Getting Lost: Place Cells and Grid Cells in Rodent Models of Alzheimer’s Disease

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder and affects 60–80% of all dementia cases in the elderly population worldwide. Getting lost is one of the earliest cognitive symptoms to emerge in AD patients. Despite decades-long efforts to identify the molecular mechanisms underlying the pathogenesis of β-amyloid (Aβ) plaques and tau neurofibrillary tangles, no effective therapeutic interventions have been developed for AD so far [1]. The early onset of amnestic deterioration occurs first in the entorhinal cortex and spreads later to the hippocampus, making the hippocampal-entorhinal loop the most vulnerable region. Thus, the hippocampal-entorhinal network has been a primary focus of AD study.

Mutated amyloid precursor protein, presenilin-1, presenilin-2, and/or tau are the mainstay for recapitulating human AD in rodent models. Over the last few decades, more than 100 transgenic rodent models of AD have contributed to our understanding of how this age-associated neurodegenerative disease leads to progressive spatial memory impairment. However, the majority of the work has focused on the identification of genes, proteins, and cell types responsible for AD neuropathology largely at the synaptic, molecular, cellular, and behavioral levels. Therefore, it has become increasingly important to uncover the underlying brain circuit mechanisms.

So far, there have been several attempts using extracellular recording from freely-moving rodent models of AD to decode spatial memory. In 2008, Cacucci and colleagues were the first to record place cells in Tg2576 mice and found that Aβ plaques are associated with degraded place cells [2]. Later, Cheng and Ji in 2013 showed that tau pathology in rTg4510 mice prevents the formation of normal place cells [3]. Zhao and colleagues in 2014 revealed less stable place cell encoding in the APP/TTA mouse model of Alzheimer’s amyloidosis [4]. An extended study of rTg4510 transgenic tauopathy by Ciupek and colleagues demonstrated in 2015 that the hippocampal CA1 place cells showed reduced ripple oscillations on linear tracks [5], and Cayzac et al. showed that CA1 place cells are impaired in APP-PS1 mice [6] in the same year. Similarly, Booth et al. [7] confirmed the same aberrant CA1 spatial coding in the rTg4510 mouse model of tauopathy in 2016 as shown by Cheng and Ji [3]. In 2017, Mably et al. reported unstable irregular place cells in 3xTg mice [8]. In the same year, Fu et al. first demonstrated that the accumulated Tau in the medial entorhinal cortex (MEC) caused impaired spatial memory and grid cell dysfunction in aged EC-Tau mice [9]. Most recently, Jun et al. went further to investigate remapping of both place cells in the hippocampus and grid cells in the MEC of an mouse model of AD [10]. This study made a major contribution to our understanding of the pathophysiology of AD by characterizing spatial deficits at the circuit level.

First, Jun and colleagues used a novel homozygous APP^NL-G-F knock-in mouse model (named APP-KI below) [10]. Since Aβ deposition in the brain begins after about 4 months and spatial memory impairment occurs after 6 months, Jun and colleagues used groups of both young (3–5 months) and old (7–13 months) APP-KI mice for their experiments. The authors went on to test whether spatial memory was impaired in young and old APP-KI mice versus age-matched wild-type (WT) mice. To test the ability of mice to distinguish two spatially distinct environments, Jun and colleagues set up a context discrimination task. Like young WT mice, young APP-KI mice exhibited normal context discrimination with intact spatial memory, whereas old APP-KI mice showed impaired discrimination, indicating that young APP-KI mice with Aβ plaques do not show deficits in spatial memory despite the presence of Aβ plaques in the cortex. Given that Aβ plaques are not exclusively expressed in CA1 and the MEC, it is unclear whether inputs are disturbed in the first place or that Aβ plaques play a primarily direct role in CA1/MEC spatial dysfunction. The generation of conditional region-specific expression of APP-KI mice might solve this problem of ubiquitous Aβ expression in the whole brain.

Next, Jun and colleagues recorded hippocampal place cells and entorhinal grid cells in a 1 m × 1 m open arena. In young APP-KI mice, CA1 neurons showed intact spatial tuning despite Aβ plaques. In contrast, MEC neurons in young APP-KI mice displayed diminished spatial tuning with a decreased percentage of grid cells. In old APP-KI mice, CA1 neurons showed less prominent firing peaks and decreased spatial tuning with the proportion of classified CA1 place cells moderately decreased as well. In contrast, MEC neurons exhibited disrupted periodic spatial tuning with a dramatic loss of grid cells in the old APP-KI mice. While CA1 spatial tuning was only mildly affected, the MEC grid cells were severely disrupted in old APP-KI mice [10], suggesting that deterioration of MEC neurons might occur at the early onset of AD.

Slow or fast gamma rhythms facilitate memory retrieval or encoding, respectively [11]. Previous work has shown disrupted rhythmic coordination of place cell spiking by slow but not fast gamma oscillations in 3xTg mice [8]. By contrast, Jun and colleagues demonstrated that fast rather than slow gamma oscillations are impaired in the hippocampal CA1 of old APP-KI mice. What can these discrepancies tell us about the temporal coupling of the gamma oscillations in APP-KI mice? This might suggest that MEC neurons in APP-KI mice and CA3 neurons in 3xTg mice selectively degenerate, in agreement with the previous report of fast gamma-mediated MEC-to-CA1 signal transfer and slow gamma-coupled CA3-to-CA1 signal routing [11]. Furthermore, both theta-fast gamma coupling and spike-fast gamma cross-frequency coupling were markedly reduced in the MEC of old APP-KI mice despite no differences in the normalized power and frequency of occurrence between WT and APP-KI mice. The authors also confirmed the diminished coherence of fast gamma oscillations between the MEC and CA1 in old APP-KI mice [10].

At the ensemble level, the firing pattern and/or firing rate of place cells and grid cells shifts when key features in the surrounding environment change either slightly or prominently, a phenomenon called rate remapping or global remapping, respectively. Global remapping shows two shifted firing patterns in two distinct environments while in rate remapping the original firing fields remain stable with only a modulated firing rate in two slightly changed environments. In a final series of experiments, Jun and colleagues intended to link the remapping of place cells and grid cells with impaired spatial memory. In young APP-KI mice, CA1 neurons showed intact spatial remapping, whereas the remapping of MEC spatial neurons was mildly disrupted. However, both hippocampal CA1 neurons and MEC neurons exhibited impaired global remapping in old APP-KI mice. Given only one successfully identified grid cell, future studies should collect a certain number of grid cells from old APP-KI mice by measuring at a wider span of progression time points from 6 to 12 months to warrant their critical remapping finding for MEC grid cells (Fig. 1).

Fig. 1.

Deterioration of spatial firing patterns in the hippocampal-entorhinal space memory circuit in an APP-KI AD mouse model. Using a novel APP^NL-G-F-KI mouse model, Jun and colleagues demonstrated that impairment of spatial firing patterns of MEC grid cells emerges earlier than that of CA1 place cells. More specifically, place cells are hardly influenced while both spatial tuning and remapping of grid cells mildly disrupted in young APP-KI mice. In old APP-KI mice, spatial remapping of place cells is severely disrupted with only moderately deteriorated spatial tuning, while both firing patterns and remapping of grid cells are diminished. Furthermore, fast gamma oscillations are impaired coherently between CA1 and MEC, leaving slow gamma oscillations in CA1 almost unaffected.

The findings of Jun and colleagues raise several important questions. Most problematically, why do CA1 place cells still display spatial tuning while MEC neurons are impaired in the old APP-KI mice? How are place cells still intact when grid cells in young APP-KI mice exhibit disrupted spatial firing? These results contradict the current grid-place transformation model, in which hippocampal place cells are generated from the linearly summated input of entorhinal grid cells. Given the unaffected spatial behavior with disrupted spatial tuning by grid cells in young APP-KI mice, it would be intriguing to determine whether any other network compensation mechanism might exist, for instance, extra-hippocampal spatial cells discovered recently in the somatosensory cortex and other cortical structures that are interconnected directly or indirectly with hippocampal CA1 bypassing MEC [12–14] to account for the grid cell loss in the MEC. A likely scenario is that the impairment of place cell remapping in APP-KI mice is a reflection of abnormal sensory input. If so, does maximizing the difference between environments perhaps rescue, to some extent, the impairment of spatial memory of context discrimination? With respect to the comparison of spatial firing between APP-KI mice and control mice, a simple argument would be that individual differences and sampling biases are involved. Thus, calcium imaging using a miniaturized microscope to capture the activity of the same group of MEC or CA1 neurons during the progressive formation of Aβ plaques in a mouse model of AD might provide more dynamic insights.

To what extent Jun and colleagues’ findings translate to AD patients remains to be addressed. It is worth noting that neither the artificial combination of three simultaneous mutations nor the homozygous mutations induced in APP-KI mice exist in familial AD patients. However, the authors’ results, together with previous reports [2–9], have made moderate contributions to understanding the complex circuit mechanism in AD. Although these collective findings [2–10] might only be the tip of the iceberg for decoding the brain network mechanism of AD, more research on dissecting the impaired spatial memory circuit could raise the potential possibility of “circuit therapy” aiming at restoring disrupted spatial memory and navigation for AD patients in the future.

Conflict of interest

The authors claim that there are no conflicts of interest.