Psychometric results
IQ. Children with ASD had significantly lower IQ than NT children (Table 1). Variability in MPI scores was high in the ASD sample and ranged from ‘very low’ to ‘higher than the average’ scores on all three scales.
Autism scores. According to each of the three parents’ questionnaires used, the severity of autism symptoms was significantly higher in the ASD compared with the NT sample (Table 1). There were high correlations between all three ‘autism severity’ scales in both the ASD (Pearson correlation coefficients for SRS/SCQlife: R=0.53, p=0.003; SRS/AQ: R=0.52, p=0.003; SCQlife/AQ, R=0.58, p=0.001) and the NT (Pearson correlation coefficients for SRS/SCQlife: R=0.60, p<0.001; SRS/AQ, 0.84, p<0.001; SCQlife/AQ, 0.51, p=0.003) participants. To construct a unified ‘Autism Score’ for neurobehavioral correlation analysis, we reduced dimensionality of the data by extracting the common variance shared by the three questionnaires. We only used data from children with ASD to construct the ‘Autism Score’. The values obtained from the three autism scales were ztransformed before performing the principle component analysis (PCA). In the subjects with missing data on some of the scales (N = 4), we substituted the missing values with the mean of the available scales. The first principle component, which accounted for 71% of the variance in the ASD group, was then used as the integrative ‘Autism Score’. Children with positive Autism Scores scored higher than the average for the ASD group on autism severity, while negative Autism Scores indicated relatively milder autism symptoms. There was a negative correlation between the ‘Autism Score’ and IQ (MPI Standard) (Pearson correlation coefficient R(31)= 0.36, p = 0.049).
The auditory response waveforms in children and adults
To investigate whether the SF is present in 7–12yearold children and whether it is homologous to the ‘adult’ SF, we compared the entire waveform of the auditory evoked response to click trains in the NT children and adults at the sensor level and at the level of cortical sources. To visualize all the main components of the auditory response – the transient components, the ASSR and SF – we set the lowpass filter at 100Hz.
Figure 2 displays the grand average sensorspace plots (A, B) and sLORETA timecourses in the SF group maxima within the left and right hemispheres (C; see Methods for details). There were marked differences in the auditory responses to click trains between the NT children and adults.
In adults, the stimulation onset evoked a sequence of transient obligatory MEG responses: a small but distinguishable P50m at around 40 ms, followed by a much more prominent N100m of the opposite polarity peaking at around 115 ms after the stimulation onset. The strength of the negative current decreased at 200 ms due to an evolving ‘positive’ P200m. These transient components overlapped with a slowly developing shift of magnetic field (SF), which had the same polarity as the N100m component, reaching its maximal strength at around 400 ms and lasting until the end of stimulation.
In children, the small P50m component at around 40 ms was followed by a second deflection of the same positive polarity at 80–84 ms (Figure 2B, C), which corresponds to the child P100m and is distinct from the P50m and N100m components in adults [37]. The absence of the N100m and the P200m peaks in auditory responses to clicks and tones is typical for children before adolescence [37, 38, 40].
The qualitative difference in the pattern of the transient auditory components between children and adults was confirmed by the presence of ‘positivity’ in children and ‘negativity’ in adults in the ~70–90 ms interval, both of which were significant in the right ROI between 74–88 msec after stimulation onset (FDR corrected p<0.01 in both children and adults).
Despite the striking developmental difference in morphology of the transient components, the SF in children resembled that in adults: in both age groups, its sources in the auditory cortex had the same direction of current and comparable magnitude (Figure 2C). Moreover, in response to contralateral stimulation, the SF dominated in the right hemispheres in both age groups. The visibly faster development of the SF in children compared with adults might be explained by the lack of the adult P200m component that masks the early segment of the SF response. In both children and adults, the 40 Hz ASSR overlapping with the SF is clearly visible on sensors (Figure 2A, B) and at the source level (Figure 2C).
Source localization of the 40Hz ASSR and SF in the auditory cortex
To compare cortical localizations of the ASSR and SF, for each subject we calculated the MNI coordinates of the sources with the maximal 40 Hz ITPC in the 200–500 ms range and those with the maximal integrated SF amplitude in the 200–500 ms range in response to contralateral stimulation. Given that localization results are highly sensitive to SNR, we excluded from the coordinate analysis those subjects who had very low ITPC values or visually undetectable SF in the hemisphere contralateral to the stimulated ear. For ASSR coordinate analysis we excluded those subjects who had ITPC equal to or lower than 0.18, the highest ITPC value observed during the baseline (see the Supplementary Figure S2). As a result, 31 NT/33 ASD and 33 NT/32 ASD children were included in the analysis of the ASSR MNI coordinates for the left and right hemispheres, respectively. For the SF coordinate analysis, we excluded subjects/hemispheres where the sustained negative deflection of the field with a rising front slope after the click onset was not detectable (see example in the Supplementary Figure S3). This resulted in the inclusion of 33 NT/29 ASD and 35 NT/35 ASD children in the analysis of the SF MNI coordinates for the left and right hemispheres, respectively. Both sets of data that fulfilled the described criteria were available for all adults for both hemispheres, for 29 NT/28 ASD children for the left hemisphere and 33 NT/32 ASD children for the right hemisphere. Except for the analysis of MNI coordinates, all the other analyses of auditory responses were performed for the full sample of participants.
The MNI coordinates of the SF and the 40 Hz ITPC maxima in the three groups of participants are shown in Table 2 and visualized in Figure 3. In the NT children, in both hemispheres, the SF source was located anterior, lateral and inferior to that of the ASSR (paired ttest, left X: t(28) = 4.8, p = 0.00005; left Y: t(28) = 3.7, p = 0.0008; left Z: t(28) = 3.1, p = 0.004; right X: t(32) = 3.4, p = 0.002; right Y: t(32) = 5.8, p = 0.000002; right Z: t(32) = 3.9, p = 0.004). The same relative positions of the ASSR and SF sources in adult participants were previously described by Keceli and colleagues [20] using single dipole modeling. For comparison purposes, Table 2 gives original Talairach and estimated MNI coordinates of the SF and ASSR, reported by Keceli et al. [20]. In our adult sample, the SF maxima were located anterior and inferior to those of the ASSR in both hemispheres (paired ttest, left Y: t(9) = 3.0, p = 0.02; left Z: t(9) = 3.3, p = 0.01; right Y: t(9) = 2.4, p = 0.04; right Z: t(9) = 1.8, p = 0.1), while the lateral shift along the X axis was not significant (left X: t(9) = 1.4, p = 0.21; right X: t(9) = 1.2, p = 0.28), possibly because of the small sample size. In children with ASD, the differences in cortical localization between the SF and ASSR sources in the right hemisphere were in the same direction as in the NT children (right X: t(30) = 3.0, p = 0.005; right Y: t(30) = 2.9, p = 0.007; right Z: t(30) = 3.1, p = 0.004), while in the left hemisphere the lateral displacement of the SF source relative to the ASSR sources was not significant (left X: t(25) = 0.65, p = 0.5; left Y: t(25) = 3.3, p = 0.003; left Z: t(25) = 1.9, p = 0.07).
Table 2. Grand average MNI coordinates of the maximal auditory steadystate response (ASSR) and sustained field (SF) sources.
Group
(N left / N right)

Left hemisphere Mean and (SD)

Right hemisphere Mean and (SD)

X
(lateralmedial)

Y
(posterior anterior)

Z
(superiorinferior)

X
(mediallateral)

Y
(posterior anterior)

Z
(superiorinferior)

ASSR

NT adults (10/10)

48.7 (6.9)

27.4 (8.6)

9.7 (6.5)

51.4 (7.8)

25.2 (7.7)

10.1 (5.5)

NT children (31/33)

45.5 (6.0)

28.5 (8.8)

11.0 (6.8)

50.2 (6.0)

25.3 (6.6)

11.1 (4.3)

ASD children (30/32)

47.8 (8.0)

25.0 (9.4)

9.0 (6.8)

51.0 (8.4)

23.4 (8.5)

10.4 (4.5)

Keceli et al. 2015#
Adults, N=11
Estimated MNI
Original Talairach

48
45

22
22

11
13

52
48.5

17
16.5

13
14

SF

NT adults (10/10)

52.1 (7.7)

12.1 (10.7)

1.2 (6.5)

55.1 (6.8)

18.5 (8.3)

6.8 (6.2)

NT children (33/35)

51.9 (6.7)

20.5 (7.3)

6.6 (3.7)

53.7 (6.6)

17.2 (6.6)

7.4 (3.4)

ASD children (29/35)

47.1 (6.0)*

19.0 (10.4)

5.7 (7.4)

55.5 (6.2)

19.1 (6.5)

7.5 (4.3)

Keceli et al. 2015#
Adults, N=12
Estimated MNI
Original Talairach

51
48

17
18

7
8.5

52
49

13
14

6
9

# The provided coordinates are approximated from Figure 3 in [20], where the authors localised the ASSR and SF responses evoked by the same periodic stimuli [20].
* Significant difference in ASD vs NT children: p=0.005.
For the 40Hz ITPC maxima in the right or left hemispheres, there were no significant differences between children with and without ASD in either X, Y or Z coordinates (ttest, left X: t(59) = 1.3, p = 0.2; left Y: t(59) = 1.5, p = 0.14; left Z: t(59) = 1.2, p = 0.2; right X: t(63) = 0.5, p = 0.6; right Y: t(63) = 1.1, p = 0.3; right Z: t(63) = 0.6, p = 0.5). The SF maximum in the left hemisphere was located significantly more medial in children with ASD than that in NT children, although the difference was relatively small (X = coordinate in NT: 51.9, ASD: 47.1, F(1,61) = 8.5, Cohen’s d = 0.74, p = 0.005, uncorrected for multiple comparisons). The multivariate Hotelling T2 test confirmed significant ASD vs NT differences in the source localization (F(3,59) = 3.1, p = 0.03, partial etasquared = 0.14). There were no group differences for the SF coordinates in the right hemisphere.
To investigate whether the SF ‘early’ interval (150–250 ms), which was visible only in children, represents the evolving SF, we compared cortical locations of the SF maxima in this interval with those in the 300–500 ms interval, where the sustained field was observed in both children and adults. There were no significant timerelated differences in X, Y or Z SF coordinates in the NT, ASD or the combined sample of children in either hemisphere (paired ttest, all ps > 0.08, incorrected for multiplr comparisons). This means that the onset interval of the SF in children originates from the same region of the auditory cortex as the rest of the SF.
Comparison of the 40 Hz ASSR in NT children and children with ASD
To analyze group differences in ASSR magnitude, we computed for each participant an average of 40 Hz zITPC and power in the common ROI, separately in the right and left hemispheres (see Methods for details). We analyzed group differences in the ASSR parameters (power and zITPC) integrated over 200500 ms interval, as well as group differences in ASSR parameters timecourses over the whole period of the stimulation.
Grand averaged ASSR waveforms (filtered between 38 and 42 Hz; FIR filter order = 100) as well ITPC and power timefrequency plots for the left and right common ROIs are presented in Figure 4. One ASD participant had extremely high ASSR power (the percentage of power change in 200500 ms interval was more than 5 SDs and 4 SDs above the group mean in the right and left hemispheres, respectively) and was excluded from the analysis (See the Supplementary Figure S4 for distribution of the % power change values and Supplementary Figure S5 for the timefrequency plots with this subject included).
In both hemispheres, contralateral to the stimulated ear, the integrated zITPC values in the common ROI were not normally distributed in NT children (Shapiro–Wilk test, NNT = 35, left hemisphere: W = 0.94, p = 0.07; right hemisphere: W = 0.93, p = 0.02) and children with ASD (Shapiro–Wilk test, NASD = 34, left hemisphere: W = 0.87, p < 0.007; right hemisphere: W = 0.89, p = 0.002). Therefore, we used nonparametric statistical analysis. In both ASD and NT participants, the zITPC was greater in the right hemisphere than in the left one (Wilcoxon matched pairs test: NNT = 35, T=57, Z = 4.2, p = 0.00002; NASD = 34, T=226, Z = 3.1, p = 0.002). There were no group differences in the zITPC for either hemisphere (Mann–Whitney U test, NNT = 35, NASD = 34; left hemisphere: ASD median = 0.63, NT median = 0.77, U = 526, p = 0.4; right hemisphere: ASD median = 0.95, NT median = 1.57, U = 477, p = 0.16). Inspection of the distributions of the zITPC values corrected for age showed a large overlap between the NT and ASD groups (Supplementary Figure S6). The findings for the integrated ASSR power in the common ROI mirrored those for zITPC and are given in the Supplementary Results.
We repeated the analysis for 30 ‘maximal vertexes’, selected individually for each subject. In this case, the zITPC distributions also deviated from normal in NT children (Shapiro–Wilk test, NNT = 35, left hemisphere: W = 0.92, p = 0.02; right hemisphere: W = 0.93, p = 0.04) and in children with ASD (Shapiro–Wilk test, NASD = 34, left hemisphere: W = 0.87, p = 0.001; right hemisphere: W = 0.88, p = 0.001). In both ASD and NT participants, the zITPC was greater in the right than in the left hemisphere (Wilcoxon matched pairs test: NNT = 35, T=71, Z = 4.0, p = 0.00006; ASD: NASD = 34, T=140, Z = 2.7, p = 0.007). There were no group differences in the zITPC in the individual maxima (Mann–Whitney test, NNT = 35, NASD = 34; left hemisphere: ASD median = 0.91, NT median = 0.95, U = 521, p = 0.38; right hemisphere: ASD median = 1.33, NT median = 1.86, U = 451, p = 0.09). As with the common ROI, the analysis of the integrated ASSR power in the individual maxima did not reveal any significant group differences (see Supplementary Results).
Figure 5 shows grand average timecourses of the zITPC and ASSR power in NT and ASD participants in the left and right hemispheres contralateral and ipsilateral to the stimulated ear. The ASD participant with an extremely high ASSR power (see above) was excluded from this analysis (see the Supplementary Figure S7 for the plot with this subject included). There were no statistically significant group differences in any of the ASSR timecourses at any time point (Wilcoxon matched pairs test, all FDR corrected ps > 0.05).
Figure 6 shows individual 40 Hz ITPC values in children with and without ASD as a function of age. The linear trends suggesting the agerelated increase of the 40 Hz ITPC were present in both groups and in both hemispheres, contralateral to the stimulated ear. A developmental increase in the 40 Hz ITPC was confirmed by analysis of the zITPC values (Spearman correlation coefficients: NASD = 35, left hemisphere: R = 0.30, p = 0.08; right hemisphere: R = 0.38, p = 0.02; NNT = 35, left hemisphere: R = 0.31, p = 0.08; right hemisphere: R = 0.34, p = 0.048). However, even at younger ages (< 9 years), the majority of children had 40 Hz ITPC values above their baseline level.
To investigate whether the 40 Hz zITPC in children with ASD correlated with their intelligence level and severity of autism, we calculated Spearman correlations. Children with more severe autism had higher 40 Hz zITPC values in the right hemisphere (Table 3). We repeated the analysis for the zITPC values after subtracting the linear age trend. This procedure increased the reliability of this result (Spearman R = 0.46, p = 0.007).
Table 3. Spearman correlation between the 40 Hz auditory steadystate response (ASSR) intertrial phase coherence (ITPC) and psychometric variables in children with autism spectrum disorder (ASD).

MPI IQ (N=32)

Autism Score (N=34)

Left hemisphere

0.15, p=0.41

0.16, p=0.35

Right hemisphere

0.17, p=0.33

0.38, p=0.026*

* Uncorrected for multiple comparisons
Comparison of the SF in children with and without ASD
As in case of the ASSR, betweengroup comparisons for the SF timecourses were performed in the source space (see Methods for details). Figure 7 shows the grand average lowpassed SF source waveforms in the left and the right hemispheres for the contra and ipsilateral ear stimulation in the ASD and the NT groups. Visually detectable SF source waveforms were present in the right hemisphere in all participants and in the left hemisphere in the majority of children from both samples, with the exception of two NT and six ASD children.
First, we analyzed the effect of age on the SF maximal amplitude using Pearson correlations. None of the correlations were significant in the NT, ASD or in the combined sample (all ps > 0.2; see Supplementary Table 1) suggesting that the SF amplitude did not change between 7 and 12 years of age.
We then examined the effects of hemisphere and contralaterality of the stimulation on the SF maximal amplitude in the NT and ASD groups. To this end, we used rmANOVA with group (ASD, NT), hemisphere (left, right) and ear of stimulation (contralateral, ipsilateral) as the factors and the maximal amplitude of the SF source current in the 150–500 ms stimulation interval as a dependent variable. There were strong effects of hemisphere (F(1,68) = 50.5, p < 0.00001, partial etasquared = 0.43) and ear of stimulation (F(1,68) = 128.4, p < 0.00001, partial etasquared = 0.65). The SF maximal amplitude was higher in response to the contra than ipsilateral stimulation, and it was generally higher in the right than in the left hemisphere (Figures 7, 8). Neither group × hemisphere, nor group × ear interaction effects were significant (both uncorrected ps > 0.1), suggesting that the auditory SF response in both groups was characterized by contralaterality and right hemisphere dominance. The SF maximal amplitude was higher in the NT compared with the ASD group (group main effect: F(1,68) = 4.6, p = 0.035, partial etasquared = 0.06). This means that the SF in children with ASD was reduced compared with NT controls in both hemispheres and in response to both contra and ipsilateral ear stimulation. Analysis of the pointbypoint group differences in the SF timecourses (Wilcoxon matched pairs test, p < 0.01, FDR corrected for multiple comparisons) revealed that the SF in children with ASD was particularly strongly attenuated in the left hemisphere approximately between 150 and 220 ms after the click train onset, suggesting an abnormally slow rise of the SF source strength in children with ASD (Figure 7).
For the further analysis of betweengroup difference in the SF source timecourse, we focused on the contralateral responses that were greater and more reliable than the ipsilateral ones (Figure 7). To test for the group differences in the SF timecourses, we divided the SF time window (150–550 ms) into four successive time intervals in respect to the click train onset (151–250, 251–350, 351–450 and 451–550 ms), with the first (151–250 ms) interval corresponding to the raising part of the SF. We then performed rmANOVA with time, hemisphere and group as the factors and the average SF source current amplitudes in the 100 ms time intervals as a dependent variable (Table 4, Figure 7).
Table 4. Group differences in the SF source timecourses: repeated measures ANOVA results.

F

p

GG epsilon

partial
etasquared

Group

5.9

0.018


0.08

Time

32.4

<1e6

0.57

0.32

Time x GR

3.3

0.04

0.57

0.05

Hemisphere

43.4

<1e6


0.39

Hem x GR

0.04

0.84


0.00

Time x Hem

9.2

0.00001

0.71

0.12

Time x Hem x GR

3.7

0.024

0.71

0.05

Children with ASD had generally lower SF source amplitude than NT children, especially during the rising part of the SF timecourse – from 150 to 250 ms after stimulation onset (Figure 8; time × group interaction: F(3,204) = 3.3, adjusted p = 0.04). The significant time × hemisphere × group interaction (F(3,204) = 3.7, adjusted p = 0.024) was due to a greater and more reliable reduction of the ‘early’ SF segment in the left hemisphere than in the right one in children with ASD (ASD vs NT left hemisphere: Bonferronicorrected p < 0.001; right hemisphere: n.s.). Except for the ‘early’ SF amplitude in the left hemisphere, none of the betweengroup differences survived after Bonferroni correction for multiple comparisons. This result is consistent with the data presented in Figure 7. The effect size (Cohen’s d = 1.01) for the difference between ASD and NT groups in the SF150–250 amplitude was classified as a large effect based on benchmarks suggested by Cohen (in [61]). To check whether the group differences could be explained by a delayed maturation of the SF in ASD children, we calculated Pearson correlations between age and the averaged SF source amplitudes in the four time intervals after the stimulation onset in the left and right auditory cortices. There were no significant correlations with age for the SF source amplitudes in either the NT, ASD or combined groups (all uncorrected ps > 0.2; see Supplementary Table 1).
To test whether the group differences in the SF were due to lower IQ in ASD children, we repeated the ANOVA for the subsample of 10 ASD and 10 NT participants who were matched both for IQ level (MPI standard score: NT = 109.7 ± 10.2; ASD = 109.4 ± 11.9) and age (age in years, mean ± SD: NT = 10.0 ± 1.4, ASD = 9.2 ± 1.5). Due to a small number of participants, we used the nonparametric Mann–Whitney U test for betweengroup comparisons. The significant difference between ASD vs NT groups was again detected only in the early time interval in the left hemisphere (SF150250, Left hemisphere, ASD median = 2.55 (a.u.), NT median = 4.94 (a.u.), U=16, Z=2.53; P=0.01; see Supplementary Figure S8). The results suggest that the slowing of neural activation in the left auditory cortex evoked by click trains characterized ASD children irrespective of the presence of cognitive difficulties.
Considering that there were small, but significant group differences in the SF localization in the left hemisphere (Table 2), we wanted to ensure that the betweengroup differences in the SF source amplitude were not driven by the choice of the SF vertices for the group analysis, which was done by averaging across the NT and ASD data (see Methods for details). For this purpose, we repeated the rmANOVA analysis for the SF source amplitude calculated in the individually chosen 30 vertices with maximal SF amplitudes. The results remained principally the same (GR: F(1,68) = 3.7, p = 0.057, partial etasquared = 0.05; time × group: F(3,204) = 3.2, adjusted p = 0.047, partial etasquared = 0.05; time × hemisphere × group: F(3,204) = 5.6, adjusted p = 0.003, partial etasquared = 0.08). The lefthemispheric reduction of the SF amplitude in the 150–250 ms range remained significant: F(1,68) = 11.2, Bonferroni corrected p < 0.05.
To summarize, in both the NT and ASD groups, the SF source strength was higher contralaterally to the stimulated ear and clearly dominated in the right hemisphere. In the ASD group, the SF was moderately reduced in both hemispheres and strongly delayed in the left one, irrespective of the stimulated ear. None of the SF parameters significantly changed with age between 7 and 12 years in either the NT or ASD group.
We performed additional correlation analysis to investigate whether the two principle findings – the bilateral reduction of the SF maximal amplitude and the lefthemispheric SF reduction in its 150–250 ms early interval (‘SF delay’) in children with ASD – are associated with psychometric variables (Table 5). Because children with lower MPI IQ scores had more severe autism (Pearson R = 0.36, p = 0.049), to estimate independent associations of these psychometric variables with the SF amplitude, we calculated partial correlations. We performed this analysis in 31 children with ASD in whom both MPI IQ and Autism Scores were available (Table 6). Although the maximal SF source amplitude (SFmax) was decreased (less negative) in children with ASD at the group level, its greater strength in the right hemisphere correlated with greater severity of autism traits (R = 0.64, Bonferronicorrected p < 0.001). The SFmax amplitude in both hemispheres also tended to be higher in the ASD participants with higher IQ, but these correlations did not survive after Bonferroni correction. Thus, the results of the correlation analysis mainly indicate that the higher SF amplitude in the right hemisphere in children with ASD is associated with greater severity of their autism symptoms.
Table 5. Partial correlations between the sustained field (SF) source current* and psychometric variables in children with ASD (N=31).
SF amplitude

MPI IQ: r, p**

Autism Score

SFmax, Left hemisphere

0.35*, p=0.06

0.22, p=0.25

SFmax, Right hemisphere

0.43, p=0.017

0.64, p=0.0001

SF150250, Left hemisphere

0.06, p=0.74

0.04, p=0.85

SF150250, Right hemisphere

0.14, p=0.47

0.41, p=0.026

* Note that the SF current is negative, and a negative correlation reflects a direct link between the SF strength and a respective psychometric variable.
** Uncorrected for multiple comparisons; p value that remained significant after Bonferroni correction is shown in bold type.