Validity of HINT-8 in evaluating health-related quality of life among cervical cancer patients.

Memory formation and organization are dynamic, ongoing processes that allow the brain to encode, update and integrate multiple experiences over time. In rodents, accumulating evidence shows that recently formed memories are initially supported by selective activation of neuronal ensembles during learning and subsequently stabilized through their reactivation during offline states such as sleep. Similar reactivation patterns have been observed in other species, including birds, bats, nonhuman primates and, more recently, in humans. Although these findings suggest that some of the hippocampal circuit mechanisms are broadly similar across species, differences in ethological demands and behavioral repertoire may shape both the content of reactivated patterns—what types of experiences are prioritized for reactivation and retention—and the organization of temporally distinct experiences into an adaptive memory system. Much of the existing work has focused on stable representations of discrete, isolated and/or recent memories. Thus, it remains unclear how the brain manages to keep track of multiple distinct experiences over time and how it (re)organizes memories originating from different points in time. This question is particularly relevant in primates, whose extended lifespans and cognitive flexibility require memory systems that support not only the preservation of individual experiences but also their integration into coherent, behaviorally adaptive representations. Although waking experience may recruit and support some of these processes, offline states—and specifically sleep—are thought to play a crucial role in this reorganization. In this study, we sought to address whether consistent activity patterns reflecting multiple spatiotemporal events (memoranda) are reactivated during sleep in macaques, and if so, how the patterns of activation and reactivation differ as a function of previous experience. Using high-density linear arrays to record wirelessly from hippocampal and extrahippocampal regions, we tracked experience-dependent changes in ensemble activity that could hold key insights into how the primate brain preserves older memories while incorporating new ones. We used high-density linear arrays to simultaneously record single-unit activity from neuronal ensembles in the hippocampus and connected extrahippocampal regions of two freely moving macaques (range: 31–273 units per session;  = 35 sessions; Fig. and Extended Data Fig. ). Task recordings took place in ‘Treehouse’, a touch-screen-enabled three-dimensional (3D) enclosure, followed by sleep recordings taken overnight in their housing room (Fig. and Extended Data Fig. ). ‘Day 1’ Treehouse sessions consisted of two types of repeated Z-shaped trajectories: a new item-context sequence with four items spanning the four touch screens in one corner and an ‘old’ sequence presented in the opposite corner that had been learned in that corner, on average, 6 weeks earlier with no intervening exposure (Fig. , and Extended Data Fig. ). This allowed simultaneous recording during the performance of two sequences that differed in their previous exposure but with the corner location of the old or new sequence types counterbalanced across sets (Fig. and Extended Data Fig. ). Individual learning results in this task have been reported previously. Accuracy was higher for old sets than for new sets, indicating long-term memory savings (Fig. ;  < 0.001; permutation test with false discovery rate (FDR) correction). ‘Day 2’ sessions occurred the following day and used the same stimulus sets and assigned locations as day 1, with the new set now termed ‘recent’. Day 2 performance on recent sets improved ( < 0.001; permutation test with FDR correction) while remaining below the level of old sets, suggesting overnight retention of the newly learned sequences and a residual benefit of long-term, remote memory for the old sequences. The recorded neural activity varied with task demands (Extended Data Fig. ). Thus, to determine whether the recorded neural activity could contribute to selective memory for these sequences, we trained a linear support vector machine (SVM) decoder to classify trial identity (old versus new or recent) based on single-unit or ensemble activity (Fig. and Extended Data Fig. ). Decoding based on ensemble activity consistently outperformed decoding that used single-unit activity and also surpassed shuffling controls (Fig. and Extended Data Fig. ). Movement-related variables during the task, including angular velocity and linear acceleration, did not differ between the new and old sequences performed in their respective corners and are therefore unlikely to account for the observed decoding performance (Extended Data Fig. ). These results indicate that the pattern of spiking activity in the recorded ensemble generally varied as a function of trial identity. , Left: 3D rendering of the Treehouse 3D task environment indicating the opposite corners w