Periodic Reporting for period 2 - GridRepresentations (Neural mechanisms, functional roles and pathophysiological relevance of human grid cell-like representations)
Reporting period: 2022-03-01 to 2023-08-31
Spatial navigation in the real world relies on a number of different strategies including allocentric navigation (i.e. navigation relying on cognitive maps), egocentric landmark-based navigation and path integration. One overarching hypothesis of the ERC project “GridRepresentations” is that these strategies can be mapped onto a limited number of relatively distinct brain systems that contain different types of spatially specific cells. Specifically, I propose that allocentric navigation depends place cells in the hippocampus, landmark-based navigation on specific cell types in retroslenial cortex and basal ganglia, and path integration on grid cells in the entorhinal cortex.
I further hypothesize that the employment of these cell types, the respective brain regions, and the corresponding strategies can be variable and depends on the one hand on the availability of spatial cues (e.g. landmark-based navigation is only possible when landmarks are available), on the other hand on the integrity of the respective brain regions. Neurological diseases which impair the function of these areas may be compensated to a certain degree by enhanced recruitment of other, non-affected brain regions, leading to relative hyperactivity in these compensatory areas and changes in navigational strategies. The entorhinal cortex is among the very first brain regions to be affected by Alzheimer’s disease (AD), and deficits of spatial navigation are among the first symptoms of AD. This suggests specific deficits of grid cell functionality and path integration in early AD. In healthy human participants, grid cells can be only measured non-invasively via specific fMRI signatures of “grid cell-like representations”. In a study published by the project PI in 2015, we could show that grid cell-like representations are indeed impaired in young AD risk carriers, and that overt navigational impairments can be prevented by enhanced recruitment of the hippocampus, leading to changes in navigational strategies (i.e. navigation closer to boundaries in an environment) (Kunz et al., Science 2015). In a follow-up study, we could show that AD risk specifically affected path integration and led to navigational deficits in an experimental condition in which no other spatial cues were available and thus, no compensatory strategies could be used (Bierbrauer et al., Science Advances 2019). In addition, we investigated the neurophysiological basis of grid cell-like representations: Using intracranial EEG recordings in presurgical epilepsy patients, we showed that they relied on oscillations in the theta frequency band (Chen et al., Current Biology 2018). However, whether and how grid cell-like representations relate to activity of single grid cells remains unclear, and different scenarios about this relationship are possible, as we described in a recent review paper (Kunz et al., TICS 2018). In addition, grid cell-like representations have not only been observed during spatial navigation in the physical world but also during “navigation” of conceptual spaces (Constantinescu et al., Science 2016) and during visual exploration of natural scenes (Killian et al., Nature 2012; Staudigl et al., Curr Biol 2018). However, the paradigms in many of these studies are relatively artificial, and the cognitive function of grid cell-like representations in these paradigms remains to be fully understood.
The main goals of the ERC project “GridRepresentations” are thus (1) to understand the neurophysiological basis of grid cell-like representations and their interaction with other spatially specific cells and representations in the human brain; (2) to investigate the cognitive function of these representations during different processes of spatial navigation and beyond; (3) to measure the relationship of impaired grid cell-like representations in early AD with AD biomarkers; and (4) to explore the causal relevance of hippocampal hyperactivity for compensation of entorhinal dysfunctions. Each of these objectives is addressed in one of the project’s work packages.
These objectives are highly relevant for both basic research and clinical application. On the one hand, understanding the relationship between neural activity at different levels of brain organization (from cells via neural assemblies characterized by specific oscillatory patterns to BOLD activity measured via fMRI) has been repeatedly described as one of the largest challenges of neuroscience (e.g. Adolph, TICS 2015). On the other hand, AD only becomes symptomatic at a relatively late stage of the disease characterized by substantial brain damage, and the relative failure of previous treatment approaches may be due to the fact that drugs have been administered too late. The development of sensitive biomarkers and cognitive markers that reflect relatively subtle dysfunction in early stages of the disease is thus pivotal for the future more successful treatment approaches.
I further hypothesize that the employment of these cell types, the respective brain regions, and the corresponding strategies can be variable and depends on the one hand on the availability of spatial cues (e.g. landmark-based navigation is only possible when landmarks are available), on the other hand on the integrity of the respective brain regions. Neurological diseases which impair the function of these areas may be compensated to a certain degree by enhanced recruitment of other, non-affected brain regions, leading to relative hyperactivity in these compensatory areas and changes in navigational strategies. The entorhinal cortex is among the very first brain regions to be affected by Alzheimer’s disease (AD), and deficits of spatial navigation are among the first symptoms of AD. This suggests specific deficits of grid cell functionality and path integration in early AD. In healthy human participants, grid cells can be only measured non-invasively via specific fMRI signatures of “grid cell-like representations”. In a study published by the project PI in 2015, we could show that grid cell-like representations are indeed impaired in young AD risk carriers, and that overt navigational impairments can be prevented by enhanced recruitment of the hippocampus, leading to changes in navigational strategies (i.e. navigation closer to boundaries in an environment) (Kunz et al., Science 2015). In a follow-up study, we could show that AD risk specifically affected path integration and led to navigational deficits in an experimental condition in which no other spatial cues were available and thus, no compensatory strategies could be used (Bierbrauer et al., Science Advances 2019). In addition, we investigated the neurophysiological basis of grid cell-like representations: Using intracranial EEG recordings in presurgical epilepsy patients, we showed that they relied on oscillations in the theta frequency band (Chen et al., Current Biology 2018). However, whether and how grid cell-like representations relate to activity of single grid cells remains unclear, and different scenarios about this relationship are possible, as we described in a recent review paper (Kunz et al., TICS 2018). In addition, grid cell-like representations have not only been observed during spatial navigation in the physical world but also during “navigation” of conceptual spaces (Constantinescu et al., Science 2016) and during visual exploration of natural scenes (Killian et al., Nature 2012; Staudigl et al., Curr Biol 2018). However, the paradigms in many of these studies are relatively artificial, and the cognitive function of grid cell-like representations in these paradigms remains to be fully understood.
The main goals of the ERC project “GridRepresentations” are thus (1) to understand the neurophysiological basis of grid cell-like representations and their interaction with other spatially specific cells and representations in the human brain; (2) to investigate the cognitive function of these representations during different processes of spatial navigation and beyond; (3) to measure the relationship of impaired grid cell-like representations in early AD with AD biomarkers; and (4) to explore the causal relevance of hippocampal hyperactivity for compensation of entorhinal dysfunctions. Each of these objectives is addressed in one of the project’s work packages.
These objectives are highly relevant for both basic research and clinical application. On the one hand, understanding the relationship between neural activity at different levels of brain organization (from cells via neural assemblies characterized by specific oscillatory patterns to BOLD activity measured via fMRI) has been repeatedly described as one of the largest challenges of neuroscience (e.g. Adolph, TICS 2015). On the other hand, AD only becomes symptomatic at a relatively late stage of the disease characterized by substantial brain damage, and the relative failure of previous treatment approaches may be due to the fact that drugs have been administered too late. The development of sensitive biomarkers and cognitive markers that reflect relatively subtle dysfunction in early stages of the disease is thus pivotal for the future more successful treatment approaches.
In WP1, we re-analyzed data from a large intracranial EEG dataset on spatial learning (for previous publications on this dataset, see Chen et al., Current Biology 2018; Kunz et al., Science Advances 2018; Chen et al., Science Advances 2020) and from an fMRI dataset of the same paradigm (Kunz et al., Science 2015). Since these two datasets employed an identical task, they are particularly well suited to address objective 1, i.e. the relationship of activities across different levels of brain organization. These data also allow addressing objective 2, because they contribute to a better understanding of the specific cognitive functions (here: navigational strategies) that are supported by neural activity in the core areas supporting the distinct spatial navigation strategies. We also started with the analysis of human single unit data that were recorded concurrently with the intracranial EEG data in a subset of patients (n=10 patients with both microelectrodes and macroelectrodes in entorhinal cortex and hippocampus).
In WP2, and in line with our objective #2, which was to unravel the cognitive function of grid cell-like representations and complementary spatially specific representations in other brain regions, we developed novel experimental paradigms of “conceptual navigation” and of visual exploration of the “spaces” of natural scenes. These paradigms capitalize on the employment of deep neural networks, which we recently described as promising models of different “representational formats” in the brain (Heinen et al., Brain Struct Funct 2023). First, we developed a paradigm that requires participants to focus on either perceptual or conceptual dimensions of a stimulus, and to employ the “cognitive map” defined by these dimensions for similarity judgments. We tested whether memory traces of individual stimuli were embedded into these maps, i.e. whether these maps influenced the memorability of the stimuli. This can be considered a prerequisite for a cognitive relevance of these maps. Specifically, we conducted 3 behavioral studies to investigate how processing of perceptual or conceptual features during different memory stages (encoding, consolidation and retrieval) affects memory performance. In study I, participants rated conceptual or perceptual similarities of a series of image pairs drawn from different categories. Subsequently, we presented the same images again, mixed with new images drawn from either the same (“new exemplars”) or new (“new concepts”) categories. Participants indicated their confidence that each image was old or new on a visual analogue scale. In study II, we used the same design but paired perceptual and conceptual ratings during encoding with condition-specific sounds. One sound was played again during post-encoding sleep to reactivate associated memories. In study III, participants made perceptual or conceptual forced-choice similarity judgements before we tested their memory on separate perceptual and conceptual retrieval tasks. In all studies, we found evidence that the distance of stimuli in task-relevant cognitive maps affected memory performance.
We also developed a novel paradigm to study the relevance of grid cell-like representations related to eye movements during visual exploration. Specifically, participants performed a visual and semantic rating task, and we tested their memory in a subsequent recognition task including old images and lures. Again, we found evidence for memory relevance of the “maps” defined by the visual space.
Finally, we conducted a 7T fMRI study in order to investigate the laminar distribution of grid cell-like representations in entorhinal cortex and their interaction with hippocampal subregions and layers. We employed a spatial learning task across 2 days in which different objects were shown at similar locations, allowing us to disentangle object and location representations. Data acquisition is completed, and the data are currently preprocessed for further analysis.
In WP2, and in line with our objective #2, which was to unravel the cognitive function of grid cell-like representations and complementary spatially specific representations in other brain regions, we developed novel experimental paradigms of “conceptual navigation” and of visual exploration of the “spaces” of natural scenes. These paradigms capitalize on the employment of deep neural networks, which we recently described as promising models of different “representational formats” in the brain (Heinen et al., Brain Struct Funct 2023). First, we developed a paradigm that requires participants to focus on either perceptual or conceptual dimensions of a stimulus, and to employ the “cognitive map” defined by these dimensions for similarity judgments. We tested whether memory traces of individual stimuli were embedded into these maps, i.e. whether these maps influenced the memorability of the stimuli. This can be considered a prerequisite for a cognitive relevance of these maps. Specifically, we conducted 3 behavioral studies to investigate how processing of perceptual or conceptual features during different memory stages (encoding, consolidation and retrieval) affects memory performance. In study I, participants rated conceptual or perceptual similarities of a series of image pairs drawn from different categories. Subsequently, we presented the same images again, mixed with new images drawn from either the same (“new exemplars”) or new (“new concepts”) categories. Participants indicated their confidence that each image was old or new on a visual analogue scale. In study II, we used the same design but paired perceptual and conceptual ratings during encoding with condition-specific sounds. One sound was played again during post-encoding sleep to reactivate associated memories. In study III, participants made perceptual or conceptual forced-choice similarity judgements before we tested their memory on separate perceptual and conceptual retrieval tasks. In all studies, we found evidence that the distance of stimuli in task-relevant cognitive maps affected memory performance.
We also developed a novel paradigm to study the relevance of grid cell-like representations related to eye movements during visual exploration. Specifically, participants performed a visual and semantic rating task, and we tested their memory in a subsequent recognition task including old images and lures. Again, we found evidence for memory relevance of the “maps” defined by the visual space.
Finally, we conducted a 7T fMRI study in order to investigate the laminar distribution of grid cell-like representations in entorhinal cortex and their interaction with hippocampal subregions and layers. We employed a spatial learning task across 2 days in which different objects were shown at similar locations, allowing us to disentangle object and location representations. Data acquisition is completed, and the data are currently preprocessed for further analysis.
We investigated the factors that contributed to theta oscillations and to fMRI BOLD activity in the hippocampus, i.e. the region showing putatively compensatory hyperactivity in the presence of entorhinal dysfunction. By leveraging the power of the rich multimodal datasets we already had collected (multiple recording techniques: iEEG and fMRI; multiple populations: healthy young adults, epilepsy patients, individuals with specific APOE genetic phenotypes), with novel analyses approaches, we were able to address the following outstanding questions in the field:
1. What is the relationship between neural activity recorded at the mesoscopic level (local field potentials as recorded by iEEG) and at the macroscopic level (BOLD activity using fMRI); specifically, what is the relationship between theta oscillations and regional changes in neural activity in the human hippocampus, during spatial navigation performance? [methodological investigation]
2. How do navigation strategies spontaneously arise over the course of exposure to an environment in which spatial memory is tested? Do participants employ one single strategy throughout the task, or does this differ depending on the relationship between goal locations and available cues in the environment, neural activity, and individual differences? How does strategy relate to overall performance? [psychological investigation]
3. How does performance, and strategy (as defined above), during navigation change as a function of genetic risk factors and disease? [cognitive health investigation]
The results of these analyses are described in section 1.1 below. To summarize, for controls and iEEG participants, recruiting the hippocampus only supports performance when participants use a distinctly hippocampal strategy - i.e. when they move on straight paths, presumably reflecting an allocentric (map-based) strategy. For APOE4 carriers, recruiting the hippocampus (via increased BOLD activity) is always beneficial for performance, regardless of the specific strategy they are using, presumably because it helps them compensate for EC dysfunction.
We also obtained important results in a novel paradigm of conceptual navigation, which are also described in section 1.1. Results from experiments in three complementary studies provide a promising basis for future investigations of grid cell-like representations in this paradigm, since they demonstrate that task-dependent cognitive maps are embedded into stimulus-specific memory traces and are thus functionally relevant. In order to test this, we conducted a follow-up fMRI study of this paradigm (n=46 participants). Similar to our behavioral studies, we found that task-relevant (but not irrelevant) distances during similarity judgments at encoding predicted subsequent recognition memory. We further showed that these distances were reflected by BOLD activity in several visual and prefrontal regions. These preliminary results were presented as a poster at the Cognitive Neuroscience Society Meeting in San Francisco 2023 and are currently further analyzed for publication. A similar paradigm has been established for recordings in epilepsy patients with intracranial EEG electrodes and with microelectrodes.
Next, we will establish a battery of tasks to conduct laminar recordings at 7T and compare grid cell-like representations and activity in other areas including the hippocampus and retrosplenial cortex across paradigms. Afterwards, we will investigate participants between 50-65 years at 7T as well as via tau- and amyloid PET in order to test whether possible alterations of grid cell-like representations, hippocampal activity and the employment of specific navigational strategies reflect early disease stages.
1. What is the relationship between neural activity recorded at the mesoscopic level (local field potentials as recorded by iEEG) and at the macroscopic level (BOLD activity using fMRI); specifically, what is the relationship between theta oscillations and regional changes in neural activity in the human hippocampus, during spatial navigation performance? [methodological investigation]
2. How do navigation strategies spontaneously arise over the course of exposure to an environment in which spatial memory is tested? Do participants employ one single strategy throughout the task, or does this differ depending on the relationship between goal locations and available cues in the environment, neural activity, and individual differences? How does strategy relate to overall performance? [psychological investigation]
3. How does performance, and strategy (as defined above), during navigation change as a function of genetic risk factors and disease? [cognitive health investigation]
The results of these analyses are described in section 1.1 below. To summarize, for controls and iEEG participants, recruiting the hippocampus only supports performance when participants use a distinctly hippocampal strategy - i.e. when they move on straight paths, presumably reflecting an allocentric (map-based) strategy. For APOE4 carriers, recruiting the hippocampus (via increased BOLD activity) is always beneficial for performance, regardless of the specific strategy they are using, presumably because it helps them compensate for EC dysfunction.
We also obtained important results in a novel paradigm of conceptual navigation, which are also described in section 1.1. Results from experiments in three complementary studies provide a promising basis for future investigations of grid cell-like representations in this paradigm, since they demonstrate that task-dependent cognitive maps are embedded into stimulus-specific memory traces and are thus functionally relevant. In order to test this, we conducted a follow-up fMRI study of this paradigm (n=46 participants). Similar to our behavioral studies, we found that task-relevant (but not irrelevant) distances during similarity judgments at encoding predicted subsequent recognition memory. We further showed that these distances were reflected by BOLD activity in several visual and prefrontal regions. These preliminary results were presented as a poster at the Cognitive Neuroscience Society Meeting in San Francisco 2023 and are currently further analyzed for publication. A similar paradigm has been established for recordings in epilepsy patients with intracranial EEG electrodes and with microelectrodes.
Next, we will establish a battery of tasks to conduct laminar recordings at 7T and compare grid cell-like representations and activity in other areas including the hippocampus and retrosplenial cortex across paradigms. Afterwards, we will investigate participants between 50-65 years at 7T as well as via tau- and amyloid PET in order to test whether possible alterations of grid cell-like representations, hippocampal activity and the employment of specific navigational strategies reflect early disease stages.