Neurophysiological network dynamics of pitch change detection

The detection of pitch changes is crucial to sound localization, music appreciation and speech comprehension, yet the brain network oscillatory dynamics involved remain unclear. We used time-resolved cortical imaging in a pitch change detection task. Tone sequences were presented to both typical listeners and participants affected with congenital amusia, as a model of altered pitch change perception. Our data show that tone sequences entrained slow (2-4 Hz) oscillations in the auditory cortex and inferior frontal gyrus, at the pace of tone presentations. Inter-regional signaling at this slow pace was directed from auditory cortex towards the inferior frontal gyrus and motor cortex. Bursts of faster (15-35Hz) oscillations were also generated in these regions, with directed influence from the motor cortex. These faster components occurred precisely at the expected latencies of each tone in a sequence, yielding a form of local phase-amplitude coupling with slower concurrent activity. The intensity of this coupling peaked dynamically at the moment of anticipated pitch changes. We clarify the mechanistic relevance of these observations in relation to behavior as, by task design, typical listeners outperformed amusic participants. Compared to typical listeners, inter-regional slow signaling toward motor and inferior frontal cortices was depressed in amusia. Also, the auditory cortex of amusic participants over-expressed tonic, fast-slow phase-amplitude coupling, pointing at a possible misalignment between stimulus encoding and internal predictive signaling. Our study provides novel insight into the functional architecture of polyrhythmic brain activity in auditory perception and emphasizes active, network processes involving the motor system in sensory integration.


46
Pitch is a fundamental perceptual feature of sound that is a form-bearing dimension of music and an important cue for understanding speech (Bregmann, 1990). Meaningful pitch changes are perceived by 5 Coupling between slow and fast neural dynamics. 136 Across participants and for both groups, we observed the strongest phase-amplitude coupling (PAC) 137 for the entire sequence in the right auditory cortex (rAud) between the phase of delta-band activity at [2,4] 138 Hz and the amplitude of neurophysiological signals in the beta frequency range at [15, 35] Hz ( Fig. 2A). 139 Time-resolved measures of PAC variations (tPAC) in rAud over the tone sequence are shown in Fig. 2B. 140 The five data points in Fig. 2B report tPAC values during the presentation of each of the five auditory 141 tones. The last two tPAC data points correspond to subsequent time windows during which there was no 142 tone presentation. We found in both groups and across all tested time windows that the strength of phase-143 amplitude coupling was above chance levels (z > 3.4, pcorrected < 0.01). Overall, coupling was stronger in 144 amusics than in typical listeners (F(1)=11.1, p < 0.001), with no effect of response accuracy (F(1)=0.02, p 145 = 0.88) or pitch deviance (F(1)=0.94, p = 0.33; Fig. 2C). There was also a main effect of time (F(6)=6.5, p 146 < 0.001): in both groups, a post-hoc analysis showed that PAC increased after the onset of the tone 147 sequence (p = 0.0006) and decreased after the occurrence of the target tone (over the three subsequent 148 time windows: p = 0.019, p = 0.013, and p < 0.0001, respectively). 149 In the right inferior frontal cortex (rIFG), phase-amplitude coupling was also the strongest between the 150 phase of regional delta activity and the amplitude of beta-band fluctuations (Fig. 2D). Time-resolved tPAC 151 analysis in that region revealed a main effect for groups (F(1)=43.95, p < 0.0001; Fig. 2E): as in rAud, 152 amusics expressed stronger PAC levels than controls (p < 0.001, Fig. 2E & F). We also observed a main 153 effect of deviance level (F(1)=5.8, p = 0.0157) and an interaction between actual deviance and accuracy 154 of pitch change detection (F(1,1)=13.1, p < 0.001). Indeed, in controls, phase-amplitude coupling was 155 stronger in rIFG when target tones were perceived as deviant than when reported as standards (pcorrected Beta bursts are temporally aligned with tone presentations in a sequence. 194 We also derived measures of phase-amplitude stimulus-to-brain coupling in the right auditory cortex. 2H: left panel). There was no significant effect of time (F(6)=1.16, p = 0.32), accuracy (F(1)=2.08, p = 0.14) 199 or pitch deviance (F(1)=0.41, p = 0.53). Overall, neurophysiological delta-to-beta phase-amplitude 200 coupling was stronger than stimulus-to-beta coupling in the tested region (t(119985)=69.45, p < 0.001).

201
For each trial, we also extracted the latency of beta amplitude bursts with respect to the corresponding 202 tone presentation in the sequence. We found in both groups that after the first tone in the sequence was 203 presented, the amplitude of beta bursts was maximal at the expected latency of auditory inputs reaching  Frequency-specific network interactions. 206 We measured the coherence between all pairs of regions of interest. We observed a main effect of the 207 pair of regions of interest (Aud-IFG presented stronger coherence than Mot-Aud and Mot-IFG, F(2) = 208 120.29, p < 0.0001), frequency band (delta-band coherence was stronger than beta-band's, F(1) = 49.7,  ROIs, frequency band, and state (resting-state, pre-and post-target tone), for controls (left) and amusics (right).

232
The distributions of dPTE values are shown for each inter-regional connection and over the baseline resting state,   In both groups, delta-beta phase-amplitude coupling was elevated in auditory and inferior frontal 299 cortices during task performance compared to baseline resting state (Fig. 2G). This observation is in line

303
A striking overall effect between groups was that delta-to-beta coupling in the auditory and inferior 304 frontal cortices was higher in amusics than in controls, both during tone-sequence presentations and at 305 baseline in the resting state. These observations of elevated ongoing phase-amplitude coupling are the 306 first observed in amusia. They contribute to converging evidence that chronically elevated PAC levels auditory cortex by words and phonemes that are less predictable in the sentence flow (Donhauser & 317 Baillet, 2020). Mechanistically, we propose that although PAC is expressed ubiquitously and dynamically 318 in the human brain (Florin & Baillet, 2015), over-expressions of PAC coupling may reflect a lack of flexibility 319 in the adjustment of the phase angle where fast frequency bursts are nested along slow frequency cycles.

320
This phase angle is related to the level of net excitability of the underlying cell assemblies and has been 321 discussed as an essential parameter for the neural registration of sensory inputs (Gips et al., 2016). High 322 levels of PAC may reduce opportunities for registering, and therefore encoding and processing, incoming 323 sensory inputs with sufficient temporal flexibility and adaptation to prediction errors (Arnal & Giraud, 2012).  Delta-to-beta coupling was stronger in auditory regions than in inferior frontal cortex in both groups 326 (Fig. 2G), which we interpret as due to the entrainment of auditory delta activity by stimulus cortical inputs, 327 which is expected to be more direct than in downstream regions. Yet, another marked difference between 328 groups was that there were modulations of delta-to-beta phase-amplitude coupling in the inferior frontal . 336 We derived time-resolved measurements of phase-amplitude coupling (tPAC) over time windows 337 around the occurrence of each of the tones in the sequence. In the auditory cortex of both groups, there 338 was an increase of cross-frequency coupling immediately after the onset of the tone sequence (Fig. 2B), 339 which culminated at the expected latency of the target tone presentation. This was confirmed by a time -340 resolved analysis of stimulus-to-beta coupling in the auditory cortex, which showed that stronger phasic 341 beta activity occurred at the expected latency of the auditory tones in the sequence (Fig. 2H). This delta-to-beta, with no temporal modulations along the presentation of the tone sequence (Fig. 2H). total of 640 tone sequences were then presented to every participant, in 10 blocks of 64 trials, which 466 resulted in a total of 320 standard tone sequences and 80 deviant trials per pitch deviance level. Trials 467 started in succession, 1 second (± < 50 ms jitter) following the subject's response to the previous trial. No 468 feedback was provided to participants on their performance.

469
Data acquisition 470 MEG data was collected during resting-state and task performance in the upright position using a 275-471 channel CTF MEG system, with a sampling rate of 2400 Hz. Simultaneous EEG data was recorded also 472 using the CTF system from four standard electrode positions: FZ, FCZ, PZ, and CZ (reference was placed 473 on right mastoid), electrode locations according to 10/20 system with 2400 Hz sampling rate (EEG data 474 shown as Supplementary Material -Fig. S1). The audio presentation, button presses, heartbeat and eye 475 movement electrophysiological signals (ECG and EOG, respectively) were also collected in 476 synchronization with MEG. Head position was monitored and controlled using three coils attached to the 477 subject's nasion). We obtained T1-weighted MRI volumes for each participant (1.5-T Siemens Sonata, 478 2018). These results emphasize how beta-band activity is expressed by the auditory tone sequence 537 instead of local delta activity in the auditory cortex, and whether beta bursts occur preferentially at the 538 expected latency of the tone presentation as a predictive form of signal.

539
Following the method used by Morillon and Baillet (2017), we generated a reference sinusoidal signal 540 at 2.85 Hz (i.e. the rate of the tone presentation every 350 ms), with its peaks aligned at the onset of each 541 tone presentation. We then estimated the tPAC cross-frequency coupling between the phase of this 542 reference signal and the amplitude of beta oscillations in the right auditory cortex. We tracked the 543 variations in time of this coupling using tPAC with a sliding window length of two cycles of the tone 544 presentation rate (700 ms) with 50% of overlap, following the specifications derived by Samiee and Baillet 545 (2017). We then identified the preferred phase of tPAC coupling along the cycle of the stimulus sinusoid 546 reference signal. Finally, we converted the corresponding phase angle into a time latency, as a fraction of 547 the 350 ms stimulus presentation cycle.

548
Functional and effective connectivity 549 We estimated frequency-specific functional connectivity between ROIs using coherence (Walter et al.,550 analyses and corrections for multiple comparisons. The distributions of event-related potentials (see 573 Supplementary Material) were tested for zero-mean using t-tests and reported with corrections for multiple 574 comparisons considering false discovery rates (FDR). tPAC values were assessed for statistical 575 significance using a non-parametric resampling approach (Samiee and Baillet, 2017): for each trial, we 576 generated 500 surrogates using block-resampling. Each surrogate was produced from selecting five time 577 points randomly in the trial epoch to subdivide the instantaneous phase signal into five blocks. These 578 blocks were then randomly shuffled and tPAC was estimated using the resulting block-shuffled phase