Ripple-locked coactivity of stimulus-specific neurons and human associative memory

Abstract

Associative memory enables the encoding and retrieval of relations between different stimuli. To better understand its neural basis, we investigated whether associative memory involves temporally correlated spiking of medial temporal lobe (MTL) neurons that exhibit stimulus-specific tuning. Using single-neuron recordings from patients with epilepsy performing an associative object–location memory task, we identified the object-specific and place-specific neurons that represented the separate elements of each memory. When patients encoded and retrieved particular memories, the relevant object-specific and place-specific neurons activated together during hippocampal ripples. This ripple-locked coactivity of stimulus-specific neurons emerged over time as the patients’ associative learning progressed. Between encoding and retrieval, the ripple-locked timing of coactivity shifted, suggesting flexibility in the interaction between MTL neurons and hippocampal ripples according to behavioral demands. Our results are consistent with a cellular account of associative memory, in which hippocampal ripples coordinate the activity of specialized cellular populations to facilitate links between stimuli.

Main

Associative memory is an essential cognitive function for everyday life that allows us to learn and remember relations between different stimuli¹. Impairments in associative memory caused by aging and memory disorders^2,3 are, thus, a growing problem for society, which makes it important to identify its underlying working principles in the brain. A large body of research has implicated the hippocampus and neighboring medial temporal lobe (MTL) regions in the encoding and retrieval of associative memories^4,5, but its neural foundations remain far from understood. We thus aimed at elucidating possible mechanisms underlying associative memory in the human MTL at the single-cell level. Given prior theories on temporally precise neural binding in perception and memory^6,7,8, we considered that the individual stimuli contributing to particular associative memories are encoded by separate sets of functionally specialized neurons and that these neurons interact transiently when individuals encode and retrieve the memories (Fig. 1a).

We examined this hypothesis on the neural basis of associative memory in the setting of object–location associations, which are particularly critical to behavioral functioning in everyday life by allowing us to know where important items are located in our spatial environments. We specifically investigated whether the encoding and retrieval of such object–location memories is correlated with the simultaneous activation of object cells, which represent specific objects^9,10, and place cells, which code for particular spatial locations¹¹. We predicted that these coactivations would occur in a temporally confined manner during hippocampal high-frequency oscillations, termed ‘ripples’^12,13,14, which are considered important for synchronizing neural activity across brain regions^15,16,17,18. Such ripple-locked coactivity of object and place cells could potentially underlie the encoding and retrieval of associative object–location memories by inducing and (re)activating synaptic connections between the object and place cells that represent the different memory elements¹⁹. In addition, ripple-locked coactivity of stimulus-specific neurons could elicit conjunctive memory representations in downstream neurons that respond only to the unique combination of all memory elements^20,21,22.

Previous studies in both animals and humans discovered that hippocampal ripples are relevant to various cognitive functions^12,13,14. Neural recordings in patients with epilepsy revealed that ripples correlate with memory encoding, retrieval and consolidation^{23,24,25,26,27,28}. Rodent studies demonstrated that ripples are linked to precisely organized multicellular activity in the service of learning, memory and planning^12,13,14. In particular, place cell sequences during hippocampal ripples were found to reflect contiguous navigation paths^29,30,31,32, showing that ripple-locked single-neuron activity directly reflects behavior in a spatiotemporally meaningful way. It has remained unknown, however, whether hippocampal ripples also play a role in interconnecting functionally different types of neurons and whether they could thus help binding different mental contents into associative memories.

To investigate this idea, we conducted single-neuron and intracranial electroencephalographic (EEG) recordings from the MTL of human patients with epilepsy performing an associative object–location memory task in a virtual environment⁹. In line with our hypothesis that human hippocampal ripples support the formation and retrieval of associative memories by defining time windows for the coactivity of stimulus-specific neurons, we show that object-specific and place-specific neurons that represent the separate elements of particular object–location associations activate together at moments close to hippocampal ripples. These findings are consistent with the idea that ripple-locked coactivity of stimulus-specific neurons provides a neural mechanism for the formation and retrieval of associative memories and, more broadly, constitutes a key property of information processing in the human brain.

Results

Human hippocampal ripples and object–location memory

To study the neural mechanisms underlying human associative memory, we recorded single-neuron activity and intracranial EEG from the MTL of patients with epilepsy (Methods, Supplementary Table 1 and Supplementary Fig. 1). During the recordings, participants performed an associative object–location memory task in a virtual environment (Fig. 1b,c). In this task⁹, participants encoded the locations of eight different objects once during an initial encoding period and then performed a series of test trials that included periods for retrieving and re-encoding the object–location associations. Each test trial started with an inter-trial interval (ITI), followed by a cue period in which the participant viewed one of the eight objects that they had encountered during initial encoding. Then, in the retrieval phase, participants navigated to the remembered location of this object and received feedback depending on their response accuracy. After feedback, in the re-encoding phase, the object appeared in its correct location, and participants traveled to this location, allowing them to update their associative memory for this object–location pair. Thirty participants contributed a total of 41 sessions and performed 103 trials per session on average (for detailed information on all statistics in the main text, see Supplementary Table 2). They successfully formed associative memories between the objects and their corresponding locations, as their memory performance increased over the course of the task (paired t-test: t(40) = −4.788, P < 0.001; Fig. 1d and Supplementary Figs. 2 and 3).

We identified human hippocampal ripples during the task by examining local field potentials (LFPs) from bipolar macroelectrode channels, which were located mostly in the anterior hippocampus (Fig. 2a–d and Supplementary Fig. 4). Following previous ripple detection algorithms^24,27,33, we recorded a total of 35,948 ripples across all sessions (Fig. 2e–h). Preceding ripple detection, we conservatively excluded interictal epileptic discharges (IEDs; Supplementary Fig. 5) to help interpret our ripples and ripple-related findings as physiological³⁴. We characterized the identified ripples with regard to various properties and confirmed that they reflected time periods with strongly elevated power at approximately 90 Hz (Fig. 2h and Supplementary Fig. 6), consistent with previous human studies^24,27,33….