Theta-phase locking of single neurons throughout human spatial reminiscence

Spatial reminiscence process

To examine numerous determinants of theta-phase locking within the human medial temporal lobe, we carried out direct neurophysiological recordings in epilepsy sufferers^63,64 throughout a spatial reminiscence process⁵⁵ (Table S1). In this “Treasure Hunt” process, members actively navigated a digital surroundings utilizing a recreation controller and had been requested to encode and bear in mind the places of objects inside the surroundings (Fig. 1; Methods). Briefly, through the encoding interval of every trial, members encountered two or three objects in treasure chests positioned at random places on a digital seaside. They had been then passively moved to an elevated recall place from the place they may oversee the seaside. After a brief distractor interval, members carried out cued recall. During location-cued object recall, they had been cued with the places and requested to talk aloud the names of the objects that they had discovered at these places. During object-cued location recall, they had been offered with the objects and indicated the objects’ remembered places on the seaside. In a full session, members geared toward encoding and remembering a complete of 100 distinctive object–location associations.

A During every navigation–encoding interval of the duty, members freely and actively navigated a digital seaside surroundings utilizing a recreation controller and successively encountered two or three distinctive objects in treasure chests at random places on the seaside. Participants geared toward encoding the objects and their related places to recall them through the retrieval interval. After the navigation–encoding interval, members had been moved to an elevated recall place the place they accomplished a distractor process to stop the members from actively rehearsing the objects and their related places. After the distractor process, members carried out two forms of reminiscence recall. B During location-cued object recall, members had been offered with a location on the seaside and recalled the related object that they had discovered at that location throughout encoding. C Histogram of session-wise reminiscence efficiency throughout object recall (left) and enchancment of object-recall reminiscence efficiency over time, i.e., through the first half of all trials versus the second half of all trials (proper; two-sided paired t-check: t(26) = –3.402, P = 0.002; n = 27 periods). We computed object-recall efficiency in every session because the ratio of the variety of efficiently recalled objects divided by the entire variety of object recollects. For occasion, in session primary, the participant efficiently recalled 9 of 33 objects, leading to an object-recall efficiency of 0.273. D During object-cued location recall, members considered an object and geared toward recalling the related location on the seaside. E Histogram of trial-wise reminiscence efficiency throughout location recall (left) and enchancment of location-recall reminiscence efficiency over time (proper; two-sided paired t-test: t(26) = –2.786, P = 0.010; n = 27 periods). We computed location-recall efficiency in every trial because the drop error (Euclidean distance) between the participant’s response location and the right location of the article, normalized by the doable distribution of drop errors. The nearer to 1, the higher reminiscence efficiency. Red dotted line, probability efficiency. Gray traces, session-wise information; blue line, common. The digital seaside surroundings was created utilizing the Unity 3D graphics engine and the Hand-painted Island pack (obtained from the Unity asset retailer below the Standard Unity Asset Store EULA; https://assetstore.unity.com/packages/3d/environments/fantasy/hand-painted-island-pack-36959#asset_quality). Source information are supplied as a Source Data file.

Eighteen epilepsy sufferers participated within the process and contributed a complete of 27 periods. Participants carried out the duty for 60 ± 2 min (imply ± SEM), and their reminiscence efficiency was much like earlier research with this process^55,56,57 (Fig. 1C, E). Participants confirmed memory-performance enhancements over time for each forms of reminiscence recall, indicating that they efficiently acquired data of the spatial surroundings (object recall: two-sided paired t-test between the primary and second half of all trials, t(26) = −3.402, P = 0.002; location recall: two-sided paired t-test, t(26) = −2.786, P = 0.010; n = 27 periods; Fig. 1C, E). Better efficiency throughout object recall was correlated with higher efficiency throughout location recall (Spearman’s rho = 0.879, P < 0.001, n = 27 periods). This process thus supplied us with the chance to analyze theta-phase locking throughout reminiscence encoding and retrieval, and as a operate of reminiscence efficiency.

Human single neurons part lock to the native theta rhythm

We carried out direct neural recordings with excessive spatiotemporal decision utilizing Behnke-Fried microelectrodes^{63,64,65,66,67} that allowed us to establish the exercise of 1025 neurons in numerous areas of the human medial temporal lobe (Fig. 2A; Fig. S1). We excluded neurons with fewer than 25 spikes (i.e., motion potentials^1,39) in whole or no spikes in additional than 80% of segments throughout any trial situation, leading to a complete variety of 666 neurons for all analyses.

A Microelectrode recordings. Left, instance post-implantation MRI. Arrow factors on the tip of the Behnke–Fried electrode. Middle, schematic of the inside finish of a Behnke–Fried electrode. Microelectrodes protrude from the tip of the macroelectrode by 3–5 mm. Top proper, native discipline potential filtered between 300 and 3000 Hz. Horizontal crimson line, threshold for spike detection. Bottom proper, spike waveforms of two items recorded on this microelectrode. B Analysis process. We filtered the native discipline potential in a broad theta-frequency vary (1–10 Hz) and computed the generalized part of the filtered hint. For every spike, we extracted its corresponding theta part. C General theta-phase locking throughout the whole experiment. Spikes pooled throughout all neurons preferentially occurred on the theta trough. Panel (A) is customized from Kunz, L., Staresina, B.P., Reinacher, P.C. et al. Ripple-locked coactivity of stimulus-specific neurons and human associative reminiscence⁹⁶. Nat Neurosci 27, 587–599 (2024). below a CC BY license: https://creativecommons.org/licenses/by/4.0/. Source information are supplied as a Source Data file.

We additionally extracted the low-frequency part of the microelectrode information in a broad theta-frequency vary (1–10 Hz) utilizing a generalized part method⁵⁹ to characterize the timing relationship between the neurons’ spiking exercise and the theta rhythm. We opted for this method with a broad filter band of 1–10 Hz as our visible inspection of the information recognized many durations during which the native discipline potential confirmed fluctuations and oscillatory exercise with quickly altering frequencies, much like earlier observations of non-stationary theta oscillations in bats and people^16,27,28,68. Our preliminary analyses with the filter-Hilbert technique together with narrow-band filtering usually failed to suit these fluctuations neatly, and we thus determined to deal with them as a coherent entity and opted for the generalized part method as a substitute. This led to match between the unique and the filtered sign within the theta-frequency vary (e.g., Fig. 2B; Fig. S2).

In addition to this higher match of the uncooked sign, our alternative of utilizing a broad 1–10 Hz frequency vary was motivated by a number of earlier research utilizing intracranial neural recordings in people. These research confirmed that behavior-associated theta oscillations in people happen throughout increased and decrease theta frequencies, which prolong past the normal 4–8 Hz theta band and are also known as excessive theta (6–10 Hz) and low theta (1–5 Hz)^{16,29,31,32,69}. For instance, earlier work discovered that elevated low-theta energy at 1–3 Hz is expounded to profitable reminiscence encoding^31,57; that 3-Hz oscillations are sometimes current throughout digital spatial navigation^16,32; and that high- and low-theta energy will increase earlier than the onset of digital motion⁶⁹. Hence, to not exclude any elements of excessive theta and low theta, we opted for the 1–10 Hz frequency vary and seek advice from fluctuations inside this vary as theta all through the manuscript. We acknowledge although that human theta has been outlined in numerous methods (e.g., as 4–7.5 Hz⁷⁰ or as 4–8 Hz⁶) and that the 1–10 Hz band contains elements of the normal delta and alpha bands⁷⁰.

To receive part estimates inside this broad frequency band, we used the just lately developed generalized part method⁵⁹. In temporary, the generalized part method entails filtering in a broad band (right here, 1–10 Hz), computing the analytic sign utilizing the Hilbert remodel, estimating the part from the analytic sign, and interpolating durations with high-frequency intrusions the place part development reverses route or part progresses with a frequency <1 Hz (Fig. S3). Due to the 1/f-like form of the ability spectrum with stronger energy in decrease as in comparison with increased frequencies, generalized part estimates are, by design, extra strongly influenced by decrease than increased frequencies. In comparability to the 5–40 Hz band of the unique research utilizing generalized part⁵⁹, our research additionally included frequencies between 1 and 5 Hz to not miss low-theta results. We thus carried out management analyses to look at whether or not our frequency band may add results associated to arousal and whether or not it could be extra liable to low-frequency intrusions (i.e., phase-distorting influences from frequencies beneath the band of curiosity). Analyzing the frequency distributions through the distractor interval, we didn’t discover proof for the concept that decrease arousal was associated to the next prevalence of 1–5 Hz exercise (Fig. S4A). Furthermore, management analyses and simulations didn’t point out that potential low-frequency intrusions biased the 1–10 Hz part estimates in a related means (Figs. S2, S3, S5). Taken collectively, computing generalized part inside the 1–10 Hz frequency vary appeared as a sound method to analyze theta-phase locking in our information.

In a primary step, we investigated common theta-phase locking of neuronal spikes throughout the whole process, with out distinguishing between durations with or with out clear theta oscillations, as carried out in prior research³⁸. To quantify every neuron’s phase-locking power, we calculated the pairwise part consistency (PPC) for the distribution of theta phases at which the neuron’s spikes occurred⁷¹ (if spike counts had been too excessive to compute PPC, we used a Rayleigh check). To assess statistical significance, we ranked the empirical PPC inside surrogate PPC values that we obtained by circularly shifting the theta phases relative to the motion potentials and recomputing the PPC (alpha stage, 0.05; Methods). We noticed many neurons (single and multi-units) for which the spikes had been strongly locked to a specific part of the native discipline potentials filtered in our broad theta-frequency vary (571 of 666 neurons; 86%; binomial check versus 5% probability, P < 0.001).

For instance, a neuron within the left entorhinal cortex of a participant preferentially spiked on the theta troughs (one-sided surrogate evaluation with round shift of part information, PPC = 0.496, P < 0.001, n = 10,261 spikes; Fig. 2B). Pooling all spikes from all neurons, we discovered that they had been typically clustered across the theta trough (Rayleigh check, z = 2.673 × 10⁴, P < 0.001, n = 7,779,085 spikes; Fig. 2C). We obtained comparable outcomes when investigating theta-phase locking in additional restricted theta-frequency bands (Figs. S6, S7, S8). When differentiating single items into putative interneurons and pyramidal cells^72,73, we discovered the next share of theta-phase locking in putative interneurons in comparison with pyramidal cells (94.1% and 83.8%, respectively; two-sided surrogate evaluation with unit-label shuffling, Δ = 10%, P = 0.029; Fig. S9). These findings replicate earlier observations of robust theta-phase locking within the human medial temporal lobe^38,40 and point out that there’s a tight temporal relationship between single-neuron exercise and native low-frequency exercise within the human medial temporal lobe.

Theta-phase locking is elevated for spikes related to excessive theta energy

Next, we geared toward figuring out whether or not theta-phase locking is modulated by primary properties of the native discipline potential. We thus requested whether or not theta-phase locking was stronger in periods of elevated theta energy. To this finish, we computed the instantaneous theta energy at every spike by squaring the magnitude of the advanced sign obtained by way of the generalized part method. We discovered that theta-phase locking was current each in periods with excessive and in periods with low theta energy utilizing a median cut up (Rayleigh checks: excessive, z = 2.851 × 10⁴, P_corr. < 0.001, n = 3,889,384 spikes; low, z = 3.918 × 10³, P_corr. < 0.001, n = 3,889,701 spikes; Bonferroni corrected for 2 checks). When we straight in contrast theta-phase locking strengths between excessive versus low theta energy, we discovered that theta-phase locking was extra strongly expressed in periods with excessive theta energy (one-sided surrogate evaluation with spike-label shuffling, Δz = 2.459 × 10⁴, P < 0.001; Fig. 3A; Fig. S10), which is according to earlier outcomes³⁸. In designing the surrogate evaluation, we ensured that the precise variety of spikes in each situations was maintained in all surrogate rounds, which is related because the Rayleigh z-value—in contrast to the PPC—will increase with the spike depend (Fig. S11).

A Theta-phase locking throughout the whole experiment as a operate of whether or not the spikes coincided with excessive or low theta energy (median cut up), pooled throughout all items. Theta-phase locking was extra strongly expressed in periods with excessive theta energy (one-sided surrogate evaluation with spike-label shuffling, Δz = 2.459 × 10⁴, P < 0.001). B Average theta-phase locking throughout excessive and low theta energy throughout items, quantified with pairwise part consistency (PPC). Results are proven for baseline (left; total session aside from encoding and recall durations; n = 666 neurons), encoding (center; n = 666), and recall (proper; n = 666). Bars present imply PPC values; dots symbolize PPC values of particular person neurons. Y-axis is expanded between 0 and 0.1 to focus on the vary containing most information factors. We in contrast pairwise part consistency between excessive and low energy utilizing t-checks and ranked the empirical t-values inside distributions of surrogate t-values (plots on the backside; one-sided surrogate evaluation: baseline, t(665) = 13.502, P_corr. < 0.001; encoding, t(665) = 12.393, P_corr. < 0.001; recall, t(665) = 13.305, P_corr. < 0.001; Bonferroni corrected for 3 checks). Source information are supplied as a Source Data file.

We additionally examined whether or not this modulation of theta-phase locking by theta energy different between totally different trial durations together with encoding, recall, and baseline (the place the baseline interval comprised the whole session aside from encoding and recall durations). In every of those three experimental durations, we separated the spikes of every unit into teams of excessive and low theta energy utilizing a median cut up. We discovered that phase-locking was persistently stronger within the high-power than within the low-power situation for baseline, encoding, and recall durations (one-sided surrogate evaluation with condition-label swapping primarily based on PPC values: baseline, t(665) = 13.502, P_corr. < 0.001; encoding, t(665) = 12.393, P_corr. < 0.001; recall, t(665) = 13.305, P_corr. < 0.001; Bonferroni corrected for 3 checks; Fig. 3B; Fig. S10). Overall, our outcomes present that the modulation of theta-phase locking by theta energy is a common phenomenon that’s constant throughout totally different reminiscence states, at the least in our spatial reminiscence process.

Phase locking to the native theta rhythm varies throughout medial temporal lobe areas

We subsequent examined whether or not theta-phase locking differed between medial temporal lobe areas, together with the amygdala, entorhinal cortex, hippocampus, parahippocampal cortex, and temporal pole. We discovered that the share of cells with important part locking different between the totally different mind areas (χ² checks: baseline, χ²(4) = 15.327, P_corr. = 0.012; encoding, χ²(4) = 16.198, P_corr. = 0.008; recall, χ²(4) = 13.729, P_corr. = 0.025; n = 666; Bonferroni corrected for 3 checks; Fig. 4A). Numerically, the share of phase-locking neurons was lowest within the hippocampus and highest within the parahippocampal cortex. To higher perceive the origin of this discovering, we analyzed the items’ PPC values and located that they differed between mind areas as nicely (ANOVA for linear mixed-effects mannequin: mounted impact of area, F(4) = 21.347, P < 0.001; Fig. 4B). Post-hoc pairwise comparisons confirmed that the regional variations had been pushed by elevated theta-phase locking within the parahippocampal cortex (Table S2). As decrease theta-phase locking is equal to the next variability of spiking-related theta phases which will encode extra info, we speculate that this consequence factors at an elevated potential for theta-phase coding within the different, non-parahippocampal medial temporal lobe areas (for an illustration of the ideas of part locking, part coding, and the way they will coexist, see Fig. S12).

A–D Theta-phase locking, variety of spikes, and theta energy in several mind areas throughout baseline (brown), encoding (blue), and recall (orange). Analyses in (A–D) had been carried out with 210 neurons from the amygdala (AMY), 130 from the entorhinal cortex (EC), 125 from the hippocampus (HC), 76 from the parahippocampal cortex (PHC), and 92 from the temporal pole (TP). A Percentages of items with important part locking. Numbers at high point out the entire variety of neurons per area. The share of cells with important part locking different between the totally different mind areas (χ² checks: baseline, χ²(4) = 15.327, P_corr. = 0.012; encoding, χ²(4) = 16.198, P_corr. = 0.008; recall, χ²(4) = 13.729, P_corr. = 0.025; n = 666; Bonferroni corrected for 3 checks). B Pairwise part consistency (PPC) throughout items. Bars present imply PPC values; dots symbolize PPC values of particular person neurons. Y-axis is expanded between 0 and 0.1 to focus on the vary containing most information factors. PPC values differed between mind areas (ANOVA for linear mixed-effects mannequin: mounted impact of area, F(4) = 21.347, P < 0.001). C Number of spikes per area, throughout neurons. Box plots present medians, twenty fifth and seventy fifth percentiles, and outliers as dots. Spike counts didn’t differ between medial temporal lobe areas (Kruskal–Wallis checks: baseline, H(4) = 8.135, P_corr. = 0.260; encoding, H(4) = 4.308, P_corr. = 1; recall, H(4) = 10.161, P_corr. = 0.113; Bonferroni corrected for 3 checks). D Log-transformed energy assigned to the spikes of every neuron. Bars present imply values; dots symbolize values of particular person neurons. Spike-associated theta energy different between medial temporal lobe areas (ANOVA for linear mixed-effects mannequin: mounted impact of area, F(4) = 204.063, P < 0.001). Source information are supplied as a Source Data file.

We geared toward ruling out that this discovering was a side-effect of different properties, comparable to increased spike counts, which might decrease PPC variance and improve the share of phase-locking neurons (Fig. S11), or elevated theta energy, which might improve theta-phase locking (Fig. 3B). We thus carried out management analyses to check whether or not spike counts or theta energy different between medial temporal lobe areas. This confirmed that spike counts didn’t differ between medial temporal lobe areas (Kruskal–Wallis checks: baseline, H(4) = 8.135, P_corr. = 0.260; encoding, H(4) = 4.308, P_corr. = 1; recall, H(4) = 10.161, P_corr. = 0.113; Bonferroni corrected for 3 checks; Fig. 4C). In distinction, spike-associated theta energy different between medial temporal lobe areas (ANOVA for linear mixed-effects mannequin: mounted impact of area, F(4) = 204.063, P < 0.001; Fig. 4D). The route of this impact couldn’t clarify the regional variations in phase-locking power, nevertheless, as we noticed highest theta energy within the hippocampus the place we noticed the bottom variety of phase-locking items (Fig. 4A; Table S3). These management analyses counsel that regional variations in theta-phase locking will not be merely reducible to variations in spike counts or theta energy.

Theta-phase locking is stronger in periods with excessive aperiodic slopes

Local discipline potentials are composed of two elements, aperiodic (non-oscillatory) and periodic (oscillatory) exercise⁶⁰, each of which can replicate and affect several types of neural processing⁷⁴. Both elements may result in adjustments in absolute theta energy^6,60 (Fig. S13) and will thus underlie our remark that increased theta-phase locking emerges in periods of elevated theta energy (Fig. 3B). We thus geared toward understanding how theta-phase locking was influenced by aperiodic, non-oscillatory and by periodic, oscillatory exercise. In temporary, we used Spectral Parameterization Resolved in Time (SPRiNT) to characterize aperiodic exercise⁶¹, and we used Bycycle to establish periodic theta oscillations⁶². These strategies allowed us to carry out two unbiased analyses the place we grouped neuronal spikes (1) as a operate of various properties of aperiodic exercise and (2) based on whether or not they occurred within the presence or absence of clear theta oscillations detected by Bycycle (Figs. 5;  6A).

A We utilized Spectral Parameterization Resolved in Time (SPRiNT⁶¹) to research energy spectra of the native discipline potentials in time steps of 500 ms. For every energy spectrum (1–40 Hz; black), SPRiNT supplied a match to its aperiodic part (blue line) and a remaining match to the unique energy spectrum (dotted crimson line). We grouped time home windows into these with excessive versus low aperiodic slopes (median cut up). We additionally grouped them into time home windows with versus with out clear theta oscillations (crimson and black, respectively) primarily based on whether or not Bycycle detected an oscillation utilizing the options of particular person cycles⁶². B Example energy spectrum for example the aperiodic and remaining match. Black line, energy spectrum; blue line, aperiodic match; dotted crimson line, remaining match; grey space, theta-frequency vary. Source information are supplied as a Source Data file.

A Schematic illustrating the unbiased analyses of theta-phase locking as a operate of aperiodic slope and clear theta oscillations detected by Bycycle. Top: excessive aperiodic slopes consequence from elevated exercise in decrease frequencies and decreased exercise in increased frequencies. These will increase and reduces needn’t consequence from rhythmic processes. Bottom: oscillations are detected if consecutive cycles have comparable amplitude and frequency. B Firing charge as a operate of aperiodic slope and the presence of clear theta oscillations. Each line, one unit (n = 666). Blue traces join median firing charges. Bottom: Empirical t-values of paired t-checks evaluating firing charges between situations (vertical traces), together with distributions of surrogate t-values (two-sided surrogate evaluation: slope, t(665) = −8.337, P < 0.001; oscillations, t(665) = 2.598, P = 0.007). C Theta-phase locking throughout excessive and low aperiodic slopes for baseline (n = 666 neurons), encoding (n = 665), and recall (n = 666). Empirical t-values of paired t-checks had been ranked in surrogate distributions of t-values (backside; one-sided surrogate evaluation: baseline, t(665) = 7.464, P_corr. < 0.001; encoding, t(664) = 2.486, P_corr. = 0.017; recall, t(665) = 3.697, P_corr. < 0.001; Bonferroni corrected). Bars, imply pairwise part consistency (PPC); dots, PPC of particular person neurons. Y-axis is expanded between 0 and 0.1. D Histogram of percentages of time with theta oscillations. E Histogram of mode frequencies of oscillations within the theta-frequency vary. F Probability distribution of frequencies of detected theta oscillations. Shaded space, normal error of the imply throughout microwires. G Theta-phase locking through the presence versus absence of clear theta oscillations, individually for baseline (n = 666 neurons), encoding (n = 666), and recall (n = 666). Bars, imply PPC; dots, PPC of particular person neurons. Y-axis is expanded between 0 and 0.1. Empirical t-values had been ranked in surrogate distributions to evaluate significance (backside; one-sided surrogate evaluation: baseline, t(665) = 9.163, P_corr. < 0.001; encoding, t(665) = 6.490, P_corr. < 0.001; recall, t(665) = 8.790, P_corr. < 0.001; Bonferroni corrected). Source information are supplied as a Source Data file.

We first requested whether or not theta-phase locking different in power as a operate of non-oscillatory, aperiodic exercise, which summarizes neural exercise that doesn’t must come up from any common, rhythmic course of⁶⁰. One can match and parameterize aperiodic exercise by way of two main parameters: the aperiodic slope and aperiodic offset of the ability spectrum⁶⁰. Here, we centered on aperiodic slopes, that are thought to replicate the steadiness between neural excitation and inhibition⁷⁴. Specifically, we examined whether or not theta-phase locking was stronger in periods with increased (i.e., extra damaging) slopes as in comparison with these with decrease slopes. This evaluation was, by design, unbiased of the query whether or not theta-phase locking was stronger in periods of clear theta oscillations (see the following part). Accordingly, durations with excessive versus low aperiodic slopes occurred each in periods with clear theta oscillations and in periods with out them (Fig. S14).

To differentiate between excessive and low aperiodic slopes over time, we utilized SPRiNT that enabled us to parameterize aperiodic exercise in a temporally resolved method⁶¹. Briefly, we estimated energy spectra in brief, 3-s time home windows each 500 ms over the course of every session (Fig. 5). For every energy spectrum, we then calculated the slope of the aperiodic exercise in log-log area (i.e., the ability legislation exponent). In preliminary analyses, we examined whether or not aperiodic slopes differed between the three process durations of encoding, recall, and baseline, and located that they had been comparable in magnitude (Fig. S15A). In common, steeper slopes had been related to decrease frequencies inside the 1–10 Hz theta frequency vary (Fig. S4, B, C). We additionally analyzed whether or not neuronal firing charges had been modulated by aperiodic slopes and located that durations with excessive aperiodic slopes had been related to decreased neuronal firing charges (two-sided surrogate evaluation with condition-label swapping, t(665) = –8.337, P < 0.001; Fig. 6B; Fig. S10), in each putative interneurons and pyramidal cells (Fig. S9). This remark is according to the concept that increased (i.e., extra damaging) aperiodic slopes are associated to elevated inhibition⁷⁴ and thus correlate with a suppression of neuronal exercise.

We then examined the influence of aperiodic slopes on neuronal theta-phase locking. We discovered that theta-phase locking was stronger for neuronal spikes related to excessive aperiodic slopes as in comparison with these related to low aperiodic slopes (median cut up). This impact was constant throughout baseline, encoding, and recall durations (one-sided surrogate evaluation with condition-label swapping: baseline, t(665) = 7.464, P_corr. < 0.001; encoding, t(664) = 2.486, P_corr. = 0.017; recall, t(665) = 3.697, P_corr. < 0.001; Bonferroni corrected for 3 checks; Fig. 6C; Fig. S10). Different firing charges throughout excessive versus low aperiodic slopes (see above) couldn’t clarify this consequence because the pairwise part consistency is unbiased of spike depend. This discovering signifies that the aperiodic slope of the native discipline potential is a significant determinant of the power of theta-phase locking.

Phase locking is stronger in periods with clear theta oscillations

So far, we used all information to analyze theta-phase locking with out excluding durations that didn’t present clear theta oscillations that had been unambiguously detected by Bycycle. This method was pushed by earlier research³⁸ and since we wished to method doable determinants of theta-phase locking as broadly as doable. We subsequent requested how the presence of theta oscillations within the native discipline potentials modulated theta-phase locking and hypothesized that theta-phase locking can be stronger in periods with clear theta oscillations as in comparison with durations with out them.

To establish theta oscillations, we remoted particular person cycles within the 1–10 Hz filtered sign and used Bycycle to categorize every as being a part of an oscillation or not (Fig. 5A). We observe that this sensible separation of the native discipline potential into durations with and with out clear theta oscillations is non-trivial and never binary, because it displays a trade-off between specificity and sensitivity decided by the parameters of the chosen oscillation detection algorithm. Here, we carried out this distinction to allow a relative comparability of theta-phase locking between durations with stronger (suprathreshold) versus weaker (subthreshold) theta oscillations. This doesn’t exclude the presence of weaker theta oscillations in periods with out clear theta oscillations, that are presumably answerable for residual theta-phase locking in these time durations.

In a preliminary step, we characterised the presence of theta oscillations in our recordings. For all 502 microwires, we computed the share of time home windows with theta oscillations and located that theta oscillations had been current about 38% of the time (Fig. 6D; Fig. S15B). To perceive whether or not theta oscillations occurred at a most popular theta frequency, we computed the mode frequency of all detected oscillations throughout wires (545,470 oscillations in whole) and noticed that they occurred most frequently at 3–5 Hz (Fig. 6E, F). Comparable frequency distributions of theta oscillations have been noticed in prior research during which members carried out comparable digital navigation duties^16,33,40. Performing this evaluation individually for microwires from totally different mind areas revealed distinct region-specific distributions of theta frequencies (Fig. S16). We moreover noticed that firing charges throughout clear theta oscillations had been increased than throughout time durations with out them (two-sided surrogate evaluation with condition-label swapping, t(665) = 2.598, P = 0.007; Fig. 6B; Fig. S10), which resembles earlier observations in people³⁸ and will relate to the discovering of elevated firing charges in periods of elevated theta energy within the membrane potential of CA1 pyramidal neurons in mice⁷⁵.

We then analyzed how the presence of clear theta oscillations modulated part locking throughout encoding, retrieval, and baseline durations. During all three durations, we discovered that part locking was stronger for neuronal spikes occurring throughout clear theta oscillations as in comparison with these outdoors clear theta oscillations (one-sided surrogate evaluation with condition-label swapping: baseline, t(665) = 9.163, P_corr. < 0.001; encoding, t(665) = 6.490, P_corr. < 0.001; recall, t(665) = 8.790, P_corr. < 0.001; Bonferroni corrected for 3 checks; Fig. 6G; Fig. S10). We obtained comparable outcomes when utilizing a 3–10 Hz frequency vary for estimating generalized part (Fig. S6). Residual theta-phase locking in periods with out clear oscillations means that some oscillatory exercise might have been missed by the oscillation-detection algorithm. In help of this interpretation, management simulations confirmed that intrinsic theta resonance of particular person neurons doesn’t result in theta-phase locking if the native discipline potential solely consists of pink noise and doesn’t include any oscillatory exercise (Fig. S17). Taken collectively, human theta-phase locking happens extra strongly throughout clear theta oscillations, however additionally it is current throughout time durations the place it’s harder to detect clear oscillations.

Human single neurons theta-phase lock throughout reminiscence encoding and retrieval

We had been curious whether or not neuronal theta-phase locking different as a operate of reminiscence state and efficiency. We thus investigated theta-phase locking individually for reminiscence encoding and retrieval. Across all 666 neurons, we discovered 306 neurons that exhibited theta-phase locking throughout encoding and 444 neurons with theta-phase locking throughout retrieval (binomial checks versus 5% probability, each P_corr. < 0.001, Bonferroni corrected for 2 checks). 273 items exhibited theta-phase locking throughout each encoding and retrieval (binomial check, P < 0.001; Fig. 7A). When solely contemplating encoding and retrieval segments with profitable reminiscence efficiency, we discovered 163 neurons that exhibited part locking throughout each encoding and retrieval, and 190 such neurons when solely utilizing segments with unsuccessful efficiency (binomial checks, each P_corr. < 0.001, Bonferroni corrected). These outcomes present that theta-phase locking is prevalent throughout each encoding and retrieval, and that the variety of phase-locking neurons is analogous in periods of profitable and unsuccessful reminiscence efficiency.

A Examples of neuronal theta-phase locking throughout encoding and retrieval. Polar histograms present the theta phases of the items’ spikes throughout encoding (left) and retrieval (proper). Red traces, imply part angles. 1–10 Hz bandpass-filtered spike-triggered native discipline potentials are proven for encoding (left) and retrieval (proper). Spike waveforms are proven as density plots. B Theta-phase locking throughout profitable (inexperienced) versus unsuccessful (crimson) encoding (n = 666 neurons). Bars present imply pairwise part consistency (PPC) values; dots symbolize PPC values of particular person neurons. Y-axis is expanded between 0 and 0.1 to focus on the vary containing most information factors. We in contrast PPC values between profitable and unsuccessful encoding utilizing a t-test and ranked the empirical t-value inside surrogate t-values (backside; one-sided surrogate evaluation: t(665) = −1.030, P_corr. = 1, Bonferroni corrected for 2 checks). C Summary plots throughout all items exhibiting their most popular theta phases and imply resultant vector lengths throughout profitable (left) and unsuccessful (proper) encoding. Black traces, imply part throughout items. D Spike-field coherence for profitable and unsuccessful encoding. Shaded areas, imply ± normal error of the imply throughout items. E Theta-phase locking throughout profitable (inexperienced) versus unsuccessful (crimson) retrieval (n = 666 neurons). Bars present imply PPC values; dots symbolize PPC values of particular person neurons. Y-axis is expanded between 0 and 0.1 to focus on the vary containing most information factors. We in contrast PPC values between profitable and unsuccessful retrieval utilizing a t-check and ranked the empirical t-value inside surrogate t-values (backside; one-sided surrogate evaluation: t(665) = 0.576, P_corr. = 1, Bonferroni corrected for 2 checks). P-value is Bonferroni corrected for testing each encoding and retrieval. F Summary plots throughout all items exhibiting their most popular theta phases and imply resultant vector lengths throughout profitable (left) and unsuccessful (proper) retrieval. Black line, imply part throughout items. G Spike-field coherence for profitable and unsuccessful retrieval. Shaded areas, imply ± normal error of the imply throughout items. SFC spike-field coherence. Source information are supplied as a Source Data file.

We subsequent examined whether or not the power of theta-phase locking was increased throughout segments with profitable versus unsuccessful reminiscence efficiency, individually for encoding and retrieval. We didn’t discover stronger theta-phase locking (i.e., increased PPC) throughout profitable encoding, indicating that the neurons’ desire towards a specific theta part throughout encoding was unbiased of whether or not encoding was profitable or not (one-sided surrogate evaluation with condition-label swapping, t(665) = –1.030, P_corr. = 1, Bonferroni corrected for 2 checks; Fig. 7B, C; Figs. S7, S8, S10). To corroborate this remark, we calculated the spike-field coherence in a frequency-resolved means between 1 and 100 Hz³⁹. We didn’t discover any frequency with stronger part locking throughout profitable versus unsuccessful reminiscence encoding (all P_corr. = 1, Bonferroni corrected for 100 frequencies and two checks; Fig. 7D). This impact was comparable throughout totally different medial temporal lobe areas (Fig. S18).

When investigating recall durations, we equally discovered that neuronal theta-phase locking didn’t differ between segments with profitable versus unsuccessful reminiscence efficiency (one-sided surrogate evaluation with condition-label swapping, t(665) = 0.576, P_corr. = 0.562, Bonferroni corrected for 2 checks; Fig. 7E, F; Figs. S7, S8, S10). A extra common spike-field coherence evaluation, as described above, didn’t reveal any frequencies between 1–100 Hz with stronger part locking throughout profitable versus unsuccessful retrieval (all P_corr. = 1, Bonferroni corrected; Fig. 7G). In distinction to earlier research with recognition and verbal free recall reminiscence duties that noticed stronger part locking throughout profitable reminiscence encoding^39,54, these outcomes counsel that the power of theta-phase locking throughout reminiscence encoding and retrieval didn’t differ as a operate of reminiscence success in our spatial reminiscence process.

We moreover examined whether or not the neurons’ most popular theta phases modified between profitable and unsuccessful reminiscence. This evaluation confirmed no important variations in most popular theta phases between segments with profitable versus unsuccessful reminiscence efficiency (surrogate evaluation with spike-label shuffling; encoding: F(665) = 0.652, P_corr. = 0.744; retrieval: F(665) = 1.352, P_corr. = 0.167, Bonferroni corrected for 2 checks; Fig. 7C, F; Fig. S10).

To present a extra complete evaluation, we then requested whether or not the connection between part locking and reminiscence efficiency could be modulated by theta energy, aperiodic exercise, or the presence of theta oscillations. We additionally distinguished between items that responded with a firing-rate improve throughout encoding (object-responsive items; Fig. S19) and items that didn’t present such object responsiveness. We individually examined items with and with out important part locking, multi-units and single items, and unit exercise throughout object recall or location recall. In all these instances, we discovered that neuronal theta-phase locking was comparable between profitable and unsuccessful segments, each for encoding and retrieval (one-sided surrogate evaluation with condition-label swapping; all P_corr. ≥ 0.073, Bonferroni corrected for performing this evaluation for encoding and retrieval and for 2 information situations in every case; Fig. 8A, B). Additional management analyses confirmed that time-frequency resolved energy, which might modulate theta-phase locking results, didn’t differ as a operate of reminiscence efficiency throughout encoding and retrieval (Fig. S20). In abstract, our findings counsel that theta-phase locking power and the popular theta phases of single neurons throughout encoding and recall didn’t rely on whether or not reminiscence retrieval was profitable or not.

A Theta-phase locking throughout profitable (inexperienced) versus unsuccessful (crimson) encoding for various information subsets. No important variations had been noticed. Recall kind refers to object-cued recall and location-cued recall, the place the encoding course of is identical, however the recall processes differ. See Fig. 7B for comparability with the total dataset outcomes. Bars present imply PPC values; dots symbolize PPC values of particular person neurons. Y-axis is expanded between 0 and 0.1 to focus on the vary containing most information factors. The variety of neurons per situation is indicated on the high. We in contrast PPC values between profitable and unsuccessful encoding utilizing a t-test and ranked the empirical t-value inside surrogate t-values (backside; one-sided surrogate evaluation, P-values are Bonferroni corrected for performing this evaluation for encoding and retrieval and for 2 information situations in every case). B Theta-phase locking throughout profitable (inexperienced) versus unsuccessful (crimson) retrieval for various information subsets. No important variations had been noticed. Bars present imply PPC values; dots symbolize PPC values of particular person neurons. Y-axis is expanded between 0 and 0.1 to focus on the vary containing most information factors. See Fig. 7E for comparability with the total dataset outcomes. The variety of neurons per situation is indicated on the high. We in contrast PPC values between profitable and unsuccessful retrieval utilizing a t-check and ranked the empirical t-value inside surrogate t-values (backside; one-sided surrogate evaluation, P-values are Bonferroni corrected for performing this evaluation for 4 checks). Source information are supplied as a Source Data file.

Some neurons shift their most popular theta phases between encoding and retrieval

We thought of that neurons may exhibit theta-phase locking to totally different theta phases throughout encoding versus retrieval (Fig. 9). Such theta-phase shifts have been predicted by the Separate Phases of Encoding And Retrieval (SPEAR) mannequin and will assist keep away from interference between encoding and retrieval processes^52,53,54,76. We thus examined for neurons that exhibited a major part distinction between encoding and retrieval (Fig. 9A) and located proof for such part shifts in some neurons of our dataset (62 of 666 neurons; 9%; binomial check versus 5% probability, P < 0.001; Fig. 9B, C, F; Fig. S21A). When solely contemplating neurons that confirmed important part locking throughout each encoding and retrieval, we discovered the same proportion of neurons with important part shifts between encoding and retrieval (26 of 273; 10%; binomial check, P = 0.001; Fig. 9C, F).

A Hypothesis that neurons shift their most popular theta phases between encoding and retrieval⁵². The illustration exhibits a hypothetical shift from an earlier part throughout encoding towards a later part throughout retrieval. B Example items with considerably shifted theta phases between encoding and retrieval. Polar histograms present theta-phase distributions of all spikes throughout encoding (blue) and retrieval (orange). Theta-filtered spike-triggered averages are proven subsequent to histograms. Left, spike-waveforms as density plots. Right, theta-phase shifts had been thought of important if the empirical F-value (vertical traces) of a Watson–Williams check exceeded the ninety fifth percentile of surrogate F-values (histograms). Units 1–3 considerably phase-locked throughout encoding and retrieval, and unit 4 didn’t. All examples present a shift from a later part throughout encoding towards an earlier part throughout retrieval. C–E Polar histograms present angular variations between theta phases throughout encoding versus retrieval for neurons with important theta-phase shifts between encoding and retrieval. Results are proven for all items and for items with important theta-phase locking. A constructive angular distinction corresponds to a shift from an earlier part throughout encoding to a later part throughout retrieval. Shifts occurred in each instructions. See Fig. S21 for plots with the precise encoding and retrieval phases. C Results for all segments (n = 62 important items from the pool of all items; n = 26 important items from the pool of considerably phase-locking items). D Results for profitable segments (n = 54 and n = 14). E Results for unsuccessful segments (n = 43 and n = 12). An angular distinction of 0° corresponds to similar theta phases throughout encoding and retrieval. F Percentages of items with important theta-phase shifts for all (clear shade) and considerably part locking items (opaque shade). Percentages are in comparison with 5% probability stage utilizing a one-sided binomial check. P-values are Bonferroni corrected for testing profitable and unsuccessful reminiscence efficiency. Gray histogram on the appropriate, empirically estimated probability stage utilizing surrogate information, confirming the a priori chosen 5% probability stage. Source information are supplied as a Source Data file.

We then examined whether or not theta-phase shifts trusted reminiscence efficiency. When solely contemplating information from profitable encoding and retrieval segments, we noticed a major variety of neurons with important part shifts between encoding and retrieval (54 phase-shifting neurons of 666 neurons; 8%; binomial check, P_corr. < 0.001, Bonferroni corrected for performing the check for profitable and unsuccessful reminiscence efficiency; Fig. 9D, F; Fig. S21B). We noticed the same pattern when solely contemplating items with important part locking throughout each encoding and retrieval (14 phase-shifting neurons of 163 neurons; 9%; binomial check, P_corr. = 0.070, Bonferroni corrected; Fig. 9D, F). When solely contemplating information from unsuccessful segments, the variety of considerably shifting items was neither important for all items (43 of 666; 6%, binomial check, P_corr. = 0.110, Bonferroni corrected; Fig. 9E, F; Fig. S21C) nor for items with important part locking throughout encoding and retrieval (12 of 190; 6%; binomial check, P_corr. = 0.488, Bonferroni corrected; Fig. 9E, F). These outcomes counsel that part shifts had been barely extra prevalent throughout profitable as in comparison with unsuccessful reminiscence, although the direct comparisons weren’t important (one-sided surrogate evaluation with condition-label shuffling: all items, Δ = 2%, P = 0.136; phase-locking items, Δ = 2%, P = 0.276).

As we observed that the shifts different in magnitude and route, we subsequent quantified the part variations of phase-shifting items between encoding and retrieval. We discovered that absolutely the part distinction was 85° ± 47° on common (imply ± round SD). Among the neurons with important part locking throughout each encoding and retrieval, absolutely the part distinction between encoding and retrieval was 36° ± 22° (imply ± SD; Fig. 9C). These results had been comparable for profitable and unsuccessful reminiscence efficiency (profitable, all neurons: 101° ± 45°; profitable, phase-locking neurons: 42° ± 18°; unsuccessful, all: 109° ± 45°; unsuccessful, phase-locking neurons: 47° ± 32°; Fig. 9D, E). These outcomes present that the extent of part shifts between encoding and retrieval in our information had been decrease than the initially proposed shifts between oscillatory peaks and troughs⁵², although the noticed shifts should still be sufficiently massive to assist separate encoding and retrieval processes.

We lastly requested whether or not theta-phase shifts different as a operate of different properties of the native discipline potential. We thus estimated the variety of items with important part shifts individually for spikes related to excessive and low theta energy, excessive and low aperiodic slope, throughout versus outdoors clear theta oscillations, and for these occurring at increased versus decrease theta frequencies (Fig. S21D–G). We discovered a major variety of phase-shifting items for top theta energy (64 of 666; 10%; binomial check, P_corr. < 0.001, Bonferroni corrected for 2 checks) however not for low theta energy (37 of 666; 6%; binomial check, P_corr. = 0.557, Bonferroni corrected; Fig. S21D). The variety of phase-shifting neurons was additionally important for spikes related to excessive aperiodic slopes (52 of 666; 7.8%; binomial check, P_corr. = 0.002, Bonferroni corrected) however not for these with low aperiodic slopes (41 of 666; 6.2%; binomial check, P_corr. = 0.206, Bonferroni corrected; Fig. S21E). The variety of phase-shifting neurons was important for spikes within the presence of clear theta oscillations (47 of 666; 7.1%; binomial check, P_corr. = 0.025, Bonferroni corrected) and of their absence, suggesting the presence of subthreshold theta oscillations adequate to elicit theta-phase shifts (50 of 666; 7.5%; binomial check, P_corr. = 0.007, Bonferroni corrected; Fig. S21F). The variety of phase-shifting neurons was important for spikes occurring at each increased and decrease frequencies inside the 1–10 Hz band (each 56 of 666; 8.4%; binomial checks, P_corr. < 0.001, Bonferroni corrected; Fig. S21G). When performing the evaluation of theta-phase shifts individually for low theta (2–5 Hz) and excessive theta (6–9 Hz) utilizing the filter-Hilbert technique, we once more noticed comparable results that had been barely stronger for low theta (Figs. S7, S8).

These outcomes point out that shifts in the popular theta part are—at the least to some extent—modulated by electrophysiological properties together with theta energy and aperiodic slope. Overall, our findings of serious part shifts between encoding and retrieval present extra, albeit restricted, proof for the SPEAR mannequin^52,54,76,77. As the vast majority of neurons confirmed comparable theta phases between encoding and retrieval, nevertheless, we recommend that each theta-phase shifts and secure theta phases might contribute to encoding–retrieval processes.