APOE ε4 influences medial temporal atrophy and tau deposition in atypical Alzheimer disease

Abstract

INTRODUCTION:

APOE ε4 is an important genetic risk factor for typical Alzheimer’s disease (AD), influencing brain volume and tau burden. Little is known about its influence in atypical presentations of AD.

METHODS:

An atypical AD cohort of 140 patients diagnosed with either posterior cortical atrophy or logopenic progressive aphasia underwent MRI and PET. Linear mixed effects models were fit to assess the influence of APOE ε4 on cross-sectional and longitudinal regional metrics.

RESULTS:

At baseline, APOE ε4 carriers had smaller hippocampal and amygdala volumes and greater tau SUVR in the hippocampus and entorhinal cortex compared to non-carriers. While, longitudinally, APOE ε4 non-carriers showed faster rates of atrophy and tau accumulation in the entorhinal cortex, with faster tau accumulation in the hippocampus.

DISCUSSION:

APOE ε4 influences patterns of neurodegeneration, tau deposition and was associated with more medial temporal involvement, although there is evidence that non-carriers may be catching up over time.

Background

Apolipoprotein E (APOE) allele ε4 is an important genetic risk factor for developing late-onset Alzheimer’s disease (AD), which is typically characterized by prominent episodic memory impairment¹.

APOE ε4 has been assessed in atypical non-amnestic predominant clinical presentations of AD but much about its influence still needs exploring. Two widely studied phenotypes include logopenic progressive aphasia (LPA) and posterior cortical atrophy (PCA)., where LPA is typically characterized by language deficits, including poor word retrieval, difficulty repeating sentences and phonological errors², and PCA by visuospatial and visuoperceptual deficits³. The risk association of APOE ε4 in these variants appears to be different than typical AD, with studies associating PCA with an increased risk^4,5, although APOE ε4 frequency is typically lower in PCA⁶ and LPA⁷ compared with typical AD⁸. Moreover, the frequency of two copies of ε4 allele has been reported to be higher in typical AD, compared to the atypical AD variants⁹. APOE ε4 status is significantly influenced by age of onset, although the relationship with onset age differs between typical and atypical AD⁸. Typical AD shows a bell-shaped curve with APOE ε4 frequency highest between the ages of 65 and 70 years, while frequencies increase with older age in atypical AD. No associations have been found between APOE ε4 and Aβ burden in atypical AD variants⁸, an association commonly found with typical AD¹. It is possible that the genetic makeup may be partially responsible for the atypical AD patients presenting differently from the typical AD patients^10,11.

In typical AD, the presence of APOE ε4 has shown greater volume loss in the hippocampus, caudate, precentral gyrus¹², temporal cortex, occipital pole¹³ and posterior cingulate cortex¹⁴ in carriers. With tau burden in entorhinal cortex, hippocampus^15,16, cingulate and insula¹⁷. It is unclear whether the relationship between APOE ε4, medial temporal tau deposition and atrophy is observed in atypical AD variants which tend to show medial temporal sparing and lower APOE ε4 frequencies⁸. A relationship between APOE ε4 and tau SUVR in the entorhinal cortex and amygdala in an AD cohort that included atypical AD patients was noted¹⁵. However, the influence of APOE genotype on atrophy and tau SUVR in atypical AD and whether APOE influences longitudinal change in these imaging metrics remains unclear.

The primary aim of this study was to investigate the association of APOE ε4 on gray matter volume loss and tau SUVR in different brain regions and their longitudinal rates of change in atypical AD patients. The secondary aim was to assess whether APOE ε4 association would be detected within LPA and PCA. Although APOE ε4 is neither necessary nor sufficient to develop AD, we hypothesized that APOE ε4 influences atrophy and tau accumulation and would be associated with greater medial temporal lobe involvement.

Methods

2.1. Patients

The atypical AD cohort consisted of 140 patients that fulfilled clinical diagnostic criteria for LPA² (n=82) or PCA³ (n=58), recruited by the Neurodegenerative Research group (NRG) at Mayo Clinic, Rochester, MN, between 2010 and 2020. Most patients (95%) were referred by Neurologists from the Department of Neurology, Mayo Clinic, with 5% self-referrals from clinicaltrials.gov. All diagnoses were rendered blinded to neuroimaging findings. All patients were enrolled regardless of their age, were beta-amyloid PET positive and underwent neurological evaluation from behavioral neurologists (KAJ or JGR), neuropsychological testing by a neuropsychologist (MMM) and blood collection for APOE testing¹⁸ and a standardized MRI protocol. Of the 140 patients, 73 were enrolled into a cross-sectional study that did not include [¹⁸F]flortaucipir PET. Sixty-seven patients were enrolled into a longitudinal study in which patients underwent [¹⁸F]flortaucipir PET (31 LPA/36 PCA); 45 of these 67 patients have undergone longitudinal MRI and PET imaging (23 LPA/22 PCA). The longitudinal cohort was similar demographically to the cross-sectional cohort (Supplementary Table 1). Thirty-eight cognitively normal controls (median age=59, 66% female) were also enrolled by NRG between 2017 and 2021 and underwent identical imaging.

2.2. Patient consent and protocols

The study was approved by the Mayo Clinic IRB. All patients gave written informed consent to participate in this study.

2.3. Clinical testing

Neurological evaluations included the Montreal Cognitive Assessment Battery (MoCA) for cognitive function¹⁹. The neuropsychological evaluations included the Boston Naming Test (BNT) for confrontational naming²⁰, Boston Diagnostic Aphasia Exam (BDAE) repetition subtest for sentence repetition²¹, Rey Auditory Verbal Learning Test – Recognition Percent Correct (AVLT-RCP) for episodic memory²², Rey-Osterrieth Complex Figure Test (Rey-O) for visuospatial constructional ability²³, Visual Object and Space Perception Battery (VOSP) Cubes and Letters for visuospatial and visuoperceptual ability²⁴.

2.4. Image acquisition

All patients underwent scanning with a 3T volumetric MRI on GE scanners at Mayo Clinic, Rochester, MN. This included a magnetization prepared rapid gradient echo (MPRAGE) sequence (TR/TE/T1 = 2300/3/900ms; 26 cm field of view, slice thickness = 1.2mm, in plane resolution = 1mm).

The tau PET scans were acquired on PET/CT scanners. Patients were injected with ~ 370MBq (range 333-407MBq) of [¹⁸F] flortaucipir. 80 mins after injection, PET acquisition was carried out for 20 mins. The scans consisted of four, 5 minutes dynamic frames after a low dose CT image. After standard corrections, PET sinograms were reconstructed into a 256mm field of view. The four individual frames were averaged for analysis²⁵.

2.5. Image processing

The tau PET images were registered to their corresponding subject-space MPRAGE T1-weighted MRI using SPM12. Regional PET values were calculated using ANTs²⁶ to propagate the Mayo Clinic Adult Lifespan Template (MCALT) atlas. Unified segmentation²⁷ determined tissue probabilities of each MPRAGE scan²⁸. Seven regions of interest (ROIs) were assessed: amygdala, hippocampus, entorhinal cortex, frontal (middle frontal, superior frontal, frontal inferior opercularis, frontal inferior triangularis), parietal (supramarginal, angular, inferior parietal, superior parietal, precuneus), lateral temporal (inferior temporal, middle temporal, superior temporal) and occipital (lingual, calcarine, cuneus, inferior occipital, middle occipital, superior occipital). Median tau PET values were calculated across gray and white matter and divided by the cerebellar crus grey matter median uptake value to generate standard uptake value ratios (SUVRs). The annualized rate of tau accumulation was calculated as the difference between the follow up and baseline SUVRs, divided by the year difference between the two scans²⁹.

Gray matter volumes were calculated and normalized to the intracranial volume. PET SUVR and gray matter images were spatially normalized to the MCALT template and smoothed for voxel-wise analyses using SPM12. For longitudinal analysis, tau-PET annualized change maps were created by subtracting the baseline from the follow-up tau-PET image and dividing by the time difference in years. For MRI longitudinal analysis, an in-house developed tensor-based morphometry using symmetric normalization (TBM-SyN) was used to generate mean annualized Jacobian images³⁰.

2.6. Statistical analysis

2.6.1. Mixed linear regression analysis

Baseline characteristics were compared between APOE ε4 carriers and non-carriers. Fisher’s exact test was used for categorial variables (sex), and Wilcoxon rank sum test for continuous variables. Longitudinal linear mixed effects models were fit to assess the effects of APOE ε4 carriers on baseline values and change over time in brain volumes and tau SUVRs. All models were fit with the log of volume or the log of tau SUVRs as response variables. Time, APOE ε4 status, age at the time of the first scan, total intracranial volume, sex, diagnosis, and the interaction between time and APOE ε4 status were included as fixed effects. The two regression parameters of primary interest were APOE ε4 status which corresponds to the mean difference between ε4 carriers and non-carriers at baseline and the interaction between time and APOE ε4 status which corresponds to the mean difference in annual rates of change between ε4 carriers and non-carriers. The model included subject-specific random intercepts and slopes. Time was expressed as years from first scan and had an origin of zero. The response variables were log-transformed to reduce skewness and address nonconstant variance or proportional errors. The regression coefficients were back-transformed to interpret the difference between carriers and non-carriers in terms of percentage differences.

The longitudinal model included data from individuals with one or more MRI or PET. This type of model simultaneously estimates both cross-sectional and longitudinal, within-subject differences by APOE ε4 status³¹. Patients with one MRI and PET scan primarily contribute information about cross-sectional differences, while patients with multiple scans contribute information about both cross-sectional and longitudinal differences. The analysis included the sum of volumes and average of SUVRs of left and right ROIs, along with individual left and right ROI analysis. Separate estimates for LPA and PCA were based on including an interaction between diagnosis and APOE ε4 in the model. We summarize the two regression parameters of interest from each model in terms of point estimates and 95% CIs displayed figures and tables. Based on the close link between CIs and hypothesis tests, 95% CIs that do not include zero correspond to a significant difference between ε4 carriers and non-carriers at p<0.05. All models were fit with the lme function in the nlme package in R version 3.6.3³².

2.6.2. Voxel-wise analysis

Multiple regression analysis was performed using SPM12 to assess voxel-wise baseline differences between disease groups and a cognitively normal control group, including age and sex as covariates. For longitudinal assessments, group averaged tau PET annualized change maps and the MRI annualized Jacobian maps were generated. Results visualized using BrainNet viewer³³.

Results

3.1. Study demographics

The demographic and clinical features of the cohort split by APOE ε4 status are shown in Table 1. APOE ε4 carriers were older at onset and baseline scan than non-carriers. There were no differences in disease duration or in the clinical tests, although LPA APOE ε4 non-carriers performed worse on the Rey-O complex figure.

Table 1.

Participants demographics at baseline by APOE ε4 status.

	Atypical AD (N=140)			LPA (N=82)			PCA (N=58)
	APOE ε4+ (N=72)	APOE ε4− (N=68)	P-value	APOE ε4+ (N=41)	APOE ε4− (N=41)	P-value	APOE ε4+ (N=31)	APOE ε4− (N=27)	P-value
Female, n (%)	40 (56%)	44 (65%)	0.30	23 (56%)	24 (59%)	>0.99	17 (55%)	20 (74%)	0.17
Education, yr	16 (12, 20)	16 (12, 20)	0.56	16 (12, 20)	15 (12, 20)	0.64	16 (12, 20)	16 (12, 20)	0.75
Age at onset, yr	64 (48, 80)	59 (44, 79)	0.003	67 (50, 80)	61 (44, 79)	0.003	60 (48, 75)	57 (51, 68)	0.22
Age at baseline scan, yr	68 (54, 85)	62 (50, 82)	0.003	70 (55, 85)	64 (50, 82)	0.004	64 (54, 77)	61 (55, 74)	0.31
Disease duration, yr	3.7 (0.5, 13.0)	3.8 (0.4, 10.3)	0.91	3.7 (0.9, 13.0)	3.2 (0.4, 9.7)	0.41	4.2 (0.5, 12.9)	4.9 (1.0, 10.3)	0.50
Participants with follow-up on MRI imaging, n (%)	19 (26%)	26 (38%)	0.15	8 (20%)	15 (37%)	0.14	11 (35%)	11 (41%)	0.79
Participants with tau PET imaging, n (%)	30 (42%)	37 (54%)	0.18	11 (27%)	20 (49%)	0.07	19 (61%)	17 (63%)	>0.99
Participants with follow-up tau PET, n (%)	19 (63%)	26 (70%)	0.61	8 (73%)	15 (75%)	>0.99	11 (58%)	11 (65%)	0.74
Baseline to follow-up scan, yr	1.0 (0.9, 1.3)	1.0 (0.9, 1.2)	0.88	1.0 (0.9, 1.1)	1.0 (0.9, 1.2)	0.63	1.0 (0.9, 1.3)	1.0 (0.9, 1.1)	0.39
MoCA	18 (1, 29)	18 (0, 25)	0.80	18 (4, 29)	16 (0, 25)	0.49	16 (1, 26)	20 (6, 25)	0.25
BNT	9 (0, 15)	10 (0, 15)	0.98	7 (0, 15)	9 (0, 15)	0.58	13 (1, 15)	12 (4, 15)	0.71
BDAE repetition	8 (3, 10)	7 (0, 10)	0.46	7 (3, 9)	6 (0, 9)	0.08	8 (4, 10)	8 (4, 10)	0.46
Rey-O MOANS	2 (1, 16)	2 (2, 17)	0.12	7 (2, 16)	3 (2, 17)	0.03	2 (1, 9)	2 (2, 2)	>0.99
VOSP cubes	8 (0, 10)	4 (0, 10)	0.16	9 (0, 10)	9 (0, 10)	0.50	2 (0, 9)	1 (0, 8)	0.37
VOSP letter	19 (0, 20)	18 (0, 20)	0.22	19 (8, 20)	19 (1, 20)	0.22	13 (0, 20)	10 (0, 20)	0.55
AVLT RCP	77 (43, 100)	73 (47, 100)	0.78	77 (47, 100)	73 (47, 97)	0.95	68 (43, 100)	70 (50, 100)	0.71

Data shown are n (%) or median (range). For continuous variables, p-values are from Wilcoxon Rank Sum test. For categorical variables, p-values are from Fisher’s Exact test.

Key; APOE ε4+, Apolipoprotein E allele ε4 carriers; APOE ε4−, Apolipoprotein E allele ε4 non-carriers; MoCA, Montreal Cognitive Assessment Battery; BNT, Boston Naming Test; BDAE, Boston Diagnostic Aphasia Exam; AVLT-RCP, Auditory Verbal Learning Test – Recognition Percent Correct; Rey-O, Rey-Osterrieth Complex Figure Test; VOSP, Visual Object and Space Perception Battery

3.2. APOE ε4 in the whole atypical AD cohort

Raw regional gray matter volumes and tau SUVRs at baseline and follow up for all patients are shown in figure 1.

Figure 1:

Spaghetti plots of regional gray matter volume and tau PET SUVRs at baseline and follow up for atypical AD cohort.

3.2.1. Gray matter volume

Analysis with mixed effects models (Figure 2 and Table 2) showed that atypical AD APOE ε4 carriers had 3.8% (95%CI: −7.4%, −0.0%) smaller hippocampal and 3.4% (95%CI: −7.1%, 0.3%) smaller amygdala volumes, versus non-carriers at baseline. Hippocampal volume loss was quite symmetric (Left: −4.0%; Right: −3.8%), while differences in amygdala tend to be greater on right than the left hemisphere (−4.8% versus −2.1%). Longitudinally, there were no clear indications of systematic difference. However, on hemispheric differences, non-carriers tend to show 2.1% (95%CI: 0.1%, 4.1%) faster rates in right entorhinal cortex and 2% (95%CI: −0.4%, 4.4%) in left parietal lobe.

Figure 2:

Mixed effects model results depicting the influence of APOE ε4 status on baseline and annualized rate of change in volume and tau PET uptake. Results are shown in forest plots indicating the difference between APOE ε4 carriers and APOE ε4 non-carriers. 95% confidence interval lines that do not cross the 0 line represent significant differences between groups at the p<0.05 level. For volume estimates, negative values indicate ε4 carriers have more volume loss or faster rates of atrophy than ε4 non-carriers. For tau PET estimates, positive values indicate ε4 carriers have more uptake or faster rates of tau accumulation than ε4 non-carriers. Estimates are shown for left, right and combined left and right hemispheres.

Table 2.

Mixed effects regression results in atypical AD.

Region	Hemisphere	Volume				Tau
		Baseline		Rate of change		Baseline		Rate of change
		Est (95% CI)	P-value	Est (95% CI)	P-value	Est (95% CI)	P-value	Est (95% CI)	P-value
Amygdala	Total	−3.4% (−7.1%, 0.3%)	0.07	0.2% (−1.5%, 2.0%)	0.81	5.4% (−1.8%, 13.1%)	0.14	−1.0% (−4.5%, 2.7%)	0.58
	Left	−2.1% (−6.1%, 2.0%)	0.31	−0.7% (−3.1%, 1.7%)	0.54	4.5% (−3.0%, 12.6%)	0.24	−1.4% (−5.4%, 2.7%)	0.49
	Right	−4.8% (−8.7%, −0.7%)	0.02	1.0% (−0.7%, 2.8%)	0.24	6.5% (−1.2%, 14.8%)	0.10	−0.9% (−4.5%, 2.9%)	0.64
Hippocampus	Total	−3.8% (−7.4%, −0.0%)	0.048	0.5% (−0.6%, 1.7%)	0.34	6.4% (−0.8%, 14.2%)	0.08	−3.7% (−7.1%, −0.2%)	0.04
	Left	−4.0% (−8.0%, 0.2%)	0.06	0.3% (−0.9%, 1.5%)	0.61	5.3% (−2.1%, 13.2%)	0.17	−2.5% (−6.1%, 1.2%)	0.17
	Right	−3.8% (−7.6%, 0.1%)	0.06	0.9% (−0.5%, 2.4%)	0.19	7.7% (0.2%, 15.7%)	0.04	−5.0% (−8.5%, −1.3%)	0.01
Entorhinal Cortex	Total	−2.4% (−6.9%, 2.4%)	0.32	0.4% (−1.1%, 1.8%)	0.63	8.2% (0.5%, 16.6%)	0.04	−3.5% (−7.0%, 0.1%)	0.06
	Left	−2.6% (−7.8%, 2.9%)	0.35	−1.3% (−3.2%, 0.7%)	0.19	7.7% (−0.6%, 16.8%)	0.07	−2.7% (−6.2%, 0.8%)	0.13
	Right	−2.5% (−7.2%, 2.5%)	0.33	2.1% (0.1%, 4.1%)	0.04	9.0% (1.0%, 17.6%)	0.03	−4.3% (−8.1%, −0.4%)	0.03
Frontal	Total	1.8% (−2.3%, 6.1%)	0.40	1.5% (−1.4%, 4.4%)	0.31	2.3% (−8.5%, 14.5%)	0.68	−2.3% (−6.0%, 1.6%)	0.25
	Left	3.2% (−1.4%, 8.0%)	0.17	2.0% (−0.8%, 4.8%)	0.17	2.0% (−9.3%, 14.8%)	0.73	−2.0% (−5.8%, 2.0%)	0.32
	Right	0.4% (−3.7%, 4.7%)	0.85	1.1% (−2.1%, 4.4%)	0.50	2.8% (−8.9%, 16.0%)	0.65	−2.6% (−6.4%, 1.3%)	0.18
Parietal	Total	1.7% (−3.1%, 6.6%)	0.49	2.0% (−0.4%, 4.4%)	0.10	−2.4% (−13.3%, 9.7%)	0.68	−1.8% (−6.1%, 2.6%)	0.41
	Left	2.3% (−3.0%, 7.9%)	0.40	2.1% (−0.1%, 4.2%)	0.06	−3.7% (−15.1%, 9.3%)	0.55	−1.7% (−6.0%, 2.8%)	0.45
	Right	1.2% (−3.8%, 6.5%)	0.63	2.0% (−1.0%, 5.1%)	0.18	−1.0% (−12.4%, 11.8%)	0.87	−2.1% (−6.6%, 2.5%)	0.35
Lateral temporal	Total	−0.6% (−4.4%, 3.3%)	0.75	1.0% (−0.5%, 2.4%)	0.20	1.4% (−8.4%, 12.3%)	0.78	−1.7% (−5.8%, 2.5%)	0.42
	Left	0.3% (−4.1%, 4.9%)	0.90	0.8% (−0.9%, 2.5%)	0.37	−0.2% (−10.4%, 11.1%)	0.97	−1.0% (−5.2%, 3.3%)	0.62
	Right	−1.4% (−5.6%, 2.9%)	0.50	1.1% (−0.4%, 2.6%)	0.15	2.9% (−7.9%, 15.0%)	0.61	−2.4% (−6.7%, 2.1%)	0.29
Occipital	Total	0.3% (−3.6%, 4.4%)	0.88	1.0% (−0.6%, 2.7%)	0.22	−1.9% (−12.3%, 9.7%)	0.74	−1.6% (−6.1%, 3.0%)	0.48
	Left	1.4% (−3.2%, 6.2%)	0.56	1.1% (−0.9%, 3.1%)	0.27	−3.5% (−13.8%, 8.1%)	0.53	−1.6% (−6.1%, 3.1%)	0.49
	Right	−0.6% (−5.0%, 4.0%)	0.79	0.9% (−0.9%, 2.8%)	0.32	0.2% (−11.5%, 13.3%)	0.98	−1.7% (−6.4%, 3.2%)	0.48

Values shown indicate the difference between APOE ε4 carriers and APOE ε4 non-carriers. For volume estimates, negative values indicate ε4 carriers have more volume loss or faster rates of atrophy than ε4 non-carriers. For tau PET estimates, positive values indicate ε4 carriers have more uptake or faster rates of tau accumulation than ε4 non-carriers.

In the voxel-level maps, both atypical AD APOE ε4 carriers and non-carriers, when compared to controls at baseline had significant (p<0.001, FWE) volume loss in the lateral temporoparietal and occipital cortices (Figure 3), with no differences identified between carriers and non-carriers on direct comparison. Longitudinally, both APOE ε4 carriers and non-carriers showed fastest rates of atrophy in the lateral temporoparietal cortex, with no differences observed between groups (Supplementary Figure 1).

Figure 3:

Cross-sectional voxel-based gray matter volume and tau PET SUVRs map by APOE ε4 status. These maps represent differences versus controls at p < 0.001 corrected for multiple comparisons using the family wise error correction. All images were generated using BrainNet Viewer.

3.2.2. Tau PET SUVR

Analysis with mixed effects models (Figure 2 and Table 2) showed that atypical AD APOE ε4 carriers had 8.2% (95%CI: 0.5%, 16.6%) higher SUVR in the entorhinal cortex and 6.4% (95%CI: −0.8%, 14.2%) higher SUVR in the hippocampus, with generally symmetric hemispheric differences. Longitudinally, there was some indication that carriers had 3.7% (95%CI: −7.1%, −0.2%) slower rates of tau accumulation in the hippocampus and 3.5% (95%CI: −7.0%, 0.1%) in the entorhinal cortex. These patterns were similar across both hemispheres.

In the voxel-level maps both atypical APOE ε4 carriers and non-carriers, when compared to controls at baseline, had significant (p<0.001, FWE) SUVRs in the lateral temporoparietal lobes, with carrier’s cortical findings being less severe (Figure 3). Longitudinally, both APOE ε4 carriers and non-carriers showed faster rates of tau accumulation in the frontal lobe (Supplementary Figure 2).

3.3. APOE ε4 in the LPA and PCA subsets

3.3.1. Gray Matter volume

Analysis with mixed effects models (Figure 2 and Table 3) showed that the effect of APOE ε4 was generally similar with perhaps a marginally larger effect in LPA. LPA APOE ε4 carriers were estimated to have 4.3% (95%CI: −8.9%, 0.6%) smaller amygdala and 3.9% (95%CI: −8.6%, 1.1%) smaller hippocampal volumes, compared to 2.2% and 3.6% in PCA. Longitudinally, the effect of APOE ε4 on atrophy was not notably different between LPA and PCA. But PCA APOE ε4 non-carriers did show 2.9% (95%CI: 0.1%, 5.8%) faster rates in right entorhinal cortex and 2% (95%CI: −0.1%, 4.0%) for right hippocampus. No significant longitudinal differences between carriers and non-carriers were observed in LPA.

Table 3.

Mixed effects regression results by diagnosis.

LPA
Region	Hemisphere	Volume				Tau
		Baseline		Rate of change		Baseline		Rate of change
		Est (95% CI)	P-value	Est (95% CI)	P-value	Est (95% CI)	P-value	Est (95% CI)	P-value
Amygdala	Total	−4.3% (−8.9%, 0.6%)	0.09	1.7% (−0.9%, 4.3%)	0.19	5.1% (−5.6%, 16.9%)	0.36	1.8% (−3.6%, 7.4%)	0.51
	Left	−3.3% (−8.5%, 2.1%)	0.22	1.0% (−2.5%, 4.7%)	0.57	5.3% (−5.9%, 17.8%)	0.36	0.4% (−5.5%, 6.7%)	0.89
	Right	−5.3% (−10.4%, 0.0%)	0.051	2.2% (−0.4%, 4.9%)	0.10	5.1% (−6.2%, 17.7%)	0.39	2.8% (−2.7%, 8.5%)	0.32
Hippocampus	Total	−3.9% (−8.6%, 1.1%)	0.12	−0.5% (−2.1%, 1.2%)	0.57	5.9% (−4.8%, 17.7%)	0.28	−1.4% (−6.6%, 4.0%)	0.59
	Left	−4.9% (−10.0%, 0.5%)	0.07	−0.5% (−2.1%, 1.1%)	0.53	4.5% (−6.4%, 16.7%)	0.43	0.1% (−5.2%, 5.7%)	0.97
	Right	−3.1% (−8.1%, 2.1%)	0.24	−0.2% (−2.3%, 1.9%)	0.85	7.4% (−3.6%, 19.7%)	0.19	−3.0% (−8.4%, 2.6%)	0.28
Entorhinal cortex	Total	−1.6% (−7.6%, 4.7%)	0.60	−0.2% (−2.3%, 2.1%)	0.88	9.7% (−2.0%, 22.8%)	0.10	−1.2% (−6.4%, 4.4%)	0.67
	Left	−2.7% (−9.4%, 4.5%)	0.45	−1.4% (−4.2%, 1.6%)	0.36	10.2% (−2.4%, 24.4%)	0.12	−1.4% (−6.6%, 4.0%)	0.59
	Right	−1.0% (−7.3%, 5.7%)	0.77	1.2% (−1.7%, 4.2%)	0.40	9.5% (−2.5%, 22.8%)	0.12	−0.9% (−6.5%, 5.1%)	0.76
Frontal	Total	2.4% (−2.9%, 8.1%)	0.38	−0.7% (−4.7%, 3.5%)	0.75	−2.9% (−17.9%, 14.8%)	0.73	−0.6% (−6.3%, 5.3%)	0.82
	Left	3.8% (−2.2%, 10.2%)	0.22	0.6% (−3.5%, 4.8%)	0.78	−5.1% (−20.3%, 13.1%)	0.56	−0.1% (−5.9%, 6.0%)	0.97
	Right	1.0% (−4.4%, 6.7%)	0.71	−1.7% (−6.2%, 3.0%)	0.46	−0.7% (−17.1%, 18.9%)	0.93	−1.4% (−7.1%, 4.6%)	0.62
Parietal	Total	4.5% (−1.8%, 11.2%)	0.16	2.2% (−1.3%, 5.8%)	0.22	−8.7% (−23.4%, 8.8%)	0.30	1.3% (−5.0%, 8.0%)	0.69
	Left	6.0% (−1.1%, 13.6%)	0.10	1.6% (−1.6%, 4.8%)	0.33	−11.7% (−26.8%, 6.5%)	0.19	2.7% (−3.6%, 9.5%)	0.40
	Right	3.2% (−3.4%, 10.3%)	0.35	2.8% (−1.6%, 7.4%)	0.21	−5.8% (−21.5%, 13.1%)	0.52	−0.3% (−6.8%, 6.7%)	0.93
Lateral temporal	Total	−1.1% (−6.0%, 4.1%)	0.67	1.4% (−0.8%, 3.6%)	0.20	2.1% (−12.4%, 19.1%)	0.78	0.8% (−5.3%, 7.4%)	0.79
	Left	−0.2% (−5.9%, 5.7%)	0.93	1.4% (−1.1%, 4.0%)	0.26	−0.4% (−15.4%, 17.3%)	0.96	2.0% (−4.2%, 8.6%)	0.53
	Right	−1.9% (−7.2%, 3.7%)	0.50	1.4% (−0.8%, 3.7%)	0.20	4.6% (−11.5%, 23.7%)	0.59	−0.3% (−6.8%, 6.6%)	0.92
Occipital	Total	0.2% (−5.0%, 5.6%)	0.95	1.1% (−1.5%, 3.6%)	0.40	−5.2% (−19.8%, 12.1%)	0.53	0.8% (−5.6%, 7.8%)	0.80
	Left	0.9% (−5.0%, 7.2%)	0.76	0.8% (−2.1%, 3.7%)	0.58	−9.2% (−23.5%, 7.7%)	0.26	2.2% (−4.2%, 9.0%)	0.51
	Right	−0.4% (−6.1%, 5.7%)	0.89	1.1% (−1.7%, 3.9%)	0.44	0.2% (−16.7%, 20.5%)	0.98	−0.7% (−7.5%, 6.7%)	0.85
PCA
Amygdala	Total	−2.2% (−7.7%, 3.6%)	0.44	−1.0% (−3.4%, 1.5%)	0.41	5.5% (−4.1%, 16.1%)	0.26	−3.2% (−8.0%, 1.7%)	0.19
	Left	−0.4% (−6.5%, 6.1%)	0.90	−2.3% (−5.6%, 1.1%)	0.17	4.1% (−5.9%, 15.1%)	0.43	−3.2% (−8.6%, 2.5%)	0.26
	Right	−4.0% (−9.8%, 2.3%)	0.21	0.2% (−2.2%, 2.7%)	0.87	7.0% (−3.3%, 18.4%)	0.19	−3.5% (−8.2%, 1.6%)	0.17
Hippocampus	Total	−3.6% (−9.1%, 2.2%)	0.21	1.3% (−0.3%, 2.9%)	0.11	7.0% (−2.7%, 17.6%)	0.16	−5.8% (−10.4%, −0.9%)	0.02
	Left	−2.7% (−8.7%, 3.7%)	0.40	0.7% (−0.8%, 2.3%)	0.35	6.3% (−3.7%, 17.3%)	0.22	−5.1% (−9.8%, −0.1%)	0.046
	Right	−4.7% (−10.3%, 1.3%)	0.12	2.0% (−0.1%, 4.0%)	0.06	7.6% (−2.4%, 18.5%)	0.14	−6.4% (−11.3%, −1.4%)	0.01
Entorhinal cortex	Total	−3.4% (−10.1%, 3.9%)	0.35	0.9% (−1.2%, 3.0%)	0.40	6.8% (−3.4%, 18.1%)	0.20	−5.4% (−10.1%, −0.4%)	0.03
	Left	−2.4% (−10.1%, 6.0%)	0.56	−1.3% (−4.0%, 1.6%)	0.37	5.6% (−5.3%, 17.7%)	0.32	−3.8% (−8.5%, 1.2%)	0.13
	Right	−4.4% (−11.4%, 3.1%)	0.24	2.9% (0.1%, 5.8%)	0.04	8.1% (−2.5%, 19.8%)	0.14	−6.9% (−11.9%, −1.7%)	0.01
Frontal	Total	1.0% (−5.1%, 7.4%)	0.76	3.4% (−0.6%, 7.6%)	0.10	6.6% (−8.3%, 23.8%)	0.40	−3.4% (−8.5%, 2.1%)	0.22
	Left	2.4% (−4.4%, 9.7%)	0.49	3.2% (−0.9%, 7.4%)	0.12	8.0% (−7.7%, 26.2%)	0.33	−3.5% (−8.7%, 2.1%)	0.21
	Right	−0.4% (−6.6%, 6.1%)	0.89	3.8% (−0.8%, 8.5%)	0.10	5.4% (−10.3%, 23.9%)	0.52	−3.3% (−8.5%, 2.3%)	0.23
Parietal	Total	−2.0% (−8.7%, 5.3%)	0.59	1.9% (−1.5%, 5.4%)	0.27	1.8% (−13.0%, 19.2%)	0.82	−3.6% (−9.3%, 2.4%)	0.23
	Left	−2.4% (−9.9%, 5.8%)	0.55	2.4% (−0.7%, 5.5%)	0.12	2.4% (−13.4%, 21.0%)	0.78	−4.7% (−10.3%, 1.2%)	0.11
	Right	−1.3% (−8.6%, 6.6%)	0.74	1.7% (−2.5%, 6.0%)	0.43	1.7% (−13.6%, 19.7%)	0.84	−2.7% (−8.7%, 3.7%)	0.39
Lateral temporal	Total	−0.3% (−6.0%, 5.8%)	0.93	0.1% (−1.9%, 2.3%)	0.89	0.4% (−12.5%, 15.2%)	0.95	−3.7% (−9.1%, 2.1%)	0.21
	Left	0.9% (−5.7%, 7.8%)	0.80	−0.3% (−2.7%, 2.1%)	0.79	0.2% (−13.4%, 15.9%)	0.98	−3.8% (−9.3%, 2.0%)	0.19
	Right	−1.0% (−7.2%, 5.5%)	0.75	0.6% (−1.6%, 2.7%)	0.59	0.5% (−13.5%, 16.7%)	0.95	−3.5% (−9.4%, 2.7%)	0.25
Occipital	Total	0.5% (−5.5%, 6.7%)	0.88	0.9% (−1.5%, 3.4%)	0.46	0.3% (−13.7%, 16.5%)	0.97	−2.8% (−8.7%, 3.6%)	0.37
	Left	2.0% (−4.9%, 9.4%)	0.58	1.2% (−1.6%, 4.1%)	0.38	0.8% (−13.5%, 17.5%)	0.91	−3.4% (−9.2%, 2.6%)	0.25
	Right	−0.9% (−7.4%, 6.1%)	0.80	0.7% (−1.9%, 3.4%)	0.58	−0.4% (−15.6%, 17.5%)	0.96	−1.8% (−8.2%, 5.1%)	0.60

Values shown indicate the difference between APOE ε4 carriers and APOE ε4 non-carriers. For volume estimates, negative values indicate ε4 carriers have more volume loss or faster rates of atrophy than ε4 non-carriers. For tau PET estimates, positive values indicate ε4 carriers have more uptake or faster rates of tau accumulation than ε4 non-carriers

In the voxel-level maps, the LPA and PCA carrier and non-carrier groups showed the expected patterns for each phenotype (Figure 3 and Supplementary Figure 1). No differences were identified between carriers and non-carriers for either LPA or PCA.

3.3.2. Tau PET SUVR

Analysis with mixed effects models (Figure 2 and Table 3) showed that for the most part the effect of APOE ε4 on tau accumulation was consistent for LPA and PCA. However, there was some indication of possibly larger effects in PCA. PCA APOE ε4 non-carriers had 5.8% (95%CI: 0.9%, 10.4%) higher rate of tau accumulation in the hippocampus and 5.4% (95%CI: 0.4%, 10.1%) higher in the entorhinal cortex. No other significant regional baseline or longitudinal differences were observed between carriers and non-carriers in LPA or PCA.

In the voxel-level maps, the LPA and PCA carrier and non-carrier groups showed the expected patterns for each phenotype. Patterns of temporoparietal uptake in LPA were more striking in the non-carriers (Figure 3). Longitudinally (Supplementary Figure 2), LPA APOE ε4 non-carriers presented with high rates of tau accumulation in the bilateral frontal lobe and right temporoparietal cortex, while carriers showed higher rates of accumulation across the cortex. The PCA APOE ε4 carriers and non-carriers presented with greatest rates of tau accumulation in the frontal lobe, with the non-carriers showing more striking rates of accumulation. No differences were observed between carriers and non-carriers on direct comparison for LPA or PCA.

Discussion

This study investigated the association of APOE ε4 on clinical and neuroimaging features in atypical AD. APOE ε4 appeared to influence cortical atrophy and tau pathology, relatively mildly, and was associated with more medial temporal involvement at baseline. However, longitudinally, APOE ε4 carriers showed slower rates of atrophy and tau accumulation in the medial temporal lobe. We found weak effects of APOE ε4 on the cognitive profile. The influence of APOE ε4 within LPA and PCA was limited to the medial temporal lobe, without having a strong effect on the syndrome’s expression patterns. The frequency of the APOE allele ε4 did not differ across the phenotypes but was associated with older age at onset in atypical AD.

The morphological and pathological heterogeneity of APOE ε4 carriers and non-carriers were explored extensively. APOE ε4 carriers showed greater volume loss in the hippocampus and amygdala, and greater tau SUVR in the hippocampus and entorhinal cortex, compared to non-carriers. Greater medial temporal involvement in the atypical AD APOE ε4 carriers are in line with previous literature^34,35 and suggests that APOE ε4 carriers are predisposed to greater involvement of the medial temporal lobe regardless of presenting syndrome. There was a tendency in our voxel-level maps for the non-carriers to show greater involvement of the cortex. Hence, a shift in pattern was seen with the carriers favoring more medial temporal involvement and non-carriers favoring a more cortical involvement. A similar shift in medial temporal to cortical involvement is seen with age in typical AD, with younger patients showing greater cortical and less medial temporal involvement than older patients^25,36,37. APOE ε4 carriers in typical AD favor greater medial temporal involvement but also have substantial cortical involvement^12,17. In our study, we account for the influence of age and, our results suggest that both age and APOE ε4 influence the relative involvement of medial temporal and cortical regions in atypical AD.

Within the LPA and PCA groups, similar patterns of greater medial temporal involvement in APOE ε4 carriers were observed. LPA APOE ε4 carriers had a marginally larger effect with smaller hippocampal and amygdalar volumes, compared to PCA. PCA APOE ε4 non-carriers had higher SUVR in the hippocampus and entorhinal cortex, with no significant differences in LPA. Even though these were only marginal differences, it is unclear why LPA showed volume differences, while PCA showed tau SUVR differences. This variability could be due to the smaller size of the subsets or asymmetric predominance of the phenotypes, which can be clearly seen in the voxel-based maps with LPA displaying a left dominant temporoparietal pattern, and PCA displaying a right dominant occipitoparietal pattern. These findings are also consistent with previous literature and in line with their phenotypic differences^2,38,39. While the medial temporal lobes are typically relatively spared in both LPA and PCA, they can still be affected^40–43, and our results suggest that APOE status could be a factor that determines the degree of medial temporal involvement. This is in line with pathological findings of lower hippocampal volumes, neurons and increased neurofibrillary tangles in APOE ε4 carriers compared to the non-carriers⁴⁴. But it is interesting to note that the influence of APOE on the medial temporal lobe was relatively symmetric in both phenotypes, perhaps suggesting an underlying disease mechanism that is independent of the mechanism driving phenotypic expression in these patients.

We did not find any differences in memory performance according to APOE status in our cohort. However, the underlying mechanism for this was not explored further, but studies have also found the presence of APOE ε4 allele to be commonly associated with memory deficits in typical AD patients^7,9,45 and less common in the visual⁹ and aphasic presentation⁷ of atypical AD. General population studies with APOE ε4 genotyping have also reported poor cognitive performance in carriers and a higher probability of developing dementia at older ages, compared to the non-carriers⁴⁶. Nonetheless, additional studies exploring these associations are needed. In the present study, the only difference we observed in clinical testing was LPA APOE ε4 non-carriers performing worse on the Rey-O complex figure. This could be attributed to the greater volume loss and tau SUVR in the frontal and temporoparietal cortices, which is in line with the regions typically associated with this test⁴⁷. Overall, our results suggest that APOE ε4 in atypical AD is not associated with greater memory impairment³⁶.

The influence of APOE ε4 was also studied longitudinally. The APOE ε4 non-carriers showed significantly higher rates of tau accumulation in hippocampus and entorhinal cortex. These patterns of faster tau accumulation could suggest that the medial temporal lobes of the non-carriers were catching up. The APOE ε4 carriers may be more predisposed to medial temporal involvement earlier in the disease, while the non-carriers show more sparing of the medial temporal lobe early in the disease but deteriorate over time. These findings differ from those observed in typical AD as the APOE ε4 carriers continue to have faster tau accumulation in the medial temporal, parietal and occipital lobes, compared to APOE ε4 non-carriers, suggesting that APOE ε4 in typical AD has a progressive effect on tau SUVR¹⁷. When assessing LPA and PCA, we only observed marginal differences in rates of medial temporal atrophy. We note that longitudinal findings may be influenced by multiple factors, such as the period of the disease being assessed, and the results may not generalize to either earlier or later disease stages. Moreover, a recent study found that tau PET uptake can plateau in patients with atypical AD in regions with high baseline uptake⁴⁸. Hence, we could hypothesize that tau burden in the medial temporal lobe in APOE ε4 carriers may have plateaued, reducing tissue capacity to accumulate tau burden or reduced blood flow for tracer delivery, while tau is still accumulating in these regions in non-carriers. Further studies are needed to better understand the underlying mechanism of these longitudinal findings.

The longitudinal patterns of atrophy and tau accumulation in each variant are consistent with previous studies, with atrophy patterns mirroring the baseline patterns, and tau accumulation occurring most prominently in the frontal lobes⁴⁹. It has been conceptualized that tau deposition precedes measurable atrophy⁵⁰, our voxel-based maps also showed higher tau SUVR in temporoparietal and occipital cortex at baseline, reflected by volume loss in the same regions. Over time, tau accumulation rapidly spread into the frontal cortex, an area yet to undergo much atrophy.

These findings help unravel the associations between APOE ε4, gray matter volume loss and tau SUVR across the brain both cross-sectionally and longitudinally. Strengths of the study include the large, atypical AD cohort, standardized imaging and clinical evaluations and the fact that our models controlled for differences in age and sex. However, there are some limitations to consider such as the absence of the other less common clinical phenotypes of atypical AD¹¹, a cohort of typical amnestic AD and cognitively normal controls. Future studies including a more heterogenous cohort in terms of race, ethnicity, and education would also be needed to better understand their associations with the APOE ε4 genotype. Lastly, our study included only two visits with an interval of 12 months so further follow-up will be needed to determine long-term effects of APOE genotype.

Conclusion

Our results highlight that the APOE ε4 genotype predisposes atypical AD patients to greater medial temporal tau deposition and neurodegeneration, although it does not lead to increased rates of worsening in these metrics, at least over the time interval assessed in this study. APOE genotype, therefore, likely plays a role in the pathophysiology of atypical AD and should be considered in future studies and clinical treatment trials assessing neuroimaging biomarkers in patients with atypical AD.

Abbreviations

APOE

Apolipoprotein E

AD

Alzheimer’s disease

LPA

Logopenic progressive aphasia

PCA

Posterior cortical atrophy

SUVRs

Standardized uptake value ratios

Aβ

β-amyloid

NRG

Neurodegenerative Research group

MoCA

Montreal cognitive assessment battery

BNT

Boston naming test

BDAE

Boston diagnostic aphasia exam

AVLT-RCP

Auditory verbal learning test – recognition percent correct

Rey-O

Rey-osterrieth complex figure test

VOSP

Visual object and space perception battery

MPRAGE

magnetization prepared rapid gradient echo

MCALT

Mayo Clinic Adult Lifespan Template

ROI

regions of interest

TBM-SyN

Tensor-based morphometry using symmetric normalization

MNI

Montreal Neurological Institute