NeuroImage. Sensory gating in intracranial recordings The role of phase locking

Transcription

1 NeuroImage 44 (2009) Contents lists available at ScienceDirect NeuroImage journal homepage: Sensory gating in intracranial recordings The role of phase locking Timm Rosburg a,, Peter Trautner a, Jürgen Fell a, Karen Anne Moxon b, Christian E. Elger a, Nash N. Boutros c a Department of Epileptology, University of Bonn, Sigmund-Freud-Str. 25, D Bonn, Germany b Drexel University, School of Biomedical Engineering, 3141 Chestnut St, Philadelphia PA, 19104, USA c Wayne State University, School of Medicine, Department of Psychiatry and Behavioral Neurosciences, UPC-Jefferson, 2751 E. Jefferson, Detroit MI, 48207, USA article info abstract Article history: Received 10 June 2008 Revised 5 August 2008 Accepted 22 September 2008 Available online 10 October 2008 Keywords: Auditory evoked potentials Electrocorticography Schizophrenia Auditory cortex Electroencephalography For patients with schizophrenia, a deficient gating (or filtering) of sensory input has been described. One major approach to study this sensory gating is to measure event-related potentials (ERPs) in response to paired clicks. In these experiments, sensory gating is quantified as amplitude reduction of the ERP components P50 and N100 from the 1st to the 2nd stimulus. In ERP studies brain electrical signals are averaged over single trials. Alterations in phase locking might be one factor contributing to the observed deficits in sensory gating, but findings have been inconclusive as yet. In particular, the contribution of different frequency bands to the deficit required further investigation. We studied N100 gating by intracranial recordings in a sample of epilepsy patients and subdivided the group into good and poor gators of the intracranial ERP component N100. Data were evaluated by frequency specific wavelet-based phase and power analyses. Poor N100 gators had an increased phase locking in the frequency range from Hz after the 2nd stimulus, as compared to good gators. Other group differences were apparent already before the 2nd stimulus. Poor gators had less phase locked beta band activity ( Hz) than good gators ms after the onset of the 1st stimulus. Within the group of poor gators, lower values of phase locking in this frequency range were also associated with lower gating ratios. The reduced beta band response in response to the 1st stimulus may reflect poorer memory encoding of the 1st stimulus in poor gators. This in turn might lead to increased demands to process the 2nd stimulus Elsevier Inc. All rights reserved. Introduction Sensory gating has been conceptualized as the ability to screen out and filter excess and trivial stimuli (Clementz et al., 1997; Freedman et al., 1997). Deficits in sensory gating, as investigated by event-related potential (ERP) studies in paired clicks experiments, have repeatedly been described in schizophrenia (Bramon et al., 2004). In these experiments, sensory gating is commonly quantified as suppression of ERP amplitudes by the stimulus repetition. The deficit in schizophrenia patients is manifested usually as a reduced suppression of the P50 component, but more recently also as a reduced suppression of the N100 component, as compared to healthy populations (Freedman et al., 1983; Blumenfeld, 1999; Young et al., 2001; Boutros et al., 2004; Brockhaus-Dumke et al., 2008). In the current study, we aimed to investigate the neurophysiological basis of poor N100 gating by intracranial recordings in epilepsy patients. In animal experiments, sensory gating was shown for single unit activity and local field potentials in various brain regions, as the Corresponding author. Fax: address: timm.rosburg@ukb.uni-bonn.de (T. Rosburg). hippocampus, prefrontal cortex and striatum, (Moxon et al., 1999; Mears et al., 2006; Cromwell et al., 2007). However, the relationship between both kinds of activity was weak or absent (Mears et al., 2006; Cromwell et al., 2007). Thus, animal data suggests that the reduction of ERPs by stimulus repetition cannot be referred primarily to changes in single unit activity. Since ERPs are, by definition, identified by averaging the single trial responses of the electroencephalography (EEG) signal after the stimulus, the amplitude of the ERP represents changes in the EEG that are phase locked to the stimulus. Impaired gating might therefore be related to alterations in phase locking. In fact, when the temporal variability of the single trial response of schizophrenia patients was compared to control subjects, there was greater temporal variability for the P50 elicited by 1st stimulus (S1) (Jin et al., 1997; Patterson et al., 2000). By calculating the phase locking of on-going EEG oscillations for single frequencies, it was found that schizophrenia patients produced significantly less phase locking after the 1st stimulus in the theta and alpha band range than healthy controls did (Jansen et al., 2004; Brockhaus-Dumke et al., 2008). Although all these studies reported a similar loss of phase locking in response to S1 in schizophrenia patients as compared to healthy controls, studies differed with regard to the reported ERP amplitudes /$ see front matter 2008 Elsevier Inc. All rights reserved. doi: /j.neuroimage

2 1042 T. Rosburg et al. / NeuroImage 44 (2009) While P50 amplitudes to S1 were reduced in patients in two of the studies (Jin et al., 1997; Patterson et al., 2000), they were virtually unaffected in the two other studies (Jansen et al., 2004; Brockhaus- Dumke et al., 2008). Furthermore, the reduction of P50 amplitudes to S1 and a deficit in gating are not necessarily associated (Freedman et al., 1983; Moxon et al., 2003) and a reduced P50 gating in schizophrenia patients can be observed in absence of a reduction of P50 amplitudes to S1. Thus, the variance in the ERP of schizophrenic patients makes it difficult to assign relationships between gating and other measures of neuronal activation. To overcome the variability when comparing schizophrenic patients to normal controls, Young et al. (2001) divided their schizophrenia patient sample into those with good and poor N100/ P200 gating in order to address this issue. Young et al. (2001) determined the N100 peak latencies in single trials. The standard deviation of the latency measurements across trials was defined as N100 latency jitter. The authors reported that subjects with good gating had less N100 latency jitter to S1 stimuli as compared to subjects with poor gating and, moreover, that this latency jitter was inversely correlated to the S1 N100/P200 amplitude. However, similar as from the above cited studies, it is difficult to deduce from this study what determines poor gating in absence of S1 deficits. Further complicating the picture, a loss of phase locked activity after S1 was found to contribute to the reduced amplitude of the evoked potential after S2: Hong et al. (2004) specifically explored evoked beta (14 26 Hz) and gamma (30 50 Hz) activity after S1 and their relationship to the P50 response after the 2nd stimulus (S2), without directly addressing the issue of phase locking. The authors found that 59% of the S2 P50 variance in patients was explained by gamma and beta activity after S1. Thus, the magnitude of the P50 S2 response in patients was determined in major parts by activity before S2 onset. In the current study, we examined the correlates of poor gating in more detail. We focused on N100 gating because of the increasing interest of clinical researchers in this functional measure and also because of the fact that N100 activity dominates auditory ERPs. In this study, we took advantage of intracranial recordings which have a superior signal-to-noise ratio as compared to scalp recordings. In addition to phase locking and the amount of event-related (phase locked) activity, we analyzed induced activity which is event-related, but not phase locked. Because of its lack of phase locking, induced activity is not reflected in the ERP. The relation between induced activity and gating of ERP components has not been explored as yet. However, it appears possible that poor gating is associated with increased levels of induced activity, indicating a systematic stimulusrelated hyperactivation. In this context we were also interested in induced high frequency gamma band activity (GBA), comprising activity up to 200 Hz, which has only been observed in intracranial recordings (e.g. Crone et al., 2001; Trautner et al., 2006). We hypothesized that induced high frequency GBA reflects ripple activity. Ripples are believed to be due to simultaneous excitation of pyramidal cells and interneuronal networks and represent IPSPs on the somata of the pyramidal cells (Buzsaki et al., 1992; Chrobak and Buzsaki, 1996). Therefore, ripples comprise an important inhibitory component and might presumably also be related to sensory gating. For the purpose of the current study, we analyzed N100 activity, recorded from the posterolateral surface of the temporal lobe of epilepsy patients who underwent invasive presurgical evaluation. We divided the sample in good and poor N100 gators. Time-frequency transforms were calculated in order to compare event-related (phase locked) activity, phase locking and induced activity between the two groups. Areas of significant differences were subjected to a correlation analysis, in order to further understand the relation between frequency specific signals and N100 gating. Methods Subjects 37 patients (21 male) with a mean age of 37.6 years (range 17 to 65 years) were selected from a larger sample of patients investigated in a study on intracranially recorded sensory gating. All patients were epilepsy (N=36) or tumor (n=1) patients and underwent presurgical evaluation by means of implanted electrodes. The exact placement of electrodes always depended on clinical considerations only. For the purpose of the current study, only patients with subdural electrodes over the superior posterolateral surface of the temporal lobe, exhibiting a P50 N100 complex, were included. Data of 16 patients have been used for a previous publication (Trautner et al., 2006). Patients were on stable anticonvulsive medication at the time of the recording and gave written informed consent. 12 of the 37 patients had a history of depression, as revealed by clinical interviews. Two of the patients had short and benign passages of auditory or visual hallucinations in their past. The study was approved by the local ethics committee of the University of Bonn. Data recording and stimulation The EEG was recorded with the digital EPAS system (Schwarzer, Munich, Germany) and its implemented Harmonie EEG software (Stellate, Quebec, Canada). The EEG was measured against a reference of left and right mastoid electrodes with a sampling rate of 1000 Hz. Electrode positions were determined by MRI recordings routinely acquired after implantation. Patients were seated on a comfortable chair in a quiet room illuminated by bright light. During the experiment the subjects were stimulated with repetitive acoustic stimuli by headphones. The stimuli used were short tone bursts of a single sine wave with 1500 Hz frequency and a duration of 6.6 ms (including rise and fall times of 1.5 ms). A set of 100 pairs of stimuli were administered with an ISI of 0.5 s and an interpair interval of 8 s. Data analysis The EEG was segmented into single trials with a total duration of 2000 ms (Brain Vision Analyzer 1.05, Brain Products, Munich, Germany). An interval of 500 ms prior to the first stimulus was included into each trial for baseline correction. Trials with activity N200 μv were rejected as artefacts. After averaging, all electrode contacts were explored for AEPs. For inspection of the AEPs, band-pass filters were applied (1 20 Hz, 12 db/oct). The N100 was expected to peak within a latency interval from ms. Data sets were only included in the analysis when the peak amplitude of N100 for S1 was N10 μv, as measured with respect to baseline. If N100 deflections were detected at several electrodes, the electrode with maximal N100 amplitude was selected. For the selected electrode, data were more thoroughly analyzed by means of continuous wavelet transform. This transform characterizes the brain activity by its dispersion in the time and frequency domain. For each subject, data of single trials were transformed, using Morlet wavelets (6 cycles, 102 logarithmically spaced scales) and routines of Torrence and Compo (1998). The transformation yields complex wavelet coefficients for each time frequency point with real part (RE, cosine components) and imaginary part (IM, sine components). From these complex data points, the power, the amplitude and the phase information can be obtained. The wavelet transformations were processed in three different ways in order to obtain representations of the event related (phase locked) activity (ERA), the induced activity (IN), and the phase correlation (ph-corr) over the trials (see also Trautner et al. 2006).

3 T. Rosburg et al. / NeuroImage 44 (2009) ERA represents all phase locked parts of the signal and, therefore, resembles in its characteristics the ERP. The calculation of ERA must allow for cancellation of non phase locked signal components. This was achieved by calculating the mean of the wavelet coefficients before calculating the power values. of the N100 (112 ms), for each frequency point within the wavelet transform. These extracted S1 and S2 values were compared by paired t-test. In order to test whether the suppression differed between frequencies, data showing a significant suppression were collapsed ERAðt; f Þ= IM 2 + RE 2 ð1þ n n The phase of a signal at a given time-frequency point is given by the complex wavelet-coefficients with ϕ=arctan(im/re), taking the special case of RE=0 and different signs of IM and RE into account so that phase-angles between π and +π result. The phase locking was quantified by calculating the mean phase-correlation (ph-corr) (Mormann et al., 2000). ph corrðt; f vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u 2 sinðþ Þ + cosðþ 2 t n n Ph-corr quantifies the degree of phase locking across trials with normalized amplitudes. Ph-corr and ERA are calculated by similar equations, as the formation of ERA requires phase locking. Phase correlation represents the phase locking of the on-going oscillatory activity independent of the power at each frequency. ERA also comprises the amplitude (power) of the individual trials. Gating could be the result of a loss of phase locking and/or the result of reduced amplitudes of phase locked activity in response to S2. Both measures were calculated in order to reveal whether there was dissociation. In contrast to ERA, the induced activity (IN) represents those parts of the signal that are not phase correlated. In this respect, IN was defined as the mean of the power values (total signal power, TSP) minus ERA. ð2þ TSPðt; f Þ= 1 n IM2 + RE 2 ð3þ INðt; f Þ= TSP ERA ð4þ Both components (ERA and IN) are calculated to represent the signal power. Amplitude values would result from calculating the square root of ERA and IN. Calculation of background activity is important in order to qualify signals attributed to stimulus processing as statistically significant. To calculate this baseline activity, 6 s of the interpair interval were used to obtain three baseline segments with the same duration as the single trials (2 s). After applying the wavelet transformation to all background segments, IN_B(t, f) and ERA_B(t, f) were calculated in the same way as IN and ERA. The statistical significance of stimulus related signals in comparison of IN_B and ERA_B was tested by calculating paired t tests. All time-frequency representations were averaged over subjects to calculate grand averages (GA). In case of ph-corr, data were first transformed by means of Fischer's z transformation and after calculating the average transformed back to correlation coefficients. All graphs show the signal in relation to the background activity that means IN_GA(t, f) IN_B_GA(t, f) and ERA_GA(t, f) ERA_B_GA(t, f). Significance levels are given on the 95% level with black lines indicating signals being higher than the background activity and white lines indicating signals being lower than the background activity. Induced high-frequency GBA was quantified from the timefrequency transforms in the frequency range from Hz and in the time range from ms after stimulus onset, or respectively for S ms after S1 stimulus onset. In order to describe general aspects of gating, individual values from the wavelet transformations were extracted at the peak latencies Fig. 1. Brain electric activity for the total group, the 1st stimulus was delivered at 0 ms, the 2nd stimulus at 500 ms: in the top the unfiltered ERP signal; each stimulus elicits an ERP complex, with an attenuated response to the 2nd stimulus; event-related activity (ERA) reflects phase locked activity, decomposed in the frequency domain; intensity is decoded in the color; of note the polarity of the ERPs is not reflected in ERA; phase correlation (ph-corr) describes the amount of phase locking between trials; this measure is closely associated to ERA, but does not contain the intensity information of single trials (see Methods); in the bottom induced activity (IN), reflecting activity which is not phase locked. IN does by definition not contain any event-related activity; activity significant larger than baseline activity is surrounded by black lines; activity significant below baseline activity is surrounded by white lines.

4 1044 T. Rosburg et al. / NeuroImage 44 (2009) into frequency bands (theta: Hz; alpha: Hz; beta 1: Hz; beta 2: Hz; gamma: Hz). Suppression ratios (S2/S1) were calculated and compared between frequency bands by means of a repeated measure analysis of variance (ANOVA). In order to compare good and poor gators, N100 amplitudes in response to the 1st and 2nd click were quantified from their ERPs at their peak latencies and the gating ratio calculated. Gating N100 =1 N100 S2 ð5þ N100 S1 Thus, large Gating N100 values reflect strong gating. If the N100 amplitude of S2 was larger than the N100 amplitude of S1, Gating N100 was set to 0 (n=2), reflecting no gating. Patients were ordered according to their N100 gating ratio. Patients with N100 gating below the median of 0.5 were regarded as poor gators; patients with N100 gating 0.5 were regarded as good gators. Grand averages of different subject groups (subjects with good N100 vs. poor N100 gating) were compared by means of two sided Student's t tests. For this analysis, IN (t, f) and ERA(t, f) (both corrected for background activity) and ph-corr (t, f) (Fischer z transformed) of the corresponding subjects were used. A group comparison of poor and good P50 gators is provided as supplementary material. For graphical representation wavelet transforms were smoothed along both the wavelet scale axis and the time axis. This was done by a smoothing operator which has a similar footprint as the wavelet used. For the Morlet wavelet a suitable smoothing operator is given by Torrence and Webster (1999). By smoothing the wavelet spectra the number of significant data points is reduced especially at higher frequencies where data points have a higher independence in the time domain (testing 2000 data points at a level of 0.05 results in 100 occurrences by chance). Thus, smoothing the wavelet spectra makes interpretation more robust. Results 33 of the 37 patients exhibited an N100 which fulfilled the outlined criteria. 4 other subjects were not included because their N100 amplitude on S1 was found to be b10 μv (n=3) or because their N100 was delayed (n=1). In the total study sample, N100 amplitudes were significantly reduced by stimulus repetition from 43.8 μv (SD 25.4) to 23.4 μv (SD 15.7) (Fig. 1 top, t 32 =5.539, pb0.001, mean gating- N100= 0.45, SD 0.25). Stronger N100 gating was associated with reduced S2 N100 amplitudes (r=0.516, p=0.002), but not significantly with increased S1 N100 amplitudes (r= 0.308, n.s.). Frequency response of N100 suppression The ERA values extracted from the time-frequency transform at the average N100 peak maximum (112 ms) decreased significantly from S1 to S2 in a frequency range from 2.3 to 45.3 Hz (7.629Nt 32 N2.213, pb0.05; Figs. 1 and 2). This decrease was confirmed when the ph-corr was examined (significant attenuation from S1 to S2 in a frequency range from 2.1 to 45.3 Hz, 8.703Nt 32 N2.065, pb0.05). The suppression differed considerably between frequency bands, with stronger suppression in the theta and alpha band as compared to the beta1, beta2 and gamma band (ERA: F 4, 128 =5.870, p=0.001, ɛ=0.791; ph-corr: F 4, 128 =4.939, p=0.003, ɛ=0.737, Fig. 2). T-values of pairwise comparisons are provided in Table 1. At the N100 peak latency, the suppression of ph-corr for each frequency band closely matched the suppression of the ERA values. In fact, both measures were highly correlated (theta: r=0.936; alpha: r=0.945; beta1: r=0.751; beta2: r=0.813; gamma: r=0.880; all pb 0.001), signifying that the gating of ERA was closely associated with a loss in ph-corr. Significant induced high-frequency GBA was detected after both S1 and S2 (Fig. 1, bottom). It started at about the N100 latency range and lasted ms after both stimuli. When the amount of induced GBA was quantified ( Hz, ms post stimulus), it was found to decrease significantly from S1 to S2 (t 32 =2.828, p=0.008). Good versus poor N100 gating Patients were grouped according to their N100 gating ratio, obtained by an ERP conventional analysis. Patients with gating N were grouped as good gators (n=17), patients with gating N100 b 0.5 as poor gators (n=16). Exemplary ERP data of good and poor N100 gators are shown in Figs. 3 and 4. Three of the 10 included patients with history of depression were poor gators, 7 were good gators. Both patients with brief passages of hallucinations were poor gators. The mean N100 gating ratios were 0.64 (SD 0.11) in the group of good gators and 0.25 (SD 0.18) in the group of poor gators. The N100 reduction in the group of poor gators was still significant (t 15 =2.295, p=0.037). The N100 amplitudes in response to S1 were somewhat smaller in the group of poor gators ( 37.7±22.0 μv vs 49.4±27.6 μv), but did not differ significantly (F 1, 32 =1.798, n.s.). The N100 amplitudes in response to S2 were significantly larger in poor gators than in good gators ( 29.5±17.3 μv vs 17.6±11.8 μv, F 1, 32 =5.306, p=0.028). Fig. 2. The suppression ratios of ERA values and ph-corr at the N100 peak latency (112 ms), averaged for frequency bands. The error bar depicts ±2 standard errors. There was a stronger suppression in the theta and alpha band as compared to the beta1, beta2 and gamma band. The suppression ratios of ph-corr and ERA had a very similar magnitude and were highly correlated.

5 T. Rosburg et al. / NeuroImage 44 (2009) Table 1 T-values of the pairwise comparison of suppression ratios between frequency bands at the N100 peak latency The half above the diagonal of the matrix refers to comparisons of ERA values, the half below the diagonal to comparisons of the phase correlation values; the differences between theta and alpha band suppression ratios on the one hand and beta1, beta2, and gamma band suppression ratios on the other hand are highlighted by the inserted frames; the level of significance is indicated by + pb0.1; pb0.05; pb0.01; pb The time-frequency data of the two groups are depicted in Fig. 5. The right column in Fig. 5 shows the significant differences in timefrequency range between groups. A major group difference was found for ERA and phase correlation in the time range from ~ ms and frequency range of ~ Hz. Good gators had significantly less ERA and less phase correlation in this time-frequency range. Extracted ERA and phase correlation values of this time-frequency range were again highly correlated (r=0.796, pb0.001). In the total group, both measures were also highly correlated with the N100 amplitude in response to S2 (r= and r= 0.792, respectively, both pb0.001). Another group difference was detected for ERA in the time range from ~ ms and frequency range of ~ Hz. The according difference in phase-correlation missed to reach significance. This dissociation of ERA and phase correlation might in parts be related to the somewhat lower correlation between the two measures at this time-frequency range (r =0.646, p b0.001). Of note, the extracted ERA and phase correlation values for this range did not correlate significantly with the N100 amplitude in response to S1 or S2. IN in the high-frequency range (N32 Hz) did not differ between groups, but group difference in IN were revealed for the timefrequency range from ~ ms and ~ Hz, as well as for the time-frequency range from ~ ms and Hz. Fig. 4. Unfiltered single trial data of one patient each with good and poor N100 gating. 30 trials each are depicted as superimposition. The grand average of the 30 trials is overlaid in black. Fig. 3. Exemplary unfiltered ERP data of five patients each with good and poor N100 gating, with each graph representing data of one subject. The individual N100 gating ratios are provided at the right upper corner of each graph.

6 1046 T. Rosburg et al. / NeuroImage 44 (2009) Fig. 5. The left and middle columns depict the ERP, ERA, ph-corr and IN of good and poor N100 gators separately (see also legend of Fig. 1); the right column shows t-values obtained by the statistical comparison of the two groups; positive values indicate that good gators had higher values than bad gators; negative values indicate the reverse; areas of significant group differences are surrounded by black or white lines depending on the algebraic sign of the t value. Neither kind of activity was significantly correlated with S1 and S2 N100 amplitudes. Correlation analysis All areas of time-frequency data showing a significant difference between good and poor gators and the extracted measure of induced high-frequency GBA were subjected to a correlation analysis in order to identify their relationship to N100 gating. This was done not only for the total group but also for good and poor gators separately. For the total group, all measures except the induced highfrequency GBA after S1 and S2 were significantly correlated with N100 gating (Table 2). Induced high-frequency GBA also did not correlate with N100 gating in the group of poor and good gators. Within subgroups, induced activity was not associated with N100 gating in any of the other frequency bands (Table 2). The other four measures were found to be associated with the degree of N100 gating in one of the two groups. In good gators, ERA and ph-corr in the time range from ms and in the frequency range from Hz were negatively correlated with N100 gating (Fig. 6 top). In poor gators, ERA and ph-corr in the time range from ms and in the frequency range from Hz were positively correlated with N100 gating (Fig. 6 bottom). These two measures were reduced in the group of poor gators as compared to

7 T. Rosburg et al. / NeuroImage 44 (2009) Table 2 Correlation between extracted time-frequency values and N100 gating Measure ERA Ph-corr ERA Ph-corr IN IN IN IN Time range [ms] Frequency [Hz] Total sample Good gators Poor gators Time frequency values are described by kind of measure (ERA, ph-corr, IN), time range, and frequency range. The only dependent variable is N100 gating. Correlation coefficients are given for the total group, and for good and poor gators separately. The level of significance is indicated by + pb0.1; pb0.05; pb0.01. good gators, and within poor gators also those patients with lower values tend to have a poorer N100 gating. Discussion The current study compared phase locked activity, phase locking and induced activity of epilepsy patients with good and poor N100 gating in a double click experiment. Poor N100 gating was characterized by increased phase locked activity and phase locking in the frequency range from Hz after S2. In addition, poor and good gators differed before S2. Poor gators had less phase locked beta band activity ( Hz) than good gators ms after S1. In contrast, induced high frequency gamma band activity was not associated with N100 gating. In the following, we discuss our findings in detail. to be characterized by both the lack of phase locking and reduced single trial power after S1. The impact of single trial power is also underlined by findings on the induced beta band activity. Poor gators exhibited also significantly less induced beta band activity after S1, as compared to good gators. Gating and phase locking after S2 Auditory ERPs in healthy subjects were reported to depend to major extent on phase locking between trials (Sayers et al., 1974), but disease related alterations of ERP components in schizophrenia subjects could in principle depend on alterations in phase locking or on alterations of single trial power. Therefore, we analyzed two measures, namely ERA and ph-corr. Ph-corr represents the phase locking of the on-going oscillatory activity independent of the power at each frequency, while ERA also comprises the power of the individual trials. In the current study, poor N100 gating was associated with increased ERA and increased ph-corr in the frequency range from Hz and time range from ms. ERA and ph-corr were highly correlated in this time-frequency range, as well as the gating ratios of ERA and ph-corr at the N100 peak latency. With other words: a large portion of the variance of ERA could be explained by ph-corr. Therefore, the excess of S2 activation in poor N100 gators appears to be due primarily to an excess of phase locking. Also, the analysis of induced activity revealed no excess of single trial power in the group of poor gators at these latencies. The only time-frequency range with higher induced activity in the group of poor gators occurred from ms in the frequency range from Hz, thus later than the S2 N100. To sum up these findings, poor N100 gating appears to be associated with increased phase locking after S2 and not with a general hyperactivation of the auditory system. As shown in an additional analysis, poor P50 gating was also related to increased phase locking after S2, but at slightly earlier latencies and slightly higher frequencies (Supplementary data). Gating and beta band activation after S1 Poor gators exhibited significantly less phase locked beta band activity after S1, as compared to good gators. The according difference for phase locking did not reach significance. However, a significant correlation between N100 gating and phase locking was found in the group of poor gators. Thus, in the beta band range poor gating appears Fig. 6. Scatter plots of the association between two extracted measures of phase correlation (Table 2) and N100 gating. Within the plots, individual values are depicted, with blue triangles indicating a data set of a person defined as good gator and a red triangle indicating a data set of a person defined as poor gator.

8 1048 T. Rosburg et al. / NeuroImage 44 (2009) Beta band oscillations have been referred to encoding and memory processes (Whittington et al., 1997; Haenschel et al., 2000; Tallon- Baudry et al., 2001, 2004; Leiberg et al., 2006). The diminished phase locked beta band activity in our group of poor gators might be regarded as an indication of deficiencies in the encoding of S1, as also proposed by Hong et al. (2004). Following this line of argumentation would, however, mean to re-interpret the increased S2 responses as consequence of encoding problems, instead of regarding them as indicator for deficient filter processes. There is evidence for this view from the study of differences in mismatch negativity (MMN) between normal subjects and schizophrenics. The reduction in the evoked response to an oddball stimulus of schizophrenia patients has been referred to as an impaired precision of the auditory sensory memory, also found in tone matching tasks (Javitt, 2000). Thus, basic auditory memory functions are impaired in schizophrenia. Additional evidence comes from a recent study reporting a correlation between MMN amplitudes and P50 gating ratios (Kisley et al., 2004). The shared variance of the two measures could mean that both refer to a common cortical function and this shared function could be the encoding of the acoustic stimuli. However, in order to establish the link between evoked beta band activity and sensory memory it would furthermore be necessary to show correlations between evoked beta band activity and the MMN amplitude, as well. To our knowledge, such a study does not exist. Gating and induced high frequency gamma band activation No association of induced high frequency GBA ( Hz) and N100 gating could be revealed. Thus, we found no evidence for our hypothesis that induced high frequency GBA could be related to inhibitory processes, also affecting the N100 generation. A previous study observed increased levels of induced high-frequency GBA for attended vs. non-attended stimuli (Ray et al., 2008). That finding also argues against the hypothesis that induced high frequency GBA might be related to inhibitory processes. An additional analysis did reveal no relation between induced high-frequency GBA and P50 gating either (Supplementary data). The exact functional role of induced high frequency GBA in auditory perception remains to be clarified. Methodological considerations For the current study, we utilized intracranial recordings rather than the more typical scalp recordings. While the drawback of this is that these signals were recorded from epilepsy patients, the signals had better signal-to-noise ratio and activity was recorded directly from brain areas involved in N100 generation (Näätänen and Picton, 1987; Rosburg et al., 2006). Furthermore, GBA can reliably be identified in intracranial recordings, as opposed to scalp recordings, which might be affected miniature saccades (Yuval-Greenberg et al., 2008). By combining this approach with a non-biased examination of the response in the time and frequency domain, we were able to identify differences between good and poor N100 gators for specific frequency ranges and time windows. The current study is assumed to provide information about neurophysiological correlates of poor N100 gating in general. Whether the correlates of poor N100 gating in schizophrenia patients are the same or different as compared to those we observed in our study remains to be shown. At the current time, we cannot rule out that also other mechanisms may contribute to a deficient gating in schizophrenia. One methodological limitation of the current findings might be that the calculation of phase locking depends on the signal-to-noise ratio in single trials. Thus, ph-corr of S2 activity could theoretically be underestimated because of a decreased single trial power. We cannot rule out completely this possibility. However, for S2 responses we observed still a clear phase locking (Fig. 1), meaning that we still had a sufficient signal-to-noise ratio in single trials. Furthermore, even in the group of poor gators (i.e. for subject having good signal-to-noise ratios for S2 responses), there was a clear decrease of phase locking from S1 to S2. Finally, good and poor N100 gators did not differ in induced activity or in total signal power (data not shown) to S2. Thus, group differences in ph-corr to S2 cannot be referred to systematically smaller signals in good gators. Conclusion We demonstrate here for the first time that poor N100 gating is a result of an important relationship between a lack of phase locked activity after S1, potentially signifying that the stimulus was not encoded properly, and increased phase locking after S2, suggesting that the S2 stimulus is considered novel if the S1 stimulus was not encoded. Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at doi: /j.neuroimage References Blumenfeld, L.D., Clementz, B.A., Hemispheric differences on auditory evoked response suppression in schizophrenia. NeuroReport 10, Boutros, N.N., Korzyukov, O., Jansen, B., Feingold, A., Bell, M., Sensory gating deficits during the mid-latency phase of information processing in medicated schizophrenia patients. Psychiatry. Res. 126, Bramon, E., Rabe-Hesketh, S., Sham, P., Murray, R.M., Frangou, S., Meta-analysis of the P300 and P50 waveforms in schizophrenia. Schizophr. Res. 70, Brockhaus-Dumke, A., Mueller, R., Faigle, U., Klosterkoetter, J., Sensory gating revisited: relation between brain oscillations and auditory evoked potentials in schizophrenia. Schizophr. Res. 99, Buzsaki, G., Horvath, Z., Urioste, R., Hetke, J., Wise, K., High-frequency network oscillation in the hippocampus. Science 256, Chrobak, J.J., Buzsaki, G., High-frequency oscillations in the output networks of the hippocampal-entorhinal axis of the freely behaving rat. J. Neurosci. 16, Clementz, B.A., Geyer, M.A., Braff, D.L., P50 suppression among schizophrenia and normal comparison subjects: a methodological analysis. Biol. Psychiatry 41, Cromwell, H.C., Klein, A., Mears, R.P., Single unit and population responses during inhibitory gating of striatal activity in freely moving rats. Neuroscience 146, Crone, N.E., Boatman, D., Gordon, B., Hao, L., Induced electrocorticographic gamma activity during auditory perception. Brazier Award-winning article, Clin. Neurophysiol. 112, Freedman, R., Adler, L.E., Waldo, M.C., Pachtman, E., Franks, R.D., Neurophysiological evidence for a defect in inhibitory pathways in schizophrenia: comparison of medicated and drug-free patients. Biol. Psychiatry. 18, Freedman, R., Coon, H., Myles-Worsley, M., Orr-Urtreger, A., Olincy, A., Davis, A., Polymeropoulos, M., Holik, J., Hopkins, J., Hoff, M., Rosenthal, J., Waldo, M.C., Reimherr, F., Wender, P., Yaw, J., Young, D.A., Breese, C.R., Adams, C., Patterson, D., Adler, L.E., Kruglyak, L., Leonard, S., Byerley, W.,1997. Linkage of a neurophysiological deficit in schizophrenia to a chromosome 15 locus. Proc. Natl. Acad. Sci. U. S. A. 94, Haenschel, C., Baldeweg, T., Croft, R.J., Whittington, M., Gruzelier, J., Gamma and beta frequency oscillations in response to novel auditory stimuli: a comparison of human electroencephalogram (EEG) data with in vitro models. Proc. Natl. Acad. Sci. U. S. A. 97, Hong, L.E., Summerfelt, A., McMahon, R.P., Thaker, G.K., Buchanan, R.W., Gamma/ beta oscillation and sensory gating deficit in schizophrenia. NeuroReport 15, Jansen, B.H., Hegde, A., Boutros, N.N., Contribution of different EEG frequencies to auditory evoked potential abnormalities in schizophrenia. Clin. Neurophysiol. 115, Javitt, D.C., Intracortical mechanisms of mismatch negativity dysfunction in schizophrenia. Audiol. Neurootol. 5, Jin, Y., Potkin, S.G., Patterson, J.V., Sandman, C.A., Hetrick, W.P., Bunney Jr., W.E., Effects of P50 temporal variability on sensory gating in schizophrenia. Psychiatry. Res. 70, Kisley, M.A., Noecker, T.L., Guinther, P.M., Comparison of sensory gating to mismatch negativity and self-reported perceptual phenomena in healthy adults. Psychophysiology 41, Leiberg, S., Lutzenberger, W., Kaiser, J., Effects of memory load on cortical oscillatory activity during auditory pattern working memory. Brain. Res. 1120, Mears, R.P., Klein, A.C., Cromwell, H.C., Auditory inhibitory gating in medial prefrontal cortex: Single unit and local field potential analysis. Neuroscience 141,

9 T. Rosburg et al. / NeuroImage 44 (2009) Mormann, F., Lehnertz, K., David, P., Elger, C.E., Mean phase coherence as a measure for phase synchronization and its application to the EEG of epilepsy patients. Physica D 144, Moxon, K.A., Gerhardt, G.A., Bickford, P.C., Austin, K., Rose, G.M., Woodward, D.J., Adler, L.E., Multiple single units and population responses during inhibitory gating of hippocampal auditory response in freely-moving rats. Brain Res. 825, Moxon, K.A., Gerhardt, G.A., Adler, L.E., Dopaminergic modulation of the P50 auditory-evoked potential in a computer model of the CA3 region of the hippocampus: its relationship to sensory gating in schizophrenia. Biol. Cybern. 88, Näätänen, R., Picton, T., The N1 wave of the human electric and magnetic response to sound: a review and an analysis of the component structure. Psychophysiology 24, Patterson, J.V., Jin, Y., Gierczak, M., Hetrick, W.P., Potkin, S., Bunney Jr., W.E., Sandman, C.A., Effects of temporal variability on p50 and the gating ratio in schizophrenia: a frequency domain adaptive filter single-trial analysis. Arch. Gen. Psychiatry 57, Ray, S., Niebur, E., Hsiao, S.S., Sinai, A., Crone, N.E., High-frequency gamma activity ( Hz) is increased in human cortex during selective attention. Clin. Neurophysiol. 119, Rosburg, T., Trautner, P., Boutros, N.N., Korzyukov, O.A., Schaller, C., Elger, C.E., Kurthen, M., Habituation of auditory evoked potentials in intracranial and extracranial recordings. Psychophysiology 43, Sayers, B.M., Beagley, H.A., Henshall, W.R., The mechanism of auditory evoked EEG responses. Nature 247, Tallon-Baudry, C., Bertrand, O., Fischer, C., Oscillatory synchrony between human extrastriate areas during visual short-term memory maintenance. J. Neurosci. 21, RC177. Tallon-Baudry, C., Mandon, S., Freiwald, W.A., Kreiter, A.K., Oscillatory synchrony in the monkey temporal lobe correlates with performance in a visual short-term memory task. Cereb. Cortex 14, Torrence, C., Compo, G.P., A practical guide to wavelet analysis. Bull. Am. Meteorl. Soc. 79, Torrence, C., Webster, P., Interdecadal changes in the ESNO-monsoon system. J. Clim. 12, Trautner, P., Rosburg, T., Dietl, T., Fell, J., Korzyukov, O.A., Kurthen, M., Schaller, C., Elger, C.E., Boutros, N.N., Sensory gating of auditory evoked and induced gamma band activity in intracranial recordings. Neuroimage 32, Whittington, M.A., Traub, R.D., Faulkner, H.J., Stanford, I.M., Jefferys, J.G., Recurrent excitatory postsynaptic potentials induced by synchronized fast cortical oscillations. Proc. Natl. Acad. Sci. U. S. A. 94, Young, K.A., Smith, M., Rawls, T., Elliott, D.B., Russell, I.S., Hicks, P.B., N100 evoked potential latency variation and startle in schizophrenia. Neuroreport 12, Yuval-Greenberg, S., Tomer, O., Keren, A.S., Nelken, I., Deouell, L.Y., Transient induced gamma-band response in EEG as a manifestation of miniature saccades. Neuron. 58,