported previously [21]. The overtly healthy subjects were recruited only by advertisement. All 79 subjects had APOE genotyping, anatomical MRI, Aβ PET with ¹¹C-PiB, cognitive screening with the Mini-Mental State Examination (MMSE) [5], and neuropsychological assessment. Additionally, the AD-MCI subjects also underwent screening with the Clinical Dementia Rating Scale (CDR) [12] to detect incipient clinical impairment. In the MCI study, a cortical:cerebellar ¹¹C-PiB uptake ratio threshold of 1.5 was used to categorise subjects as being with or without abnormal Aβ deposition [13]. Subjects were classified as cognitively normal (CN-Aβ-) if they did not have any cognitive complaints. Subjects were classified as preclinical (pAD-Aβ + ) if they had no memory or other cognitive complaints but showed abnormal Aβ deposition on their PET scans. Subjects were classified as having MCI due to AD (AD-MCI-Aβ + ), if they had a CDR score of 0.5, memory complaints corroborated by an informant, abnormal Aβ deposition on a PET scan, and were able to function independently.

2.2. APOE genotyping

A blood sample was obtained from each subject in order to determine their APOE genotype. Anticoagulated whole blood was collected in EDTA-containing tubes, and frozen to allow DNA extraction in batches. DNA was extracted using the DSP mini kit on the QiaSymphony SP platform (Qiagen). We used polymerase chain reaction (PCR) and Sanger sequencing to determine APOE genotypes. The PCR products were sequenced using BigDye Terminator version 1.1 (Thermofisher) and separated on an Applied Biosystems 3500XL Genetic Analyzer (Thermofisher). Sequence traces were aligned to the APOE reference sequence NM_000041 using SeqScape version 2.7 (Thermofisher). This allowed us to determine if the APOE genotype was ε2/ε2, ε2/ε3, ε3/ε3, ε3/ε4, or ε4/ε4, with ε1/ε3 and ε2/ε4 being interchangeable and therefore not possible to distinguish.

2.3. Positron Emission Tomography (PET)

¹¹C-PiB PET was performed with a High-Resolution Research Tomograph (HRRT, CTI/Siemens, Knoxville, TN) scanner according to a previously described scan protocol [21]. A mean dose of 394 (+/- 39) MBq ¹¹C-PiB was injected through the subject’s intravenous catheter followed by a 10 ml saline flush, after which subjects rested for 30 min before being placed supine in the scanner. A 6 min transmission scan was performed prior to each subject’s scan. List-mode PET scan was acquired at 40–90 min post injection. Subsequently, the data was rebinned into five frames of 10 min each. Experienced nuclear medicine physicians (JFA/PP) reviewed all PET scans.

2.4. Magnetic resonance imaging (MRI)

MRI was performed on a Skyra 3 T (Siemens, Erlangen, Germany) scanner. A T1 MP2RAGE (magnetisation prepared rapid gradient echo with two gradient echo sequence images) sequence was obtained along with a T2 and T2 FLAIR (fluid attenuated inversion recovery) sequence. The T1 MP2RAGE image was used for co-registration of MRI with PET, normalisation into MNI standard space, and generation of grey matter masks. The T2 FLAIR image was used to exclude structural and space-occupying pathology, e.g., tumours, chronic subdural hematoma etc. An experienced neuroradiologist (RBD) reviewed all MRI scans.

2.5. Neuropsychology

The subjects were assessed with a neuropsychological test battery of different cognitive tests. All subjects were assessed, except 5 subjects from the pAD-Aβ + group due to these subjects withdrawing from the pAD study. The test battery included: MMSE, Rey Auditory Verbal Learning Test (RAVLT; [27]), Trail Making test, part A and B [22], Digit span – Forwards, Backwards, and Ranking [33], Golden Stroop – Words, Colours, and Interference [7], Block designs [33], Boston Naming test [16]; Controlled Oral Word Association Test – category and phonemic [3], and Similarities subtest [33]. Assessments were overseen by an experienced psychologist (MFD/PLK) and performed by either psychologists or trained psychology research assistants. Additionally, the AD-MCI subjects were assessed with the CDR by trained interviewers.

2.6. Image analysis

MINC Software (http://en.wikibooks.org/wiki/MINC McConnell Brain Imaging Centre, Montreal, Canada), was used to perform the initial PET-to-MRI co-registration on each individual subject, segment MRI into grey matter (GM), white matter (WM), and cerebrospinal fluid (CSF), and to spatially normalise the images. A probabilistic atlas [10] was used in concert with GM masks to define regions of interest (ROIs) on each subject’s GM template.

The spatially normalised ¹¹C-PiB PET images were summed from 60 to 90 min, and voxel signals were divided by the mean signal from cerebellar GM and smoothed 8 mm. to generate ¹¹C-PiB PET SUVR images. These images were used as the basis for subsequent computations.

A composite cortical ¹¹C-PiB PET SUVR was used to determine Aβ status in previous studies [21]. This was computed as a volume-weighted average of six bilateral regions (frontal, parietal, lateral, posterior temporal, posterior cingulate, and precuneus).

Data points for 15 cortical regions were plotted to determine which cortical regions showed the most Aβ asymmetry. An ¹¹C-PiB PET SUVR sum score of five bilateral regions (frontal, medial temporal, precuneus, posterior cingulate, and the parietooccipital region) was then computed to identify those regions that showed the most Aβ asymmetry. As opposed to the composite score, the sum score may not indicate regions that have the highest Aβ load, but rather reflect the highest degree of regional Aβ asymmetry, i.e., the biggest difference in Aβ load in the left as compared to the right hemisphere. Thus, the composite score was primarily used to determine Aβ status and the sum score was used to examine Aβ AI.

The Asymmetry Index (AI) was calculated from the ¹¹C-PiB PET SUVR images as a percent score as: [AI %] = (L-R)/(L + R) × 200. In order to show asymmetry regardless of lateralisation, images showing more Aβ in the right hemisphere were flipped, giving the appearance of all images having more Aβ in the left hemisphere.

In order to examine each group as a whole, an average ¹¹C-PiB PET image was computed separately for each of the individual groups. ¹¹C-PiB PET difference images were also calculated from the average ¹¹C-PiB PET images. As ¹¹C-PiB has a high affinity for Aβ, but is also lipophilic [18], the WM signal can be high in non-AD cases. Thus, in order to better distinguish the GM signal from the WM signal in the pAD and AD-MCI subjects the WM signal in the CN-Aβ- subjects was subtracted for one figure (see Fig. 1). Difference images were created by subtracting the individual pAD and MCI subjects’ ¹¹C-PiB PET images by the CN-Aβ- group’s average ¹¹C-PiB PET images.

Fig. 1.

Areas with a high ¹¹C-PiB PET signal. Difference image for the MCI-Aβ + group. The WM signal has been subtracted from the GM signal in the CN-Aβ- group prior to calculating the difference image. The average ¹¹C-PiB SUVR of the whole MCI-Aβ + group subtracted by the average ¹¹C-PiB SUVR of the whole CN-Aβ- group. The image is co-registered on to a T1 average brain and thresholded at SUVR > 1.5.

2.7. Statistics

Statistical Package for Social Sciences ver. 28 (SPSS 28) (IBM, New York, United States) was used for statistical analyses. Demographics were assessed with descriptive statistics. Data were described using either means (M) and standard deviations (SD) or medians (Mdn) and interquartile ranges (IQR). Between-group comparisons on the imaging data were performed using the Kruskal-Wallis test and follow-up Mann-Whitney tests. Between-group comparisons on the cognitive data were performed using Independent T-tests. Correlations were performed using the Kendall’s Tau test. Significance was determined as p < 0.05.

MINC software was used for analyses of imaging variables and to produce image figures.

3. Results

Demographics are presented in Table 1. There were 34 CN-Aβ-, 24 pAD-Aβ+, and 21 AD-MCI-Aβ + included in the study (mean age: 69.1, range: 54–85 years). The subjects in the AD-MCI-Aβ + group were significantly older than the subjects in the CN-Aβ- and the pAD-Aβ + groups, however, considering that AD is an age-related, progressive disease, this is to be expected. Surprisingly, there were more men than women in the pAD-Aβ + and AD-MCI-Aβ + groups, despite women being statistically more likely to develop AD, but this may be a consequence of our relatively small sample size. The subjects did not differ significantly in MMSE score, although the AD-MCI-Aβ + group had a lower score than the CN-Aβ- and pAD-Aβ + group.

Table 1.

Demographics of 79 included subjects.

	CN-Aβ- (n = 34)	pAD-Aβ+(n = 24)	AD-MCI-Aβ+(n = 21)
Demographics (M ± SD)
Age (years)	63.0 ± 5.8	70.3 ± 3.4	74.2 ± 5.7
Sex (% women)	73 %	44 %	34 %
APOE ε4 (%)	97 %	100 %	62 %
MMSE score	28.7 ± 1.3	28.7 ± 0.8	26.5 ± 1.6
GDS-15	0.95 ± 1.4	0.89 ± 1.0	1.57 ± 1.7
11C-PiB composite SUVR	1.4 ± 0.1	1.9 ± 0.4	2.6 ± 0.4
Fazekas score	1.11 ± 1.1	1.52 ± 0.6	1.71 ± 0.9

*M = Mean, SD = Standard Deviation.

MMSE = Mini Mental State Examination, GDS-15 = 15-item Geriatric Depression Scale.

¹¹C-PiB = ¹¹C-Pittsburgh Compound B, SUVR = Standard uptake value ratio.

In this sample, when Aβ was present, the results showed that the ¹¹C-PiB binding to Aβ was higher in the precuneus, the posterior cingulate cortex, and the inferior frontal cortex, than it was for other regions. When the ¹¹C-PiB PET WM activity was subtracted, and the image thresholded at 1.5, only the binding in these cortical areas survived, suggesting a higher Aβ load in these areas (see Fig. 1).

A Kruskal-Wallis test showed that the Aβ AI differed significantly between groups H(2) = 9.945, p = 0.007. Post-hoc Mann-Whitney U tests revealed that the pAD-Aβ + group (Mdn = 0.008, IQR = 0.057) had higher AI compared to the CN-Aβ- group (Mdn = -0.007, IQR = 0.017) (p = 0.002) as well as a higher AI compared to the AD-MCI-Aβ + group (Mdn = 0.002, IQR = 0.044). The difference between the pAD-Aβ + and CN-Aβ- group AI was statistically significant (p = 0.002). However, the difference between the pAD-Aβ + and AD-MCI-Aβ + group AI was not significant (p = 0.126), and neither was the difference between the AD-MCI-Aβ + and CN-Aβ- group AI (p = 0.135). Furthermore, the Aβ AI range was greater in the pAD-Aβ + group (range = 0.089) than in the CN-Aβ- group (range = 0.051) and the AD-MCI-Aβ + group (range = 0.063). Thus, the median as well as the range was higher in the pAD-Aβ + group than in the other two groups (see Fig. 2), indicating that this group as a whole had more asymmetry, but also that it included more extreme cases than the other two groups.

Fig. 2.

Degree of asymmetry in each group. Boxplot of the AI for an ¹¹C-PiB SUVR sum score of five bilateral regions (frontal lobe, precuneus, posterior cingulate, medial temporal lobe, and parietoeoccipital lobe) for each group. Positive values indicate more Aβ in the right hemisphere, and negative values indicate more Aβ in the left hemisphere.

Despite previous conflicting findings regarding lateralization in Aβ deposition in early AD cases, we did not find that Aβ was particularly lateralized to either side, although it could be argued that there was a slight, albeit non-significant tendency towards Aβ being more left-lateralized in the AD-MCI-Aβ + group (see Fig. 2).

Given the lack of consistent lateralization, PET images showing higher right hemisphere PiB uptake were flipped so that all the images were transformed to have more Aβ in the left hemisphere. This was done in order to be able to examine the degree of asymmetry in Aβ, regardless of whether there was more Aβ in the right or left hemisphere. Despite flipping some of the images for comparison, there was no apparent asymmetry in the CN-Aβ- group. However, there was clear asymmetry in both the pAD-Aβ + group and the AD-MCI-Aβ + group, with the pAD-Aβ + group having more asymmetry than the MCI-Aβ + group (see Fig. 3).

Fig. 3.

Aβ load and Aβ asymmetry for each group. Axial ¹¹C-PiB PET average images and ¹¹C-PiB PET SUVR images for each of the three groups. Those images showing the most Aβ in the right hemisphere were flipped. Thus, all images were made to appear as having the most Aβ in the left hemisphere. A voxel asymmetry was performed on all images for values > 1.5. AI was calculated as a percent score using the formula (L-R)/(L + R) × 200 [25]. Subsequently, average images were created for each group individually. A) Average SUVR ¹¹C-PiB PET images for each of the three groups. B) Left-flipped asymmetry ¹¹C-PiB PET images for each of the three groups.

Further, we investigated the relationship between Aβ load and Aβ AI. Using SPSS 28, a scatter plot was produced to show the relationship between Aβ load and asymmetry. A quadratic fit line (R² Quadratic = 0.145) was added to visualise the trend in this relationship. Our findings suggest that when Aβ initially accumulates in the preclinical stage, it first appears in an asymmetrical manner, but then as Aβ spreads across both hemispheres during the MCI stage, it becomes more symmetrically distributed (see Fig. 4).

Fig. 4.

Relationship between ¹¹C-PiB PET composite SUVR and sum score AI. Scatter plot showing the ¹¹C-PiB PET SUVR score plotted against the ¹¹C-PiB PET AI for all subjects with an added quadratic fit line.

Finally, we investigated whether there was an association between Aβ AI and cognition as well as between Aβ lateralization and cognition, respectively.

As expected, CN-Aβ- group performed better than both the pAD-Aβ + and AD-MCI-Aβ + groups on most of the cognitive tests, with the exception of Verbal Fluency – Phonemic. The pAD-Aβ + group performed better than AD-MCI-Aβ + group on all of the cognitive tests (see Table 2 in the Appendix).

To examine associations between Aβ AI and cognitive test performance, correlations were run between Aβ AI and individual cognitive tests using Kendall’s Tau. This test was chosen due to the high number of tied ranks for the different tests. Correlations were carried out at group-level. The correlations between Aβ AI and cognitive test results in the CN-Aβ- or pAD-Aβ + groups did not reach or show a trend towards significance (p > 0.1). There was one significant correlation between Aβ AI and cognitive test results in the AD-MCI-Aβ + group, namely for Aβ AI and Digit span – Backwards (Tau = 0.68, p = 0.03), but no other correlations showed a trend towards significance (p > 0.2). No clear pattern could be identified in the correlations between Aβ AI and cognitive test results across or within groups, with no single tests showing a stronger trend towards significance in all groups (see Table 3 in the Appendix). Correlations were also carried out between the CN-Aβ- group and the combined pAD-Aβ + and AD-MCI-Aβ + groups, since the latter can be thought of as representing an early AD continuum, however, there were no significant correlations at the p > 0.05 two-tailed significance level between Aβ AI and cognitive test results in this combined group.

To assess the association between Aβ lateralization and cognitive test performance, the subjects were categorised into two groups based on whether they showed more left- or right- Aβ lateralization. Only the subjects in the pAD-Aβ + and AD-MCI-Aβ + were assessed. Once the subjects were sorted into one of the two groups, an independent t-test was run on each of the cognitive tests. There were no significant differences between Aβ lateralization and cognitive test results in the pAD-Aβ + nor in the AD-MCI-Aβ + group at the p > 0.05 two-tailed significance level (see Table 4, Table 5 in the Appendix).

4. Discussion

In the present study, we explored Aβ load and distribution in subjects with preclinical AD as compared to cognitively normal and AD-MCI subjects using ¹¹C-PiB PET. We found that while there was no obvious asymmetry in the ¹¹C-PiB PET signal distribution in the CN individuals, there was clear asymmetry in both the pAD and the AD-MCI individuals. The pAD group had a significantly higher Aβ AI than the CN group as well as a higher, albeit non-significant AI than the AD-MCI group. The AD-MCI group, however, did not have a significantly higher AI than the CN-Aβ- group. In subjects with asymmetrical Aβ-load, there was no predilection for either the left nor right side. Finally, we found no significant correlations between the degree of Aβ asymmetry and cognition and no significant differences between left versus right Aβ lateralization and cognition.

The results from this study provide additional evidence that Aβ is asymmetrically distributed in the earliest stages of AD. Previous studies have shown a high degree of asymmetric Aβ distribution in AD-MCI compared to clinical AD [6], [34], however, our results suggest that there is an even higher degree of asymmetry in preclinical AD. The result showing that the pAD group had a significantly higher Aβ AI than the CN group, and a higher, though not significantly higher AI than the AD-MCI group, while the AD-MCI group did not have a significantly higher AI than the CN-Aβ- group supports the notion that pAD and AD-MCI taken together represent a continuum of early AD, the Aβ distribution becomes less symmetrical once AD-MCI manifests. Preclinical AD has garnered increasing attention within the field in recent years, as it is believed to be the earliest detectable disease stage [29]. These results provide evidence that asymmetric Aβ distribution is, in fact, a distinguishing feature of this very early stage.

The subjects in our study were primarily APOE ε4 carriers. The general effects of APOE ε4 on AD risk are still not fully understood, however, there is substantial evidence that it is associated with onset of Aβ accumulation. However, while APOE ε4 carriage is known to increase risk of Aβ accumulation in the elderly brain (Lim et al., 2017), there is no pathological evidence to suggest that it has an effect on Aβ distribution. Thus, APOE ε4 carriage is unlikely to have altered the nature of Aβ distribution in our subjects.

The results from the study do not indicate that Aβ is lateralized to either hemisphere in preclinical AD. However, they could potentially reflect a tendency towards Aβ being more left-lateralized in AD-MCI. Previous studies have suggested that Aβ is more left-lateralized in AD, although findings are conflicting [4], [32], and it has been suggested that left- and right-lateralization of Aβ may be associated with specific cognitive symptoms. Our results did not reveal significant correlations between either Aβ asymmetry and cognition in the CN-Aβ-, pAD-Aβ+, or AD-MCI-Aβ + group, except for one significant correlation between Aβ AI and Digit Span – Backwards in the MCI-Aβ + group. This result is surprising, given that there were no other significant correlations and that this type of test, which primarily measures executive functioning [9], is not typically the most impaired in MCI related to AD. This single, positive result is likely to be spurious, given the small sample size, and would not survive a multiple comparison correction.

Additionally, the results did not reveal any differences between left- and right Aβ lateralization and cognition in neither the pAD-Aβ + nor AD-MCI-Aβ + group, although there was a slight tendency towards left-lateralization in the AD-MCI-Aβ + group. Other studies have found some significant correlations between Aβ lateralization and cognition in MCI due to AD. One study found left-lateralized pathology to be associated with more severe language impairment, right-lateralized pathology to be associated with more severe visuospatial impairment, and symmetrically distributed pathology to be associated with predominant memory impairment [6], whereas in another study, right-lateralized pathology was found to be associated with language deficits [17]. These conflicting results could be attributable the presence of atypical AD variants that are known to have more lateralized pathology and symptoms such as logopenic progressive aphasia [34]. It may also be attributable to the fact that numerous cognitive tests measure more than one localized cognitive domain. Some memory tests not only depend on memory function but also on language functions, executive functions, and visuospatial functions. Many cognitive tests are in themselves also language-based and may therefore intrinsically measure language integrity along with other functions. No other studies have found an association between left- or right-lateralized pathology and cognitive impairment in preclinical AD. This is not surprising, however, as preclinical AD is defined by the presence of AD biomarkers in the absence of overt cognitive decline [28]. Some studies using particularly sensitive paired association cognitive tests have shown an association between Aβ load and cognitive deficits in preclinical AD [23]but an association between Aβ lateralization and cognitive deficits in preclinical AD has so far not been investigated. Further, since the increased Aβ load is typically at the lower in preclinical AD, it may, in fact, be difficult to find associations between certain facets of Aβ distribution and cognitive deficits in this early stage.

The results pertaining to the relationship between Aβ load and Aβ AI indicate that Aβ begins in an asymmetrical matter and then becomes increasingly symmetrical as the diseases progresses. This shift in Aβ distribution is occurring during the prodromal AD-MCI stage where previous studies have reported that tau tangle pathology and activation of microglia are also present [14], [21]. Our findings reinforce the concept that pathological trajectories in AD brain begin many years before symptoms present [28]. We found that Aβ deposition typically appeared first in the precuneus, the posterior cingulate cortices, and the frontal cortices – regions that are nodes in the default mode network [20]. Thus, these regions may be especially prone to forming Aβ plaques. Additionally, we saw a tendency that when Aβ was present in a certain region in one hemisphere, e.g., inferior frontal gyrus, it was more likely to appear in the same region in the contralateral hemisphere followed by other regions. This could suggest that Aβ in AD starts in a small region of one hemisphere and spreads via the connectome leading to initial distribution in the ipsilateral hemisphere. In the human brain, only 1 % of axons are commissural and for each anatomical region, these axons project mainly to the anatomical homolog of the contralateral hemisphere. This mechanism of so-called connectome-dependent propagation has recently been proposed as an explanation of the well-known asymmetry seen in Parkinson’s disease pathology [19]. Larger longitudinal studies with not only pAD and AD-MCI subjects but also established AD subjects are needed to confirm if AD pathology follows the same propagation.

Limitations in this study are small sample size, differences in age and gender between the different groups, the differences in age and gender between the different groups, the cross-sectional design, and the lack of clinical AD patients, which unfortunately does not allow for direct examinations of how Aβ spreads along the AD continuum over time. This study was intended to examine Aβ distribution in only the very early stages of AD, and a larger, longitudinal study based on a wider sample size including subjects with preAD and AD-MCI as well as AD may be able to establish the exact Aβ distribution pattern corresponding to each known disease stage. We hope such a study will be performed in the future. This would be of great value to the field, as it would not only further our understanding of AD as a disease, but also ultimately assist in diagnosis and management.

5. Conclusion

Previous studies have shown that the degree of Aβ deposition asymmetry is higher in MCI due to AD than in AD itself. This study, however, indicates that the highest degree of Aβ asymmetry may, in fact, be present in pAD. Additionally, while this study indicates that the degree of Aβ asymmetry is high in pAD, it does not indicate that Aβ is necessarily lateralised to either the left or right side, or that it is associated with specific types of cognitive impairements in this early stage. Finally, this study suggests that Aβ may begin asymmetrically in the brain, and then spread possibly via the connectome to become a more symmetrical pattern over time. Larger, longitudinal studies are needed to confirm such a pattern along the AD continuum. This may aid in furthering our understanding and management of AD, even from a very early stage.

Funding

This study was funded by the European Union’s Horizon 2020 research and innovation programme – Fast Track to Innovation (FTI), [grant agreement 820636].

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Appendix.

Table 2.

Neuropsychological test results.

	CN-Aβ- (n = 34)	pAD-Aβ+ (n = 19)	MCI-Aβ+ (n = 21)
RAVLT (Immediate Recall)	M = 50.91, SD = 8.58	M = 42.67, SD = 11.83	M = 30.57, SD = 9.83
RAVLT (Delayed Recall)	M = 10.06, SD = 2.93	M = 7.56, SD = 4.09	M = 3.95, SD = 3.32
RAVLT (Recognition)	M = 13.75, SD = 1.28	M = 12.94, SD = 2.18	M = 12.05, SD = 2.63
Trail Making A	M = 38.62, SD = 10.97	M = 41.17, SD = 15.83	M = 51.01, SD = 14.57
Trail Making B	M = 82.94, SD = 27.09	M = 112.39, SD = 42.51	M = 170.39, SD = 101.39
Stroop – Words	M = 87.97, SD = 16.09	M = 95.06, SD = 9.94	M = 74.26, SD = 12.78
Stroop – Colours	M = 70.18, SD = 10.11	M = 66.94, SD = 11.66	M = 51.80, SD = 9.631
Stroop – Interference	M = 39.91, SD = 8.94	M = 33.28, SD = 10.10	M = 21.10, SD = 10.50
Digit span – Forwards	M = 8.35, SD = 1.65	M = 8.50, SD = 1.54	M = 8.05, SD = 1.88
Digit span – Backwards	M = 8.79, SD = 2.08	M = 8.72, SD = 1.84	M = 6.62, SD = 1.49
Digit span – Ranking	M = 8.82, SD = 1.86	M = 8.56, SD = 1.247	M = 5.62, SD = 2.03
Verbal Fluency – Category	M = 24.23, SD = 4.85	M = 25.41, SD = 4.86	M = 17.43, SD = 4.52
Verbal Fluency – Phonemic	M = 15.72, SD = 4.76	M = 16.65, SD = 5.77	M = 12.67, SD = 4.36
Boston Naming	M = 28.65, SD = 5.87	M = 27.56, SD = 2.83	M = 25.08, SD = 3.45
Block Designs	M = 43.18, SD = 9.50	M = 38.44, SD = 9.68	M = 32.19, SD = 12.36
Similarities	M = 23.50, SD = 4.81	M = 26.22, SD = 4.18	M = 18.88, SD = 4.27

M = Mean, SD = standard deviation.

Table 3.

Correlation between Aβ AI and cognitive test performance.

	CN-Aβ- (n = 34)	pAD-Aβ+ (n = 19)	MCI-Aβ+ (n = 21)
MMSE	Tau = 0.143 (p = 0.254)	Tau = -0.019 (p = 0.909)	Tau = 0.126 (p = 0.455)
RAVLT – Immediate Recall	Tau = 0.74 (p = 0.529)	Tau = -0.098 (p = 0.523)	Tau = -0.106 (p = 0.505)
RAVLT – Delayed Recall	Tau = -0.008 (p = 0.947)	Tau = -0.118 (p = 0.452)	Tau = -0.165 (p = 0.313)
RAVLT – Recognition	Tau = 0.171 (p = 0.177)	Tau = -0.141(p = 0.373)	Tau = -0.102 (p = 0.538)
Trail Making A	Tau = 0.082 (p = 0.479)	Tau = -0.239 (p = 0.121)	Tau = 0.167 (p = 0.290)
Trail Making B	Tau = 0.157 (p = 0.865)	Tau = -0.061 (p = 0.690)	Tau = -0.091 (p = 0.566)
Stroop – Words	Tau = -0.20 (p = 0.795)	Tau = 0.086 (p = 0.576)	Tau = 0.144 (p = 0.365)
Stroop – Colours	Tau = 0.083 (p = 0.471)	Tau = 0.028 (p = 0.852)	Tau = 0.111 (p = 0.486)
Stroop – Interference	Tau = 0.170 (p = 0.145)	Tau = -0.045 (p = 0.770)	Tau = 0.058 (p = 0.717)
Digit Span – Forwards	Tau = -0.090 (p = 0.457)	Tau = -0.139 (p = 0.376)	Tau = 0.005 (p = 0.975)
Digit Span – Backwards	Tau = 0.062 (p = 0.606)	Tau = -0.145 (p = 0.359)	Tau = 0.368 (p = 0.030)*
Digit Span – Ranking	Tau = 0.035 (p = 0.770)	Tau = -0.061 (p = 0.702)	Tau = -0.041 (p = 0.806)
Verbal Fluency – Category	Tau = 0.098 (p = 0.416)	Tau = 0.008 (p = 0.957)	Tau = 0.059 (p = 0.714)
Verbal Fluency – Phonemic	Tau = 0.064 (p = 0.598)	Tau = 0.000 (p = 1.000)	Tau = 0.049 (p = 0.761)
Block Designs	Tau = 0.084 (p = 0.471)	Tau = -0.032(p = 0.832)	Tau = -0.034 (p = 0.832)
Boston Naming	Tau = 0.131 (p = 0.278)	Tau = 0.047 (p = 0.767)	Tau = 0.128 (p = 0.429)
Similarities	Tau = 0.194 (p = 0.098)	Tau = -0.127 (p = 0.409)	Tau = 0.073 (p = 0.649)

Tau = Kendall’s Tau statistic, p = significance level.

* Significant at p < 0.05.

Table 4.

Differences between Aβ lateralization and cognitive test performance in the pAD-Aβ + group.

	Left-lateralized (n = 10)	Right-lateralized (n = 9)	Group difference (Two-Sided p)
MMSE	M = 28.64, SD = 1.027	M = 28.67, SD = 0.651	t = -0.085, p = 0.933
RAVLT – Immediate Recall	M = 46.33, SD = 8.411	M = 39.00, SD = 14.027	t = 1.345, p = 0.197
RAVLT – Delayed Recall	M = 8.33, SD = 4.183	M = 6.78, SD = 4.086	t = 0.798, p = 0.437
RAVLT – Recognition	M = 13.67, SD = 2.179	M = 12.22 SD = 2.048	t = 1.449, p = 0.167
Trail Making A	M = 44.89, SD = 19.290	M = 37.44, SD = 11.38	t = 0.997, p = 0.334
Trail Making B	M = 116.11, SD = 41.47	M = 108.67, SD = 45.72	t = 0.362, p = 0.722
Stroop – Words	M = 92.33, SD = 9.083	M = 97.78, SD = 10.53	t = -1.174, p = 0.257
Stroop – Colours	M = 66.44, SD = 11.91	M = 67.44, SD = 12.12	t = -0.177, p = 0.862
Stroop – Interference	M = 34.44, SD = 10.19	M = 32.11, SD = 10.49	t = 0.479, p = 0.639
Digit Span – Forwards	M = 8.44, SD = 1.130	M = 8.56, SD = 1.944	t = -0.148, p = 0.884
Digit Span – Backwards	M = 9.00, SD = 1.936	M = 8.44, SD = 1.810	t = 0.629, p = 0.538
Digit Span – Ranking	M = 8.56, SD = 1.667	M = 8.56, SD = 0.926	t = 1.376, p = 0.175
Verbal Fluency – Category	M = 26.44, SD = 5.725	M = 24.25, SD = 3.694	t = 0.925, p = 0.370
Verbal Fluency – Phonemic	M = 16.89, SD = 6.153	M = 16.38, SD = 5.731	t = 0.177, p = 0.862
Block Designs	M = 37.00, SD = 10.235	M = 39.98, SD = 9.480	t = -0.621, p = 0.543
Boston Naming	M = 27.89, SD = 1.833	M = 27.22, SD = 2.667	t = 0.488, p = 0.632
Similarities	M = 27.00, SD = 3.808	M = 25.44, SD = 4.613	t = 0.780, p = 0.447

M = Mean, SD = standard deviation.

Table 5.

Differences between Aβ lateralization and cognitive test performance in the MCI-Aβ + group.

	Left-lateralized (n = 9)	Right-lateralized (n = 12)	Group difference (Two-Sided p)
MMSE	M = 26.67, SD = 1.803	M = 26.33 SD = 1.497	t = 0.463, p = 0.649
RAVLT – Immediate Recall	M = 32.56, SD = 10.93	M = 29.08, SD = 9.130	t = 0.793, p = 0.438
RAVLT – Delayed Recall	M = 3.56, SD = 4.275	M = 4.25, SD = 2.563	t = -0.464, p = 0.648
RAVLT – Recognition	M = 12.22, SD = 2.863	M = 11.92 SD = 2.575	t = 0.257, p = 0.800
Trail Making A	M = 52.61, SD = 19.70	M = 49.82, SD = 9.625	t = 0.425, p = 0.675
Trail Making B	M = 176.06, SD = 125.77	M = 166.14, SD = 84.51	t = 0.217, p = 0.831
Stroop – Words	M = 70.67, SD = 12.845	M = 76.95, SD = 12.60	t = -1.122, p = 0.276
Stroop – Colours	M = 50.22, SD = 11.53	M = 52.98, SD = 8.271	t = -0.639, p = 0.530
Stroop – Interference	M = 21.00, SD = 13.65	M = 21.17, SD = 8.077	t = -0.035, p = 0.972
Digit Span – Forwards	M = 8.44 SD = 2.007	M = 7.75, SD = 1.815	t = 0.830, p = 0.417
Digit Span – Backwards	M = 6.00, SD = 1.000	M = 7.08, SD = 1.676	t = -1.717, p = 0.102
Digit Span – Ranking	M = 5.67, SD = 2.500	M = 5.58, SD = 1.730	t = 0.090, p = 0.929
Verbal Fluency – Category	M = 16.67, SD = 4.924	M = 18.00, SD = 4.328	t = -0.659, p = 0.518
Verbal Fluency – Phonemic	M = 12.78, SD = 5.403	M = 12.58, SD = 3.655	t = 0.099, p = 0.923
Block Designs	M = 33.44, SD = 11.77	M = 31.25, SD = 13.23	t = 0.393, p = 0.698
Boston Naming	M = 24.00, SD = 4.387	M = 25.88, SD = 2.468	t = -1.251, p = 0.226
Similarities	M = 19.38, SD = 4.357	M = 18.50, SD = 5.143	t = 0.412, p = 0.685

M = Mean, SD = standard deviation.