Abstracts

Rationale: The hippocampus is a key region affected by temporal lobe epilepsy and a frequent target for ablative and neuromodulatory treatments. Disruption of hippocampal function may contribute to memory impairments common in patients with epilepsy. This region is thought to support episodic memory via a process called pattern separation, in which representations of similar experiences are orthogonalized to reduce interference during memory retrieval. However, there is limited evidence that the hippocampus is necessary for pattern separation in humans. It is also unclear how similar memories are maintained over time after pattern separation has occurred. To address these questions, we temporarily disrupted hippocampal circuits during pattern separation using direct electrical stimulation (DES), while recording local field potentials (LFPs) from the same contacts in a time-resolved, spatially precise manner.

Methods: Adults with epilepsy who have responsive neurostimulation (RNS) implants (10 bilateral hippocampus, 11 outside of medial temporal lobe [non-MTL]) first encoded a series of unique object images. They then completed two blocks of a memory test in which they viewed old items from encoding (targets), objects highly similar to encoded items (lures), and novel items (foils), with the instruction to label each as “old”, “similar”, or “new”, respectively. During one test block (order counterbalanced), we applied 1s of DES at stimulus onset; during the other test block, no DES was applied as a within-subjects control. Afterwards, participants again viewed lures from the test blocks and their similar counterparts from encoding. LFPs were recorded during all non-stimulated periods. We conducted similarity analyses on high-frequency activity patterns from these recordings to determine distinctiveness of similar item representations throughout the task.

Results: Hippocampal DES during the test phase reliably impaired discrimination of similar lures from targets and foils, while DES in the non-MTL group did not impair performance (Figure 1). After the test blocks, neural representations of correctly identified lures were anticorrelated from their similar encoding items 1.5-2s after stimulus onset; this anticorrelation, or below-baseline similarity, is an extreme form of pattern separation called differentiation (Figure 2). Differentiation was asymmetric: the representation of the initially encoded item changed more than that of the lure. The greatest differentiation occurred when the initially encoded item was moderately reactivated during lure discrimination.

Conclusions: Successfully applying DES and neural recording in the same task opens the door to more fine-grained, causal studies of cognition. Using this novel approach, we found the first within-subjects causal evidence in humans that the hippocampus is necessary for discriminating similar memories. We also showed that successful discrimination drives differentiation of hippocampal memory representations. These findings advance our understanding of hippocampal function, with implications for characterizing memory deficits associated with neurological disorders such as epilepsy.

Funding: Swebilius Foundation Grant (to IHQ)