Older, non-demented apolipoprotein ε4 carrier males show hyperactivation and structural differences in odor memory regions: a blood-oxygen-level-dependent and structural magnetic resonance imaging study.

The current study sought to examine the interaction of sex and Apolipoprotein ε4 status on olfactory recognition memory within non-demented, older individuals. We separated 39 participants into groups based on ε4 status and sex. Each participant completed an olfactory memory recognition task during 2 functional magnetic resonance imaging scans and 1 structural scan. The ε4 carriers had greater functional recruitment of memory regions during false positives relative to ε4 non-carriers. During hits, the male ε4 carriers showed greater functional recruitment compared to female ε4 carriers. The ε4 carriers had larger bilateral putamen volumes relative to ε4 non-carriers. Neuroimaging data were significantly associated with Dementia Rating Scale scores solely in males. Results suggest differential olfactory memory processing in relation to sex and ε4 status. Male ε4 carriers in particular, demonstrated hyperactivation during recognition memory, which we suspect reflects neuronal compensation to maintain functional performance. Future studies should consider examining underlying mechanisms that contribute to these sex differences within ε4 carriers.

Introduction

The presence of the Apolipoprotein (ApoE) ε4 allele is the most well-established genetic risk factor for Alzheimer’s disease (AD), a devastating neurodegenerative disease with no cure (Alzheimer’s Association, 2018; Bu, 2009). In humans, the ApoE gene exists as 3 different polymorphic alleles (ε2, ε3, ε4; Bu, 2009; Farrer et al., 1997; Lind et al., 2006; Yu et al., 2013). Under normal physiological conditions, ApoE supports cholesterol and lipid transport and performs membrane repair processes to protect neuronal integrity (Mahley and Huang, 2012). The ε4 variant generally exhibits poorer cholesterol transport and amyloid-β (Aβ) plaque clearance, as well as increased neurofibrillary tangle formation relative to the ε3 allele (Mahley and Rall, 2000). The accumulation of neurofibrillary Aβ plaques and tau tangles is a pathological marker in AD and associated with cognitive decline (Blennow et al., 2006; Braak and Braak, 1991, 1996; Hedden et al., 2013). These accumulations are often reported in medial temporal lobe (MTL) regions (e.g., entorhinal cortex [ERC], hippocampus), which are areas associated with odor recognition memory (Cerf-Ducastel and Murphy, 2009).

Sex differences in odor identification (odor ID) and AD development have been noted in previous literature. Older women tend to outperform older men in odor ID tasks (Murphy et al., 2002; Wehling et al., 2016). Males’ ability to identify odors may deteriorate earlier (approximately 20 years) in the lifespan relative to females (Ship et al., 1996). The ApoE ε4 allele tends to have a more pronounced effect on late-onset AD development in women compared to men, although some attribute this to women’s longer lifespan (Altmann et al., 2014; Bartrés-Faz et al., 2002; Holland et al., 2013; Hyman et al., 1996; Payami et al., 1996; Poirier et al., 1993; Ungar et al., 2014). Women homozygous for ε4 have a greater risk than heterozygous women, but male homozygous and heterozygous for the ε4 allele do not differ significantly (Payami et al., 1996). Despite the apparent sex differences in AD progression, it is relatively overlooked in the existing literature (Ungar et al., 2014).

Participants with and at-risk for AD have demonstrated poor recognition memory overall, with olfactory recognition memory being the most compromised (Moberg et al., 1987). Individuals at-risk for AD but who have not yet developed clinical dementia demonstrate deficits in olfactory functioning, particularly in olfactory memory, recall, and recognition (Albers et al., 2015; Murphy, 2019; Murphy et al., 1999; Nordin and Murphy, 1996, 2006; Olofsson et al., 2010). Participants with selective odor memory deficits are significantly more likely to possess the ε4 allele (Albers et al., 2016). Odor recognition memory and odor ID are most impaired in ε4 carriers relative to other memory tasks (e.g., picture ID, facial recognition memory; Calhoun-Haney and Murphy, 2005; Gilbert and Murphy, 2004).

The brain regions that are impacted by AD overlap with brain regions that are critical for olfaction. Early projections of the olfactory system involve MTL brain regions that are involved in the encoding and retrieval of episodic memories, such as the ERC and hippocampus (Cerf-Ducastel and Murphy, 2009; Haase et al., 2013). Those with odor memory deficits demonstrate significantly reduced ERC thickness relative to those without deficits (Albers et al., 2016). Atrophy and plaque and tangle accumulation in the MTL are early pathological changes in AD progression (Albers et al., 2015; Attems et al., 2005, 2014; Braak and Braak, 1991, 1996, 1997; Esiri and Wilcock, 1984; Frisoni et al., 2010; Hyman, 1997; Murphy et al., 2003; Price et al., 1991; Stoub et al., 2006; Visser et al., 2002).

Although there are studies exploring neuroimaging data within ε4 carriers and ε4 non-carriers during episodic memory tasks and resting-state functional magnetic resonance imaging (fMRI) (Chen et al., 2015, 2016; Dowell et al., 2016; Filippini et al., 2009; Mondadori et al., 2007; Ungar et al., 2014), information is absent on the interaction of sex and ApoE ε4 status on functional and structural neuroimaging during olfactory recognition memory processing.

In order to fill this gap in the current literature, we examined differential olfactory recognition memory processing within a sample of non-demented, older ApoE ε4 carriers and ε4 non-carriers using structural neuroimaging and blood-oxygen-level-dependent fMRI. We further examined how sex modifies olfactory recognition memory processing within this population.

Hypotheses

Our first hypothesis is that ε4 carriers will demonstrate significantly less volume and thickness within brain regions associated with olfactory memory, such as the ERC and hippocampus, after adjusting for age and ICV. A second hypothesis is that while all ε4 carriers will demonstrate significantly greater functional activation in brain regions associated with odor memory during FPs and significantly less activation during hits, this activation pattern will be most pronounced in male ε4 carriers. Our third hypothesis is that neuroimaging data will be associated with Dementia Rating Scale (DRS) scores within all groups.

Materials and methods

The current study sample was taken from an archival dataset. The Institutional Review Boards both at San Diego State University and the University of California, San Diego had approved the research. The only study published to date on this dataset was Haase et al. (2013) which reported behavioral and functional connectivity analyses.

Participants

Participants (N = 39) ranged in age from 64 to 88 and were divided into one of the 4 groups based on ε4 status and sex: male ε4 carriers (n = 9), female ε4 carriers (n = 9), male ε4 non-carriers (n = 10), and female ε4 non-carriers (n = 11). ApoE ε4 status was determined through genomic testing. We performed a 2 (ε4 status: carrier, non-carrier) × 2 (sex: male, female) multivariate analysis of variance on demographic variables, which included age, education in years, odor threshold (Murphy et al., 1990), odor ID (San Diego Odor Identification Test; Murphy et al., 2002), or DRS scores, a global measure of cognitive function (Mattis, 1998). There were no significant differences between groups in demographics based on E4 status, sex, or the E4 × sex interaction (p > 0.05; Table 1).

Table 1

Participant demographics

Demographics	Mean (SD)				F			Significance
	ε4 non-carriers		ε4 carriers
	Males (n = 10)	Females (n = 11)	Males (n = 9)	Females (n = 9)	ε4 status	Sex	ε4 × sex
Age (y)	69.6 (3.7)	75 (7.1)	73 (7.7)	72.4 (8.9)	0.04	1.14	1.72	p > 0.05
Education (y)	15.5 (2.3)	15.3 (3.5)	15.8 (3.4)	15.3 (3.5)	0.02	0.09	0.03	p > 0.05
Odor threshold	5.5 (1.8)	5.0 (1.4)	4.4 (1.7)	5.8 (1.6)	0.05	0.85	3.10	p > 0.05
Odor ID	4.7 (1.3)	5.8 (1.4)	5.0 (1.5)	5.0 (1.8)	0.29	1.36	0.25	p > 0.05
DRS	140.6 (2.0)	138.0 (4.1)	139.4 (3.5)	138.8 (5.0)	0.02	1.78	0.44	p > 0.05

Key: DRS, Dementia Rating Scale; SD, standard deviation.

Neuroimaging procedure

Prior to scanning, participants were presented with 16 familiar odors, in random order, corresponding to the list A of the California Odor Learning Test (Murphy et al., 1997). During the scan, participants were presented with labels of odors and their task was to indicate if the label was an odor that was presented to them prior to the scan (target) or if it was not (foil) using a button box (Cerf-Ducastel and Murphy, 2009, 2006; Haase et al., 2013). Each participant completed 2 functional runs (6 minutes each) and a structural run. Target periods consisted of 7 targets and 2 foils. Foil periods consisted of 7 foils and 2 targets. This paradigm was adapted from Stark and Squire (2000a, 2000b). Participants discriminated between odors using a button box, pressing 1 if they recognized the odor as having been presented before the scan and 2 if not. For the purposes of our current study, we focused on hit (correctly identifying a target as a target) and false positive (FP; incorrectly identifying a foil as a target) responses.

Imaging acquisition

Functional images were collected first using a standard gradient echo EPI pulse sequence to acquire T2-weighted functional images (30 axial slices, field of view = 25 cm, resolution 4 × 4 × 4 mm3, repetition time = 4 seconds, echo time = 30 ms, flip angle = 90°). Parameters used to acquire structural images were as follows: T1-weighted whole-brain fast spoiled gradient echo magnetic resonance imaging sequence, field of view = 25 cm, resolution = 1 × 1 × 1 mm3, repetition time = 16 seconds, echo time = 4.4 m, flip angle = 18°.

Structural neuroimaging data

T1-weighted structural scans were processed using standard FreeSurfer automated processing procedures within the FreeSurfer image analysis suite, version 5.2.0 (http://surfer.nmr.mgh.harvard.edu; Dale and Sereno, 1993; Fischl and Dale, 2000; Fischl et al., 2001; Fischl et al., 2004a; Fischl et al., 1999a; Dale et al., 1999; Fischl et al., 1999, 2002, 2004; Han et al., 2006; Jovicich et al., 2006; Kuperberg et al., 2003; Reuter et al., 2010, 2012; Reuter and Fischl, 2011; Rosas et al., 2002; Salat et al., 2004; Ségonne et al., 2004, 2007; Sled et al., 1998). Cortical thickness and volumetric estimates of regions of interest were extracted from automatic surface parcellation labels using the Desikan/Killiany Atlas (Desikan et al., 2006).

Functional neuroimaging data

Imaging data were processed using FMRIB Software Library (Analysis Group, FMRIB, Oxford, UK) and Analysis of Functional NeuroImage (open source software), using 3dDeconvolve, on each participant’s concatenated runs based on the specified contrast (e.g., activation during hits and FPs; Cox, 1996; Cox and Hyde, 1997; Gold et al., 1998; Smith et al., 2004; Zald and Pardo, 2000). The output from 3dDeconvolve contains fit coefficients (i.e., beta weights) for each voxel, indicating the amplitude of the signal model for each contrast, and corresponding t-statistics.

Statistical analyses

Task performance

We performed a 2 (sex: male, female) × 2 (ε4 status: ε4 non-carrier, ε4 carrier) multivariate analysis of covariance (MANCOVA) on the hit rate sand FP rate for the 2 independent runs and their average performance (Table 2). The MANCOVA controlled for age.

Table 2

Olfactory memory task performance rates

Memory measure	Mean (SD)				F (significance)
	ε4 non-carriers		ε4 carriers
	Males (n = 10)	Females (n = 11)	Males (n = 9)	Females (n = 9)	ε4 status	Sex	ε4 × sex
Hit rates
Run 1	0.42 (0.13)	0.57 (0.11)	0.59 (0.12)	0.64 (0.17)	7.597[a] (p = 0.01)	4.829[a] (p = 0.04)	1.390 (p > 0.05)
Run 2	0.44 (0.17)	0.55 (0.14)	0.53 (0.21)	0.60 (0.15)	2.025 (p > 0.05)	2.234 (p > 0.05)	0.135 (p > 0.05)
FP rates
Run 1	0.32 (0.15)	0.30 (0.11)	0.36 (0.12)	0.29 (0.09)	0.189 (p > 0.05)	1.27 (p > 0.05)	0.384 (p > 0.05)
Run 2	0.29 (0.15)	0.30 (0.14)	0.30 (0.16)	0.30 (0.11)	0.018 (p > 0.05)	0.012 (p > 0.05)	0.002 (p > 0.05)

Covariates appearing in the model are evaluated at the following values: age = 72.56.

Key: FP, false positives; SD, standard deviation.

Denotes significance.

We performed a 2 (sex: male, female) × 2 (ε4 status: ε4 non-carrier, ε4 carrier) MANCOVA on left/right hippocampal volumetric and left/right ERC thickness measurements (Table 3). The MANCOVA controlled for age and intracranial volume (ICV).

Table 3

Hippocampal volume (mm3) and entorhinal cortex thickness (mm)

Structural measurements	Mean (SD)				F (significance)
	ε4 non-carriers		ε4 carriers
	Males (n = 10)	Females (n = 11)	Males (n = 9)	Females (n = 9)	ε4 status	Sex	ε4 × sex
Left hippocampus (mm3)	2897.30 (869.75)	2957.45 (396.96)	2742.11 (656.94)	3400.44 (519.24)	0.16 (p > 0.05)	5.76[a] (p = 0.02)	0.949 (p > 0.05)
Right hippocampus(mm3)	2923.40 (649.00)	3240.64 (538.18)	2994.78 (569.51)	3507.33 (462.88)	1.263 (p > 0.05)	8.59[a] (p = 0.006)	0.008 (p > 0.05)
Left ERC (mm)	2.45 (0.55)	2.64 (0.62)	2.22 (0.94)	2.61 (0.39)	0.532 (p > 0.05)	3.24 (p > 0.05)	0.003 (p > 0.05)
Right ERC (mm)	2.83 (0.49)	2.61 (0.18)	2.497 (0.39)	2.91 (0.39)	0.002 (p > 0.05)	0.66 (p > 0.05)	5.934[a] (p = 0.02)

Covariates appearing in the model are evaluated at the following values: age = 72.56, intracranial volume = 1,343,211.69.

Key: ERC, entorhinal cortex; SD, standard deviation.

Denotes significance.

We performed a set of independent voxel-wise samples t-tests in whole-brain functional activations between groups during hits and FPs. Each analysis controlled for age and ICV. In an attempt to control for type I error in all group analyses, we thresholded the individual voxels at p ≤ 0.015 and group statistical maps were corrected for multiple comparisons at the cluster level using the Analysis of Functional NeuroImage program ClustSim to protect a whole-brain probability of FPs at an overall alpha of 0.05 (Zald and Pardo, 2000). For an overall alpha level of 0.05, a cluster threshold of 21 contiguous voxels was applied.

Partial correlations

We performed a series of partial correlational analyses to examine associations between DRS scores and neuroimaging data. For correlations examining the association between structural measurements and DRS scores, partial correlations controlled for age and ICV. These correlations used a corrected alpha level of 0.004, which was calculated by dividing the original alpha level of 0.05 by the product of the number of groups (4) and the number of predictors (3).

If we examined the relationship between beta coefficients and DRS scores, each partial correlation controlled for age, ICV, and volume of the structure we were evaluating. These partial correlations used a corrected alpha level of 0.003, which was calculated by dividing the original alpha level of 0.05 by 16, the product of the number of groups (4) and the number of predictors (4).

Results

Task performance

Females demonstrated a significantly higher hit rate in run 1 when compared to males (MF = 0.605 vs. MM = 0.500, p = 0.04). The ε4 carrier group also demonstrated significantly higher hit rates in run 1 when compared to ε4 non-carriers (M+ = 0.616 vs. M− = 0.501; p = 0.01).

We found no significant main effects of sex or ε4 status on run 2 hit rates or FP rates. There were also no significant interaction effects of the ε4 × sex interaction on hit or FP rates (p > 0.05). The results of this multivariate analysis of variance are summarized in Table 2.

Structural neuroimaging data

Table 3 summarizes group differences in hippocampal volume and ERC thickness. Females demonstrated significantly larger left hippocampal volumes (MF = 3156.80 vs. MM = 2823.79; p = 0.022) and right hippocampal volumes relative to males (MF = 3360.65 vs. MM = 2957.21; p = 0.006). For the right ERC thickness dependent variable, the ε4 status × sex interaction effect was statistically significant (p = 0.020). However, when simple effects tests were conducted to probe the interaction, we found no statistically significant (p > 0.0125) simple effects of ε4 status at any level of sex or of sex at any level of ε4 status.

Functional neuroimaging data

Hits

When functional activations of ApoE ε4 carriers were subtracted from functional activations of ApoE ε4 non-carriers during hits, no significant differences were found between groups.

When functional activations during hits of ε4 carrier females were subtracted from functional activations of ε4 carrier males, males demonstrated significantly greater activation than females in the ApoE ε4 carrier group in the anterior cingulate cortex (ACC), Brodmann’s area (BA) 10, cuneus, precuneus, middle temporal gyrus, caudate, and putamen (Fig. 1A; Supplementary Materials A).

Fig. 1.

(A) Whole-brain activation during hits when functional activation of ε4 carrier females was subtracted from ε4 carrier males. Orange indicates the areas where ε4 carrier males had greater activation compared to ε4 carrier females. (B) Whole-brain activation during false positives when functional activation of ε4 non-carriers was subtracted from ε4 carriers. Yellow indicates the areas where ε4 carriers had greater activation than ε4 non-carriers. (C) Whole-brain activation during false positives when functional activation of ε4 carrier females was subtracted from ε4 carrier males. Blue indicates the areas where ε4 carrier females had greater functional activation than ε4 carrier males.

When functional activation of ApoE ε4 non-carrier males was subtracted from ApoE ε4 non-carrier females, there were no statistically significant differences between groups.

False positives

When functional activations of ApoE ε4 non-carriers were subtracted from functional activations of ApoE ε4 carriers during FPs, ε4 carriers demonstrated significantly greater activation than ε4 carriers in the middle temporal gyrus, cuneus, precuneus, and right posterior cingulate cortex (PCC; Fig. 1B; Supplementary Materials B).

When functional activations of ApoE ε4 carrier females were subtracted from ApoE ε4 carrier males, females demonstrated significantly greater functional activation in areas associated with visual processing, such as the inferior and middle occipital gyrus (Fig. 1C; Supplementary Materials C).

We found no significant differences between males and females in the ε4 non-carrier group during FPs.

Partial correlational analyses: neuroimaging data

We found no significant (p > 0.004) associations between beta coefficients and DRS scores in any of the 4 groups. Male ε4 non-carriers demonstrated a significant, negative association between right hippocampal volume and DRS scores (r = −0.888, p = 0.003; Fig. 2B).

Fig. 2.

Association between right hippocampal volume (mm3) and Dementia Rating Scale scores in male ε4 non-carriers. Age and ICV were included as covariates. Abbreviation: ICV, intracranial volume.

Post hoc analyses

We performed exploratory analyses in order to further observe differences between groups. A 2 (sex: male, female) × 2 (ε4 status: non-carrier, carrier) between-subjects MANCOVA was performed on 8 dependent variables: left/right caudate volume, left/right putamen volume, left/right isthmus cingulate thickness, and left/right parahippocampal thickness (Table 4). The MANCOVA controlled for age and ICV.

Table 4

Post hoc structural measurements

Structural measurements	Mean (SD)				F (significance)
	ε4 non-carriers		ε4 carriers
	Males (n = 10)	Females (n = 11)	Males (n = 9)	Females (n = 9)	ε4 status	Sex	ε4 × sex
Left caudate (mm3)	3493.60 (439.56)	3531.55 (620.92)	4170.78 (666.75)	3616.56 (486.29)	3.47 (p > 0.05)	2.02 (p > 0.05)	2.10 (p > 0.05)
Right caudate (mm3)	3492.30 (415.74)	3628.82 (960.68)	4219.78 (387.79)	3704.22 (541.19)	1.91 (p > 0.05)	0.58 (p > 0.05)	2.05 (p > 0.05)
Left putamen (mm3)	4679.30 (824.49)	4658.55 (607.08)	5502.22 (741.86)	4930.44 (620.86)	5.52[a] (p = 0.03)	0.96 (p > 0.05)	2.61 (p > 0.05)
Right putamen (mm3)	4424.10 (670.16)	4401.27 (668.75)	5419.67 (625.01)	4889.22 (628.16)	11.96[a] (p = 0.002)	1.22 (p > 0.05)	2.27 (p > 0.05)
Left parahippocampal gyrus (mm)	2.39 (0.34)	2.52 (0.35)	2.01 (0.38)	2.62 (0.44)	1.54 (p > 0.05)	9.65[a] (p = 0.004)	2.95 (p > 0.05)
Right parahippocampal gyrus (mm)	2.32 (0.24)	2.40 (0.36)	2.36 (0.38)	2.44 (0.16)	0.36 (p > 0.05)	0.77 (p > 0.05)	0.04 (p > 0.05)
Left isthmus cingulate gyrus (mm)	2.67 (0.21)	2.51 (0.23)	2.36 (0.15)	2.47 (0.17)	3.10[a] (p = 0.09)	0.57 (p > 0.05)	4.54[a] (p = 0.041)
Right isthmus cingulate gyrus (mm)	2.68 (0.24)	2.47 (0.35)	2.46 (0.24)	2.46 (0.19)	1.12 (p > 0.05)	0.90 (p > 0.05)	0.78 (p > 0.05)

MANCOVA controlled for age and ICV.

Key: ICV, intracranial volume; SD, standard deviation.

Denotes significance.

The caudate and putamen make up the dorsal striatum, which is affected by amyloid and tau pathology (Alexander et al., 1986; Beach et al., 2012; Braak and Braak, 1990; Rudelli et al., 1984). Higher plaque accumulation in the striatum was highly sensitive (95.8%) and moderately (75.7%) specific to Braak neurofibrillary tangle stage V or VI. Furthermore, a higher striatal plaque density score had 85.6% sensitivity and 86.2% specificity for the presence of dementia and clinicopathological AD (Beach et al., 2012). The parahippocampal gyrus is located in an MTL region that projects onto the hippocampus and is activated during olfactory encoding, working memory, and successful recognition memory (Cerf-Ducastel and Murphy, 2009; Haase et al., 2013; Luzzi et al., 2007; Stoub et al., 2006). The isthmus of the cingulate gyrus connects the cingulate gyrus to the parahippocampal gyrus and has shown long-term volume reduction in response to stressful life events and dementia within older populations (Calati et al., 2018; Sener, 1997; Yang et al., 2016).

Females demonstrated significantly greater thickness measurements in the left parahippocampal gyrus, relative to males (MF = 2.56 vs. MM = 2.21; p = 0.004). The ε4 carrier group, relative to the ε4 non-carriers, demonstrated significantly larger volumes in the left putamen (M+ = 5216.33 vs. M− = 4668.43; p = 0.03) and right putamen (M+ = 5154.44 vs. M− = 4412.14; p = 0.002).

The ε4 status × sex interaction effect was also significant (p = 0.041; Fig. 2). When simple effects were examined, male ε4 non-carriers demonstrated significantly greater thickness measurements relative to male ε4 carriers (M = 2.666 vs. M = 2.359; p = 0.002).

Post hoc partial correlation analyses

We performed partial correlational analyses, controlling for age and ICV, between neuroimaging data and DRS scores using the same methods described in Section 2.4.4. We found a significant, positive association between DRS scores and functional activation in the right parahippocampal gyrus during FPs in male ε4 carriers (r = 0.952, p = 0.003; Fig. 3).

Fig. 3.

Partial regression plots using age, ICV, and structure volume as covariates. Association between Dementia Rating Scale scores and functional activation in the right parahippocampal gyrus during false positives in male ε4 carriers. Abbreviation: ICV, intracranial volume.

Discussion

Structural neuroimaging

Our first hypothesis, that ε4 carriers would demonstrate smaller volumetric and thickness measurements in memory regions relative to ε4 non-carriers, was not upheld. We found that ε4 carriers demonstrated significantly larger bilateral putamen volumes. The larger volumetric measurements may be related to inflammation, plaque accumulation, and shape abnormalities previously associated with striatal areas (Beach et al., 2012; Braak and Braak, 1990; de Jong et al., 2008, 2011; Pievani et al., 2013; Rudelli et al., 1984). Inflammation and Aβ plaque accumulation have been suggested as possible explanations for increased thickness or volumetric measurements (Fox et al., 2005). An fluorodeoxyglucose (FDG)-positron emission tomography (PET) study found that the putamen of cognitively normal ApoE ε4 homozygotes had the highest amyloid deposition relative to other brain regions (Pardo and Lee, 2018). Similarly, another fluorodeoxyglucose (FDG)-positron emission tomography (PET) study found that amyloid deposition occurs very early in the striatum, which is comprised of the caudate and putamen, and these deposits are often not associated with clinical symptoms (Klunk et al., 2007). Furthermore, an X-ray micro-diffraction study demonstrated that there can be structural heterogeneity of amyloid such that subjects with different clinical histories may contain different ensembles of fibrillar structures (Liu et al., 2016). Polymorphism in the distribution of amyloid in plaques may be related to different disease states and variable manifestations of clinical symptoms (Liu et al., 2016; Lu et al., 2013).

The greater putamen volumes may also be attributed to neuronal compensation, which refers to a dissociation between brain pathology and behavioral change during the early and prodromal stages of neurodegenerative diseases (Gregory et al., 2017; Scheller et al., 2014). It has been proposed that in conjunction with pathological loss of brain tissue, there is a structural reorganization in the brain to compensate for these losses, which enables prodromal patients to perform functionally at the same level as those not at risk for AD (Gregory et al., 2017). It is the lack of behavioral differences between groups that indicates neuronal compensation (Gregory et al., 2017). Task performance data suggest that there were no differences between ε4 carriers and non-carriers, except for run 1, when ε4 carriers made more hits relative to ε4 non-carriers. Larger putamen volumes have also been found in other developmental or neurodegenerative disorders. When adults with autism spectrum disorder and typical developed controls were compared, the putamen was found to be significantly larger in the autism spectrum disorder group after controlling for age, sex, and ICV (Sato et al., 2014). Similarly, larger putamen volumes have been reported in individuals with bipolar disorder, schizophrenia, and obsessive-compulsive disorders (Luo et al., 2019). Hyperactivation in dopamine pathways within the striatum has been suggested to be the cause of this enlargement of the dopamine-rich putamen (Luo et al., 2019). We therefore suspect that the greater putamen volumes may be indicative of overcompensation which allows ε4 carriers to maintain functional performance in response to changes imposed by aging and ε4 status.

We found sex effects on thickness measurements in the left parahippocampal gyrus and left isthmus cingulate. Increasing asymmetry in these regions is associated with the transition of mild cognitive impairment (MCI) to AD (Long et al., 2013). A recent longitudinal study reported that when comparing ε4 carriers and ε4 non-carriers, ε4 carriers demonstrated greater rates of volume loss in the hippocampus and parahippocampal gyrus (Reiter et al., 2017). The isthmus cingulate, which connects the cingulate gyrus to the parahippocampal gyrus, was significantly reduced in ε4 carriers and in male ε4 carriers relative to ε4 non-carriers and male ε4 non-carriers, respectively (Sener,1997). This finding is supported by previous literature reporting greater atrophy and increased rate of cortical thinning in the isthmus cingulate in AD (Hayata et al., 2015; Mak et al., 2015; Vasconcelos et al., 2014).

Functional neuroimaging

Our second hypothesis, that ε4 carriers would demonstrate less functional activation during hits and greater functional activation during FPs, was partially upheld. Furthermore, we hypothesized that this activation would be especially pronounced in male ε4 carriers; this was strongly upheld. The ε4 carrier group, relative to ε4 non-carriers, and male ε4 carriers, relative to female ε4 carriers, showed greater recruitment during FPs and hits, respectively in brain regions associated with memory and decision-making (e.g., BA10, precuneus, PCC, ACC). The BA10 is an area associated with encoding the incentive value of a stimulus during decision-making and is activated by a reward stimulus when already activated by working memory processing (O’Doherty et al., 2001). The precuneus is an integral region associated with source memory retrieval and the PCC demonstrates activation during autobiographical memory retrieval (Bonnì et al., 2015; Maddock et al., 2001). Lateralization in the ACC has been activated in error processing and conflict monitoring, suggesting that the ACC is vital in making correct choices (Lütcke and Frahm, 2008).

The precuneus and PCC are not only associated with memory and information processing, but also affected early in AD progression. Hyperactivation in the left precuneus and right cingulate gyrus was found in early AD patients, relative to healthy controls, during a time representation task (Leyhe et al., 2009). During a semantic memory task, increased activation in the bilateral PCC and precuneus has been reported in asymptomatic ε4 carriers relative to controls (Seidenberg et al., 2009). Atrophy in tracts connecting the hippocampus to the precuneus and PCC differentiates normal controls from MCI and AD patients (Palesi et al., 2012). Furthermore, abnormal hypoperfusion in the PCC and precuneus distinguishes MCI patients from normal controls (Bradley et al., 2002; Rombouts et al., 2005).

It is noteworthy that despite no significant differences in FP rates, ε4 carriers demonstrated significantly greater cognitive expenditure during FPs when compared to non-carriers. Hyperactivation has been previously noted in AD populations and is often understood as a compensatory response to functional impairment and accumulating AD pathology, possibly even acting as a protective factor to help ε4 carriers maintain their cognitive abilities in the short-term (Dickerson et al., 2004; Leyhe et al., 2009; Seidenberg et al., 2009; Woodard et al., 2010). This compensatory response may occur to support task performance in response to reduced communication between brain regions that typically function together (Seidenberg et al., 2009; Zhu et al., 2015). Age-related activation increases have been correlated with poorer behavioral performance and lower white matter integrity (Zhu et al., 2015). Furthermore, excessive hyperactivation over time can lead to excitotoxicity and ultimately result in synapse degeneration and death (Mattson and Chan, 2003; Poirier et al., 1993; Woodard et al., 2010; Yanker, 1996) and has been suggested as one mechanism contributing to olfactory dysfunction in AD (Jacobson et al., 2019; Murphy, 2019). Interestingly, ApoE ε4 carrier mice show hyperactivation in olfactory areas that has been associated with olfactory deficits, suggesting that subtle, early olfactory deficits may presage future abnormalities in olfactory circuitry, and further supporting the hypothesis that hyperactivation is associated with olfactory dysfunction (East et al., 2018; Peng et al., 2017). This hyperactivation has been suggested to precede clinical symptoms in the AD pathological timeline (Gregory et al., 2017; Murphy, 2019).

Partial correlations: neuroimaging data

The third hypothesis, that neuroimaging data and DRS scores would be associated in all groups, was partially upheld. Partial correlations revealed significant associations between neuroimaging data and DRS scores within males. For male ε4 carriers, activation in the right parahippocampal gyrus was associated with higher DRS scores after age, ICV, and right parahippocampal gyrus thickness were partialed out. Higher DRS scores suggest greater global cognitive functioning (Mattis, 1998). These results support the hypothesis that hyperactivation may be protective against cognitive decline in the short term. In the male ε4 non-carrier group, larger right hippocampal volume was associated with lower DRS scores after controlling for age and ICV. This association suggests that within this sample, greater right hippocampal volume does not necessarily mean better global cognitive functioning.

Conclusions

The present study is the first to investigate how the interaction of sex and ApoE ε4 status impacts both structural MRI and fMRI activation during an olfactory recognition memory task in a non-demented, older population. The study demonstrates differential neuroimaging data and olfactory memory processing in relation to sex and ApoE ε4 status. No study to our knowledge has reported both structural and functional neuroimaging data within older, non-demented ε4 carriers and non-carriers during an olfactory recognition memory task. Moreover, neuroimaging studies reporting on sex differences present in olfactory recognition memory processing are lacking. Our results expand preceding literature suggesting hyperactivation patterns in those at risk for AD and provide valuable information regarding structural differences between ε4 carriers and ε4 non-carriers before AD symptoms manifest. Results suggest that prior to developing AD symptomatology, male ε4 carriers demonstrate hyperactivation during odor recognition memory relative to other groups, which may ultimately lead to future brain atrophy and decreased cognitive functioning.

There were limitations to our study. First, the study would have benefitted from larger sample sizes. Although we considered structural neuroimaging data and fMRI, we did not include data on amyloid and plaque accumulation through other imaging methods (e.g., amyloid-PET imaging, diffusion tensor imaging), which allow for examination of pre-mortem amyloid plaque burden (Hedden et al., 2013; Klunk et al., 2004). Future studies may consider these methods in conjunction with fMRI and structural MRI.

Together, our findings underscore the potential of incorporating MRI methodologies, genetics, and sex differences to identify those at risk for AD. As clinical trial research progresses to develop measures or interventions intended to slow or prevent AD clinical progression, it is imperative that we consider these factors to design interventions that are suitable, effective, and most beneficial for the population.