Alzheimer's disease (AD) is the most prevalent form of dementia. In a small percentage of cases, however, patients that carry specific mutations in the amyloid precursor protein (APP), presinilin-1 (PSEN1), or presinilin-2 (PSEN2) genes will develop an autosomal dominant form of AD known as familial AD (FAD) (Campion et al., 1999; Bateman et al., 2012; Cruchaga et al., 2012; Long and Holtzman, 2019). However, most cases of AD are sporadic of unknown cause (Hampel et al., 2021). It is thought that AD develops decades before cognitive manifestations (Jack et al., 2013), making it difficult to understand the early processes that occur in the disease. Thus, identifying early biomarkers of AD would be important for both understanding the progression of the disease as well as for clinical translation.

Previous studies have suggested that disrupted connectivity in the default mode network, in particular the hippocampus, is one of the early functional changes in AD (Greicius et al., 2004; Allen et al., 2007; Li et al., 2018), leading to the network degeneration hypothesis, which proposes that pathology begins in select vulnerable regions, leading to synaptic loss and dysfunction that then spreads to other anatomically-related brain regions (Delbeuck et al., 2003; Seeley et al., 2009; Oh et al., 2015). In the hippocampus, there are also differences between the left and right hemispheres. These include differences in receptor expression, physiology, and spatial memory, among other features (Jordan, 2020). Furthermore, interhemispheric connectivity is important for some cognitive tasks, which has been shown to be defective in AD patients (Lakmache et al., 1998), and has been shown to act as a potential compensatory mechanism, e.g., after sleep deprivation (Zhu et al., 2016). However, it is still not fully understood how functional connectivity (FC) changes across age, and if there are asymmetric differences in those changes, particularly in the context of AD.

Here, we used in vivo resting-state functional magnetic resonance imaging (rs-fMRI) to investigate changes in hippocampal connectivity across age in wild-type and App^{NL-G-F/NL-G-F} knock-in (APPKI) mice harboring the Swedish, Arctic, and Beyreuther/Iberian mutations in the amyloid precursor protein (App) gene, which results in the production of Aβ plaques at approximately 4 months of age and leads to learning and memory deficits starting at approximately 6 months of age (Saito et al., 2014). In contrast to transgenic mouse models of FAD, this model does not encounter potential artifacts due to overexpression of one or more transgenes (Saito et al., 2014).

We observed strengthening in interhemispheric connectivity across age, but this effect was weaker in the APPKI mice relative to the wild-type. Conversely, we observed reduced intrahemispheric connectivity with age. However, this reduction was not symmetric, with the right hemisphere showing a greater decline with age relative to the left hemisphere, and differences between wild-type and APPKI specifically in the right hemisphere. Together, these results help to improve our understanding of the early functional changes that occur in the hippocampus in mice with AD pathology, and how FC is affected by age.

In this study, we observed age-related alterations in FC of the hippocampus in the App^{NL-G-F/NL-G-F} mouse model of AD. Evidence from previous studies suggests that functional and structural interhemispheric connectivity is disrupted in AD compared to healthy aging controls (Lakmache et al., 1998; Wang et al., 2015; Li et al., 2018), as well as reduced network connectivity (Allen et al., 2007), particularly in the default mode network (DMN) (Raichle et al., 2001; Greicius et al., 2004), and atrophy of major white matter fibers, such as the corpus callosum (Li et al., 2018). A previous study by Wang et al. (2015) found that interhemispheric FC is reduced in AD patients, and was correlated with reduced integrity of the corpus callosum measured via DTI, as well as correlated with reduced cognitive scores. Further, it has been observed that disconnection of the corpus callosum, e.g., for the treatment of epilepsy, leads to significant loss of interhemispheric FC (Johnston et al., 2008; Wang et al., 2015). Since myelin and oligodendrocytes are also known to be affected in AD (Bartzokis, 2011), this suggests that structural integrity of white matter connections may also be important for preservation of functional connectivity. Many AD patients also experience disruptions in their circadian rhythm (Webster et al., 2014), and interhemispheric connectivity can also be modulated by sleep deprivation and changes in circadian rhythm (Zhu et al., 2016). While the precise mechanisms are still not well understood, taken together, this suggests that there is a complex interaction between brain structure, function, and environmental factors in AD pathology.

This reduction in resting-state interhemispheric and DMN connectivity in AD may be coupled with a complementary increase in task-induced activity. With altered resting-state connectivity, it may be more difficult for the brain to inhibit activity not relevant for a given task (Celone et al., 2006; Sheline and Raichle, 2013), referred to as task-induced deactivation (TID). Indeed, previous studies have shown that more cognitively demanding tasks are also associated with higher TID (Daselaar et al., 2004). Early MCI patients showed increased hippocampal activity during an episodic memory task and less TID relative to controls (Celone et al., 2006), in this context referred to as hippocampal “hyperexcitation.” The study by Bakker et al. (2012) observed hyperexcitation in the CA3/DG in mild cognitive impairment (MCI) patients relative to controls during memory-related tasks. Treatment with anti-epileptic medication, which effectively reduced the hyperactivity, resulted in increased memory performance in these patients (Bakker et al., 2012, 2015). Furthermore, hippocampal hyperexcitation has been shown to precede the onset of the cognitive symptoms. Analysis of the hyperexcitation index (HI) in cognitively healthy middle-aged APOE4 carriers and non-carriers showed that female APOE4 carriers have an increased HI relative to non-carriers (Fortel et al., 2020).

At the neurobiological level, this may be due to induced neuronal expression of APOE, leading to APOE4-mediated toxicity of GABAergic interneurons in the hippocampus, which could lead to wider network hypersynchrony (Najm et al., 2019). While the majority of GABAergic neurons synapse locally, previous studies have identified long-range GABAergic projections, in the hippocampal commissure from the hilus and from the CA3 and CA1 (Ribak et al., 1986). Notably, in the hAPP-J20 mouse model, which expresses human APP with FAD mutations (Mucke et al., 2000), imbalance between GABAergic and glutamatergic transmission has been reported to compromise hippocampal neurogenesis, which regulates the activity level of the DG and hippocampus (Sun et al., 2009). In addition, evidence from mouse stereological studies suggests that the DG, CA1, and CA3 each have specific GABAergic to glutamatergic neuron ratios, with ≈1.5 % in the DG, ≈11 % in the CA1, and ≈10 % in the CA3 in the dorsal hippocampus. In the ventral hippocampus, the GABAergic ratio is larger, with ≈4 % in the DG, ≈22 % for the CA1, and ≈21 % for the CA3. Thus, the CA1 and CA3 have comparable GABAergic ratios, while the DG has the lowest (Jinno et al., 1998; Jinno and Kosaka, 2009). Furthermore, previous studies suggest that there may be age-related changes in both glutamatergic and GABAergic signaling with age and AD (Stephens et al., 2011; Albuquerque et al., 2015; Hollnagel et al., 2019; Kumar and Foster, 2019). Here, we observed larger changes in interhemispheric FC with age primarily in the CA3 and CA1 subregions. Given that the CA1 and CA3 have a relatively larger GABAergic ratio relative to the DG, this may suggest that there are alterations in the excitation-inhibition balance in these regions. Consistent with this, a recent study by Arroyo-García et al. (2021) reported disruption in γ oscillations in fast-spiking interneurons in the CA3 of App^{NL-G-F/NL-G-F} mice by 2 months of age, prior to the onset of amyloid plaque formation. They also observed that γ oscillations in CA3 pyramidal cells become disrupted by 6 months of age, the age at which Saito et al. (2014) had previously observed cognitive impairments to begin. This suggests that these early changes in excitation-inhibition balance may be due to soluble Aβ. Additionally, the septum, an important regulator of hippocampal θ oscillations, as well as the entorhinal cortex, contain long-range GABAergic connections with the hippocampus. These long-range projections also primarily synapse to other GABAergic neurons, suggesting that they may function to regulate synchronous activity in the hippocampus (Caputi et al., 2013). Thus, changes in the functional connectivity in the hippocampus in AD may be due in part to disinhibition of GABAergic regulatory circuits that lead to hyperexcitation-induced toxicity (Najm et al., 2019).

The inter/intrahemispheric FC changes observed may be influenced by the neurobiological asymmetries known to exist in the hippocampus (Jordan, 2020). For example, there is left–right asymmetric expression of synaptic receptors. Left ipsi- and contralateral CA3–CA1 projections primarily express the GluN2B receptor, whereas the right ipsi- and contralateral CA3–CA1 projections express the GluR1 receptor (Kawakami et al., 2003; Shinohara et al., 2008; Jordan, 2020). Consequently, these projections have distinct synaptic plasticity properties. These asymmetries also extend to the behavioral level. Based on data from previous studies, the discrete–continuous model of spatial processing suggests that the left CA3 may encode salient features, whereas the right CA3 processes more continuous navigation (Jordan, 2020). These studies highlight that there are molecular and functional asymmetries in the hippocampus and throughout the brain, which may be important for cognitive performance. For example, a study by Shimbo et al. (2018) showed that knockout of the β2-microglobulin, which helps to establish hippocampal circuit left-right asymmetry (Kawahara et al., 2013; Shimbo et al., 2018), resulted in slower learning of non-spatial cognitive tasks compared to control mice, suggesting that intact asymmetry is important for healthy cognitive function. In humans, Lakmache et al. (1998) observed that AD patients engaging in tasks requiring interhemispheric connectivity performed worse compared to healthy control patients, while tasks requiring only one hemisphere were not nearly as impaired. Together, the combined effects of aging and AD pathology may be responsible for reducing these functional asymmetries, leading to reduced memory performance and cognitive deficits seen in AD patients.

Resting-state fMRI has become increasingly used for studying rodent models (Mandino et al., 2020). Previous studies in App^{NL-G-F/NL-G-F} and App^NL-F/NL-F mouse models have observed increased bilateral hippocampal activity (Shah et al., 2018) and prefrontal synchrony (Latif-Hernandez et al., 2019) around 3–4 months of age, with subsequent decline after 7 months of age, which is consistent with our observations, suggesting that between 4 and 6 months of age there is a shift from increased to decreased interhemispheric connectivity in the hippocampus. Around this age point, 5×FAD mice have also been shown to have alterations in their functional connectome network (Kesler et al., 2018). In older age groups, reduced hippocampal FC has been observed by 10 months of age in longitudinal studies of TgF344 (APPSwe; PSEN1ΔE9) rats (Anckaerts et al., 2019), and reduced interhemispheric FC was reported in 18-month-old APP/PS1 mice (Shah et al., 2013). Reduced FC has also been observed in aged APOE mouse models (Zerbi et al., 2014), suggesting that disruptions in FC observed in mouse models with APP and presinilin FAD mutations may also generalize to other AD genetic risk factors. Together, these studies and ours suggest that in early age there is higher functional activity in AD rodent models that decreases around middle age and continues with increasing age, and appears in both sexes and in mouse models with multiple AD-related mutations.

There are some limitations to this study. While only female mice were used in this study, this was translationally relevant in that females are significantly more likely to develop AD (Li and Singh, 2014; Alzheimer's Association, 2022), as well as controls for sex-specific differences in the current study. However, future studies with male mice will be needed to determine if the functional dynamics we observed in female mice may be different in male mice. Finally, there are also potential confounds of anesthesia that should be considered, since this is a crucial distinction between human and animal rs-fMRI. Previous studies have described the effects of various anesthetics on functional networks in the context of animal imaging (Bukhari et al., 2017; Mandino et al., 2020). While consistency of isoflurane administration is taken, it is possible there are variable effects of isoflurane on neurovascular coupling with aging or between different individuals. With this caveat in mind, however, while higher concentrations of isoflurane may impact interhemispheric connectivity (Mandino et al., 2020), previous studies have shown that interhemispheric FC is preserved with low doses of isoflurane comparable to that observed in awake mice (Jonckers et al., 2014; Zerbi et al., 2014).

In summary, we used in vivo resting-state fMRI to measure the FC of wild-type and App^{NL-G-F/NL-G-F} mice in different age groups. We observed higher interhemispheric FC, followed by later decrease in interhemispheric FC in the hippocampus of App^{NL-G-F/NL-G-F} mice compared to wild-type mice. Furthermore, we observed no laterality in intrahemispheric FC in the App^{NL-G-F/NL-G-F} mice. Together, these results suggest that FC in the context of AD is affected by both aging and hemispheric asymmetry, and that early alterations in hippocampal activity may be an important biomarker and for understanding the progression of the disease.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The animal study was reviewed and approved by Animal Care Committee (ACC) at the University of Illinois at Chicago.

ZM, OL, and AL designed research. ZM, JG, and WL performed research. TSaid and TSait contributed unpublished reagents/analytic tools. ZM, LZ, IF, and AL analyzed data. ZM wrote the paper. ZM, JG, LZ, WL, IF, AB, SM, OA, OL, and AL edited the paper. All authors contributed to the article and approved the submitted version.