Supplementary Material

Transcription

1 Supplementary Material Orthogonal representation of sound dimensions in the primate midbrain Simon Baumann, Timothy D. Griffiths, Li Sun, Christopher I. Petkov, Alex Thiele & Adrian Rees Methods: Animals The data was obtained from 28 fmri sessions with three male macaque monkeys (Macaca mulatta) weighting 9 16 kg. For two animals we acquired three sessions for the tonotopy experiment and seven sessions for the periodotopy experiment. In a third animal, the data consist of one tonotopy session and seven periodotopy sessions. The animals were previously habituated to the scanner environment and were scanned while awake. A custom made primate chair was used to position the animal in the bore of the scanner and head movements were minimised with a head holder. Details of the positioning procedures are given in Baumann et al. 11. All experiments were carried out in accordance with the UK, Animals (Scientific Procedures) Act (1986), European Communities Council Directive 1986 (86/609/EEC) and the US National Institutes of Health Guidelines for the Care and Use of Animals for Experimental Procedures, and were performed with great care to ensure the well-being of the animals. Sound stimuli and presentation Sound stimuli were created in MATLAB 7.1 (MathWorks, Natick, USA) with a sample rate of 44.1 khz and 16 bit resolution. Stimuli for characterising the BOLD response to spectral frequencies (tonotopy-experiment) were based on a random-phase noise carrier with three different pass-bands, khz, 2 4 khz and 8 16 khz resulting in three different stimuli that encompassed different spectral ranges. The carriers were amplitude modulated with a 1

2 sinusoidal envelope of 90% depth at 10 Hz to achieve a robust response in the brain stem. The stimuli for characterising the temporal frequencies in the amplitude modulation experiment (periodotopy-experiment) were also based on random-phase noise carrier but had a flat broadband spectrum from 25 Hz to 16 khz. This carrier was amplitude modulated at six different frequencies, 0.5 Hz, 2 Hz, 8 Hz, 32 Hz, 128 Hz and 512 Hz resulting in six different stimuli that cover the approximate range of amplitude modulation rates to which neurons in the midbrain are responsive 10. The duration of all the stimuli was 6 s which included at least three cycles of the modulation in the case of the lowest temporal frequency. This duration is also sufficient for the BOLD response in the brain stem of macaques to reach the plateau 11. The on- and off-set of the stimulus were smoothed by a linear ramp of 50 ms. We presented the stimuli in the scanner at an RMS sound pressure level of 75 db using custom adapted electrostatic headphones based on a Nordic NeuroLab system (NordicNeuroLab, Bergen, Norway). These headphones feature a flat frequency transfer function up to 16 khz and are free from harmonic-distortion at the applied sound pressure level. Sound pressure levels were verified using an MR-compatible condenser microphone B&K Type 4189 (Bruel&Kjaer, Naerum, Denmark) connected by an extension cable to the sound level meter Type 2260 from the same company. MRI hardware and imaging Data were recorded in an actively shielded, vertical 4.7 T MRI scanner (Bruker Biospec 47/60 VAS) equipped with a Bruker GA-38S gradient system with an inner-bore diameter of 38 cm (Bruker Medical, Ettlingen, Germany). The applied RF transmitter-receiver coil (Bruker) was of a volume array design that covered the entire head of the animals. Functional and structural data were acquired from identical 2 mm thick slices in a plane orientated roughly perpendicular to the axis of the brainstem. Two slices that covered the inferior colliculus (IC) were used for this study. The slices were selected with the help of an additional structural scan that covered the brainstem in sagittal orientation. The approximate location and orientation of the scanned slices are displayed in Fig 1a. 2

3 Functional scan parameters: Single-shot gradient-recalled echo-planar imaging sequences were optimised for each subject sharing an in-plane resolution of 1 x 1 mm 2 and a volume acquisition time (TA) of 1s. Typical acquisition parameters were: TE: 21 ms, flip angle (FA): 90º, spectral bandwidth: 200 khz, Field of view (FOV): 9.6 x 9.6 cm 2, with an acquisition matrix of 96 x 96. Each volume acquisition was separated by a 9 s gap to avoid recording the BOLD response to the gradient noise of the previous scan ( sparse design ). In combination with the TA of 1s this results in a repetition time (TR) of 10 s. The stimuli were presented during the last 6 s of the silent gap. The detailed timing was based on a previous BOLD response time course characterisation in the auditory system of macaques 11. Before every other volume acquisition no stimulus was presented to obtain data for a silent baseline. For the tonotopy-experiment a total 720 volumes were acquired per session. This resulted in 120 volumes per stimulus per session (half of the volumes served for the baseline) or 360 volumes per stimulus in total for the three sessions. For the periodotopy-experiment 540 volumes per session were acquired resulting in 45 volumes per stimulus per session and 315 volumes per stimulus for the combined seven sessions. Structural scan parameters: Structural images (T1-weighted) used the same slice geometry as the functional scans to simplify coregistration. The imaging parameters of the MDEFT (Modified Driven Equilibrium Fourier Transform) sequence were: TE: 6 ms, TR: 2240 ms, FA: 30º, FOV 9.6 x 9.6 cm 2 using an encoding matrix of 256 x 256 to result in an in-plane resolution of x mm 2 per voxel. Correct overlap of structural and functional images has been confirmed for all maps. Data analysis For preprocessing and general linear model analysis we employed the SPM5 software package ( implemented in Matlab 7.1. The data acquired from each animal were analysed separately. Image volumes from each session were realigned to the first volume and the sessions of the two experiments were subsequently realigned to each other before smoothing the data with a kernel of 2 mm full-width half-maximum. The time-series were high 3

4 pass filtered with a cut-off of 300 s to account for slow signal drifts and the data was adjusted for global signal fluctuations (global scaling). In a general linear model analysis for the combined sessions of each experiment, the voxel-wise response estimate coefficients (beta-values) and t- values for the contrast of the different stimuli versus the silent baseline were calculated. Fig 1b shows the map of t-values for the combined AM stimuli versus baseline for animal Cr. Significant auditory responses based on t-maps matched the predicted location and extent of the IC derived from anatomical features in the corresponding structural scans. Further analysis and data display was performed using custom designed Matlab scripts. The two slices of the response estimate coefficient maps covering the IC were collapsed in to one by calculating the mean, and the resulting maps were masked retaining only the voxels that showed significant values for the combined stimuli versus baseline contrast for each of the two experiments (FWE corrected for multiple comparisons). Given the relatively small number of voxels (20 30) that remained for each of the two ICs we collapsed the maps of the six temporal stimuli obtained for each animal into three by taking the mean of the 0.5 Hz and 2 Hz, the 8 Hz and 32 Hz, and the 128 Hz and 512 Hz maps. This left us with a low rate (lr), a middle rate (mr) and a high rate (hr) map of the response estimate coefficients (beta-values) for each animal for the periodotopy experiment. In the case of the tonotopy-experiment the three different stimulus conditions that were tested (0.5 1 khz, 2 4 khz and 8 16 khz spectral bands) already resulted in three response estimation coefficient maps for low frequencies (lf), middle frequencies (mf) and high frequencies (hf). These six response estimate maps for the temporal and the spectral stimuli for each of the three animals were used for further analysis of the detailed spectro-temporal organisation in the IC. Best frequency/rate response maps (BF/R-map): The best frequency/rate response maps were calculated by identifying voxel by voxel for each experiment and animal which of the three spectral frequencies or temporal rates showed the highest response estimate coefficients. The resulting maps represent the preferred frequency or rate for each voxel (Suppl. Fig. S1). 4

5 Frequency/rate response difference maps (Diff-map): In order to highlight the gradual change of frequency or rate preference for the different voxels we subtracted voxel by voxel the response estimate coefficents of the low frequencies (lf) or rates (lr) from the high frequencies (hf) or rates (hr). The resulting maps represent the degree of preference for high or low frequencies/rates. Direction and quantification of frequency/rate gradients: The observed one-dimensional gradients for spectral frequencies and temporal rates in two-dimensional maps were tested by a linear, two-dimensional multi-regression analysis. Effectively, a plane was fitted to the values of the Diff-map (Fig. 2) and regression coefficients (R 2 ), p-values for the fit were obtained. The application of a linear (planar) regression was used as the simplest model to test a one dimensional frequency or AM rate versus distance trend, respectively. The results do not exclude that a higher order model would better explain the data. The direction of gradient (or steepest angle) of the plane represent the estimated direction of the frequency/rate preference gradient or the direction of the tonotopy and periodotopy axis, respectively. The angles of the gradients relative to the dorsal-ventral axis were calculated for both ICs in all three animals for the tonotopy- and periodotpy-experiments. From these angles we calculated the relative angles between the tonotopy gradient and the periodotopy gradient for the IC nuclei. 5

6 Supplementary Figure S1 Example waveforms and spectra for sound stimuli. Each panel shows (above) a 1-s excerpt of the stimulus waveform (temporal envelope) and (below) its spectrum. Panels A and B show example stimuli from the spectral experiment. The frequency range is 500 Hz to 1 khz in A and 8 khz to 16 khz in B; in both cases the temporal envelope is fixed at 10 Hz. Panels C and D show example stimuli from the temporal experiment. The spectrum is fixed at 25 Hz 16 khz, while the amplitude modulation rate is 2 Hz in panel C and 32 Hz in panel D. 6

7 Supplementary Figure S2 Maps of stimulus types that show maximal BOLD response in the inferior colliculi of three animals (IC). The maps in the left column show which frequency range gives the strongest BOLD response in each voxel. The color code for low frequencies (lf), middle frequencies (mf) and high frequencies is displayed in the bottom panel. The maps in the right column show the best range of amplitude modulation rates for each voxel. The colour code for low, middle and high rates (lr, mr and hr) is located in the right bottom panel. 7

8 Supplementary Figure S3 8

9 Regression planes and gradient direction maps for all subjects. The figure shows data equivalent to data in Figure 2 but for all subjects. The first two columns show data from the right IC s and the second two columns show data from the left IC s. Tonotopic experiments are represented by column 1, 3 and periodotopic experiments are represented by column 2, 4. Abbreviations Ws, Dl and CR indicate subjects. 9

10 Supplementary Table 1 Animal L R relative angle [degree] periodotopy tonotopy periodotopy tonotopy L R Direction Ws [degree] Dl Cr average std. dev R 2 coeff. Ws Dl Cr p value Ws <10 5 <10 6 < *10 4 Dl <10 12 <10 7 <10 6 <10 6 Cr <10 6 <10 6 < *10 4 N Ws [voxel] Dl Cr Direction and significance values from regression analysis for tonotopy and periodotopy maps. Gradient quantification data from the tontotopy and periodotopy experiment for the left (L) and right (R) ICs of the three animals. Gradient directions are shown relative to the dorso-ventral axis (0 ). The relative angles between the gradients from the two experiments are shown on the right side with the average and standard deviation of the direction values immediately below. Regression coefficients (R 2 ), p-values, and the number of data points per map (N) are shown for each gradient analysis. 10