Hierarchies of cross-frequency phase synchronization during human visual working memory maintenance Felix Siebenhühner, Matias Palva, Satu Palva Neuroscience Center, University of Helsinki References Introduction Materials and Methods CF phase synchrony during VWM retention Cross-frequency networks and subsystem interactions MEAN LOAD low-α to β, ɶ all ratios, T- high-α to β 1:2, T+  Visual Working Memory (VWM) sustains information online transiently.  Oscillations and synchronization have been suggested to coordinate the anatomically distributed processing during VWM maintenance.  Inter-areal 1:1-phase synchrony in α- (8-15 Hz), β- (15-30 Hz) and ɶ- (30–90 Hz) frequency bands characterizes the WVM retention period [1].  Synchronization in lower frequency bands may coordinate attentional and executive functions, and in higher bands contribute to the representation of sensory information in VWM [1-5].  how are these distinct networks coordinated and integrated? Hypothesis: Cross-frequency (CF) phase synchrony underlies integration and coordination of spatially and spectrally distributed networks of coherent oscillations [4-6] Aim: To investigate whether cross-frequency n:m-phase synchrony is prominent in VWM retention period and behaviourally relevant n:m-phase synchrony (here 1:3) 1:1 phase synchrony f1 f1 f2 f1  Delayed matching-to-sample task used in [1]  Object loads varied from 1 to 6, ~300 trials / load  12 healthy right-handed subjects (29 ± 6 y, 7 female)  Concurrent MEG (306-ch Elekta Neuromag ) and 60-channel EEG recordings  anatomical MRIs for source localization Task and Stimuli Behavioural Performance Data analysis  Maxfilter software (Elekta) used for noise suppression and co-localization of sensor level data  Freesurfer [7] software used for cortical surface reconstruction from individual MRIs.  Source models, forward and inverse operators obtained with MNE software [8] after filtering MEG data with Morlet wavelets into 25 bands from 3 to 90 Hz.  Source time series were then collapsed into optimized cortical parcellations [9] of 400 patches  1:m CF phase synchrony was computed as phase locking value between all patches and among all frequencies for the ratios m ϵ {2,3,4,5,6}  Statistical analyses were performaned by collapsing data into a coarser parcellation of 148 patches and computing the average over and the correlation with memory load  To quantify subsytem interaction, all patches were assigned to 7 subsystems per hemisphere as defined by Yeo et al. [10] for analyses We identified the fraction κ of significant edges of all possible edges for two statistical contrasts: • MEAN: average of memory loads 1-6 (one-sided t-test, p < 0.05, FDR corr., retention period vs. pre-stimulus baseline) • LOAD: correlation across loads 1-6 in the retention period (Spearman rank test, p < 0.05, FDR corr.) We also quantified how consistently lower frequencies f 1 couple to higher frequencies f 2 across 1:m ratios. Harmonic CF synchrony is prominent in the VWM retention period [1] Palva JM et al.: Neuronal synchrony reveals working memory networks and predicts individual memory capacity, Proc Natl Acad Sci 107: 7580 –7585, 2010. [2] Fries P: Neuronal Gamma-Band Synchronization as a Fundamental Process in Cortical ComputationAnnu. Rev. Neurosci. 32:209–24, 2009. [3] Sauseng P et al.: Control mechanisms in working memory: A possible function of EEG theta oscillations, Neurosci Biobehav. Reviews 34:1015-1022, 2010. [4] Palva JM et al.: Phase Synchrony among Neuronal Oscillations in the Human Cortex, Journal of Neuroscience, 25(15):3962–3972, 2005. [5] Palva S et al. : Discovering oscillatory interaction networks with M/EEG: challenges and breakthroughs, Trends in Cognitive Sciences, vol. 16, no. 4, pp. 219-230 , 2012. [6] Canolty RT et al.: The functional role of cross-frequency coupling, Trends in cognitive sciences, vol. 14, no. 11, pp. 506-515 , 2010. [7] Freesurfer: http://surfer.nmr.mgh.harvard.edu/ [8] MNE: www.martinos.org/mne/ [9] Korhonen O et al.: Sparse weightings for collapsing inverse solutions to cortical parcellations optimize M/EEG source reconstruction accuracy, J Neurosci Meth 226: 147–160, 2014. [10] Yeo BT et al.: The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol 106(3):1125-65, 2011. high-α to ɶ 1:3 – 1:6, T+ Acknowledgements This work was supported by: • Academy of Finland, • Helsinki University Research Funds, • Doctoral Program of Brain & Mind of the University of Helsinki. Conclusions Correlation with Accuracy We compared strength in task-pos. and -neg. networks between correct and incorrect trials for all subjects, ratios  network strength increased for CF synchrony from θ (t-test, p <0.01) , high-α (p < 10 -4 ) in task-positive networks  network strength decreased for CF synchrony from ɷ (p < 0.05), low-α (p < 0.001) in task-negative networks  Harmonic CF-phase synchrony among θ, high-α, β and ɶ bands is enhanced during VWM maintenance  observed among visual and fronto-parietal regions  correlated with VWM accuracy and task performance  Strengthening of task-relevant CF connections enables the integration of attentional and representational processing  Harmonic CF-phase synchrony is suppressed among ɷ, low-α, β and ɶ bands  not correlated with task performance  prominent in visual – FP network but also in DMN  Uncoupling of task-irrevelant CF assemblies? Flexible frequency-specific CF synchrony characterizes VWM retention period Correlation with VWM Capacity We tested whether the strength of CF-synchrony across all ratios in task-positive networks is correlated with individual VWM capacity, where: Capacity (i) = HitRate (i) * i , for each load i  strength of task-positive CF networks is correlated (Spearman rank test, p < 10 -4 ) with individual capacity 99% CI ”Task-positiǀe” netǁorks: T+ ɷ ↔ α, β Hα↔β,ɶ θ↔Hα,β,ɶ ”task-negatiǀe” netǁorks: T- Lα↔β,ɶ θ to α, β all ratios, T+ Networks of increased (T+) and decreased (T-) CF phase synchrony. Hubs of the lower frequency are displayed with red borders, high-frequency hubs with blue borders. subsys. interact. str. 0 max subsystem low freq. subsystem low freq. subsystem low freq. subsystem low freq. subsystem low freq. subsystem low freq. subsystem low freq. subsystem low freq. subsystem high freq. subsystem high freq. subsystem high freq. subsystem high freq. subsystem high freq. subsystem high freq. subsystem high freq. subsystem high freq. subsys. interact. str. 0 max se. F1000 Posters: Use Permitt mmons License. F1000 Posters: Use Permitted under Cre r Creative Commons License. F1000 Posters: Use Permitted under Creative Common ermitted under Creative Commons License. F1000 Posters: Use Permitted under Creative Commons Lic osters: Use Permitted under Creative Commons License. F1000 Posters: Use Permitted under Creative Posters: Use Permitted under Creative Commons License. F1000 Posters: Use Permitted un Permitted under Creative Commons License. F1000 Posters: Use er Creative Commons License. F1000