Relationships between GABA, Glutamate, and GABA/Glutamate and Social and Olfactory Processing in Children with Autism Spectrum Disorder

Abstract

Theories of altered inhibitory/excitatory signaling in autism spectrum disorder (ASD) suggest that gamma amino butyric acid (GABA) and glutamate (Glu) abnormalities may underlie social and sensory challenges in ASD. Magnetic resonance spectroscopy was used to measure Glu and GABA+ levels in the amygdala-hippocampus region and cerebellum in autistic children (n=30), a clinical control group with sensory abnormalities (SA) but not ASD (n=30), and children with typical development (n=37). All participants were clinically assessed using the Autism Diagnostic Interview-Revised, the Autism Diagnostic Observation Scale-2, and the Child Sensory Profile-2. The Social Responsiveness Scale-2, Sniffin Sticks Threshold Test, and the University of Pennsylvania Smell Identification Test were administered to assess social impairment and olfactory processing. Overall, autistic children showed increased cerebellar Glu levels compared to TYP children. Evidence for altered excitatory/inhibitory signaling in the cerebellum was more clear-cut when analyses were restricted to male participants. Further, lower cerebellar GABA+/Glu ratios were correlated to more severe social impairment in both autistic and SA males, suggesting that the cerebellum may play a transdiagnostic role in social impairment. Future studies of inhibitory/excitatory neural markers, powered to investigate the role of sex, may aid in parsing out disorder-specific neurochemical profiles.

1. Introduction

Atypical responses to sensory stimuli are frequently associated with the neurodiversity observed in autism spectrum disorder (ASD) (Leekam et al., 2007),(Tomchek and Dunn, 2007) and can impose lifelong challenges to autistic individuals (Suarez, 2012). Sensory processing symptoms associated with ASD can be very severe and, although heterogeneous in their presentation (Fernández-Andrés et al., 2018),(Tillmann et al., 2020), are predictive of both social dysfunction (St. John et al., 2022),(Sweigert et al., 2020) and repetitive behaviors (Sweigert et al., 2020). Although sensory differences have not been as well characterized as social impairments, atypical sensory processing has been observed as young as 9 months of age in infants who later develop ASD (Gliga et al., 2015) suggesting that these may emerge concurrently with other social impairment and restricted and repetitive symptoms. Atypical responses to sensory features such as smell, texture, color, temperature, and sound can contribute to food selectivity (Chistol et al., 2018), avoidance of certain play materials, refusal to participate in family activities (Reynolds et al., 2011), and academic underachievement (Ashburner et al., 2008).

Sensory features were reported in the earliest descriptions of autism (Kanner, 1943), yet physiological studies of basic sensory processing in ASD have yielded mixed results (See (Robertson and Baron-Cohen, 2017) for review). Frequently cited studies have found evidence of sensory enhancement, such as enhanced pitch sensitivity and auditory discrimination (Bonnel et al., 2003), but also, for example, reduced speech-specific tone perception in Mandarin-speaking autistic children and reduced auditory attention deployment in young autistic adults (Wang et al., 2017),(Emmons et al., 2022). Studies of tactile processing have found increased sensitivity to vibration (Cascio et al., 2008) and pain (Cascio et al., 2008),(Riquelme et al., 2016) abnormal tactile perceptual thresholds (Blakemore et al., 2006),(Puts et al., 2014), and higher intensity ratings (Blakemore et al., 2006),(Haigh et al., 2016), but no differences in detecting light pressure on the skin (Cascio et al., 2008). Olfactory testing has shown that odor detection thresholds are not impacted in autistic children but that odor identification may be impaired (Sweigert et al., 2020),(Bennetto et al., 2007) and related to increased ASD severity (Bennetto et al., 2007),(Lane et al., 2010),(May et al., 2011).

A growing body of literature supports the theory that sensory, social, and emotional systems in ASD may be impacted by an inhibition/excitation (I/E) imbalance (Rubenstein and Merzenich, 2003), specifically between glutamate (Glu) and gamma amino butyric acid (GABA). Research over the past decades has provided compelling evidence that differences in both GABA and Glu systems are present in autism. GABA, the most abundant inhibitory neurotransmitter, appears early in development and is critical for synapse and network formation. Genetic variations at loci encoding for GABA receptor subtypes have been widely reported (Buxbaum et al., 2002),(McCauley et al., 2004) and reduction of GABA receptor proteins and binding (Mendez et al., 2013),(Blatt et al., 2001),(Fatemi et al., 2011) lower GABA concentration in the supplementary motor area (SMA) (Umesawa et al., 2020) sensory hyperactivity with GABAergic interneuron dysfunction (Chen et al., 2020), and associations of GABA/Glu ratio with neuroinflammation, ASD symptoms, and functional connectivity (El-Ansary and Al-Ayadhi, 2014),(Zhang et al., 2020),(Hegarty et al., 2018) demonstrate functional differences.. Evidence for alterations in the Glutamatergic system include atypical levels of glutamate receptors and associated functions (Fatemi et al., 2011),(Fatemi et al., 2018),(Purcell et al., 2001) elevated glutamate concentrations in amygdala-hippocampal and parietal regions (Page et al., 2006), Glu modulation differences (Trobiani et al., 2018), and abnormalities in enzymes responsible for converting Glu to GABA (Yip et al., 2007),(Yip et al., 2008).

This study measured regional brain GABA+, Glu, and GABA+/Glu ratios in autistic children, compared to groups of age and sex-matched children having sensory abnormalities (SA) or neurotypical development (TYP). In addition, we tested whether odor detection, odor identification, and social impairment were related to an imbalance of inhibitory/excitatory brain circuitry in 1) the amygdala-hippocampal region, a neural hub for both socioemotional and primary olfactory processing (Soudry et al., 2011) that is altered in ASD and, 2) in the cerebellum, which supports a range of social behaviors (Stoodley and Tsai, 2021) develops atypically in ASD (Fetit et al., 2021), (Fatemi et al., 2012) and has been implicated in altered sensory processing timing in subgroups of ASD children (Welsh et al., 2023).

2. Methods

We enrolled 151 children ages 8–13 following an initial telephone screening to determine eligibility and group assignment. Each participant was evaluated by a licensed psychologist (TSJ) or by a psychology graduate student under the supervision of TSJ in order to verify group assignment. Clinical best estimate diagnosis of ASD was determined using DSM-5 ASD criteria (American Psychiatric, 2013), the Autism Diagnostic Interview-Revised (ADI-R; (Lord et al., 1994)), and the Autism Diagnostic Observation Schedule, 2nd Edition (ADOS-2; (Lord et al., 2012)). Children in the SA group were characterized by the presence of clinically significant sensory symptoms, without meeting diagnostic criteria for ASD. The sensory criterion was operationalized by requiring participants to have at least one elevated quadrant on the Child Sensory Profile, Second Edition (CSP-2; (Dunn, 2014))(Seeking/Seeker, Avoiding/Avoider, Sensitivity/Sensor, Registration/Bystander), defined as 2 or more standard deviations above the mean. All children in the ASD group met criteria for ASD. No children in the TYP or SA groups met criteria for ASD. Children in the TYP, SA and ASD groups were administered the Kiddie Schedule for Affective Disorders and Schizophrenia (K-SADS; (Kaufman et al., 1997)) to detect other neurodevelopmental or psychiatric conditions. All children were administered the Wechsler Abbreviated Scale of Intelligence (WASI; (Wechsler, 1999)) to measure IQ. Adaptive functioning, a collection of pragmatic abilities related to communication, social skills, and self-care, was measured using the Vineland Adaptive Behavior Scales, Second Edition parent-report form (Vineland-II; (Sparrow et al., 2012)). The Vineland-II yields four domain standard scores: Communication, Daily Living Skills, Socialization, and Motor Skills, and one overall composite score, the Adaptive Behavior Composite. Higher scores indicate better adaptive functioning in the measured domains.

Children from all groups were excluded if the measured FSIQ fell below 70. MRI contraindications and a prior diagnosis and/or indication of severe hyposmia were also exclusionary. Exclusion criteria for ASD and SA included histories of psychotic disorders or inherited genetic disorders (fragile x syndrome, PKU, and Tuberous Sclerosis). SA group exclusion criteria additionally included a history of ASD in the child or a first-degree relative. Children were excluded from the TYP group if diagnosed with attention deficit/hyperactivity disorder (ADHD), another developmental disorder or psychiatric disorder, history of ASD in the child or a first-degree relative, or any elevated score on the CSP-2. Eighteen participants were excluded following enrollment based on the above criteria. Exclusion events included IQ < 70 (2 ASD, 2 SA), not meeting ASD criteria (4 ASD) or sensory processing profile criteria (3 SA, 1 TYP) at the time of evaluation, no show (1 SA, 1 ASD), incomplete participation due to receptive language deficits (1 ASD), presence of ADHD symptoms (2 TYP), and clinically significant olfactory dysfunction (1 TYP). Additional participants were excluded following MRI quality assurance procedures (described below). Demographic characteristics of participants with a minimum of one useable magnetic resonance spectroscopy (MRS) data set are presented in Table 1.

Table 1.

Participant Characteristics

	ASD (n=30)		SA (n=30)		TYP (n=37)		ANOVA		ASD v SA	ASD v TYP	SA v TYP
	M	(SD)	M	(SD)	M	(SD)	F	p-value	p-value	p-value	p-value
Male and Female
Sex (Male:Female)	26:4		25:5		30:7
Race (White:Non-White)	25:5		24:6		27:10
Age at MRI (years)	10.48	(1.71)	9.86	(1.62)	10.24	(1.47)	1.15	.321	.156	.548	.316
WASI FSIQ	113.43	(16.51)	117.87	(18.70)	118.30	(11.49)	0.94	.395	.334	.161	.908
WASI PIQ	113.90	(13.89)	113.50	(17.10)	113.97	(13.13)	0.01	.991	.921	.982	.899
WASI VIQ	110.13	(19.44)	118.07	(18.67)	118.62	(12.16)	2.52	.086	.112	.033	.884
ADOS-2 Total Score	11.27	(3.63)	5.77	(3.62)	2.11	(2.01)	72.24	<.001	<.001	<.001	<.001
ADOS-2 Social Affect	9.20	(3.46)	4.93	(3.67)	1.81	(1.79)	49.91	<.001	<.001	<.001	<.001
ADOS-2 RRB	2.07	(1.11)	0.83	(0.75)	0.30	(0.70)	35.86	<.001	<.001	<.001	<.001
ADI-R Communication	14.00	(5.29)	6.40	(4.61)	2.16	(2.92)	62.69	<.001	<.001	<.001	<.001
ADI-R RRB	5.59	(2.43)	3.27	(1.84)	0.51	(0.77)	69.39	<.001	<.001	<.001	<.001
ADI-R Social	15.83	(6.72)	9.47	(6.04)	2.16	(2.18)	57.29	<.001	<.001	<.001	<.001
CSP-2 Seeking/Seeker	41.17	(16.13)	48.77	(14.93)	22.95	(7.68)	34.86	<.001	.063	<.001	<.001
CSP-2 Avoiding/Avoider	55.83	(14.56)	58.07	(16.12)	26.73	(7.24)	63.35	<.001	.575	<.001	<.001
CSP-2 Sensitivity/Sensor CSP-2	51.53	(13.80)	52.57	(13.65)	22.05	(6.96)	76.44	<.001	.772	<.001	<.001
Registration/Bystander	52.47	(15.15)	53.67	(17.32)	25.19	(9.00)	45.80	<.001	.776	<.001	<.001
Vineland ABC	81.00	(11.29)	86.79	(12.15)	106.92	(12.17)	43.20	<.001	.068	<.001	<.001
Vineland Communication	84.40	(11.91)	89.55	(11.38)	106.89	(12.03)	33.40	<.001	.095	<.001	<.001
Vineland Daily Living Skills	85.52	(13.98)	92.48	(15.24)	106.22	(12.22)	19.29	<.001	.075	<.001	<.001
Vineland Socialization	70.93	(30.04)	77.17	(34.35)	137.46	(40.33)	36.60	<.001	.457	<.001	<.001
Male only
ADOS-2 Social Affect CSS	6.96	(1.80)	4.12	(2.55)	2.23	(1.22)	43.45	<.001	<.001	<.001	.001
Social Responsiveness Scale-2	72.42	(9.65)	68.48	(12.01)	43.60	(5.01)	82.91	<.001	.201	<.001	<.001
UPSIT (% correct)	65.48	(15.69)	68.20	(11.85)	76.00	(8.34)	5.71	.005	.487	.002	.006
Threshold Test - Roses	7.48	(4.23)	7.41	(4.20)	9.18	(3.68)	1.75	.181	.954	.119	.101
Threshold Test - Vanillin	6.43	(2.27)	6.52	(3.18)	7.47	(2.90)	1.16	.319	.910	.157	.129

Notes. WASI = Wechsler Abbreviated Scale of Intelligence, ADOS = Autism Diagnostic Observation Scale, ADI-R = Autism Diagnostic Interview - Revised, CSP-2 = Child Sensory Profile - 2, UPSIT = University of Pennsylvania Smell Identification Test

Participants were recruited from the Seattle metropolitan area via flyers, local online groups, and utilizing a research registry at the University of Washington. There were no group differences in age, sex, race, or IQ scores measured by the WASI.

All TYP children were unmedicated. Prescription medications reported by ASD and SA participants with valid MRI data included selective serotonin reuptake inhibitors (SSRI; n=10), serotonin and norepinephrine reuptake inhibitors (SNRI; n=1), stimulants (n=14), atypical antipsychotics (n=4), anticonvulsants (n=3), and central alpha-2 agonists (n=13). Of the ASD children, 14/30 took 1 or more of the medications listed above. In the SA group, 13/30 took 1 or more of the medications listed above. See Supplementary Table 1 for additional detail.

All study procedures were approved by the University of Washington Human Subjects Division Institutional Review Board. Written informed consent from the parent and verbal assent were obtained from each participant.

2.1. Measures

2.1.1. Social Functioning

The Social Responsiveness Scale, 2nd Edition (SRS-2) is a 65-item parent-report measure that assesses ASD specific impairments in reciprocal social behavior and restricted and repetitive behaviors (Constantino and Gruber, 2012). T-scores are provided by the SRS-2 (M=50, SD=10). Higher scores suggest greater impairment in reciprocal social behavior. ADOS-2 social affect calibrated severity score (ADOS-2 SA CSS) is a measure of social-communication severity. Calibrated severity scores are less influenced by child characteristics such as language level and age than the ADOS-2 total scores that are used to provide cut-offs for autism classification purposes (Hus et al., 2014).

2.1.2. Olfactory Processing

Olfactory processing was selected as our physiological measure because it is unique among sensory processes in that olfactory stimulation directly activates amygdala neurons without the thalamus providing a relay function (Zald and Pardo, 1997). Olfactory testing was conducted to assess odor detection and odor identification. Participants were screened prior to olfactory testing for sources of nasal congestion including allergies and active rhinitis, and only allowed to consume water for the 30 minutes prior to testing. No additional odors were present in the testing room and ventilation was maintained.

Odor detection thresholds were measured using the Sniffin’ Sticks Threshold Test with two odorants: phenylethyl alcohol (Roses) and vanillin (Vanilla). Participants were blindfolded and asked to indicate when a specified odor was detected, beginning from the lowest concentration Sniffin’ Sticks pen and increasing with each trial. These odorants have minimal trigeminal effects and are generally considered pure olfactory stimulants and ideal for determining olfactory detection thresholds (Hädel et al., 2013). One ASD participant’s data were excluded because they provided the same response for every trial.

Odor identification was evaluated using a 40-item, multiple choice test called the University of Pennsylvania Smell Identification Test (UPSIT; (Doty et al., 1984)). This “scratch and sniff” test contains microcapsules on each page with odors that are released upon scratching. Each participant was asked to scratch the microcapsule, and then identify the target odor from the list of 4 potential odors. Supra-threshold concentrations of each odorant are contained in the microcapsules to discriminate odor identification performance from odor detection thresholds.

2.2. MRS/MRI Data acquisition

2.2.1. MRS

All MRS data were acquired on a Philips 3T Achieva 5.1.7 scanner using a 32-channel head coil. An MPRAGE [magnetization – prepared rapid gradient – echo], T₁–weighted anatomical image was acquired for voxel placement and segmentation using 3.6 msec echoes to prepare 1 mm³, isotropic images with effective TR = 7.9 msec. In order to maximize magnetic field homogeneity, allowing for gradient field settling, scans were scheduled at least an hour after rapid imaging sequences (e.g., echo planar imaging) were collected. Spectra were acquired in each of two 30×30×30 mm³ voxels, one centered on the left amygdala-hippocampal region and the other located in the left cerebellar hemisphere. See Figure 1 for voxel placement. A standard shim procedure based upon FASTMAP was applied to optimize field homogeneity for the individually localized voxels. For measurement of glutamate, a PRESS sequence with TE set to 80 (Schubert et al., 2004) and TR 2000 msec, was used to acquire 64 averages of 2000 Hz bandwidth into 2048 points. PRESS data were acquired with and without water suppression. GABA-edited spectra were acquired using a MEGA-PRESS pulse sequence (Mescher et al., 1998). MEGA-PRESS localization was achieved with minimum–phase amplitude–modulated excitation pulses (2-kHz bandwidth) and amplitude–modulated refocusing pulses (1.3 kHz bandwidth) (Mullins et al., 2014). Spectra were edited with 14 msec, sinc–Gaussian pulses applied at 1.9 ppm in the ‘on’ experiment and 7.46 ppm in the ‘off’ experiment. This editing scheme co-edits approximately 50% macromolecules at 3 ppm, which are coupled to inverted spins at 1.7 ppm. All measured GABA values reported here thus reflect GABA+ macromolecules. Additional MEGA-PRESS acquisition parameters were: TR/TE of 2 s/68 ms; 320 transients with on – off scans alternating every 2 transients; a 16-step phase cycle (with steps repeated for on and off); 2048 datapoints acquired at a spectral width of 2 kHz; and variable pulse power and optimized relaxation delays (VAPOR) water suppression (Harris et al., 2015). Sixteen transients of water data were also acquired for quantification using the same acquisition variables, without suppression.

Figure 1.

Example voxel placement for the cerebellum (A) and amygdala-hippocampal region (B). LCModel output for PRESS spectrum at TE = 80 (C). The black line is the acquired MRS data and red line is the LCModel fit. The prominent glutamate peak, centered around 2.35 ppm, is indicated by an arrow. Example spectra of fitted GABA+ (D), water and creatine (E). The blue line is the acquired data and the red line is the model fit.

2.2.2. MRS data processing

PRESS data were processed with the LCModel software package (version 6.3–1B; (Provencher, 1993)). No filtering was applied to the raw FID signals. Quantification of glutamate concentrations was performed by referencing the water-suppressed glutamate signal to the unsuppressed water signal.

GABA spectra were analyzed in the Gannet 2.0 toolbox (Edden et al., 2014) within MATLAB 2017a using the toolbox-standard approach, including automated frequency and phase correction, artifact rejection (frequency correction >3 SD above the mean), and 3 Hz exponential line broadening. To calculate the concentration of GABA+, a Gaussian was fit to the peak at 3 ppm. The area under the fitted curve served was scaled relative to water as the measure for metabolite concentration; the unsuppressed water peak was fit with a mixed Gaussian – Lorentzian.

All glutamate and GABA+ values were corrected for CSF partial volume within each voxel. CSF partial volume were calculated from CSF maps produced through segmentation of the T1-weighted anatomical scan using FSL FAST (FMRIB’s Automatic Segmentation Tool).

Concentrations for GABA+ and Glu are reported in institutional units (i.u.) standardized to the unsuppressed water peak. See Figure 1 for example GABA+ and glutamate spectra.

2.2.3. MRS Quality Assurance Procedures

Success of acquisition and of editing data was assessed by visual inspection of both trends and jumps in frequency traces of each acquisition. Criteria for rejecting data provided by load function was used to assess both movement and movement compensation. Greater than 9 load rejects and/or large jumps in the signal were taken to denote excess movement. Similarly, data seen to possess large excursions such as drifts, resets, steps, spikes, and Cr chemical shift, especially if poorly realigned [Δδ> 0.03 ppm], were eliminated. Results of fitting functions likewise provided spectral quality and fit information. MEGA-PRESS data were rejected if the Gannet signal fit error was greater than 10% for spectra (as in (Wang et al., 2016; Liu et al., 2017)). Li ewise R SS data were excluded if the Cram r-Rao lower bound for glutamate provided by LCModel was greater than 15%, or if the LCModel-processed spectrum had a FWHM greater than 0.1. Following quality assurance procedures, 41 amygdala-hippocampal data sets (ASD=12, SA=13, TYP=16) and 46 cerebellar data sets were excluded (ASD=10, SA=21, TYP=15).

2.3. Statistical Analyses

All statistical analyses were completed using SPSS Version 27. Between-group differences in the descriptive statistics reported in Table 1 and GABA+, Glu, and GABA+/Glu levels were analyzed with one-way ANOVA (for differences between the three groups) and independent-sample t-tests for between-group differences. Because sex-differences are known to impact the GABAergic system (Spurny-Dworak et al., 2022),(De Bondt et al., 2015),(Epperson et al., 2002) and increasing evidence of sex-based differences in neuroimaging findings in ASD (Walsh et al., 2021), secondary analyses limited to male participants were conducted to assess group differences. Too few female participants were enrolled to assess separately or to include sex as an independent variable in the statistical model. (See Supplementary Table 2 for descriptive statistics of the female participants)

Pearson correlation analyses were conducted with the male-only ASD and SA groups separately to explore the relationships between social impairment, olfactory processing and GABA+/Glu imbalance. Follow-up multiple linear regression analyses were conducted to assess the effect of diagnosis (ASD-Male, SA-Male) on the relationships between social impairments (ADOS Social CSS, SRS), olfactory processing (UPSIT, Threshold Test) and GABA+/Glu. These interaction effects were tested using models that included group (ASD-Male, SA-Male), behavioral measure (ADOS Social CSS, SRS-2, UPSIT, or Threshold Tests), and the group by behavioral measure interaction term.

3. Results

3.1. MR Spectroscopy – Amygdala-Hippocampal and Cerebellar Regions

No diagnostic group differences in GABA+, Glu, or GABA+/Glu were found in the amygdala-hippocampal region. In the cerebellar region, the ASD group demonstrated significantly elevated Glu compared to the TYP group (ASD M = 14.28, SD = .99, TYP M = 13.60, SD = 1.19, p < .05). No other diagnostic group differences were found for GABA+, Glu, or GABA+/Glu in the cerebellum. See Table 2.

Table 2.

Group Differences in MRS metabolite concentrations

	ASD		SA		TYP		ANOVA		ASD v TYP	ASD v SA	SA v TYP
	M	(SD)	M	(SD)	M	(SD)	F	p-value	p-value	p-value	p-value
All
Amyg-Hippo	n=25		n=25		n=29
GABA+	1.24	(0.27)	1.16	(0.21)	1.21	(0.39)	0.44	.647	.727	.244	.574
Glu	13.12	(1.53)	13.11	(1.89)	12.54	(1.36)	1.21	.303	.143	.984	.200
GABA+/Glu	0.095	(0.02)	0.090	(0.02)	0.097	(0.03)	0.60	.553	.820	.337	.330
Cerebellum	n=23		n=16		n=25
GABA+	1.58	(0.21)	1.63	(0.11)	1.59	(0.18)	0.34	.714	.897	.389	.451
Glu	14.28	(0.99)	13.97	(1.20)	13.60	(119)	2.19	.121	.038	.381	.343
GABA+/Glu	0.111	(0.02)	0.117	(0.01)	0.117	(0.02)	1.17	.318	.178	.217	.952
Male only
Amyg-Hippo	n=21		n=23		n=24
GABA+	1.28	(0.25)	1.15	(0.18)	1.17	(0.36)	1.41	.251	.252	.048	.792
Glu	13.14	(1.58)	13.02	(1.94)	12.41	(1.36)	1.31	.303	.103	.823	.217
GABA+/Glu	0.095	(0.02)	0.090	(0.02)	0.095	(0.03)	0.76	.471	.675	.134	.469
Cerebellum	n=20		n=11		n=19
GABA+	1.56	(0.21)	1.67	(0.09)	1.60	(0.19)	1.16	.320	.531	.125	.311
Glu	14.33	(0.99)	13.54	(0.89)	13.31	(1.17)	4.82	.012	.007	.042	.579
GABA+/Glu	0.109	(0.02)	0.124	(0.01)	0.121	(0.02)	3.81	.029	.037	.015	.687

Note: GABA+ = gamma amino butyric acid + macromolecules; Glu = glutamate; Amyg-Hippo = amygdala-hippocampal region

Secondary analyses were conducted for male participants. ASD males had significantly higher levels of amygdala-hippocampal GABA+ than the SA males (ASD M = 1.24, SD = 0.27; SA M = 1.15, SD = 0.18, p = .048). No other sex-based diagnostic group differences were observed in the amygdala-hippocampus region. In the left, lateral cerebellar hemisphere, ASD males had significantly higher Glu than both the TYP males and the SA males, and significantly lower GABA+/Glu ratios than both the TYP and SA males. See Table 2.

3.2. Correlations between GABA+/Glu and Social and Olfactory Processing in Male ASD and SA Participants

We tested the correlations between the ratio of GABA+/Glu concentrations and social impairment (ADOS-2 SA CSS and the SRS), odor identification, and odor detection thresholds in male ASD and SA participants. See Table 3. Neither the ASD nor SA males showed a significant relationship between amygdala-hippocampal GABA+/Glu and the measures of social impairment. However, a greater imbalance in cerebellar GABA+/Glu was correlated with more severe social impairment as measured by the ADOS-2 SA CSS (r = −.461, p = .04) in the ASD males. The male SA group showed a significant correlation between cerebellar GABA+/Glu and the SRS-2 Total Score (r=−.625, p = .04). In addition, the ASD males showed a negative correlation between smell detection for the vanillin odorant and amygdala-hippocampal GABA+/Glu (r = −.570, p = .011). See Figure 2.

Table 3.

Correlations between neurometabolites and psychological measures in male participants

	ADOS-2	SRS-2	UPSIT	Threshold Test
Group	Social CSS	Total Score	% correct	Roses	Vanillin
ASD
Amygdala-hippo a
GABA+/Glu	.119	−.058	−.419	−.158	−.570*
Cerebellum b
GABA+/Glu	−.461*	−.339	.265	.352	.225
SA
Amygdala-hippo c
GABA/Glu	−.204	−.024	.054	−.027	.220
Cerebellum d
GABA+/Glu	−.067	−.625*	−.099	−.392	.475

^*Correlation is significant at the p < .05 level (2-tailed).

^an=21

^bn=23

^cn=20

^dn=11

Note. ADOS-2 = Autism Diagnostic Observation Schedule −2; SRS-2 = Social Responsiveness Scale-2, UPSIT = University of Pennsylvania Smell Identification Test

Figure 2.

Scatter plot depicting the relationship between I/E ratios in the cerebellum and social impairment in ASD and SA male participants. ASD participants are noted using blue, filled circles. SA participants are noted using red, unfilled circles. TYP participants, noted with black unfilled circles, are shown for illustrative purposes. Test statistics for the depicted ASD and SA correlations are reported in Table 3.

3.3. Interaction Effects between Diagnostic Group and Behavioral Measures in Male ASD and SA Participants

Tests of interaction effects of diagnostic group and ADOS SA CSS on amygdala-hippocampal GABA+/Glu (t = .918, p = .364) and cerebellar GABA+/Glu (t = −1.385, p = .177) were not significant. Tests of interaction effects of diagnostic group and SRS-2 on amygdala-hippocampal GABA+/Glu (t = −.137, p = .892) and cerebellar GABA+/Glu (t = .213, p = .833) were not significant. Tests of interaction effects of diagnostic group and Threshold Test – Vanillin were significant for amygdala-hippocampal GABA+/Glu (t = −2.754, p = .009) but not cerebellar GABA+/Glu (t = .025, p = .980). Tests of interaction effects of group and Threshold Test – Roses were not significant for cerebellar GABA+/Glu (t = 1.77, p = .089) nor amygdala-hippocampal GABA+/Glu (t = −0.397, p = .694). Tests for interaction effects of group and UPSIT were not significant for amygdala-hippocampal GABA+/Glu (t = −1.394, p = .171) nor cerebellar GABA+/Glu (t = .797, p = .432).

4. Discussion

The present study investigated amygdala-hippocampal and cerebellar region GABA+ and Glu concentrations, and their ratio (GABA+/Glu) as an inhibitory-excitatory index (I/E), in school-aged autistic children compared with age and sex-matched SA and TYP children. Additional secondary analyses were conducted to address differences between diagnostic groups in males only. These measures were then related to individual differences in social and olfactory processing. Overall, our data support the hypothesis that Glu is increased, both absolutely and relative to GABA+, in the cerebellum, but not in the amygdala-hippocampal region in male autistic children. These results add to a growing body of evidence demonstrating regional differences in I/E imbalance in ASD and provide compelling evidence that the cerebellum may play a critical, trans-diagnostic role in social impairment.

4.1. Amygdala-Hippocampus

Contrary to our hypothesis, we found no differences in amygdala-hippocampal GABA+, Glu, and GABA+/Glu levels between ASD and TYP children. These results were in the opposite direction of our hypothesis of a greater imbalance in GABA+/Glu in autistic children, and inconsistent with several previous post-mortem (Blatt et al., 2001),(Guptill et al., 2007),and spectroscopic imaging reports(O’Neill et al., 2020),(Page et al., 2006). Though our results diverge from two MRS studies that reported elevated Glu+glutamine in the amygdala of autistic individuals ages 4–60 (O’Neill et al., 2020) and in the amygdala-hippocampal region of a sample with a mean age 35(Page et al., 2006), they are consistent with a recent study of autistic adults that reported no differences in Glu in the amygdala (Shirayama et al., 2022), a high field (4T) pilot study of Glu in the medial temporal lobe in autistic teenagers (Joshi et al., 2013) and a positron emission tomography study that found no differences in GABA_A or GABA_A α5 subunit availability in the amygdala or hippocampus (Horder et al., 2018). To our knowledge, no previous spectroscopy studies have measured GABA+ in the medial temporal lobe. Given prior work and our null findings, we posit that ASD symptomatology in school-aged children may not be a result of an overall imbalance of amygdala-hippocampal GABA+ and Glu concentrations, but rather reflects functional differences within the GABAergic and glutamatergic systems.

4.2. Cerebellum

In the cerebellum, male autistic children demonstrated significantly elevated Glu and reduced GABA+/Glu ratios compared to TYP and SA male children. This was consistent with our hypothesis, and corroborates prior work demonstrating that ASD may be associated with imbalanced I/E neurotransmitters in the cerebellum, or of Glu excitotoxicity, and/or a combination of these factors. Specifically, previous work has reported abnormalities in cerebellar Glu neurometabolite levels (Hassan et al., 2013), (Ito et al., 2017) but see (Tebartz van Elst et al., 2014), receptor density (Fatemi et al., 2009),(Purcell et al., 2001) glutamate receptor binding (Fatemi et al., 2018), and GAD65 and GAD67 levels, which are proteins responsible for converting Glu to GABA (Yip et al., 2007),(Yip et al., 2008),(Yip et al., 2009). Our cerebellar voxel was positioned in the left lateral hemisphere corresponding to lobules VI, VII (Crus I/II). As neural abnormalities of this cerebellar region have been previously reported in individuals with ASD, including reduced volume (D’Mello et al., 2015) and Purkinje cell reductions (Fatemi et al., 2002), elevated Glu with resultant excitotoxicity could provide a possible mechanism (Wegiel et al., 2014),(Whitney et al., 2009).

Our study makes a significant contribution to the I/E literature in evaluating autistic children and SA because we collected both GABA+ and Glu spectroscopic measurements on high functioning, unsedated children. Few studies have measured both GABA+ and Glu in the cerebellum using spectroscopy and some present interpretation challenges because of the use of sedation. DeVito et al. (DeVito et al., 2007) reported lower Glu in the cerebellum, but a large subset of ASD participants (18/26) were sedated with midazolam, which inhibits glutamate release (Sakai and Amaha, 2000). In contrast, Ito et al. (Ito et al., 2017) reported reduced GABA+ and increased Glu, in the cerebellum in ASD participants despite using the sedative triclofos sodium, which is also a GABA agonist, prior to scanning. However, all participants including typically developing controls were sedated in that study, which may have mitigated this potential confound. A recent, large (N=343) multi-site study of children with neurodevelopmental disorders, including ASD, ADHD, obsessive compulsive disorder, and typical development was conducted to determine whether cerebellar GABA+ levels would differentiate the participants (ages 2–22; mean age approximately 12 years) with neurodevelopmental disorders from typical development or between each other (Pang et al., 2023). Consistent with our null findings for GABA+ alone (p = .531) and previous work showing no significant difference in GABA A binding in the cerebellum (Horder et al., 2018), no significant differences in GABA+ levels were found (Pang et al., 2023). Unfortunately, Glu was not reported, therefore, it is unknown whether alterations in the ratio of GABA+ to Glu were present in this large, well-characterized sample.

4.3. Sex Differences in GABA and Glu

We reported our statistical analyses with and without female participants because prior work suggests that sex impacts GABA and Glu measurements and there are sex-based neuroimaging differences in ASD. Sex differences in GABA levels have been reported in the thalamus (Fung et al., 2021), dorsolateral prefrontal cortex (O’Gorman et al., 2011),(Grachev and Apkarian, 2000) and the insula (Spurny-Dworak et al., 2022). Sex differences in Glu and glutamate+glutamine have been reported in the hippocampus (Hädel et al., 2013), thalamus (O’Neill et al., 2020) and dorsolateral prefrontal cortex (O’Gorman et al., 2011).

Further, with the caveat that there is little data on this topic, sex differences do not appear to follow neurotypical patterns in ASD. O’Neill et al. (O’Neill et al., 2020) and Fung et al., (Fung et al., 2021) reported sex-by-diagnosis interaction effects in Glu and GABA+ respectively. In addition, the relationship between metabolite levels and autism symptomology appears to differ between autistic males and females. Kirkovski et al. (Kirkovski et al., 2018) found a positive correlation between GABA+ concentration in the superior temporal sulcus and ASD-related social impairments in women but not men. Fung et al. reported a significant interaction between sex, thalamic GABA+ levels and score on the Autism Quotient, with female ASD participants showing a positive correlation while male ASD participants showed a negative correlation between GABA+ and Autism Quotient scores. These results suggest that there may be sex differences in the way the GABAergic system is impacted in ASD, and that these differences are region-specific.

4.4. GABA and Glu imbalance are related to social and olfactory processing in children with ASD and SA

To better understand the relationship between GABA+/Glu imbalance and social and olfactory processing, pearson correlations were conducted, revealing several significant neurochemical-behavioral associations in male autistic and SA children. Notably GABA+/Glu imbalance in the cerebellum, but not the amygdala-hippocampal region, was associated with social impairment for the ASD and SA male children. Abnormal cerebellar pathophysiology in ASD is one of the most replicated findings in the field, yet conceptualizing cerebellar abnormalities in relation to social deficits is uncommon because of the historical view of the cerebellum as a brain region that is primarily involved in motor functions. More recent theories include the cerebellum as a key node in social processes including implicit social preference, theory of mind, social learning, and behavioral flexibility (Stoodley and Tsai, 2021). Our cerebellar voxel was positioned in the left lateral hemisphere corresponding to lobules VI, VII (Crus I/II), which are cerebellar regions known to be involved in social behaviors through their connections with neocortical circuitry (Strick et al., 2009),(Stoodley and Tsai, 2021),(Van Overwalle et al., 2020) and structural impairments in this region have been associated with core autism symptoms (D’Mello et al., 2015). Damage to this region may impair the cerebellum’s contribution to social cognition via prediction and adaptation (Courchesne and Allen, 1997),(Stoodley and Tsai, 2021). We do not suggest that the relationship between GABA+/Glu imbalance and social impairment is specific to the cerebellum, and it is important to note that connections between the cerebellum and the amygdala arise from the vermis (Pierce and Péron, 2020), not the lateral hemisphere, and therefore consistent findings between these structures would not be predicted. Future research interrogating the cerebellum and structurally connected prefrontal regions may more comprehensively reveal the link between cerebrocerebellar circuity, I/E imbalance and social dysfunction.

The relationship between GABA and/or glutamate and physiological measures of sensory sensitivity have been tested in the thalamus (He et al., 2021), sensorimotor cortex (He et al., 2021),(Puts et al., 2017), occipital cortex (Puts et al., 2017),(Kolodny et al., 2020) and sensorimotor cortex (Kolodny et al., 2020) in individuals with ASD. However, this is the first report of which we are aware that directly tested relationships between olfaction and GABA+/Glu imbalance. Of note, olfaction is strongly interconnected with social-emotional pathways that are implicated in ASD. As expected, no relationships between olfactory measures and GABA+/Glu were observed in the cerebellum. However, we also found limited evidence that amygdala-hippocampal GABA+/Glu ratios were correlated with olfactory processing in male participants with ASD. I/E imbalance was correlated to smell detection for the vanillin odorant but not the roses odorant, and in ASD but not SA males. No relationship was observed between smell identification and I/E imbalance. Other groups who have tested the relationship between physiological measures of sensory processing and I/E imbalance have reported inconsistent findings as well. For example, Puts et al. found that higher levels of GABA+ in the sensorimotor cortex were correlated with higher detection thresholds for sub-threshold tactile stimulation (Puts et al., 2017) in children with ASD. However, a follow-up paper by the same group reported that lower glutamate+glutamine (Glx) in the thalamus and sensorimotor cortex was associated with higher detection thresholds for sub-threshold stimulation but no relationship with GABA+ (He et al., 2021). Given the null result for group differences in the amygdala-hippocampus, it is unlikely that olfactory processing differences are driven by I/E imbalance in the amygdala-hippocampal region. Future studies should investigate other olfactory brain regions to help clarify the neural basis of atypical olfactory processing.

Our prior behavioral studies of social and sensory functioning revealed similar levels of parent-reported sensory processing atypicalities in autistic children compared to a control group of children with SA, for whom ASD had been ruled out (Sweigert et al., 2020),(St. John et al., 2022). Both groups reported significantly higher levels of parent-reported sensory symptoms compared to typically developing children. SA children were rated significantly higher than ASD children on the Seeking and Touch scales of the sensory profile, but no other significant differences in sensory reactivity were reported. Olfactory testing showed that autistic children had intact odor detection with reduced odor identification ability compared to TYP children, and that odor identification was significantly correlated with ASD symptom severity (Sweigert et al., 2020). The SA children had significantly more difficulty with both smell detection and smell identification than the TYP children but did not perform significantly differently than the ASD children. Similarly, our previous work found that children with SAs demonstrated more impaired social functioning than children with TYP and comparable social functioning to autistic children (St. John et al., 2022). These results suggest that in addition to atypical sensory processing, children with SAs experience clinically significant problems with social functioning despite not meeting diagnostic criteria for ASD. Additional research is needed to better understand whether social dysfunction and sensory abnormalities are best understood as transdiagnostic neurobehavioral constructs, or whether diagnostic category provides important etiological information regarding the neural basis of phenotypic symptom expression.

4.5. Limitations

Small sample sizes having valid MRS data were a limitation of this study. GABA scanning protocols are time intensive, and a limitation when studying difficult to scan populations. We obtained fewer valid data sets in the cerebellum than the amygdala because the cerebellum was collected last, and some children had increasing difficulty holding still or requested to stop. Further, the SA group was disproportionately impacted by invalid/missing data in the cerebellum. Though the current study prioritized MRS data quality through stringent exclusion criteria, the tradeoff was high levels of data loss, and other investigators might have reasonably loosened criteria to increase sample size. For this approach to be useful in a clinical context, faster MRS acquisition protocols that result in higher percentages of useable data would be required.

Diagnostic group differences were accentuated or revealed when only male participants were included in the analyses, reinforcing the importance of assessing sex-based differences for understanding ASD. Our results, however, cannot be generalized as we were insufficiently powered to separately examine autistic girls. To more fully understand the role of I/E imbalance in ASD, future studies must have sufficient female participants to systematically test sex-based relationships.

Half of our participants with ASD or SA were taking psychotropic medication when enrolled in the research. While this is a limitation of the current study, it is important to note that the results did not change, although mean differences between the ASD and TYP groups were larger when the children taking medication were removed from the sample (see Supplementary Table 3). It is possible that psychotropic medication contributed to normalizing GABA+ and Glu levels in those children.

Additionally, the amygdala-hippocampal voxel included structures outside the region of interest. It is possible that error related to partial voluming effects may be an alternative explanation for the negative findings in this region.

Lastly, the SA and ASD groups included children with not only olfactory sensitivities, but also with other sensory sensitivities (auditory, tactile, etc.) that were not directly assessed. In addition, children with different processing profiles (seeking vs avoiding, hyper vs hypo sensitivities, etc.) were combined instead of separated into subgroups. These limitations impact the generalizability of our results. Further, children with severe hyposmia were excluded due to the inability to provide valid data for assessments that required the ability to perceive odors, However, it is possible that these children are the most impacted and may have exhibited greater MRS effects. Future studies should explore the effects of processing profiles to define the generalizability of these results.

4.6. Conclusion

To our knowledge, this study is the first to measure GABA+ and Glu levels in ASD children compared to TYP and SA children and relate them to olfactory processing. These results provide new insights into ASD-specific regional brain differences in inhibitory and excitatory neurotransmitter concentrations (GABA+ Glu). Metabolite differences were observed despite limited phenotypic differences between autistic children and SA on odor tasks or parent-reports of individual sensory processing profiles. We additionally found a relationship between decreased cerebellar GABA+/Glu and greater social impairment in both ASD and SA males. These results add to a growing body of evidence demonstrating regional brain differences in I/E imbalance in ASD and provide important evidence that the cerebellum may play a critical, transdiagnostic role in social impairment.

Highlights.

• Altered inhibitory and excitatory signaling was observed in male autistic children compared to typically developing male peers and clinical control group of boys with sensory processing abnormalities.

• Severity of social dysfunction was inversely correlated with lower levels of GABA relative to glutamate in male autistic children and children with sensory abnormalities.

• Alterations in inhibitory and excitatory signaling were not observed in the amygdala-hippocampal region.