Pone D 22 34955 - R1

PLOS ONE
Dorsolateral Prefrontal Cortex Metabolic Profiles in Autism Spectrum Disorder

Correlate Atypically with Nonverbal IQ and Typically with Attention Switching
--Manuscript Draft--
Manuscript Number: PONE-D-22-34955R1
Article Type: Research Article
Full Title: Dorsolateral Prefrontal Cortex Metabolic Profiles in Autism Spectrum Disorder
Correlate Atypically with Nonverbal IQ and Typically with Attention Switching
Short Title: DLPFC Metabolites in Autism Correlate Atypically with Nonverbal IQ and Typically with
Attention Switching
Corresponding Author: Akila S Weerasekera

Athinoula A Martinos Center for Biomedical Imaging
Charlestown, MA UNITED STATES
Keywords: MRS; choline; glutamate; Glx; Executive function
Abstract: The neurometabolic profile associated with autism spectrum disorder (ASD) has been
reported to be abnormal by some studies showing region specific metabolite levels in
ASD, while others report no group differences. The neurometabolic profile of the left
dorsolateral prefrontal cortex (DLPFC) is of particular interest due to the DLPFC
relevance to cognitive and executive function, and to ASD. We used 1H-MRS to
investigate neurometabolic profiles in the DLPFC of ASD and sex/IQ-matched typically
developing (TD) children (ages 9-13). We focused on levels of Glutamate and
Glutamine (Glx) due to many reported Glx abnormalities ASD, and of Choline (Cho)
because of its relationship to intelligence quotient (IQ) and to attentional re-orienting
difficulties. While no significant group differences were observed in concentrations
(absolute or creatine referenced), metabolite levels were correlated with the behavioral
phenotype of ASD children. In the ASD group but not the TD group, nonverbal IQ
(NVIQ) was negatively associated with Cho (r=-0.59, p=0.026) and positively
associated with Glx/Cr (r=0.66, p=0.011). Furthermore, attentional-switching scores in
the ASD group correlated negatively with Cho (r=-0.69, p=0.009), and positively with
Glx to creatine ratio (Glx/Cr) for both the ASD (r=0.73, p=0.004) and TD (r=0.54,
p=0.040) groups. Cho and Glx/Cr have different neurometabolic roles in modulating
NVIQ in ASD compared to TD children, while their role in attentional switching seems
preserved in ASD. Elucidating the role of neurometabolites in ASD also in the absence
of significant group differences in absolute levels is important, as sometimes even in
the absence of group differences in means, differences in the ASD group might still
emerge that are lost when only group means are considered. Understanding how
individual neurophysiological markers might correlate with ASD specific traits similarly
to, or differently from, correlations with the same traits in TD populations, and in this
specific case NVIQ, is therefore a step towards understanding and mapping the neural
correlates of ASD.
Order of Authors: Akila S Weerasekera
Adrian Ion-Margineanu
Nicole M McGuiggan
Nandita M Shetty
Robert M Joseph
Shantanu Ghosh
Mohamad M Alshikho
Martha R Herbert
Tal M Kenet
Eva-Maria M Ratai
Opposed Reviewers:
Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation
Response to Reviewers: Response to reviews
We would like to begin by thanking the reviewers for the valuable comments on this
research article. We are grateful for the constructive suggestions on how to further
improve our work and strengthen our findings. Below, we address the comments point
by point. For convenience, the response to reviewers’ comments is written in blue, and
changes to the manuscript have been highlighted.
Reviewers' comments:
Reviewer #1:
In this cross sectional study of patients with Autism Spectrum Disorder (ASD), authors
collected single-voxel MRS data in DLPFC cortical region of the brain in 14 ASD males
and 16 Typically Developing males who were age matched. Authors used MRS data to
estimate Choline (Cho), Glutamate-Glutamine (Glx), and Creatine (Cr) levels. Authors
also collected symptom severity measures in patients, and NVIQ, IQ, attentional re-
orienting, and attentional switching in patients and healthy. Metabolite levels did not
differ between ASD from TD. And metabolite levels did not correlate with symptom
severity. However, Cho levels were inversely associated with NVIQ in ASD but not in
TD participants. Attentional-switching scores were inversely correlated with Cho levels
in the in ASD group; and Glx/Cr levels were positively correlated with attentional-
switching scores the in the ASD and the TD groups.
There are several conceptual problems with the manuscript and analyses
(1) Authors utilized anatomical MRI to estimate gray, white, and CSF fraction in each
voxel; and using these fractions to estimate metabolites levels corrected for partial
volume. With MRS data collected for single voxel, correction for partial volumes is not
possible. If partial volume correction was done, then no details have been provided.
Without this, it is impossible to interpret the findings in the manuscript.
We thank the reviewer for the opportunity to clarify this. In this study, partial volume
correction was performed as now stated in lines 210-217
“All metabolite levels were adjusted for gray matter (GM), white matter (WM), and
cerebrospinal fluid (CSF) contributions as follows: the MEMPRAGE images and the
voxel coordinates from the Siemens RDA files were used in Gannet toolkit version 3.1
(Edden et al., 2014) to generate binary masks of the voxel location. These masks were
then used in SPM 12 (Penny et al., 2007) to calculate the partial volumes of GM, WM
and CSF percentages within the voxel. The segmented tissue fractions were then used
to correct for metabolite concentrations using LC Model (see S1 Appendix). Metabolite
levels are reported relative to the non-suppressed water signal and creatine. The
metabolite values are in institutional units (for absolute concentrations).”
”
(2) The manuscript seems to suggests that they used LCmodel software for absolute
quantification of metabolite levels. If so, then the details for the absolute quantification
are missing. How was absolute quantification done with acquisition of any reference
scan?
Absolute quantification was performed using the water unsuppressed reference scan.
Information of this scan is now added to manuscript (line 187):
“In addition, unsuppressed water signal (number of averages = 4) was acquired for
absolute concentration calculation.”
(3) The authors claim to use Glx/Cr ratio to account for the correlations of Glx with age.
This seems strange. The ratio is computed to normalize metabolite levels to Cr levels
so that they can be compared across participants. Even if Glx is correlated with age, Cr
has to be similarly correlated with age to remove the age effects in the ratio; that too, if
Glx and Cr depend upon only age, which is not true. So the justification provided for
the use of Glx/Cr ratio is not clear.
Glx was not correlated with age in this study, but a trend was observed in the ASD
group (p=0.052). Cr was also not correlated with age in either group or no trends were
observed (p=0.524 [ASD], p=0.310 [TD]). Because of the Glx trend to correlate with
age, we tested whether the Glx/Cr ratio was correlated with age, following the
approach used in several other studies of Glx in children in general {Benamor 2014;
Yang et al., 2015; Perlov et al., 2009}, and children with ASD {Kubas et al., 2012;
Doyle-Thomas et al., 2014; Ito et al., 2017}. This information was added to the
manuscript (lines 233-241).
Kubas, B., Kulak, W., Sobaniec, W., Tarasow, E., Lebkowska, U., and Walecki, J.
(2012). Metabolite alterations in autistic children: a 1H MR spectroscopy study. Adv.
Med. Sci. 57, 152–156. Doi: 10.2478/v10039-012-0014-x
Ito, H., Mori, K., Harada, M., Hisaoka, S., Toda, Y., Mori, T., Goji, A., Abe, Y., Miyazaki,
M., & Kagami, S. (2017). A Proton Magnetic Resonance Spectroscopic Study in Autism
Spectrum Disorder Using a 3-Tesla Clinical Magnetic Resonance Imaging (MRI)
System: The Anterior Cingulate Cortex and the Left Cerebellum. Journal of child
neurology, 32(8), 731–739. https://doi.org/10.1177/0883073817702981
Doyle-Thomas KA, Card D, Soorya LV, Wang AT, Fan J, Anagnostou E. Metabolic
mapping of deep brain structures and associations with symptomatology in autism
spectrum disorders. Res Autism Spectr Disord. 2014 Jan;8(1):44-51. doi:
10.1016/j.rasd.2013.10.003. PMID: 24459534; PMCID: PMC3897261.
Benamor L. (2014). (1)H-Magnetic resonance spectroscopy study of stimulant

medication effect on brain metabolites in French Canadian children with attention
deficit hyperactivity disorder. Neuropsychiatric disease and treatment, 10, 47–54.
https://doi.org/10.2147/NDT.S52338
Yang, Z. Y., Quan, H., Peng, Z. L., Zhong, Y., Tan, Z. J., & Gong, Q. Y. (2015). Proton
magnetic resonance spectroscopy revealed differences in the glutamate +
glutamine/creatine ratio of the anterior cingulate cortex between healthy and pediatric
post-traumatic stress disorder patients diagnosed after 2008 Wenchuan earthquake.
Psychiatry and clinical neurosciences, 69(12), 782–790.
https://doi.org/10.1111/pcn.12332
Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., Hesslinger, B.,
Buechert, M., Henning, J., Ebert, D., & Tebartz Van Elst, L. (2009). Spectroscopic
findings in attention-deficit/hyperactivity disorder: review and meta-analysis. The world
journal of biological psychiatry : the official journal of the World Federation of Societies
of Biological Psychiatry, 10(4 Pt 2), 355–365.
https://doi.org/10.1080/15622970802176032
(4) Why was Cho levels used without normalizing it to either background noise levels to
account for gain factor or to Cr levels? If Cho levels were absolute quantities in mmol,
than how was the absolute quantification done?
In this study Cho levels are reported in absolute concentration, referenced to water.
This is now clarified in the methods section.
(5) Partial volume correction was done using anatomical MRI data that were acquired
in addition to the single voxel MRS data. The correction procedure assumes that the
patient does not move between anatomical and MRS scan, which likely is not the case.
Patient movement between two scans will lead to error in locating where the MRS
voxel is location in the anatomical space. No steps have been proposed to minimize
motion effects.
A mock scanner was used to prepare (reduced anxiety and motion) the children for
actual MRI scan session. Next, MRS was performed immediately following the
anatomical scan. We used QA (visual inspection) following the scan acquisition to
exclude participants with excessive motion or other artefacts. 3 ASD and 1 TD
participant data were excluded due to poor data quality.
(6) Several statements are made in the abstract and discussion that do not really
provide any information and leave a person wondering what is being conveyed. For
example, in the abstract
Elucidating the apparently divergent role of neurometabolites in ASD in the absence of

significant group differences in absolute levels is an important step towards
understanding and mapping the neural correlates of ASD.
Why is elucidating divergent role is an important step forward? And why it leads to
understanding and mapping the neural correlates of ASD is metabolites in those
regions are neither different than TD nor correlate with symptom severity? What is
being conveyed here by the authors?
We agree that the statement is indeed too vague. We changed it to the following
statement:
“Elucidating the role of neurometabolites in ASD also in the absence of significant

group differences in absolute levels is important, as sometimes even in the absence of
group differences in means, differences in the ASD group might still emerge that are
lost when only group means are considered. Understanding how individual
neurophysiological markers might correlate with ASD specific traits similarly to, or
differently from, correlations with the same traits in TD populations, and in this specific
case NVIQ, is therefore a step towards understanding and mapping the neural
correlates of ASD.”
(7) Similarly, the next statement in the abstract
These results are also relevant in the context of the significant cognitive function
heterogeneity associated with the ASD phenotype, as they suggest possible underlying
neural mechanisms that do not overlap with those expected from typical development.
What is “cognitive function heterogeneity”? And even if there differences across

participants, findings do not support any bases for heterogeneity. And how do the
findings suggest a neural mechanism for this heterogeneity that does not overlay with
TD individuals? The statement as written does not convey any valid thought at all.
We apologize for this vague statement. We believe that the revised version of the prior
statement is now clear enough as a standalone statement and replaces both of these
previous sentences. Therefore, this statement is now removed.
(8) Authors claim that they focused on only 2 metabolites because of small sample
size. How does focusing on 2 metabolites over the problems of small sample size?
We thank the reviewer for the opportunity to clarify this. Since each added metabolite
needs to be corrected for additional comparisons, thus decreasing power further in this
already small sample, we decided to focus on only two metabolites.
(9) Authors state that the patients had a prior diagnosis of ASD. What does prior
diagnosis mean? Did they have or did not have autism at the scan time?
Clearly this phrasing created confusion. We apologize for this. We meant to say that
the participants had gotten their diagnosis before being recruited. The word “prior” was
removed. All diagnoses were current at the time of recruitment, and further verified as
noted.
(10) Authors used multiple and not multivariate linear regression analysis. If they did
use multivariate, then they need to explain what were the dependent variables in the
analyses.
The regression equation provided
Y_Behavior= β0 + β1[met1]_group + β2[met2]_group
is not clear. What are the terms in the equation?
We thank the reviewer for pointing this out. We have now changed this to multiple
linear regression. This is now explained in the manuscript (lines 226-230).
“YBehavior= β0 + β1[met1]group + β2[met2]group, where dependent variable Y is the

predicted behavior measure (i.e., NVIQ, ICS-S), and met1/met2 are continuous
predictor variables (metabolite levels), β0 is the estimate of the model intercept and β1/
β2 are the regression coefficients.“
(11) Why did they use Bonferroni correction for 10 comparisons? There seems to be
many more analyses that were run.
We thank the reviewer for the opportunity to clarify this point. As noted in the
manuscript, we actually did not apply any correction for multiple comparisons, following
the recent best practices (Lu and Belitskaya-Levy, 2015; Amrhein et al., 2019; Lowe,
2019), that advise to simply note p values as they are without corrections, due to the
dependencies masked by corrections. Instead, we indicated using * the significance of
the p-value, with ** indicating a p value 10 times more significant than 0.05. We now
removed the sentence “, i.e. p values that would survive a Bonferroni correction for 10
comparisons”, since it is indeed confusing. The number of comparisons can be
assessed by the reader within each figure. Two groups x 3 metabolites in figure 2, 2
groups x 4 behavioral measures in figure 3 and in figure 4. Note also that a Bonferroni
correction would be too strict, as some of these measures, e.g. the behavioral
measures, are not actually completely independent. Hence the practice of not
correcting, allowing the reader to assess significance using uncorrected values,
(12) Lines 387-388, Discussion Section
This may suggest that DLPFC Glx in ASD may have a different glutamatergic
transmission than that of TD children.
What does different glutamatergic transmission means? And why is it different because
of different correlations with NVIQ?
The absence of a group difference in the concentrations of Glx or Glx/Cr indicate that
Glx levels themselves are not altered in the autistic DLPFC. However, Glx/Cr predicted
the dynamics of NVIQ only in ASD. This study is clearly not equipped to answer these
important questions (what does it mean and why), but we know this is not the first time
such differences have been observed (Coplan et al., 2018). ASD and anxiety can
impact development, and we know neurotransmission is impacted by development.
How and why exactly the changes occur this way in this and other studies seem to
indicate will hopefully be explored in future studies.
(13) Line 415, discussion section
Therefore, longitudinal MRS studies are clearly necessary.
Why are longitudinal MRS studies clearly needed?
This sentence is now replaced with: “Because ASD is a neurodevelopmental disorder,

it is possible that the function of different metabolites is impacted differently due to
differing maturation trajectories throughout development. Longitudinal studies of
metabolites and their correlation with specific traits at different points in time could help
to elucidate this important question.” (line 436-439)
(14) Line 418 and 419, discussion section
There were no significant between-group differences in voxel brain matter and CSF,
metabolite concentrations were tissue corrected, and the results are consistent with
prior metabolites studies of ASD and of other related disorders.
Prior studies have reported significant metabolite differences in ASD. So how are the
reported findings of no difference in this manuscript consistent with those?
Prior studies show that the neurometabolic changes among ASD and healthy controls
is somewhat mixed (Review: Ford TC and Crewther DP 2016). For example, within this
review, 4 studies employing TD and ASD children (Corrigan et al., 2013; Friedman et
al., 2006; Fayed and Modrego 2005; DeVito et al., 2007) found a group difference in
Cho levels in the frontal regions, while 8 studies (Friedman et al., 2003; Levitt et al.,
2003; Zeegers et al., 2007; Hisaoka et al., 2001; Kubas et al., 2012; Fujii et al., 2010;
Vasconcelos et al., 2008; Endo et al., 2007) did not. The differences across studies
could stem from cohort differences, methodological differences, age differences, power
differences, etc. Thus, we consider the important finding here to be not the lack of
difference in group means, which may be a power issue, but the correlations between
metabolite levels and behavioral traits. This has now been added to the manuscript
(line 444).
This suggests that Cho and Glx/Cr may have different neurometabolic roles in ASD
compared to TD children in overall cognitive function.
Different neurometabolic role? Really? Cho and Glx are mediating different molecular
processes in ASD than in TD? If so, then at is the evidence for making such a claim?
We agree that this sentence was written poorly. It is now revised to read:
“The results suggests that the extent to which Cho and Glx/Cr levels in the DLPFC
mediate VIQ and inhibition of attention, may differ for ASD children relative to TD
children.” (line 463)
Reviewer #2:
In this work, Weerasekera et al investigated choline and glutamatergic compounds in

the left dorsolateral prefrontal cortex of 14 ASD and 16 TD children. While there were
no significant between group metabolite differences, correlations between metabolite
levels and intelligence and cognitive scores were observed. NVIQ was negatively
correlated with Cho in the ASD group but positively correlated in the TD group,
whereas Glx/Cr was positively correlated with NVIQ in ASD but not correlated in the
TD group. On the other hand, attention switching showed similar patterns of
correlations across groups, it was negatively correlated with NVIQ and positively
correlated with Glx/Cr. Hence the authors conclude that Cho and Glx/Cr have different
roles in NVIQ modulation in ASD, while their role in attention switching seems
preserved in ASD, and state that these results are relevant in the context of cognitive
heterogeneity as they suggest possible neural mechanisms in ASD that do not overlap
with the expected from typical development.
I have the following comments and suggestions:
Abstract:
1. The authors state that no significant differences were observed for absolute
concentrations but there were both absolute and relative measures reported. Please
correct.
We thank the Reviewer for point this out, this now changed as; “While no significant
group differences were observed in concentrations (absolute or creatine referenced)”
(line 44)
2. The findings of the submitted work to not speak to the cognitive heterogeneity within
ASD, only to a different relationship between metabolites and cognitive profile between
ASD and non-ASD groups. The statement regarding heterogeneity should be removed
as it is not supported by the data.
We thank the reviewer for the suggestion. We have now removed this statement from
the abstract.
Introduction:
3. Please revise the references throughout the manuscript, a few examples in the
introduction are:
Line 66: APA DSM 5 ref missing and the others are not adequate.
Line 68: ref missing
Line 69: Ref Howlin 2021 does not cover neuropathology of ASD, hence it’s not
adequate.
Line 102: Ref Lam et al 2006 is out of context
We have now revised the references as suggested by the reviewer.
Methods and materials
4. What was the medication status in the ASD group?
Seven of the 14 boys in the ASD group used psychotropic medication (Ritalin,
Citalopram, Focalin, Adderall, Sertraline, Concerta and Risperidone). None of the
participants in the TD group were on medication. This information is now added to the
methods section (line 166).
5. What do the ICS-I and ICS-S scores mean, do higher scores mean worse/better
performance? It would be helpful to briefly explain this assessment.
We thank the reviewer for the opportunity to clarify this.
-The NEPSY-II Inhibition Contrast Scaled Score (ICS-I): measures the ability to
voluntarily inhibit attention.
- NEPSY-II Switching Contrast Scaled Score (ICS-S): measures the ability for
switching attention between competing stimuli.
The higher the ICS-I and ICS-S scores, the better the performance. This information is
now included in Figure (3-4) legends.
6. There’s no a priori sample size calculation and, as noted by the authors, the study
sample is small. Since an a priori sample size was not calculated, the authors should at
least discuss their finds compared to other similar studies regarding their sample sizes.
We thank the reviewer for the suggestion. Now we have added:
“Previous studies using smaller ASD samples (n=7-12, ages 8-17), reported
differences in Glx and Glx/Cr in the frontal regions compared to TD groups (Bejjani et
al., 2012; Kubas et al., 2012; Joshi et al., 2013). However, no group-level differences
were found in Glx or Glx/Cr in our study. Yet, we found a significant positive correlation
between Glx/Cr and nonverbal intelligence only in the ASD group.” (line 403).
7. How were the metabolite absolute concentrations estimated? It seems that no water
reference was acquired for that purpose. If that’s the case, ideally all values reported
should be relative, for instance, to creatine. Please, report the analysis using Cho/Cr in
addition to Cho.
We apologize for not including this information. The absolute concentrations are
estimated using an unsuppressed water signal. This information is now added to the
methods section (line 187).
8. While it is stated that metabolites levels were adjusted for grey matter, white matter
and CSF, later on it is stated that CSF content was used (line 222). Please clarify
which correction was made and provide the formula used.
We have now added S1 Appendix which contains the detailed information of this
process. This is referenced in the methods section.
We used LC Model (LCM) to estimate the corrected metabolite concentrations. LCM

uses the following formula:
Cmet = Ratioarea × (2/N1Hmet) × (ATTH2O/attmet) × WCONC
Cmet = corrected metabolite concentration in mmol/l

Ratioarea= ratio of area of metabolite/area of unsuppressed water
N1Hmet = number of equivalent protons contributing to the resonance
ATTH2O & attmet are the factors (< 1) by which the water and metabolite resonance
areas are attenuated by relaxation. Default; ATTH2O = 0.7, attmet = 1
WCONC = water concentration (in mM) in the voxel.
WCONC = (43300fgm + 35880fwm + 55556fcsf) / (1 – fcsf), fgm, fwm and fcsf are the
volume fractions segmented using the T1-mprage.
The product: [Ratioarea × (2/N1Hmet) × (ATTH2O/attmet)] is the default output of

metabolite levels from LCM and by multiplying this product with WCONC, we get the
corrected metabolite concentration.
9. State that values are in institutional units (for absolute concentrations) as no other
corrections were made (e.g. relaxation times).
This is now added to the methods section (line 217).
10. Line 243 – number 5 should be removed here as there were only 4 behavioural
measures for TD.
We have now removed number 5 from the sentence.
11. Some MRSinMRS reporting standards (Lin et al 2021, NMR in Biomedicine) have
been followed (e.g. visualisation of all spectra), but please provide as much
recommended details as possible for acquisition, quantification, and quality assurance,
for instance, which water suppression method was used, the basis set used, a table
with SNR, FWHM and CRLB mean values and range for each metabolite (could be
supp material), etc.
This information is now added to the methods section and a table with quality
parameters is in the supplementary materials.
ParameterTDASDp-value
SNR17.2±4.117.1±3.30.940
FWHM 6.8±2.2 7.9±2.20.203
ChoCRLB 2.9±0.6 2.8±0.50.697
NAACRLB 2.9±0.5 3.0±0.90.666
CrCRLB 2.5±0.5 2.4±0.50.551
GlxCRLB 7.1±1.6 6.9±0.90.706
Results:
12. What is the test value for the slope comparison? Only p value is reported.
This information is now added to the manuscript (lines: 292, 321 and figure 3-4
legends)
13. Line 305 – Combining the two groups decreased the p value, but the correlation
coefficient was smaller, so it did not strengthen, this last part of the sentence should be
removed.
We have now revised this sentence as follows: combining the two groups slightly
decreased the correlation coefficient but increased the statistical significance of the
correlation (r=-0.56, p=0.002). [line: 312].
14. What does the multivariate analysis add to the work? This has not been addressed.
We thank the reviewer for pointing this out and now below statement is added to the
discussion (line 416). The multivariate analysis was added to test whether combining
measures improved predictions. While r and p values for predicting NVIQ and ICS-S
were somewhat improved, the improvement was not statistically significant. This is now
added in the results (lines 325).
Discussion
15. Line 348: Again, the correlation is not stronger as the r value is smaller. Please
remove.
Now the sentence reads:
“Glx/Cr ratio also showed a positive association with the ICS-S score for attention
switching for both ASD and TD groups, and the combination of the two groups
strengthened the statistical significance since the power (i.e., N) was increased.” (line
389).
16. In the second paragraph the authors discuss the findings in light of the ASD group
correlations, but the Cho findings in the TD group don’t seem to fit here. The
discussion is mixing findings in clinical and non-clinical populations. Please rewrite this
part of the discussion in the light of both ASD and TD findings and previous literature
regarding Cho to be clearer on what is the point of the paragraph.
This paragraph is now re-written as requested by the reviewer (line 369-385).
“Choline containing compounds are key components of cell membranes and the
myeline sheets and associated with white matter density and speculated to reflect
excessive neuronal connectivity or abnormal myelination (Laycock et al., 2008). Our
observation of a significant association between low Cho levels and higher NVIQ in the
ASD group is similar to the findings of a previous study where an inverse association
with the left DLPFC Cho/Cr and full-scale IQ (FSIQ) was reported in generalized
anxiety disorder (Coplan et al., 2018b). During disease states, elevated Cho is
associated with membrane breakdown and inflammation (Davie et al., 1995),
processes which are likely to impair executive functioning. However, we also observed
a significant positive association between DLPFC Cho and NVIQ in the control group,
where Coplan et al reported no significant association. Another previous study reported
an inverse relationship between Cho and performance IQ in healthy adults (Jung et al.,
1999). A major difference between our study and aforementioned previous studies is
that those included adult populations. Progressive myelination of white matter
pathways during typical development of children and adolescents has significant
changes in cognitive abilities, due to more rapid neural communication (Buyanova and
Arsalidou, 2021). Therefore, dynamic myelin turnover, which is indicative of active Cho
metabolism in typically developing children may explain the positive association with
cognitive performance. “
17. Line 388-389: “It also suggests a possible neural mechanism within the DLPFC for
the heterogeneity observed in cognitive function in ASD.” – This statement is not
supported by the data and should be removed.
This statement is now removed.
18. The first paragraph on page 13 (starting on line 368) is missing references for the
glutamine-glutamate-GABA cycle.
Appropriate references are now added.
19. Paragraph starting on line 390:

First sentence is missing references. The authors discuss literature reporting opposite
findings regarding glutamate correlations and reading performance, but it’s not clear
how this discussion relates to the current manuscript findings, please clarify.
20. In the limitations sections the authors mention that future studies with larger sample
sizes are needed, aren’t there previous studies with larger sample sizes published
already, at least for metabolite between group comparisons? Current findings should
be discussed in the light of previous works.
We have re-worded this sentence as follows:
Because ASD is a neurodevelopmental disorder, it is possible that the function of

different metabolites is impacted differently due to differing maturation trajectories
throughout development. Longitudinal studies of metabolites and their correlation with
specific traits at different points in time could help to elucidate this important question.
(lines 436-439).
Reviewer #3:
The authors brought novel ideas of examining neurometabolic profiles and their
relatedness with behavioral measures in autism. It was also clearer that the authors
worked hard and put a great deal of effort in data collection, and data analyses
sections. However, the study would become more interpretable if it was conceptualized
better, such that, inclusion of right frontal cortex, wider range of executive function
measures, and full-scale IQ.
We thank the reviewer for the encouraging comments. We acknowledge Reviewer’s
concern regarding the conceptualization of the study. The data for this study was
collected a while ago, as a part of a multimodal study, therefore due to time limitation
MRS data was limited to a single location.
Additional Information:
Question Response
Financial Disclosure This study was supported by the following: The National Institutes of Health
(R01NS048455, MH); the Nancy Lurie Marks Foundation (MH); MIT-MGH Strategic
Enter a financial disclosure statement that Partnership grant from the Executive Committee on Research (ECOR) at
describes the sources of funding for the Massachusetts General Hospital (EMR); the National Institute of Child Health and
work included in this submission. Review Development (R01HD073254, TK); the National Institute of Mental Health
the submission guidelines for detailed (R01MH117998, TK)
requirements. View published research
articles from PLOS ONE for specific
examples.
This statement is required for submission

and will appear in the published article if
the submission is accepted. Please make
sure it is accurate.
Unfunded studies
Enter: The author(s) received no specific
funding for this work.
Funded studies
Enter a statement with the following details:
• Initials of the authors who received each
award
• Grant numbers awarded to each author
• The full name of each funder
• URL of each funder website
• Did the sponsors or funders play any role in
the study design, data collection and
analysis, decision to publish, or preparation
of the manuscript?
• NO - Include this sentence at the end of
your statement: The funders had no role in
study design, data collection and analysis,
decision to publish, or preparation of the
manuscript.
• YES - Specify the role(s) played.
* typeset
Competing Interests The authors have no conflicts of interest to declare in relation to this work. E.M.R.
serves on the Scientific Advisory Board of BrainSpec Inc.
Use the instructions below to enter a
competing interest statement for this
submission. On behalf of all authors,
disclose any competing interests that
could be perceived to bias this
work—acknowledging all financial support
and any other relevant financial or non-
financial competing interests.
This statement is required for submission

and will appear in the published article if
the submission is accepted. Please make
sure it is accurate and that any funding
sources listed in your Funding Information
later in the submission form are also
declared in your Financial Disclosure
statement.
View published research articles from

PLOS ONE for specific examples.
NO authors have competing interests
Enter: The authors have declared that no

competing interests exist.
Authors with competing interests
Enter competing interest details beginning

with this statement:
I have read the journal's policy and the

authors of this manuscript have the following
competing interests: [insert competing
interests here]
* typeset
Ethics Statement Parents of the participants provided informed consent according to protocols approved
by the MGH Institutional Review Board (IRB). Participant assent was also provided in
Enter an ethics statement for this addition to parent consent for participants aged 14-17.
submission. This statement is required if
the study involved:
• Human participants
• Human specimens or tissue
• Vertebrate animals or cephalopods
• Vertebrate embryos or tissues
• Field research
Write "N/A" if the submission does not

require an ethics statement.
General guidance is provided below.

Consult the submission guidelines for
detailed instructions. Make sure that all
information entered here is included in the
Methods section of the manuscript.
Format for specific study types
Human Subject Research (involving human

participants and/or tissue)
• Give the name of the institutional review
board or ethics committee that approved the
study
• Include the approval number and/or a
statement indicating approval of this
research
• Indicate the form of consent obtained
(written/oral) or the reason that consent was
not obtained (e.g. the data were analyzed
anonymously)
Animal Research (involving vertebrate
animals, embryos or tissues)

• Provide the name of the Institutional Animal
Care and Use Committee (IACUC) or other
relevant ethics board that reviewed the
study protocol, and indicate whether they
approved this research or granted a formal
waiver of ethical approval
• Include an approval number if one was
obtained
• If the study involved non-human primates,
add additional details about animal welfare
and steps taken to ameliorate suffering
• If anesthesia, euthanasia, or any kind of
animal sacrifice is part of the study, include
briefly which substances and/or methods
were applied
Field Research
Include the following details if this study
involves the collection of plant, animal, or
other materials from a natural setting:

• Field permit number
• Name of the institution or relevant body that
granted permission
Data Availability Yes - all data are fully available without restriction
Authors are required to make all data

underlying the findings described fully
available, without restriction, and from the
time of publication. PLOS allows rare
exceptions to address legal and ethical
concerns. See the PLOS Data Policy and
FAQ for detailed information.
A Data Availability Statement describing
where the data can be found is required at
submission. Your answers to this question
constitute the Data Availability Statement
and will be published in the article, if
accepted.
Important: Stating ‘data available on request

from the author’ is not sufficient. If your data
are only available upon request, select ‘No’ for
the first question and explain your exceptional
situation in the text box.
Do the authors confirm that all data

underlying the findings described in their
manuscript are fully available without
restriction?
Describe where the data may be found in Data available on request from the authors
full sentences. If you are copying our
sample text, replace any instances of XXX
with the appropriate details.
• If the data are held or will be held in a

public repository, include URLs,
accession numbers or DOIs. If this
information will only be available after
acceptance, indicate this by ticking the
box below. For example: All XXX files
are available from the XXX database
(accession number(s) XXX, XXX.).
• If the data are all contained within the
manuscript and/or Supporting
Information files, enter the following:
All relevant data are within the
manuscript and its Supporting
Information files.
• If neither of these applies but you are
able to provide details of access
elsewhere, with or without limitations,
please do so. For example:
Data cannot be shared publicly because

of [XXX]. Data are available from the
XXX Institutional Data Access / Ethics
Committee (contact via XXX) for
researchers who meet the criteria for
access to confidential data.
The data underlying the results

presented in the study are available
from (include the name of the third party
and contact information or URL).
• This text is appropriate if the data are
owned by a third party and authors do
not have permission to share the data.
* typeset
Additional data availability information:
Cover Letter
MASSACHUSETTS GENERAL HOSPITAL

HARVARD MEDICAL SCHOOL
Eva-Maria Ratai, Ph.D.

Associate Professor of Radiology HMS
Director of Clinical Spectroscopy MGH
A.A. MARTINOS CENTER FOR BIOMEDICAL IMAGING
13TH STREET, BUILDING 149, ROOM 2301, CHARLESTOWN, MA 02129
VOICE 617-726-1744 FAX 617-726-7422
EMAIL eratai@mgh.harvard.edu
Dr. George Vousden

Deputy-Editor-in-Chief
PLOS ONE
December 21st, 2022
Dear Dr. Vousden:
Please find attached our manuscript “Dorsolateral Prefrontal Cortex Metabolic Profiles in Autism Spectrum
Disorder Correlate Atypically with Nonverbal IQ and Typically with Attention Switching” submitted for
your consideration for publication in PLOS ONE.
The neurometabolic profile associated with autism spectrum disorder (ASD) has been studied widely using non-
invasive methods, with mixed results. While some studies show abnormalities in neurometabolite profiles in
ASD, others do not. Here, we took the approach of focusing on correlations between behavioral measures and
neurometabolite profile, rather than just on neurometabolite concentrations. Specifically, we focused on the left
dorsolateral prefrontal cortex (DLPFC) in children with and without ASD, due to its role in cognitive and
executive function, and its high relevance to ASD.
While we observed no significant group differences in absolute concentration of the two neurometabolites we
chose to focus on – Choline (Cho) and Glutamate/Glutamine (Glx), we found that levels of both of these
metabolites correlated with the behavioral phenotype of ASD children differently than they did with the
behavioral phenotype of typically developing matched children.
Our findings support the idea that choline and glutamate may have different neurometabolic roles in ASD
compared to typically developing children when it comes to overall cognitive function, while the roles of these
metabolites for executive function skills may be preserved in ASD. Interestingly, these findings are in line with
some prior findings in individuals with anxiety, suggesting a potential common underlying mechanism.
We believe this article will be of interest to the readership of PLOS ONE. It offers a potential view of the
importance of elucidating the apparently divergent role of neurometabolites in ASD to better understand and map
the neural correlates of ASD.
We confirm that the work described in the manuscript has not been published previously or not under
consideration for publication elsewhere. The manuscript is approved by all authors and if accepted, it will not be
published elsewhere in the same form.
Thank you for your time and consideration of our manuscript.
Sincerely,
Eva-Maria Ratai, Ph.D. Tal Kenet, Ph.D.
–2–
Manuscript Click here to
access/download;Manuscript;Manuscript_PO_Weerasekera et
1 Dorsolateral Prefrontal Cortex Metabolic Profiles in Autism Spectrum Disorder Correlate

2 Atypically with Nonverbal IQ and Typically with Attention Switching
4 Akila Weerasekera1, Adrian Ion-Mӑrgineanu2, Nicole M. McGuiggan3, Nandita Shetty3, Robert

5 M. Joseph4, Shantanu Ghosh3,5, Mohamad Alshikho3, Martha R. Herbert 3,6, Tal Kenet3*, Eva-
6 Maria Ratai1*
7
1
8 Department of Radiology, and Athinoula A. Martinos Center for Biomedical Imaging,
9 Massachusetts General Hospital, Harvard Medical School, Boston, MA, USA.
2
10 Department of Electrical Engineering (ESAT), STADIUS Center for Dynamical Systems, Signal
11 Processing and Data Analytics, KU Leuven, Leuven, Belgium.
3
12 Department of Neurology, and Athinoula A. Martinos Center for Biomedical Imaging,
4
14 Boston University School of Medicine, Department of Anatomy and Neurobiology, Boston, MA,
15 USA.
5
16 Psychoacoustics and Neural Dynamics Lab, Akal University, Raman Road, Talwandi Sabo
17 151302, Punjab, India.
6
18 Higher Synthesis Foundation, Cambridge, MA, USA
19
20
21
22
23 *Share senior authorship
24
25
26 Corresponding Author: Eva-Maria Ratai

27 Athinoula A. Martinos Center for Biomedical Imaging
28 Massachusetts General Hospital, Department of Radiology, Neuroradiology Division
29 149 13th Street, Suite 2318
30 Charlestown, MA 02129, USA
31 Email: eratai@mgh.harvard.edu
32 Phone: +1-617-726-1744
33 Fax: (617) 726-7422
34
1
35 ABSTRACT
36 The neurometabolic profile of associated with autism spectrum disorder (ASD) has been reported
37 to be abnormal by some studies showing region specific metabolite levels in ASD, while others
38 report no group differences. The neurometabolic profile of the left dorsolateral prefrontal cortex
39 (DLPFC) is of particular interest due to the DLPFC relevance to cognitive and executive function,
40 and to ASD. We used 1H-MRS to investigate neurometabolic profiles in the DLPFC of ASD and
41 sex/IQ-matched typically developing (TD) children (ages 9-13). We focused on levels of
42 Glutamate and Glutamine (Glx) due to many reported Glx abnormalities ASD, and of Choline
43 (Cho) because of its relationship to intelligence quotient (IQ) and to attentional re-orienting
44 difficulties. While no significant group differences were observed in absolute concentrations,
45 metabolite levels were correlated with the behavioral phenotype of ASD children. In the ASD
46 group but not the TD group, nonverbal IQ (NVIQ) was negatively associated with Cho (r=-0.59,
47 p=0.026) and positively associated with Glx/Cr (r=0.66, p=0.011). Furthermore, attentional-
48 switching scores in the ASD group correlated negatively with Cho (r=-0.69, p=0.009), and
49 positively with Glx to creatine ratio (Glx/Cr) for both the ASD (r=0.73, p=0.004) and TD (r=0.54,
50 p=0.040) groups. Cho and Glx/Cr have different neurometabolic roles in modulating NVIQ in ASD
51 compared to TD children, while their role in attentional switching seems preserved in ASD.
52 Elucidating the apparently divergent role of neurometabolites in ASD in the absence of significant
53 group differences in absolute levels is an important step towards understanding and mapping the
54 neural correlates of ASD. These results are also relevant in the context of the significant cognitive
55 function heterogeneity associated with the ASD phenotype, as they suggest possible underlying
56 neural mechanisms that do not overlap with those expected from typical development.
57
58
59
60 Key words: MRS, choline, glutamate, Glx, executive function
61
2
62 1. INTRODUCTION
63
64 Autism spectrum disorder (ASD) is an early onset neurodevelopmental disorder

65 characterized by deficits in social interaction and communication and limited and repetitive
66 behaviors and interests (Margari et al., 2018; Hiremath et al., 2021). Studies of ASD have
67 shown a wide range of abnormalities in brain function and structure (Müller and Fishman, 2018;
68 Forde et al., 2020), as well as abnormalities in the levels of brain metabolites. However, many
69 questions about the underlying neuropathology of ASD remain unanswered (Blatt, 2012; Howlin,
70 2021).
71 In this study, we sought to gain a better understanding of how metabolite imbalances in the
72 frontal cortex may contribute to ASD symptomatology. The association of the frontal cortex in
73 the neurobiology of ASD has long been recognized (Courchesne et al., 2011a; Stockman, 2013;
74 Stoner et al., 2014). The frontal regions of the brain are known to play a key role in executive
75 and socioemotional function, two of the cognitive processes known to be impaired in ASD (Kim
76 et al., 2015). In particular, structural and metabolic irregularities have been reported in the
77 dorsolateral prefrontal cortex (DLPFC) in ASD, and the DLPFC has long been associated with
78 deficits in executive function and social cognition (Alexander et al., 1986; Haznedar, 2006;
79 Schmitz et al., 2007; Kalbe et al., 2010). Postmortem studies suggests that children with ASD
80 have an abnormally larger number of neurons in the DLPFC than typical developing children,
81 and that the DLPFC contain areas of immature cells that do not exhibit the regular layered
82 organization of the cerebral cortex (Courchesne et al., 2011b; Lainhart and Lange, 2011).
83 Complementing these findings, neuroimaging studies of ASD individuals have shown atypical
84 structural and functional connections between the PFC and other brain areas (García
85 Domínguez et al., 2013; Supekar et al., 2013).
86 Metabolic abnormalities in ASD have been mapped primarily using postmortem histological
87 methods, as well as non-invasively using 1H-magnetic resonance spectroscopy (1H-MRS,
88 henceforth MRS), a technique that allows the detection and quantification of absolute and
89 relative concentration of neurometabolites. Postmortem studies of frontal cortex abnormalities in
90 ASD have found GABAergic, glutamatergic, mitochondrial, and microglial dysfunction (Blatt,
91 2012; Wei et al., 2014; Varghese et al., 2017; Fetit et al., 2021). In parallel, MRS based studies
92 of ASD have shown lower GABA levels and higher glutamate levels in ASD children, in line with
93 postmortem studies. Interestingly, alterations in glutamate or combined glutamate and
94 glutamine (Glx) levels in ASD appear to vary between children and adults, with fewer
95 abnormalities, particularly in GABA levels, detected in adults, suggesting age or disorder
96 associated changes over the lifespan (Naaijen et al., 2015; Ajram et al., 2019a).
97 Many of the metabolic abnormalities documented in the DLPFC in ASD or other related
98 disorders also found associations with behavioral phenotypes. For instance, one study reported
99 a negative association between left DLPFC concentrations of glutamate with perspective taking
100 scores in ASD adults (Montag et al., 2008). In parallel, in electrophysiological studies,
101 cholinergic pathways have been associated with atypical social interaction and behaviors, as
102 well as with ASD symptom severity, orientation of attention and sensory integration (Lam et al.,
103 2006; Orekhova and Stroganova, 2014a). Furthermore, a recent 1H-MRS study conducted in
3
104 individuals with generalized anxiety disorder (GAD), a common comorbid diagnosis in ASD,
105 reported left DLPFC Cho/Cr levels inversely predicted IQ and predicted the severity of anxiety in
106 individuals diagnosed with generalized anxiety, but this was not the case in the neurotypical
107 control group, further highlighting the relationship between left DLPFC and cognitive functioning,
108 and underscoring the need for further research into behavioral correlates of choline containing
109 metabolites (Coplan et al., 2018a). Given the role of the DLPFC in cognitive function, as well as
110 its putative role in ASD, a better understanding of the metabolic profile of the DLPFC in ASD
111 might deepen our understanding of how documented metabolic DLPFC abnormalities might
112 contribute to the cognitive profile of ASD individuals (Smith et al., 2004a; Barbey et al., 2013a).
113 Here, we used 1H-MRS to investigate the neurometabolic profiles in the left DLPFC of 14
114 children with ASD ages 9-13, and 16 age-, sex-, and IQ- matched typically developing children.
115 Due to the small sample size of the study, we minimized the number of variables by focusing
116 only on two metabolites: Glx due to widely documented abnormalities in Glx in ASD (Cochran et
117 al., 2015a; Ajram et al., 2017, 2019b), and Cho, because of its previously mapped relationship
118 to IQ in the DLPFC (Barbey et al., 2013b; Coplan et al., 2018b), and to attention re-orienting
119 difficulties due to deficits in cholinergic arousal system (Deutsch et al., 2010; Anand et al., 2011;
120 Orekhova and Stroganova, 2014b). We were interested both in possible differences in absolute
121 levels and creatine ratios of these metabolites in ASD, and in any association these metabolites
122 may have with assessments of intellectual and executive function (specifically attention), as well
123 as severity of ASD. More specifically, since IQ and attentional switching have been shown to
124 correlate with metabolites in the DLPFC (Lauber et al., 1996; Smith et al., 2004b; Barbey et al.,
125 2013a), we hypothesized that DLPFC Cho and Glx levels would show significant associations
126 with these neuropsychological scores and may show distinct association patterns in ASD and
127 TD children.
128
129 2. METHODS & MATERIALS
130 2.1 Study participants and behavioral assessments
131 We recruited 17 male children with ASD and 17 TD children matched on IQ, age, sex, and
132 handedness, for this study. Of these, the MRS data of 3 of 17 ASD children and 1 of 17 TD
133 children did not pass quality control, as detailed below. The remaining sample therefore
134 consisted of 14 children with ASD and 16 TD children. Parents of the participants provided
135 informed consent according to protocols approved by the MGH Institutional Review Board (IRB).
136 Participant assent was also provided in addition to parent consent for participants aged 14-17.
137 Phenotypic data collected from all participants who passed MRS data quality control are
138 summarized in Table 1. The age range was 6–17 and 7–17 years in the TD and ASD groups,
139 respectively, with a mean age of 13.3 and median age of 13. All participants were right-handed,
140 with the exception of two ASD individuals, determined from information collected using the Dean
141 Questionnaire (Piro, 1998). Participants with ASD had a prior clinical diagnosis of ASD and met
142 ASD criteria on the Autism Diagnosis Observation Schedule, Version 2 (ADOS-2) (Lord et al.,
143 2012; Hus et al., 2014) administered by a trained research assistant with inter-rater reliability.
144 The Social Communication Questionnaire - Lifetime Version (SCQ Lifetime) (Rutter Bailey, A., &
4
145 Lord, C., 2003) was administered to further confirm ASD in ASD participants and rule out ASD
146 in TD participants. ASD participants who did not meet a cutoff of >15 on the SCQ or who had a
147 borderline score on the ADOS-2 were further evaluated by expert clinician and co-author Dr.
148 Robert Joseph to confirm the ASD diagnosis. Individuals with autism-related medical conditions
149 (e.g., Fragile-X syndrome, tuberous sclerosis) and other known risk factors (e.g., gestation <36
150 weeks) were excluded from the study. All TD participants were below the threshold on the SCQ
151 Lifetime questionnaire. Parent-questionnaires were administered to confirm that participants
152 were free of any neurological or psychiatric conditions and substance use in the past 6 months.
153 For ASD, the Social Responsiveness Scale (SRS-2) was used to assess the severity of the
154 ASD symptoms (Bruni, 2014). Verbal IQ (VIQ) and nonverbal IQ (NVIQ) were assessed using
155 the Differential Ability Scales – II (Beran, 2007) for all participants. The TD and ASD groups did
156 not differ on VIQ and showed a non-statistically significant trend towards a group difference on
157 NVIQ. Lastly, all participants completed the INN (Inhibition-Naming), INI (Inhibition-Inhibition),
158 and INS (Inhibition-Switching) sections of the NEPSY-II neurocognitive evaluation. Derived from
159 these sections, the Inhibition Contrast Scaled Score (ICS-I), which measures the ability to
160 voluntarily inhibit attention, and the Switching Contrast Scaled Score (ICS-S), which measures
161 the ability for switching attention between competing stimuli, were computed for each
162 participant.
163 Table 1. Characterization of the participants.
ASD (n = 14 males) TD (n = 16 males)
Mean (SD) Range Mean (SD) Range p-value
Age 11 (2) 8-14 10 (2) 7-14 0.36
NVIQ 100.9 (16.4) 70-127 112.4 (15.3) 91-149 0.06
VIQ 108.7 (19.4) 66-141 116.4 (11.4) 98-142 0.21
ICS-I 9.2 (3.2) 1-16 10.0 (3.7) 2-17 0.68
ICS-S 9.5 (3.7) 8-31 9.3 (4.0) 0-13 0.89
SRS 83.5 (9.6) 67-100 -- -- --

164
165 The p-values are from two-sample t-tests for the difference in means between the ASD and TD
166 groups. NVIQ: nonverbal IQ; VIQ: verbal IQ; ICS-I: inhibition of attention, as measured using the
167 NEPSY-II; ICS-S: attentional switching as measured using the NEPSY-II; SRS: Social
168 Responsiveness Scale.
169
170 2.2 Brain imaging data acquisition and processing
171 Brain imaging was performed using a 3T Siemens Trio MR scanner (Siemens Healthineers,
172 Erlangen Germany) equipped with a 12-channel head coil. In all participants, a high-resolution
173 multi-echo Magnetization Prepared - RApid Gradient Echo MEMPRAGE (T1-weighted structural
174 MRI) volume was also acquired (TR/TE1/TE2/TE3/TE4 = 2530/1.69/3.55/5.41/7.27 ms, flip
5
175 angle = 7°, voxel size = 1 mm isotropic), for the purpose of anatomical localization, MRS voxel
176 placement and the correction for partial volume effects of cerebrospinal fluid (CSF).
1
177 H-MRS protocol: Single voxel MRS was acquired using a conventional Point RESolved
178 Spectroscopy (PRESS) sequence (TE=30ms, TR=2.5s, bandwidth=1.2 kHz, and 96 averages,
179 1024 sample points). A 20x20x20 mm voxel was placed in the left DLPFC. The center of the
180 voxel was positioned in the axial slice above the superior margin of the left lateral ventricle.
181 (Figure 1A).
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204 Figure 1: Voxel placement, representative spectrum and brain matter fractions. (A) Voxel
205 placement on left DLPFC and representative spectrum. (B) Superimposed individual spectra for
206 ASD and TD groups. (C) Gray and white matter (GM, WM) fractions within the 8 cm3 left
207 DLPFC-voxel in ASD and TD subjects.
208
209 MR spectra were processed off-line using the LCModel software, version 6.3 (Provencher,
210 2001) for quantitative assessment of the following neurometabolites: N-acetylaspartate (NAA),
211 Creatine (Cr), choline (Cho), myo-inositol (mI), and glutamate+glutamine
212 (Glu+Gln). LCModel analysis was conducted on spectra within the chemical shift range 0.5–4.1
213 ppm. Spectra were excluded when the signal to noise ratio (SNR), estimated
214 by LCModel (defined as peak height of NAA divided by the root mean square of the noise of
215 the LCModel fit) was less than 5 and the Cramér-Rao lower bounds (CRLB) higher than 20%.
6
216 All metabolite levels were adjusted for gray matter (GM), white matter (WM), and cerebrospinal
217 fluid (CSF) contributions as follows: the MEMPRAGE images and the voxel coordinates from
218 the Siemens RDA files were used in Gannet toolkit version 3.1 (Edden et al., 2014) to generate
219 binary masks of the voxel location. These masks were then used in SPM 12 (Penny et al.,
220 2007) to calculate the partial volumes of GM, WM and CSF percentages within the voxel. The
221 segmented tissue fractions were then used to correct for metabolite concentrations quantified
222 using LC Model for CSF content according to the literature (Gasparovic et al., 2006).
223 2.3. Statistical analyses and multiple comparisons
224 Statistical analysis was performed using Prism GraphPad v9 (GraphPad, La Jolla, California),
225 since data were normally distributed (D'Agostino & Pearson test), unpaired, two-tailed t-tests for
226 the group comparisons (ASD vs TD). Pearson correlation coefficients were calculated to
227 assess the relationship between neurometabolite concentrations and each behavioral measure.
228 We used t-statistics to test for significant difference between slopes of the two lines, using the
229 slope, standard error, and sample size for each line. Additionally, we assessed effects of
230 combined metabolite levels in ASD and TD with a multivariate linear regression model in
231 GraphPad (least squares method: YBehavior= β0 + β1[met1]group + β2[met2]group, where dependent
232 variable Y is the behavior measure and continuous predictor variables are metabolite levels).
233 Since the two groups were age- matched, there was no significant difference in age between
234 the two groups. Therefore, age was not used as a covariate for the group analyses. For the
235 correlation analysis, we used creatine normalized (Glx/Cr) because based on the literature Glx
236 is age dependent in the age range used in this study, and in our own results there was a trend
237 towards a correlation with age. Therefore, Glx/Cr, which was not age-dependent in our study
238 was used in the analysis.
239 We tested for group differences in ASD versus TD in mean concentrations of Cho, Glx, Cr, for 3
240 tests in all; no significant group differences emerged, and so we did not correct the results for
241 multiple comparisons. We also tested the following correlations within each group: Cho
242 correlations with VIQ, NVIQ, SRS (ASD only), ICS-S, ICS-I, and Glx to creatine ratio (Glx/Cr)
243 with the same 5 behavioral metrics, for 10 correlations tests in total in the ASD group, and 8 in
244 the TD group. Lastly, we tested for significant differences between slopes for the ASD vs TD
245 groups only for correlations between Cho and NVIQ, and Glx/Cr and NVIQ.
246 When considering a correction for multiple comparisons, we first considered potential
247 dependencies across the behavioral measures. As expected, NVIQ and VIQ were significantly
248 correlated (combined groups, r=0.52, p=0.003), and NVIQ and ICS-S were also significantly
249 correlated (combined groups, r=0.53, p=0.004) ICS-I and SRS scores were not correlated with
250 ICS-S, ICS-I, VIQ or NVIQ. Because the behavioral measures are not statistically independent,
251 it is challenging to account for such dependencies when correcting for multiple comparisons.
252 We therefore chose to report uncorrected p-values, following recent best practices (Lu and
253 Belitskaya-Levy, 2015; Amrhein et al., 2019; Lowe, 2019). We used * to mark p values of 0.05
254 or below, and ** to mark p-values of 0.005 or below, i.e. p values that would survive a
255 Bonferroni correction for 10 comparisons.
256
7
257 3. RESULTS
258 3.1 Tissue composition within the MRS voxel.
259 We began by testing whether the voxels chosen across the two groups had similar
260 compositions. As expected, the groups did not differ significantly in mean grey matter, white
261 matter, or CSF fraction within the voxel (Figure 1B) (p > 0.05 for all).
262 3.2 Neurometabolite levels and age effects
263 Next, we investigated group differences in absolute metabolite levels, for Glx and Cho. No
264 significant differences in absolute levels of these two neurometabolites were observed between
265 groups (Figures 2A-B, and Table 2). Because maturation trajectories of metabolites in this age
266 range are not well understood (Nelson et al., 2019; Porges et al., 2021), we then tested for age
267 effects of metabolite concentrations within our sample. No significant age-related metabolite
268 changes were observed for either group (p>0.05) for Cho (Figure 2D), while a negative trend
269 was observed for Glx in the ASD group (r= -0.53, p=0.052) (Figure 2E). Given this trend, we
270 opted to normalize Glx levels with Creatine (Cr) levels, as is common practice (Wilson et al.,
271 2019). The Cr referenced Glx levels still did not differ between groups (Figure 2C), and the
272 trend with age in the ASD group was indeed removed by this normalization (Figure 2F).
273
274 Figure 2: Metabolite concentrations between groups, and correlation with age. Top Row: Group
275 comparisons for (A) Cho, (B) Glx, (C) Glx/Cr. Bottom row: Correlations between age and
276 metabolites levels for (D) Cho, (E) Glx, (F) ratio of Glx/Cr.
277
278
8
279 Table 2: Average metabolite levels
Diagnosis Cho Glx Glx/Cr

(mmol/l) (mmol/l)
ASD (n=14) 1.57±0.30 8.15±1.86 1.71±0.34
TD (n=16) 1.54±0.22 7.72±1.14 1.80±0.30
280 Values are the mean ± SD
281
282 3.3 Correlations between Cho and behavioral measures
283 We then investigated the relationship between Cho levels in the left DLPFC and VIQ/NVIQ,
284 ASD severity, and measures of attention. Between VIQ and NVIQ, significant correlations were
285 found only with NVIQ (Figure 3A). In the ASD group, absolute Cho levels showed a negative
286 correlation with NVIQ (r= -0.59, p=0.026). In contrast, in the TD group, Cho levels correlated
287 positively with NVIQ (r= -0.68, p=0.004). There was a significant group difference in the slopes,
288 indicating a significant group difference in the interaction between NVIQ and Cho.
289 No significant correlations were found between Cho and ASD SRS scores, which assess ASD
290 severity (Figure 3B).
291
9
292 Figure 3: Correlation between behavioral measures, and Cho. (A) NVIQ. r and p values are
293 shown for each group separately. A statistical ASD vs TD comparison of difference in slopes
294 showed a highly significant difference across the two groups (black text, p=0.0004) (B) SRS. (C)
295 inhibition of attention. (D) Attentional switching. Since both groups showed a similar trend, we
296 also computed r and p values for both groups combined (purple text). Red: ASD. Blue: TD.
297 Purple: ASD and TD groups combined. Dotted blue and red lines indicate the 95% confidence
298 intervals for regression lines. Dashed purple lines indicate the slope of ASD and TD combined.
299 r=correlation coefficient (Pearson’s). p=uncorrected p-values.
300 Lastly, we looked at two behavioral measures of attention: inhibition of attention (ICS-I) and
301 attentional switching (ICS-S). While no significant correlations were found between ICS-I and
302 Cho (Figure 3C), ICS-S values in both groups correlated significantly with Cho (Figure 3D), and
303 this significant correlation was especially pronounced in the ASD group (r=-0.69, p=0.009).
304 There was no group difference in the direction of this correlation, and indeed, combining the two
305 groups strengthened the correlation (r=-0.56, p=0.002).
306
307 3.4 Correlations between Glx/Cr and behavior measures
308 The same analyses were then repeated with Glx/Cr instead of Cho (Section 3.3). As with Cho,
309 VIQ did not correlate with Glx/Cr in either group. As with Cho, Glx/Cr was significantly correlated
310 with NVIQ in the ASD group, albeit in the opposite direction – positively rather than negatively;
311 in contrast to the finding with Cho, there was no significant correlation between Glx/Cr and
312 NVIQ in the TD group (Figure 4A).
313 No significant association was found between Glx/Cr levels and the SRS scores (ASD group)
314 (Figure 4B). Glx/Cr did not correlate with ICS-I (Figure 4C) and was positively correlated with
315 attentional switching (Figure 4D) for both ASD (r=0.73, p=0.004) and TD (r=0.54, p=0.040)
316 groups. Indeed, when the ASD and TD groups were combined, we again observed an increased
317 statistical significance for the correlation between ICS-S and Glx/Cr (r=0.61, p=0.0005).
318 Lastly, a multivariate linear regression model: Y[Beh] ~ Intercept (β0) + β1[Glx/Cr][TD/ASD] + β2
319 [Cho][TD/ASD], showed that the combined Cho and Glx/Cr predicted both behavioral measures,
320 NVIQ (r=0.70, p=0.023) and ICS-S (r=0.80, p=0.005) among ASD children and NVIQ (r=0.72,
321 p=0.009) among TD children (Figure 5).
322
10
323
324 Figure 4: Correlation between behavioral measures, and Glx/Cr. (A) NVIQ. r and p values
325 are shown for each group separately. A statistical ASD vs TD comparison of difference in slopes
326 showed a significant difference across the two groups (black text, p=0.044) (B) SRS. (C)
327 inhibition of attention. (D) Attentional switching. Since both groups showed a similar trend, we
328 also computed r and p values for both groups combined (purple text). Red: ASD. Blue: TD.
329 Purple: ASD and TD groups combined. Dashed purple lines indicate the slope of ASD and TD
330 combined. Dotted blue and red lines indicate the 95% confidence intervals for regression lines.
331 r=correlation coefficient (Pearson’s). p=uncorrected p-values.
332
11
333
334 Figure 5. Multivariate linear regression model in ASD and TD. Performance of combined
335 Cho+Glx/Cr predicting (A) NVIQ and (B) ICS-S
336
337
338 4. DISCUSSION
339
340 In this study, no significant differences in the left DLPFC metabolite levels, either
341 absolute or creatine referenced, were found between children with ASD and TD children. In
342 spite of no differences in absolute levels of neurometabolites, in agreement with the initial
343 hypotheses, we found significant associations between neuropsychological profiles and
344 metabolites of interest. Specifically, we observed a significant negative correlation between Cho
345 and NVIQ for children with ASD versus a significant positive correlation between these values
346 for the TD group. Choline also showed a significant negative correlation with the ICS-S score
347 (attentional switching) within the ASD group. Even though this association was not significant for
348 the TD group, combining both groups yielded stronger negative statistical significance, meaning
349 both groups followed a similar trend for this association. NVIQ was also positively correlated
350 with the Glx/Cr ratio in the TD group, but not in the ASD group. Additionally, the combination of
351 Cho and Glx/Cr predicted both NVIQ and ICS-S in the ASD group, but only NVIQ in the TD
352 group.
353 Choline containing compounds are key components of cell membranes, and essential
354 for the synthesis of acetylcholine (ACh), a neurotransmitter linked with learning, memory and
355 attention, that has also been associated with cognitive dysfunction and anxiety severity
356 (Ferguson et al., 2003; Moon and Jeong, 2015). Additionally, Cho levels are also associated
357 with the dopaminergic neuronal transduction and synaptic activity and overall dopaminergic and
358 cholinergic balance, involved in attention, memory, and academic achievements (Wiguna et al.,
359 2012). A previous study of individuals with generalized anxiety disorder reported an inverse
360 association with the left DLPFC Cho/Cr and full-scale IQ (FSIQ), while no such relationship was
361 observed for the control group (Coplan et al., 2018b). Furthermore, Cho levels have also been
362 associated with white matter density and speculated to reflect excessive neuronal connectivity
363 or abnormal myelination (Laycock et al., 2008). Our finding of a significant association between
12
364 low Cho levels and higher NVIQ and better performance on attentional switching in ASD is
365 consistent with the classical view. It is also consistent with prior studies (Jung et al., 1999;
366 Coplan et al., 2018b) that found lower Cho levels were associated with increased IQ, and higher
367 Cho levels were associated with increased anxiety Index.
368 In parallel to the findings of associations with Cho levels, we also found a significant
369 positive correlation between Glx/Cr ratio and NVIQ in the ASD group, but not in the TD group.
370 Glx/Cr ratio also showed a positive association with the ICS-S score for attention switching for
371 both ASD and TD groups, and the combination of the two groups showed a much stronger
372 statistical significance. Glutamate is synthesized from glutamine in neurons and released into
373 the synapse, which is converted back into glutamine in glial cells, by the enzyme glutamine-
374 synthetase. Glutamate to GABA conversion is catalyzed by the neuronal enzyme glutamate
375 decarboxylase. Glutamate and GABA have been found as reliable markers of cortical excitability
376 and inhibition, and thus critical for the mechanisms of neuroplasticity and learning. Moreover,
377 brain excitation and inhibition levels are thought to be critical for triggering the onset of sensitive
378 periods for cognitive skill acquisition by shaping plastic responsiveness of underlying neural
379 systems in response to environmental stimulation (Werker and Hensch, 2015). MRS studies of
380 Glx in neurodevelopmental disorders, including ASD and ADHD, have reported abnormal
381 glutamate or Glx levels in patient populations relative to typically developing subjects (Carrey et
382 al., 2007; Brown et al., 2013; Cochran et al., 2015b). Higher levels of glutamate have been
383 hypothesized to indicate hyperexcitability in these disorders that affect neuronal-network
384 dynamics involved in learning, and memory consolidation (Pugh et al., 2014a). No group-level
385 differences were found in Glx or Glx/Cr in our study. Yet, we found a significant positive
386 correlation between Glx/Cr and nonverbal intelligence only in the ASD group. This may suggest
387 that DLPFC Glx in ASD may have a different glutamatergic transmission than that of TD
388 children. It also suggests a possible neural mechanism within the DLPFC for the heterogeneity
389 observed in cognitive function in ASD.
390 There is a multitude of evidence suggesting a neurochemical basis of higher-

391 level cognitive skills, yet the exact mechanisms remain largely unknown. Across developmental
392 disorders such as ASD and attention-related disorders, neurometabolites profiles has been
393 found to vary compared to typically developing age-matched counterparts (Perlov et al., 2009;
394 Baruth et al., 2013). In a 1H-MRS study using children with reading disabilities (RD), it was
395 shown that RD children had elevated Cho and glutamate relative to TD children. They further
396 indicated that Cho and glutamate levels inversely correlated with reading and related language
397 measures such that increased concentrations were associated with poorer performance (Pugh
398 et al., 2014b). A more recent RD study reported a negative association between Cho and
399 reading ability of children aged between 6 and 8 (Del Tufo et al., 2018). The same study also
400 observed a positive relationship between glutamate concentration and reading performance.
401 Lastly, the fact that in both the ASD and TD groups both Cho and Glx/Cr levels were correlated
402 with attentional switching scores, but not with attentional inhibition, is consistent with the
403 putative role of the prefrontal cortex in attention, since its role for attentional switching, but not
404 inhibition of attention, is well established (Rossi et al., 2009). Attentional switching happens in
405 the twin anticorrelated functional networks involving the DLPFC as a task-positive functional
406 site, demonstrated by Fox and colleagues (Fox et al., 2005). The task-switching aspect and its
13
407 relationship with neurometabolites in ASD is worth exploring as a biomarker for further
408 investigation.
409 Several limitations of the present study should be noted. Our final analysis included just
410 14 children with ASD and 16 TD children. Future studies with larger samples are needed to gain
411 greater statistical power and validate our findings. Since we included males between the ages of
412 7-14 years, our results in this study are representative for only males in this age
413 group. Furthermore, the cross-sectional design of our study limited the power to detect age-
414 driven changes in the neurometabolic profile which could hinder ability to identify disorder
415 specific biomarkers in ASD. Therefore, longitudinal MRS studies are clearly necessary. Despite
416 these limitations it should be noted that it is unlikely that our findings can be fully explained by
417 potential confounds, such as differences in voxel tissue composition, age- or IQ. There were no
418 significant between-group differences in voxel brain matter and CSF, metabolite concentrations
419 were tissue corrected, and the results are consistent with prior metabolites studies of ASD and
420 of other related disorders. Furthermore, note that the results of Figure 4A in the ASD group, the
421 correlation between NVIQ and Glx/Cr, were primarily driven by one outlier participant who had
422 not only the lowest NVIQ score in the ASD cohort, but also the lowest VIQ and total IQ scores.
423 This underscores the need to replicate and extend such studies to cohorts of ASD participants
424 with greater IQ ranges, and especially to ASD participants with below average IQ scores.
425 In summary, the main findings revealed no significant group difference in Cho or Glx/Cr
426 metabolite levels between groups, alongside relevant correlations with behavioral measures.
427 Specifically, we documented an inverse relationship between both Cho and Glx/Cr and
428 nonverbal intelligence in ASD children, which was opposite in direction when compared with
429 age-, IQ- matched TD children. This suggests that Cho and Glx/Cr may have different
430 neurometabolic roles in ASD compared to TD children in overall cognitive function. In contrast,
431 for both groups, attention switching scores showed an inverse relationship with Cho and a
432 positive correlation with Glx/Cr, suggesting further that the roles of metabolites for executive
433 function skills, and specifically attentional switching and inhibition of attention, are preserved in
434 ASD. These results are especially interesting in the context of the significant heterogeneity
435 associated with the ASD phenotype (Lenroot and Yeung, 2013), since they suggest possible
436 underlying neural mechanisms for the cognitive heterogeneity in particular, which do not overlap
437 which those expected from typical development. Although the mechanisms connecting these
438 neurometabolites to atypical cognitive functioning alongside typical executive functioning remain
439 to be elucidated, the overall finding that the role of neurometabolites in ASD does not follow a
440 consistent pattern relative to TD children may provide valuable information on potential
441 neurobiological pathways of atypical development, and the heterogeneities associated with
442 cognitive function in ASD in particular. Indeed, a better understanding of the
443 neurobiological/chemical underpinnings of how heterogeneity in cognitive and executive
444 function may correlated with metabological underpinnings is critical for advancing our
445 understanding of the neural bases that underlie heterogeneity in ASD.
446
447
14
448 ACKNOWLEDGMENT
449 We like to thank our participants and their families. This study was supported by the following:
450 The National Institutes of Health (R01NS048455, MH); the Nancy Lurie Marks Foundation (MH);
451 MIT-MGH Strategic Partnership grant from the Executive Committee on Research (ECOR) at
452 Massachusetts General Hospital (EMR); the National Institute of Child Health and Development
453 (R01HD073254, TK); the National Institute of Mental Health (R01MH117998, TK)
454 Author contributions
455 EMR, TK, MRH, AW: Conceptualization; EMR, MRH, NMM, TK: Data curation; AW, AIM: Data
456 analysis; EMR, TK, MRH: Methodology; Project administration; Supervision; EMR, TK, AW:
457 Validation; Visualization; Writing: AK, EMR and TK wrote the manuscript, all other authors
458 reviewed and or edited the manuscript.
459 CONFLICT OF INTEREST
460 The authors have no conflicts of interest to declare in relation to this work. E.M.R. serves on the
461 Scientific Advisory Board of BrainSpec Inc.
462
463 REFERENCES
464 Ajram, L. A., Horder, J., Mendez, M. A., Galanopoulos, A., Brennan, L. P., Wichers, R. H., et al.
465 (2017). Shifting brain inhibitory balance and connectivity of the prefrontal cortex of adults
466 with autism spectrum disorder. Transl Psychiatry. doi: 10.1038/tp.2017.104.
467 Ajram, L. A., Pereira, A. C., Durieux, A. M. S., Velthius, H. E., Petrinovic, M. M., and McAlonan,
468 G. M. (2019a). The contribution of [1H] magnetic resonance spectroscopy to the study of
469 excitation-inhibition in autism. Prog Neuropsychopharmacol Biol Psychiatry. doi:
470 10.1016/j.pnpbp.2018.09.010.
472 G. M. (2019b). The contribution of [1H] magnetic resonance spectroscopy to the study of
474 10.1016/j.pnpbp.2018.09.010.
475 Alexander, G. E., DeLong, M. R., and Strick, P. L. (1986). Parallel organization of functionally
476 segregated circuits linking basal ganglia and cortex. Annu Rev Neurosci. doi:
477 10.1146/annurev.ne.09.030186.002041.
478 Amrhein, V., Greenland, S., Mcshane, B., Wasserstein, R. L., Schirm, A. L., and Lazar, N. A.
479 (2019). Moving to a World Beyond “ p < 0.05.” Am Stat.
480 Anand, R., A., S., Ponath, G., I., J., Nasir, M., and B., S. (2011). “Nicotinic Acetylcholine
481 Receptor Alterations in Autism Spectrum Disorders – Biomarkers and Therapeutic
482 Targets,” in Autism - A Neurodevelopmental Journey from Genes to Behaviour doi:
483 10.5772/20752.
15
484 Barbey, A. K., Colom, R., and Grafman, J. (2013a). Dorsolateral prefrontal contributions to
485 human intelligence. Neuropsychologia. doi: 10.1016/j.neuropsychologia.2012.05.017.
486 Barbey, A. K., Colom, R., and Grafman, J. (2013b). Dorsolateral prefrontal contributions to
488 Baruth, J. M., Wall, C. A., Patterson, M. C., and Port, J. D. (2013). Proton Magnetic Resonance
489 Spectroscopy as a Probe into the Pathophysiology of Autism Spectrum Disorders (ASD): A
490 Review. Autism Research. doi: 10.1002/aur.1273.
491 Beran, T. N. (2007). Elliott, C. D. (2007). Differential Ability Scales (2nd ed.). San Antonio, TX:
492 Harcourt Assessment. Can J Sch Psychol. doi: 10.1177/0829573507302967.
493 Blatt, G. J. (2012). The Neuropathology of Autism. Scientifica (Cairo). doi:
494 10.6064/2012/703675.
495 Brown, M. S., Singel, D., Hepburn, S., and Rojas, D. C. (2013). Increased glutamate
496 concentration in the auditory cortex of persons with autism and first-degree relatives: A 1H-
497 MRS study. Autism Research. doi: 10.1002/aur.1260.
498 Bruni, T. P. (2014). Test Review: Social Responsiveness Scale–Second Edition (SRS-2). J
499 Psychoeduc Assess. doi: 10.1177/0734282913517525.
500 Carrey, N. J., MacMaster, F. P., Gaudet, L., and Schmidt, M. H. (2007). Striatal creatine and
501 glutamate/glutamine in attention-deficit/ hyperactivity disorder. J Child Adolesc
502 Psychopharmacol. doi: 10.1089/cap.2006.0008.
503 Cochran, D. M., Sikoglu, E. M., Hodge, S. M., Edden, R. A. E., Foley, A., Kennedy, D. N., et al.
504 (2015a). Relationship among glutamine, γ-aminobutyric acid, and social cognition in autism
505 spectrum disorders. J Child Adolesc Psychopharmacol. doi: 10.1089/cap.2014.0112.
507 (2015b). Relationship among glutamine, γ-aminobutyric acid, and social cognition in autism
509 Coplan, J. D., Webler, R., Gopinath, S., Abdallah, C. G., and Mathew, S. J. (2018a).
510 Neurobiology of the dorsolateral prefrontal cortex in GAD: Aberrant neurometabolic
511 correlation to hippocampus and relationship to anxiety sensitivity and IQ. J Affect Disord.
512 doi: 10.1016/j.jad.2017.12.001.
513 Coplan, J. D., Webler, R., Gopinath, S., Abdallah, C. G., and Mathew, S. J. (2018b).
516 doi: 10.1016/j.jad.2017.12.001.
517 Courchesne, E., Mouton, P. R., Calhoun, M. E., Semendeferi, K., Ahrens-Barbeau, C., Hallet,
518 M. J., et al. (2011a). Neuron number and size in prefrontal cortex of children with autism.
519 JAMA - Journal of the American Medical Association. doi: 10.1001/jama.2011.1638.
521 M. J., et al. (2011b). Neuron number and size in prefrontal cortex of children with autism.
16
523 Del Tufo, S. N., Frost, S. J., Hoeft, F., Cutting, L. E., Molfese, P. J., Mason, G. F., et al. (2018).
524 Neurochemistry predicts convergence of written and spoken language: A proton magnetic
525 resonance spectroscopy study of cross-modal language integration. Front Psychol. doi:
526 10.3389/fpsyg.2018.01507.
527 Deutsch, S. I., Urbano, M. R., Neumann, S. A., Burket, J. A., and Katz, E. (2010). Cholinergic
528 abnormalities in autism: Is there a rationale for selective nicotinic agonist interventions?
529 Clin Neuropharmacol. doi: 10.1097/WNF.0b013e3181d6f7ad.
530 Edden, R. A. E., Puts, N. A. J., Harris, A. D., Barker, P. B., and Evans, C. J. (2014). Gannet: A
531 batch-processing tool for the quantitative analysis of gamma-aminobutyric acid-edited MR
532 spectroscopy spectra. Journal of Magnetic Resonance Imaging. doi: 10.1002/jmri.24478.
533 Ferguson, S. M., Savchenko, V., Apparsundaram, S., Zwick, M., Wright, J., Heilman, C. J., et al.
534 (2003). Vesicular Localization and Activity-Dependent Trafficking of Presynaptic Choline
535 Transporters. Journal of Neuroscience. doi: 10.1523/jneurosci.23-30-09697.2003.
536 Fetit, R., Hillary, R. F., Price, D. J., and Lawrie, S. M. (2021). The neuropathology of autism: A
537 systematic review of post-mortem studies of autism and related disorders. Neurosci
538 Biobehav Rev. doi: 10.1016/j.neubiorev.2021.07.014.
539 Forde, N., Llera, A., Dell’Acqua, F., Ecker, C., Buitelaar, J. K., and Beckmann, C. F. (2020).
540 P.124 Linking functional and structural brain organisation in autism spectrum disorder.
541 European Neuropsychopharmacology. doi: 10.1016/j.euroneuro.2020.09.103.
542 Fox, M. D., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., and Raichle, M. E.
543 (2005). The human brain is intrinsically organized into dynamic, anticorrelated functional
544 networks. Proc Natl Acad Sci U S A. doi: 10.1073/pnas.0504136102.
545 García Domínguez, L., Stieben, J., Pérez Velázquez, J. L., and Shanker, S. (2013). The
546 Imaginary Part of Coherency in Autism: Differences in Cortical Functional Connectivity in
547 Preschool Children. PLoS One. doi: 10.1371/journal.pone.0075941.
548 Gasparovic, C., Song, T., Devier, D., Bockholt, H. J., Caprihan, A., Mullins, P. G., et al. (2006).
549 Use of tissue water as a concentration reference for proton spectroscopic imaging. Magn
550 Reson Med. doi: 10.1002/mrm.20901.
551 Haznedar, M. (2006). Volumetric Analysis and Three-Dimensional Glucose Metabolic Mapping
552 of the Striatum and Thalamus in Patients With Autism Spectrum Disorders. American
553 Journal of Psychiatry. doi: 10.1176/appi.ajp.163.7.1252.
554 Hiremath, C. S., Sagar, K. J. V., Yamini, B. K., Girimaji, A. S., Kumar, R., Sravanti, S. L., et al.
555 (2021). Emerging behavioral and neuroimaging biomarkers for early and accurate
556 characterization of autism spectrum disorders: a systematic review. Transl Psychiatry. doi:
557 10.1038/s41398-020-01178-6.
558 Howlin, P. (2021). Adults with Autism: Changes in Understanding Since DSM-111. J Autism Dev
559 Disord. doi: 10.1007/s10803-020-04847-z.
560 Hus, V., Gotham, K., and Lord, C. (2014). Standardizing ADOS domain scores: Separating
561 severity of social affect and restricted and repetitive behaviors. J Autism Dev Disord. doi:
562 10.1007/s10803-012-1719-1.
17
563 Jung, R. E., Brooks, W. M., Yeo, R. A., Chiulli, S. J., Weers, D. C., and Sibbitt, W. L. (1999).
564 Biochemical markers of intelligence: A proton MR spectroscopy study of normal human
565 brain. Proceedings of the Royal Society B: Biological Sciences. doi:
566 10.1098/rspb.1999.0790.
567 Kalbe, E., Schlegel, M., Sack, A. T., Nowak, D. A., Dafotakis, M., Bangard, C., et al. (2010).
568 Dissociating cognitive from affective theory of mind: A TMS study. Cortex. doi:
569 10.1016/j.cortex.2009.07.010.
570 Kim, S. Y., Choi, U. S., Park, S. Y., Oh, S. H., Yoon, H. W., Koh, Y. J., et al. (2015). Abnormal
571 activation of the social brain network in children with autism spectrum disorder: An fMRI
572 study. Psychiatry Investig. doi: 10.4306/pi.2015.12.1.37.
573 Lainhart, J. E., and Lange, N. (2011). Increased neuron number and head size in autism. JAMA
574 - Journal of the American Medical Association. doi: 10.1001/jama.2011.1633.
575 Lam, K. S. L., Aman, M. G., and Arnold, L. E. (2006). Neurochemical correlates of autistic
576 disorder: A review of the literature. Res Dev Disabil. doi: 10.1016/j.ridd.2005.03.003.
577 Lauber, E. J., Meyer, D. E., Evans, J. E., Rubinstein, J., Gmeindl, L., Junck, L., et al. (1996).
578 The brain areas involved in the executive control of task switching as revealed by PET.
579 Neuroimage. doi: 10.1016/s1053-8119(96)80249-7.
580 Laycock, S. K., Wilkinson, I. D., Wallis, L. I., Darwent, G., Wonders, S. H., Fawcett, A. J., et al.
581 (2008). Cerebellar volume and cerebellar metabolic characteristics in adults with dyslexia.
582 in Annals of the New York Academy of Sciences doi: 10.1196/annals.1416.002.
583 Lenroot, R. K., and Yeung, P. K. (2013). Heterogeneity within autism spectrum disorders: What
584 have we learned from neuroimaging studies? Front Hum Neurosci. doi:
585 10.3389/fnhum.2013.00733.
586 Lord, C., Petkova, E., Hus, V., Gan, W., Lu, F., Martin, D. M., et al. (2012). A multisite study of
587 the clinical diagnosis of different autism spectrum disorders. Arch Gen Psychiatry. doi:
588 10.1001/archgenpsychiatry.2011.148.
589 Lowe, N. K. (2019). The Push to Move Health Care Science Beyond p <.05. JOGNN - Journal
590 of Obstetric, Gynecologic, and Neonatal Nursing 48, 493–494. doi:
591 10.1016/j.jogn.2019.07.005.
592 Lu, Y., and Belitskaya-Levy, I. (2015). The debate about p-values. Shanghai Arch Psychiatry.
593 doi: 10.11919/j.issn.1002-0829.216027.
594 Margari, L., De Giacomo, A., Craig, F., Palumbi, R., Peschechera, A., Margari, M., et al. (2018).
595 Frontal lobe metabolic alterations in autism spectrum disorder: A1h-magnetic resonance
596 spectroscopy study. Neuropsychiatr Dis Treat. doi: 10.2147/ndt.s165375.
597 Montag, C., Schubert, F., Heinz, A., and Gallinat, J. (2008). Prefrontal cortex glutamate
598 correlates with mental perspective-taking. PLoS One. doi: 10.1371/journal.pone.0003890.
599 Moon, C. M., and Jeong, G. W. (2015). Functional neuroanatomy on the working memory under
600 emotional distraction in patients with generalized anxiety disorder. Psychiatry Clin
601 Neurosci. doi: 10.1111/pcn.12295.
18
602 Müller, R. A., and Fishman, I. (2018). Brain Connectivity and Neuroimaging of Social Networks
603 in Autism. Trends Cogn Sci. doi: 10.1016/j.tics.2018.09.008.
604 Naaijen, J., Lythgoe, D. J., Amiri, H., Buitelaar, J. K., and Glennon, J. C. (2015). Fronto-striatal
605 glutamatergic compounds in compulsive and impulsive syndromes: A review of magnetic
606 resonance spectroscopy studies. Neurosci Biobehav Rev. doi:
607 10.1016/j.neubiorev.2015.02.009.
608 Nelson, M. B., O’Neil, S. H., Wisnowski, J. L., Hart, D., Sawardekar, S., Rauh, V., et al. (2019).
609 Maturation of brain microstructure and metabolism associates with increased capacity for
610 self-regulation during the transition from childhood to adolescence. Journal of
611 Neuroscience. doi: 10.1523/jneurosci.2422-18.2019.
612 Orekhova, E. V., and Stroganova, T. A. (2014a). Arousal and attention re-orienting in autism
613 spectrum disorders: Evidence from auditory event-related potentials. Front Hum Neurosci.
614 doi: 10.3389/fnhum.2014.00034.
615 Orekhova, E. V., and Stroganova, T. A. (2014b). Arousal and attention re-orienting in autism
617 doi: 10.3389/fnhum.2014.00034.
618 Penny, W., Friston, K., Ashburner, J., Kiebel, S., and Nichols, T. (2007). Statistical Parametric
619 Mapping: The Analysis of Functional Brain Images. doi: 10.1016/B978-0-12-372560-
620 8.X5000-1.
621 Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., et al. (2009).
622 Spectroscopic findings in attention-deficit/hyperactivity disorder: Review and meta-analysis.
623 World Journal of Biological Psychiatry. doi: 10.1080/15622970802176032.
624 Piro, J. M. (1998). Handedness and intelligence: Patterns of hand preference in gifted and
625 nongifted children. Dev Neuropsychol. doi: 10.1080/87565649809540732.
626 Porges, E. C., Jensen, G., Foster, B., Edden, R. A. E., and Puts, N. A. J. (2021). The trajectory
627 of cortical gaba across the lifespan, an individual participant data meta-analysis of edited
628 mrs studies. Elife. doi: 10.7554/eLife.62575.
629 Provencher, S. W. (2001). Automatic quantitation of localized in vivo 1H spectra with LCModel.
630 NMR Biomed. doi: 10.1002/nbm.698.
631 Pugh, K. R., Frost, S. J., Rothman, D. L., Hoeft, F., Del Tufo, S. N., Mason, G. F., et al. (2014a).
632 Glutamate and choline levels predict individual differences in reading ability in emergent
633 readers. Journal of Neuroscience. doi: 10.1523/JNEUROSCI.3907-13.2014.
634 Pugh, K. R., Frost, S. J., Rothman, D. L., Hoeft, F., Del Tufo, S. N., Mason, G. F., et al. (2014b).
637 Rossi, A. F., Pessoa, L., Desimone, R., and Ungerleider, L. G. (2009). The prefrontal cortex and
638 the executive control of attention. in Experimental Brain Research doi: 10.1007/s00221-
639 008-1642-z.
19
640 Rutter Bailey, A., & Lord, C., M. (2003). Manual of the Social Communication Questionnaire.
641 Los Angeles, CA.
642 Schmitz, N., Daly, E., and Murphy, D. (2007). Frontal anatomy and reaction time in Autism.
643 Neurosci Lett. doi: 10.1016/j.neulet.2006.07.077.
644 Smith, A. B., Taylor, E., Brammer, M., and Rubia, K. (2004a). Neural Correlates of Switching
645 Set as Measured in Fast, Event-Related Functional Magnetic Resonance Imaging. Hum
646 Brain Mapp. doi: 10.1002/hbm.20007.
647 Smith, A. B., Taylor, E., Brammer, M., and Rubia, K. (2004b). Neural Correlates of Switching
650 Stockman, J. A. (2013). Neuron Number and Size in Prefrontal Cortex of Children With Autism.
651 Yearbook of Pediatrics. doi: 10.1016/j.yped.2011.12.022.
652 Stoner, R., Chow, M. L., Boyle, M. P., Sunkin, S. M., Mouton, P. R., Roy, S., et al. (2014).
653 Patches of Disorganization in the Neocortex of Children with Autism. New England Journal
654 of Medicine. doi: 10.1056/nejmoa1307491.
655 Supekar, K., Uddin, L. Q., Khouzam, A., Phillips, J., Gaillard, W. D., Kenworthy, L. E., et al.
656 (2013). Brain Hyperconnectivity in Children with Autism and its Links to Social Deficits. Cell
657 Rep. doi: 10.1016/j.celrep.2013.10.001.
658 Varghese, M., Keshav, N., Jacot-Descombes, S., Warda, T., Wicinski, B., Dickstein, D. L., et al.
659 (2017). Autism spectrum disorder: neuropathology and animal models. Acta Neuropathol.
660 doi: 10.1007/s00401-017-1736-4.
661 Wei, H., Alberts, I., and Li, X. (2014). The apoptotic perspective of autism. International Journal
662 of Developmental Neuroscience. doi: 10.1016/j.ijdevneu.2014.04.004.
663 Werker, J. F., and Hensch, T. K. (2015). Critical periods in speech perception: New directions.
664 Annu Rev Psychol. doi: 10.1146/annurev-psych-010814-015104.
665 Wiguna, T., Guerrero, A. P. S., Wibisono, S., and Sastroasmoro, S. (2012). Effect of 12-week
666 administration of 20-mg long-acting methylphenidate on Glu/Cr, NAA/Cr, Cho/Cr, and
667 mI/Cr ratios in the prefrontal cortices of school-age children in Indonesia: A study using 1H
668 Magnetic Resonance Spectroscopy (MRS). Clin Neuropharmacol. doi:
669 10.1097/WNF.0b013e3182452572.
670 Wilson, M., Andronesi, O., Barker, P. B., Bartha, R., Bizzi, A., Bolan, P. J., et al. (2019).
671 Methodological consensus on clinical proton MRS of the brain: Review and
672 recommendations. Magn Reson Med. doi: 10.1002/mrm.27742.
673
20
Manuscript Click here to
access/download;Manuscript;Manuscript_Weerasekera et


6 Maria Ratai1*
7
1
2
3
4
15 USA.
5
6
19
20
21
22
24
25

32 Phone: +1-617-726-1744
33 Fax: (617) 726-7422
34
1
35 ABSTRACT
36 The neurometabolic profile associated with autism spectrum disorder (ASD) has been reported
39 (DLPFC) is of particular interest due to the DLPFC relevance to cognitive and executive
40 function, and to ASD. We used 1H-MRS to investigate neurometabolic profiles in the DLPFC of
41 ASD and sex/IQ-matched typically developing (TD) children (ages 9-13). We focused on levels
42 of Glutamate and Glutamine (Glx) due to many reported Glx abnormalities ASD, and of Choline
44 difficulties. While no significant group differences were observed in concentrations (absolute or
45 creatine referenced), metabolite levels were correlated with the behavioral phenotype of ASD
46 children. In the ASD group but not the TD group, nonverbal IQ (NVIQ) was negatively
47 associated with Cho (r=-0.59, p=0.026) and positively associated with Glx/Cr (r=0.66, p=0.011).
48 Furthermore, attentional-switching scores in the ASD group correlated negatively with Cho (r=-
49 0.69, p=0.009), and positively with Glx to creatine ratio (Glx/Cr) for both the ASD (r=0.73,
50 p=0.004) and TD (r=0.54, p=0.040) groups. Cho and Glx/Cr have different neurometabolic roles
51 in modulating NVIQ in ASD compared to TD children, while their role in attentional switching
52 seems preserved in ASD. Elucidating the role of neurometabolites in ASD also in the absence
53 of significant group differences in absolute levels is important, as sometimes even in the
54 absence of group differences in means, differences in the ASD group might still emerge that are
55 lost when only group means are considered. Understanding how individual neurophysiological
56 markers might correlate with ASD specific traits similarly to, or differently from, correlations with
57 the same traits in TD populations, and in this specific case NVIQ, is therefore a step towards
58 understanding and mapping the neural correlates of ASD.
59
60
61
62
64
2
65 1. INTRODUCTION
66

69 behaviors and interests (Cooper, 2017; Hodges et al., 2020). Studies of ASD have shown a
70 wide range of abnormalities in brain function and structure (Müller and Fishman, 2018; Forde et
71 al., 2020), as well as abnormalities in the levels of brain metabolites (Ford and Crewther, 2016).
72 However, many questions about the underlying neuropathology of ASD remain unanswered
73 (Blatt, 2012; Fetit et al., 2021a).
79 et al., 2015). In particular, structural, and metabolic irregularities have been reported in the
94 2012; Wei et al., 2014; Varghese et al., 2017; Fetit et al., 2021b). In parallel, MRS based studies
3
105 well as with ASD symptom severity, orientation of attention and sensory integration (Stroganova
106 et al., 2013; Orekhova and Stroganova, 2014a). Furthermore, a recent 1H-MRS study conducted
107 in individuals with generalized anxiety disorder (GAD), a common comorbid diagnosis in ASD,
130 TD children.
131
4
165 participant.
5
166 Seven of the 14 boys in the ASD group used psychotropic medication (Ritalin, Citalopram,
167 Focalin, Adderall, Sertraline, Concerta and Risperidone). None of the participants in the TD
168 group were on medication.
169
Age 11 (2) 8-14 10 (2) 7-14 0.36
NVIQ 100.9 (16.4) 70-127 112.4 (15.3) 91-149 0.06
VIQ 108.7 (19.4) 66-141 116.4 (11.4) 98-142 0.21
ICS-I 9.2 (3.2) 1-16 10.0 (3.7) 2-17 0.68
ICS-S 9.5 (3.7) 8-31 9.3 (4.0) 0-13 0.89
SRS 83.5 (9.6) 67-100 -- -- --

171
176
1
186 1024 sample points) with VAPOR (Variable Power and Optimized Relaxation Delays) water
187 suppression method. A 20x20x20 mm voxel was placed in the left DLPFC. In addition,
188 unsuppressed water signal (number of averages = 4) was acquired for absolute concentration
189 calculation. The center of the voxel was positioned in the axial slice above the superior margin
190 of the left lateral ventricle. (Figure 1A).
6
191
192
193
194
195 Figure 1: Voxel placement, representative spectrum, and brain matter fractions. (A) Voxel
199
201 2001) with a simulated basis set for quantitative assessment of the following neurometabolites:
202 N-acetylaspartate (NAA), Creatine (Cr), choline (Cho), myo-inositol (mI), and
203 glutamate+glutamine (Glu+Gln). LCModel analysis was conducted on spectra within the
204 chemical shift range 0.5–4.1 ppm. Spectra were excluded when the signal to noise ratio (SNR),
205 estimated by LCModel (defined as peak height of NAA divided by the root mean square of the
206 noise of the LCModel fit) was less than 5 and the Cramér-Rao lower bounds (CRLB)
207 higher than 20%. Detailed information regarding acquisition, quantification, and quality
208 assurance can be found in Supplementary Table 1.
209
7
215 segmented tissue fractions were then used to correct for metabolite concentrations using LC
216 Model (see S1 Appendix). Metabolite levels are reported relative to the non-suppressed water
217 signal and creatine. The metabolite values are in institutional units (for absolute concentrations).
218
226 combined metabolite levels in ASD and TD with a multiple linear regression model in GraphPad
227 (least squares method: YBehavior= β0 + β1[met1]group + β2[met2]group, where dependent variable Y
228 is the predicted behavior measure (i.e., NVIQ, ICS-S), and met1/met2 are continuous predictor
229 variables (metabolite levels), β0 is the estimate of the model intercept and β1/ β2 are the
230 regression coefficients.
231
233 the two groups. Therefore, age was not used as a covariate for the group analyses. Glx was not
234 correlated with age in this study, but a trend was observed in the ASD group (p=0.052). Cr was
235 also not correlated with age in either group or no trends were observed (p=0.524 [ASD],
236 p=0.310 [TD]). Because of the Glx trend to correlate with age, we tested whether the Glx/Cr
237 ratio was correlated with age (p=0.307 [ASD], p=0.825 [TD]), following the approach used in
238 several other studies of Glx in children in general (Perlov et al., 2009a; BenAmor, 2014; Yang et
239 al., 2015) and children with ASD (Kubas et al., 2012; Doyle-Thomas et al., 2014; Ito et al.,
240 2017). Therefore, for the correlation analysis in this study, we used creatine normalized Glx
241 (Glx/Cr).
242
247 with the same behavioral metrics, for 10 correlations tests in total in the ASD group, and 8 in the
8
248 TD group. Lastly, we tested for significant differences between slopes for the ASD vs TD groups
249 only for correlations between Cho and NVIQ, and Glx/Cr and NVIQ.
250
259 or below, and ** to mark p-values of 0.005 or below.
260
261 3. RESULTS
266
278
279
280
9
284

(mmol/l) (mmol/l)
ASD (n=14) 1.57±0.30 8.15±1.86 1.71±0.34
TD (n=16) 1.54±0.22 7.72±1.14 1.80±0.30
287
294 indicating a significant group difference in the interaction between NVIQ and Cho (t-Value = 3.8,
295 degrees of freedom = 26, p=0.0004). No significant correlations were found between Cho and
296 ASD SRS scores, which assess ASD severity (Figure 3B).
10
297
300 showed a highly significant difference across the two groups (black text, t-Value = 3.8, degrees
301 of freedom = 26, p=0.0004) (B) SRS. (C) inhibition of attention. (D) Attentional switching. Since
302 both groups showed a similar trend, we also computed r and p values for both groups combined
303 (purple text). Red: ASD. Blue: TD. Purple: ASD and TD groups combined. Dotted blue and red
304 lines indicate the 95% confidence intervals for regression lines. Dashed purple lines indicate the
305 slope of ASD and TD combined. r=correlation coefficient (Pearson’s). p=uncorrected p-values.
306 ICS-I: The NEPSY-II Inhibition Contrast Scaled Score: measures the ability to voluntarily inhibit
307 attention. ICS-S: The NEPSY-II Switching Contrast Scaled Score: measures the ability for
308 switching attention between competing stimuli. The higher the ICS-I and ICS-S scores, the
309 better the performance.
314 There was no group difference in the direction of this correlation, and combining the two groups
11
315 slightly decreased the correlation coefficient but increased the statistical significance of the
316 correlation (r=-0.56, p=0.002).
317
323 NVIQ in the TD group. Similar to Cho, there was also a significant group in the interaction
324 between NVIQ and Glx/Cr (t-Value = 2.2, degrees of freedom = 26, p=0.044) (Figure 4A).
325 No significant association was found between Glx/Cr levels and the SRS scores (ASD
326 group) (Figure 4B). Glx/Cr did not correlate with ICS-I (Figure 4C) and was positively
327 correlated with attentional switching (Figure 4D) for both ASD (r=0.73, p=0.004) and TD
328 (r=0.54, p=0.040) groups. Indeed, when the ASD and TD groups were combined, we again
329 observed an increased statistical significance for the correlation between ICS-S and Glx/Cr
330 (r=0.61, p=0.0005).
331 Lastly, a multiple linear regression model: Y[Beh] ~ Intercept (β0) + β1[Glx/Cr][TD/ASD] + β2
332 [Cho][TD/ASD], was used to test whether combining cho and Glx/Cr predicted behavioral measures
333 in either group better than using each value in isolation. The results showed that the combined
334 Cho and Glx/Cr indeed predicted both behavioral measures, NVIQ (r=0.70, p=0.023) and ICS-S
335 (r=0.80, p=0.005) among ASD children and NVIQ (r=0.72, p=0.009) among TD children (Figure
336 5), and while the predictions were improved, the improvement was not statistically significant.
337
12
338
341 showed a significant difference across the two groups (black text, t-Value = 2.2, degrees of
342 freedom = 26, p=0.044) (B) SRS. (C) inhibition of attention. (D) Attentional switching. Since both
343 groups showed a similar trend, we also computed r and p values for both groups combined
344 (purple text). Red: ASD. Blue: TD. Purple: ASD and TD groups combined. Dashed purple lines
345 indicate the slope of ASD and TD combined. Dotted blue and red lines indicate the 95%
346 confidence intervals for regression lines. r=correlation coefficient (Pearson’s). p=uncorrected p-
347 values. ICS-I: The NEPSY-II Inhibition Contrast Scaled Score: measures the ability to voluntarily
348 inhibit attention. ICS-S: The NEPSY-II Switching Contrast Scaled Score: measures the ability for
351
352
13
353
354 Figure 5. Multiple linear regression model in ASD and TD. Performance of combined
356
357
358 4. DISCUSSION
359
370 with the Glx/Cr ratio in the TD group, but not in the ASD group.
371 Choline containing compounds are key components of cell membranes and the myeline
372 sheets and associated with white matter density and speculated to reflect excessive neuronal
373 connectivity or abnormal myelination (Laycock et al., 2008). Our observation of a significant
374 association between low Cho levels and higher NVIQ in the ASD group is similar to the findings
375 of a previous study where an inverse association with the left DLPFC Cho/Cr and full-scale IQ
376 (FSIQ) was reported in generalized anxiety disorder (Coplan et al., 2018b). During disease
377 states, elevated Cho is associated with membrane breakdown and inflammation (Davie et al.,
378 1995), processes which are likely to impair executive functioning. However, we also observed a
14
379 significant positive association between DLPFC Cho and NVIQ in the control group, where
380 Coplan et al reported no significant association. Another previous study reported an inverse
381 relationship between Cho and performance IQ in healthy adults (Jung et al., 1999). A major
382 difference between our study and aforementioned previous studies is that those included adult
383 populations. Progressive myelination of white matter pathways during typical development of
384 children and adolescents has significant changes in cognitive abilities, due to more rapid neural
385 communication (Buyanova and Arsalidou, 2021). Therefore, dynamic myelin turnover, which is
386 indicative of active Cho metabolism in typically developing children may explain the positive
387 association with cognitive performance.
391 both ASD and TD groups, and the combination of the two groups strengthened the statistical
392 significance since the power (i.e., N) was increased. Glutamate is synthesized from glutamine in
393 neurons and released into the synapse, which is converted back into glutamine in glial cells, by
394 the enzyme glutamine-synthetase (Bak et al., 2006; Hertz, 2013). Glutamate to GABA
395 conversion is catalyzed by the neuronal enzyme glutamate decarboxylase. Glutamate and
396 GABA have been found as reliable markers of cortical excitability and inhibition, and thus critical
397 for the mechanisms of neuroplasticity and learning. Moreover, brain excitation and inhibition
398 levels are thought to be critical for triggering the onset of sensitive periods for cognitive skill
399 acquisition by shaping plastic responsiveness of underlying neural systems in response to
400 environmental stimulation (Werker and Hensch, 2015). MRS studies of Glx in
401 neurodevelopmental disorders, including ASD and ADHD, have reported abnormal glutamate or
402 Glx levels in patient populations relative to typically developing subjects (Carrey et al., 2007;
403 Brown et al., 2013; Cochran et al., 2015b). Higher levels of glutamate have been hypothesized
404 to indicate hyperexcitability in these disorders that affect neuronal-network dynamics involved in
405 learning, and memory consolidation (Pugh et al., 2014a). Previous studies using smaller ASD
406 samples (n=7-12, ages 8-17), reported differences in Glx and Glx/Cr in the frontal regions
407 compared to TD groups (Bejjani et al., 2012; Kubas et al., 2012; Joshi et al., 2013). However,
408 no group-level differences were found in Glx or Glx/Cr in our study. Yet, we found a significant
409 positive correlation between Glx/Cr and nonverbal intelligence only in the ASD group. This may
410 suggest that DLPFC Glx in ASD may have a different glutamatergic transmission than that of
411 TD children.
15
413 level cognitive skills, yet the exact mechanisms remain largely unknown (Ross and Sachdev,
414 2004; Jung et al., 2009; Logue and Gould, 2014; Cohen Kadosh et al., 2015; Zacharopoulos et
415 al., 2021). Across developmental disorders such as ASD and attention-related disorders,
416 neurometabolites profiles has been found to vary compared to typically developing age-matched
417 counterparts (Perlov et al., 2009b; Baruth et al., 2013). In a 1H-MRS study using children with
418 reading disabilities (RD), it was shown that RD children had elevated Cho and glutamate
419 relative to TD children. They further indicated that Cho and glutamate levels inversely correlated
420 with reading and related language measures such that increased concentrations were
421 associated with poorer performance (Pugh et al., 2014b). A more recent RD study reported a
422 negative association between Cho and reading ability of children aged between 6 and 8 (Del
423 Tufo et al., 2018). The same study also observed a positive relationship between glutamate
424 concentration and reading performance. Lastly, the fact that in both the ASD and TD groups
425 both Cho and Glx/Cr levels were correlated with attentional switching scores, but not with
426 attentional inhibition, is consistent with the putative role of the prefrontal cortex in attention,
427 since its role for attentional switching, but not inhibition of attention, is well established (Rossi et
428 al., 2009). Attentional switching happens in the twin anticorrelated functional networks involving
429 the DLPFC as a task-positive functional site, demonstrated by Fox and colleagues (Fox et al.,
430 2005). The task-switching aspect and its relationship with neurometabolites in ASD is worth
431 exploring as a biomarker for further investigation.
438 specific biomarkers in ASD. Because ASD is a neurodevelopmental disorder, it is possible that
439 the function of different metabolites is impacted differently due to differing maturation
440 trajectories throughout development. Longitudinal studies of metabolites and their correlation
441 with specific traits at different points in time could help to elucidate this important question.
442 Despite these limitations it should be noted that it is unlikely that our findings can be fully
443 explained by potential confounds, such as differences in voxel tissue composition, age- or IQ.
444 There were no significant between-group differences in voxel brain matter and CSF, metabolite
445 concentrations were tissue corrected, and the results are consistent with prior metabolites
16
446 studies of ASD and of other related disorders. Note that prior studies on neurometabolic
447 changes among ASD and healthy controls is somewhat mixed [Review: (Ford and Crewther,
448 2016)]. For example, within this review, 4 studies employing TD and ASD children (Fayed and
449 Modrego, 2005; Friedman et al., 2006; DeVito et al., 2007; Corrigan et al., 2013) found a group
450 difference in Cho levels in the frontal regions, while 8 studies (Hisaoka et al., 2001; Friedman et
451 al., 2003; Levitt et al., 2003; Endo et al., 2007; Zeegers et al., 2007; Vasconcelos et al., 2008;
452 Fujii et al., 2010; Kubas et al., 2012) did not. The differences across studies could stem from
453 cohort differences, methodological differences, age differences, power differences, etc. Thus,
454 we consider the important finding here to be not the lack of difference in group means, which
455 may be a power issue, but the correlations between metabolite levels and behavioral traits.
456 Furthermore, note that the results of Figure 4A in the ASD group, the correlation between NVIQ
457 and Glx/Cr, were primarily driven by one outlier participant who had not only the lowest NVIQ
458 score in the ASD cohort, but also the lowest VIQ and total IQ scores. This underscores the need
459 to replicate and extend such studies to cohorts of ASD participants with greater IQ ranges, and
460 especially to ASD participants with below average IQ scores.
465 age-, IQ- matched TD children. The results suggests that the extent to which Cho and Glx/Cr
466 levels in the DLPFC mediate VIQ and inhibition of attention, may differ for ASD children relative
467 to TD children. In contrast, for both groups, attention switching scores showed an inverse
468 relationship with Cho and a positive correlation with Glx/Cr, suggesting further that the roles of
469 metabolites for executive function skills, and specifically attentional switching and inhibition of
470 attention, are preserved in ASD. These results are especially interesting in the context of the
471 significant heterogeneity associated with the ASD phenotype (Lenroot and Yeung, 2013), since
472 they suggest possible underlying neural mechanisms for the cognitive heterogeneity in
473 particular, which do not overlap which those expected from typical development. Although the
474 mechanisms connecting these neurometabolites to atypical cognitive functioning alongside
475 typical executive functioning remain to be elucidated, the overall finding that the role of
476 neurometabolites in ASD does not follow a consistent pattern relative to TD children may
477 provide valuable information on potential neurobiological pathways of atypical development, and
478 the heterogeneities associated with cognitive function in ASD in particular. Indeed, a better
479 understanding of the neurobiological/chemical underpinnings of how heterogeneity in cognitive
17
480 and executive function may correlated with metabological underpinnings is critical for advancing
481 our understanding of the neural bases that underlie heterogeneity in ASD.
482
483
484 ACKNOWLEDGMENT
485 We would like to thank our participants and their families. This study was supported by the
486 following: The National Institutes of Health (R01NS048455, MH); the Nancy Lurie Marks
487 Foundation (MH); MIT-MGH Strategic Partnership grant from the Executive Committee on
488 Research (ECOR) at Massachusetts General Hospital (EMR); the National Institute of Child
489 Health and Development (R01HD073254, TK); the National Institute of Mental Health
490 (R01MH117998, TK)
491
500
501
502 REFERENCES
503
510 10.1016/j.pnpbp.2018.09.010.
514 10.1016/j.pnpbp.2018.09.010.
18
517 10.1146/annurev.ne.09.030186.002041.
523 10.5772/20752.
524 Bak, L. K., Schousboe, A., and Waagepetersen, H. S. (2006). The glutamate/GABA-glutamine
525 cycle: Aspects of transport, neurotransmitter homeostasis and ammonia transfer. J
526 Neurochem 98. doi: 10.1111/j.1471-4159.2006.03913.x.
534 Bejjani, A., O’Neill, J., Kim, J. A., Frew, A. J., Yee, V. W., Ly, R., et al. (2012). Elevated
535 glutamatergic compounds in pregenual anterior cingulate in pediatric autism spectrum
536 disorder demonstrated by1H MRS and 1H MRSI. PLoS One 7. doi:
537 10.1371/journal.pone.0038786.
538 BenAmor, L. (2014). 1H-Magnetic resonance spectroscopy study of stimulant medication effect
539 on brain metabolites in French Canadian children with attention deficit hyperactivity
540 disorder. Neuropsychiatr Dis Treat 10. doi: 10.2147/NDT.S52338.
544 10.6064/2012/703675.
550 Buyanova, I. S., and Arsalidou, M. (2021). Cerebral White Matter Myelination and Relations to
551 Age, Gender, and Cognition: A Selective Review. Front Hum Neurosci 15. doi:
552 10.3389/fnhum.2021.662031.
19
562 Cohen Kadosh, K., Krause, B., King, A. J., Near, J., and Cohen Kadosh, R. (2015). Linking
563 GABA and glutamate levels to cognitive skill acquisition during development. Hum Brain
564 Mapp 36. doi: 10.1002/hbm.22921.
565 Cooper, R. (2017). Diagnostic and statistical manual of mental disorders (DSM). Knowledge
566 Organization 44. doi: 10.5771/0943-7444-2017-8-668.
570 doi: 10.1016/j.jad.2017.12.001.
574 doi: 10.1016/j.jad.2017.12.001.
575 Corrigan, N. M., Shaw, D. W. W., Estes, A. M., Richards, T. L., Munson, J., Friedman, S. D., et
576 al. (2013). Atypical developmental patterns of brain chemistry in children with autism
577 spectrum disorder. JAMA Psychiatry 70. doi: 10.1001/jamapsychiatry.2013.1388.
584 Davie, C. A., Feinstein, A., Kartsounis, L. D., Barker, G. J., McHugh, N. J., Walport, M. J., et al.
585 (1995). Proton magnetic resonance spectroscopy of systemic lupus erythematosus
586 involving the central nervous system. J Neurol 242. doi: 10.1007/BF00867424.
590 10.3389/fpsyg.2018.01507.
20
594 DeVito, T. J., Drost, D. J., Neufeld, R. W. J., Rajakumar, N., Pavlosky, W., Williamson, P., et al.
595 (2007). Evidence for Cortical Dysfunction in Autism: A Proton Magnetic Resonance
596 Spectroscopic Imaging Study. Biol Psychiatry 61. doi: 10.1016/j.biopsych.2006.07.022.
597 Doyle-Thomas, K. A. R., Card, D., Soorya, L. V., Ting Wang, A., Fan, J., and Anagnostou, E.
598 (2014). Metabolic mapping of deep brain structures and associations with symptomatology
599 in autism spectrum disorders. Res Autism Spectr Disord 8. doi:
600 10.1016/j.rasd.2013.10.003.
604 Endo, T., Shioiri, T., Kitamura, H., Kimura, T., Endo, S., Masuzawa, N., et al. (2007). Altered
605 Chemical Metabolites in the Amygdala-Hippocampus Region Contribute to Autistic
606 Symptoms of Autism Spectrum Disorders. Biol Psychiatry 62. doi:
607 10.1016/j.biopsych.2007.05.015.
608 Fayed, N., and Modrego, P. J. (2005). Comparative study of cerebral white matter in autism and
609 attention-deficit/hyperactivity disorder by means of magnetic resonance spectroscopy.
610 Acad Radiol 12. doi: 10.1016/j.acra.2005.01.016.
611 Fetit, R., Hillary, R. F., Price, D. J., and Lawrie, S. M. (2021a). The neuropathology of autism: A
614 Fetit, R., Hillary, R. F., Price, D. J., and Lawrie, S. M. (2021b). The neuropathology of autism: A
617 Ford, T. C., and Crewther, D. P. (2016). A comprehensive review of the 1H-MRS metabolite
618 spectrum in autism spectrum disorder. Front Mol Neurosci 9. doi:
619 10.3389/fnmol.2016.00014.
626 Friedman, S. D., Shaw, D. W., Artru, A. A., Richards, T. L., Gardner, J., Dawson, G., et al.
627 (2003). Regional brain chemical alterations in young children with autism spectrum
628 disorder. Neurology 60. doi: 10.1212/WNL.60.1.100.
21
629 Friedman, S. D., Shaw, D. W. W., Artru, A. A., Dawson, G., Petropoulos, H., and Dager, S. R.
630 (2006). Gray and white matter brain chemistry in young children with autism. Arch Gen
631 Psychiatry 63. doi: 10.1001/archpsyc.63.7.786.
632 Fujii, E., Mori, K., Miyazaki, M., Hashimoto, T., Harada, M., and Kagami, S. (2010). Function of
633 the frontal lobe in autistic individuals: A proton magnetic resonance spectroscopic study.
634 Journal of Medical Investigation 57. doi: 10.2152/jmi.57.35.
641 Hertz, L. (2013). The glutamate-glutamine (GABA) cycle: Importance of late postnatal
642 development and potential reciprocal interactions between biosynthesis and degradation.
643 Front Endocrinol (Lausanne) 4. doi: 10.3389/fendo.2013.00059.
644 Hisaoka, S., Harada, M., Nishitani, H., and Mori, K. (2001). Regional magnetic resonance
645 spectroscopy of the brain in autistic individuals. Neuroradiology 43. doi:
646 10.1007/s002340000520.
647 Hodges, H., Fealko, C., and Soares, N. (2020). Autism spectrum disorder: Definition,
648 epidemiology, causes, and clinical evaluation. Transl Pediatr 9. doi:
649 10.21037/tp.2019.09.09.
652 10.1007/s10803-012-1719-1.
653 Ito, H., Mori, K., Harada, M., Hisaoka, S., Toda, Y., Mori, T., et al. (2017). A Proton Magnetic
654 Resonance Spectroscopic Study in Autism Spectrum Disorder Using a 3-Tesla Clinical
655 Magnetic Resonance Imaging (MRI) System: The Anterior Cingulate Cortex and the Left
656 Cerebellum. J Child Neurol 32. doi: 10.1177/0883073817702981.
657 Joshi, G., Biederman, J., Wozniak, J., Goldin, R. L., Crowley, D., Furtak, S., et al. (2013).
658 Magnetic resonance spectroscopy study of the glutamatergic system in adolescent males
659 with high-functioning autistic disorder: A pilot study at 4T. Eur Arch Psychiatry Clin
660 Neurosci 263. doi: 10.1007/s00406-012-0369-9.
664 10.1098/rspb.1999.0790.
665 Jung, R. E., Gasparovic, C., Chavez, R. S., Caprihan, A., Barrow, R., and Yeo, R. A. (2009).
666 Imaging intelligence with proton magnetic resonance spectroscopy. Intelligence 37. doi:
667 10.1016/j.intell.2008.10.009.
22
670 10.1016/j.cortex.2009.07.010.
674 Kubas, B., Kułak, W., Sobaniec, W., Tarasow, E., Łebkowska, U., and Walecki, J. (2012).
675 Metabolite alterations in autistic children: A 1H MR spectroscopy study. Adv Med Sci 57.
676 doi: 10.2478/v10039-012-0014-x.
681 Neuroimage. doi: 10.1016/s1053-8119(96)80249-7.
687 10.3389/fnhum.2013.00733.
688 Levitt, J. G., O’Neill, J., Blanton, R. E., Smalley, S., Fadale, D., McCracken, J. T., et al. (2003).
689 Proton magnetic resonance spectroscopic imaging of the brain in childhood autism. Biol
690 Psychiatry 54. doi: 10.1016/S0006-3223(03)00688-7.
691 Logue, S. F., and Gould, T. J. (2014). The neural and genetic basis of executive function:
692 Attention, cognitive flexibility, and response inhibition. Pharmacol Biochem Behav 123. doi:
693 10.1016/j.pbb.2013.08.007.
699 10.1016/j.jogn.2019.07.005.
701 doi: 10.11919/j.issn.1002-0829.216027.
23
709 10.1016/j.neubiorev.2015.02.009.
716 doi: 10.3389/fnhum.2014.00034.
719 doi: 10.3389/fnhum.2014.00034.
722 8.X5000-1.
723 Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., et al. (2009a).
726 Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., et al. (2009b).
742 Ross, A. J., and Sachdev, P. S. (2004). Magnetic resonance spectroscopy in cognitive
743 research. Brain Res Rev 44. doi: 10.1016/j.brainresrev.2003.11.001.
24
746 008-1642-z.
762 Stroganova, T. A., Kozunov, V. V., Posikera, I. N., Galuta, I. A., Gratchev, V. V., and Orekhova,
763 E. V. (2013). Abnormal Pre-Attentive Arousal in Young Children with Autism Spectrum
764 Disorder Contributes to Their Atypical Auditory Behavior: An ERP Study. PLoS One 8. doi:
765 10.1371/journal.pone.0069100.
768 Rep. doi: 10.1016/j.celrep.2013.10.001.
771 doi: 10.1007/s00401-017-1736-4.
772 Vasconcelos, M. M., Brito, A. R., Domingues, R. C., Da Cruz, L. C. H., Gasparetto, E. L.,
773 Werner, J., et al. (2008). Proton magnetic resonance spectroscopy in school-aged autistic
774 children. Journal of Neuroimaging 18. doi: 10.1111/j.1552-6569.2007.00200.x.
25
782 Yang, Z. Y., Quan, H., Peng, Z. L., Zhong, Y., Tan, Z. J., and Gong, Q. Y. (2015). Proton
783 magnetic resonance spectroscopy revealed differences in the glutamate +
784 glutamine/creatine ratio of the anterior cingulate cortex between healthy and pediatric post-
785 traumatic stress disorder patients diagnosed after 2008 Wenchuan earthquake. Psychiatry
786 Clin Neurosci 69. doi: 10.1111/pcn.12332.
787 Zacharopoulos, G., Sella, F., Kadosh, K. C., Hartwright, C., Emir, U., and Kadosh, R. C. (2021).
788 Predicting learning and achievement using GABA and glutamate concentrations in human
789 development. PLoS Biol 19. doi: 10.1371/journal.pbio.3001325.
790 Zeegers, M., Van Der Grond, J., Van Daalen, E., Buitelaar, J., and Van Engeland, H. (2007).
791 Proton magnetic resonance spectroscopy in developmentally delayed young boys with or
792 without autism. J Neural Transm 114. doi: 10.1007/s00702-006-0501-y.
793
26
Figure 1 Click here to access/download;Figure;Figure_1.tif
Figure 5 Click here to access/download;Figure;Figure_5.TIF
Supporting Information
Click here to access/download

Supplementary Table 1.docx
Click here to access/download

S1 Appendix.docx
Revised Manuscript with Track Changes


6 Maria Ratai1*
7
1
2
3
4
15 USA.
5
6
19
20
21
22
24
25

32 Phone: +1-617-726-1744
33 Fax: (617) 726-7422
34
1
35 ABSTRACT
36 The neurometabolic profile associated with autism spectrum disorder (ASD) has been reported
39 (DLPFC) is of particular interest due to the DLPFC relevance to cognitive and executive
40 function, and to ASD. We used 1H-MRS to investigate neurometabolic profiles in the DLPFC of
41 ASD and sex/IQ-matched typically developing (TD) children (ages 9-13). We focused on levels
42 of Glutamate and Glutamine (Glx) due to many reported Glx abnormalities ASD, and of Choline
44 difficulties. While no significant group differences were observed in concentrations (absolute or
45 creatine referenced), metabolite levels were correlated with the behavioral phenotype of ASD
46 children. In the ASD group but not the TD group, nonverbal IQ (NVIQ) was negatively
47 associated with Cho (r=-0.59, p=0.026) and positively associated with Glx/Cr (r=0.66, p=0.011).
48 Furthermore, attentional-switching scores in the ASD group correlated negatively with Cho (r=-
49 0.69, p=0.009), and positively with Glx to creatine ratio (Glx/Cr) for both the ASD (r=0.73,
50 p=0.004) and TD (r=0.54, p=0.040) groups. Cho and Glx/Cr have different neurometabolic roles
51 in modulating NVIQ in ASD compared to TD children, while their role in attentional switching
52 seems preserved in ASD. Elucidating the role of neurometabolites in ASD also in the absence
53 of significant group differences in absolute levels is important, as sometimes even in the
54 absence of group differences in means, differences in the ASD group might still emerge that are
55 lost when only group means are considered. Understanding how individual neurophysiological
56 markers might correlate with ASD specific traits similarly to, or differently from, correlations with
57 the same traits in TD populations, and in this specific case NVIQ, is therefore a step towards
58 understanding and mapping the neural correlates of ASD.
59
60
61
62
64
2
65 1. INTRODUCTION
66

69 behaviors and interests (Cooper, 2017; Hodges et al., 2020). Studies of ASD have shown a
70 wide range of abnormalities in brain function and structure (Müller and Fishman, 2018; Forde et
71 al., 2020), as well as abnormalities in the levels of brain metabolites (Ford and Crewther, 2016).
72 However, many questions about the underlying neuropathology of ASD remain unanswered
73 (Blatt, 2012; Fetit et al., 2021a).
79 et al., 2015). In particular, structural, and metabolic irregularities have been reported in the
94 2012; Wei et al., 2014; Varghese et al., 2017; Fetit et al., 2021b). In parallel, MRS based studies
3
105 well as with ASD symptom severity, orientation of attention and sensory integration (Stroganova
106 et al., 2013; Orekhova and Stroganova, 2014a). Furthermore, a recent 1H-MRS study conducted
107 in individuals with generalized anxiety disorder (GAD), a common comorbid diagnosis in ASD,
130 TD children.
131
4
165 participant.
5
166 Seven of the 14 boys in the ASD group used psychotropic medication (Ritalin, Citalopram,
167 Focalin, Adderall, Sertraline, Concerta and Risperidone). None of the participants in the TD
168 group were on medication.
169
Age 11 (2) 8-14 10 (2) 7-14 0.36
NVIQ 100.9 (16.4) 70-127 112.4 (15.3) 91-149 0.06
VIQ 108.7 (19.4) 66-141 116.4 (11.4) 98-142 0.21
ICS-I 9.2 (3.2) 1-16 10.0 (3.7) 2-17 0.68
ICS-S 9.5 (3.7) 8-31 9.3 (4.0) 0-13 0.89
SRS 83.5 (9.6) 67-100 -- -- --

171
176
1
186 1024 sample points) with VAPOR (Variable Power and Optimized Relaxation Delays) water
187 suppression method. A 20x20x20 mm voxel was placed in the left DLPFC. In addition,
188 unsuppressed water signal (number of averages = 4) was acquired for absolute concentration
189 calculation. The center of the voxel was positioned in the axial slice above the superior margin
190 of the left lateral ventricle. (Figure 1A).
6
191
192
193
194
195 Figure 1: Voxel placement, representative spectrum, and brain matter fractions. (A) Voxel
199
201 2001) with a simulated basis set for quantitative assessment of the following neurometabolites:
202 N-acetylaspartate (NAA), Creatine (Cr), choline (Cho), myo-inositol (mI), and
203 glutamate+glutamine (Glu+Gln). LCModel analysis was conducted on spectra within the
204 chemical shift range 0.5–4.1 ppm. Spectra were excluded when the signal to noise ratio (SNR),
205 estimated by LCModel (defined as peak height of NAA divided by the root mean square of the
206 noise of the LCModel fit) was less than 5 and the Cramér-Rao lower bounds (CRLB)
207 higher than 20%. Detailed information regarding acquisition, quantification, and quality
208 assurance can be found in Supplementary Table 1.
209
7
215 segmented tissue fractions were then used to correct for metabolite concentrations using LC
216 Model (see S1 Appendix). Metabolite levels are reported relative to the non-suppressed water
217 signal and creatine. The metabolite values are in institutional units (for absolute concentrations).
218
226 combined metabolite levels in ASD and TD with a multiple linear regression model in GraphPad
227 (least squares method: YBehavior= β0 + β1[met1]group + β2[met2]group, where dependent variable Y
228 is the predicted behavior measure (i.e., NVIQ, ICS-S), and met1/met2 are continuous predictor
229 variables (metabolite levels), β0 is the estimate of the model intercept and β1/ β2 are the
230 regression coefficients.
231
233 the two groups. Therefore, age was not used as a covariate for the group analyses. Glx was not
234 correlated with age in this study, but a trend was observed in the ASD group (p=0.052). Cr was
235 also not correlated with age in either group or no trends were observed (p=0.524 [ASD],
236 p=0.310 [TD]). Because of the Glx trend to correlate with age, we tested whether the Glx/Cr
237 ratio was correlated with age (p=0.307 [ASD], p=0.825 [TD]), following the approach used in
238 several other studies of Glx in children in general (Perlov et al., 2009a; BenAmor, 2014; Yang et
239 al., 2015) and children with ASD (Kubas et al., 2012; Doyle-Thomas et al., 2014; Ito et al.,
240 2017). Therefore, for the correlation analysis in this study, we used creatine normalized Glx
241 (Glx/Cr).
242
247 with the same behavioral metrics, for 10 correlations tests in total in the ASD group, and 8 in the
8
248 TD group. Lastly, we tested for significant differences between slopes for the ASD vs TD groups
249 only for correlations between Cho and NVIQ, and Glx/Cr and NVIQ.
250
259 or below, and ** to mark p-values of 0.005 or below.
260
261 3. RESULTS
266
278
279
280
9
284

(mmol/l) (mmol/l)
ASD (n=14) 1.57±0.30 8.15±1.86 1.71±0.34
TD (n=16) 1.54±0.22 7.72±1.14 1.80±0.30
287
294 indicating a significant group difference in the interaction between NVIQ and Cho (t-Value = 3.8,
295 degrees of freedom = 26, p=0.0004). No significant correlations were found between Cho and
296 ASD SRS scores, which assess ASD severity (Figure 3B).
10
297
300 showed a highly significant difference across the two groups (black text, t-Value = 3.8, degrees
301 of freedom = 26, p=0.0004) (B) SRS. (C) inhibition of attention. (D) Attentional switching. Since
302 both groups showed a similar trend, we also computed r and p values for both groups combined
303 (purple text). Red: ASD. Blue: TD. Purple: ASD and TD groups combined. Dotted blue and red
304 lines indicate the 95% confidence intervals for regression lines. Dashed purple lines indicate the
305 slope of ASD and TD combined. r=correlation coefficient (Pearson’s). p=uncorrected p-values.
306 ICS-I: The NEPSY-II Inhibition Contrast Scaled Score: measures the ability to voluntarily inhibit
307 attention. ICS-S: The NEPSY-II Switching Contrast Scaled Score: measures the ability for
314 There was no group difference in the direction of this correlation, and combining the two groups
11
315 slightly decreased the correlation coefficient but increased the statistical significance of the
316 correlation (r=-0.56, p=0.002).
317
323 NVIQ in the TD group. Similar to Cho, there was also a significant group in the interaction
324 between NVIQ and Glx/Cr (t-Value = 2.2, degrees of freedom = 26, p=0.044) (Figure 4A).
325 No significant association was found between Glx/Cr levels and the SRS scores (ASD
326 group) (Figure 4B). Glx/Cr did not correlate with ICS-I (Figure 4C) and was positively
327 correlated with attentional switching (Figure 4D) for both ASD (r=0.73, p=0.004) and TD
328 (r=0.54, p=0.040) groups. Indeed, when the ASD and TD groups were combined, we again
329 observed an increased statistical significance for the correlation between ICS-S and Glx/Cr
330 (r=0.61, p=0.0005).
331 Lastly, a multiple linear regression model: Y[Beh] ~ Intercept (β0) + β1[Glx/Cr][TD/ASD] + β2
332 [Cho][TD/ASD], was used to test whether combining cho and Glx/Cr predicted behavioral measures
333 in either group better than using each value in isolation. The results showed that the combined
334 Cho and Glx/Cr indeed predicted both behavioral measures, NVIQ (r=0.70, p=0.023) and ICS-S
335 (r=0.80, p=0.005) among ASD children and NVIQ (r=0.72, p=0.009) among TD children (Figure
336 5), and while the predictions were improved, the improvement was not statistically significant.
337
12
338
341 showed a significant difference across the two groups (black text, t-Value = 2.2, degrees of
342 freedom = 26, p=0.044) (B) SRS. (C) inhibition of attention. (D) Attentional switching. Since both
343 groups showed a similar trend, we also computed r and p values for both groups combined
344 (purple text). Red: ASD. Blue: TD. Purple: ASD and TD groups combined. Dashed purple lines
345 indicate the slope of ASD and TD combined. Dotted blue and red lines indicate the 95%
346 confidence intervals for regression lines. r=correlation coefficient (Pearson’s). p=uncorrected p-
347 values. ICS-I: The NEPSY-II Inhibition Contrast Scaled Score: measures the ability to voluntarily
348 inhibit attention. ICS-S: The NEPSY-II Switching Contrast Scaled Score: measures the ability for
351
352
13
353
354 Figure 5. Multiple linear regression model in ASD and TD. Performance of combined
356
357
358 4. DISCUSSION
359
370 with the Glx/Cr ratio in the TD group, but not in the ASD group.
371 Choline containing compounds are key components of cell membranes and the myeline
372 sheets and associated with white matter density and speculated to reflect excessive neuronal
373 connectivity or abnormal myelination (Laycock et al., 2008). Our observation of a significant
374 association between low Cho levels and higher NVIQ in the ASD group is similar to the findings
375 of a previous study where an inverse association with the left DLPFC Cho/Cr and full-scale IQ
376 (FSIQ) was reported in generalized anxiety disorder (Coplan et al., 2018b). During disease
377 states, elevated Cho is associated with membrane breakdown and inflammation (Davie et al.,
378 1995), processes which are likely to impair executive functioning. However, we also observed a
14
379 significant positive association between DLPFC Cho and NVIQ in the control group, where
380 Coplan et al reported no significant association. Another previous study reported an inverse
381 relationship between Cho and performance IQ in healthy adults (Jung et al., 1999). A major
382 difference between our study and aforementioned previous studies is that those included adult
383 populations. Progressive myelination of white matter pathways during typical development of
384 children and adolescents has significant changes in cognitive abilities, due to more rapid neural
385 communication (Buyanova and Arsalidou, 2021). Therefore, dynamic myelin turnover, which is
386 indicative of active Cho metabolism in typically developing children may explain the positive
387 association with cognitive performance.
391 both ASD and TD groups, and the combination of the two groups strengthened the statistical
392 significance since the power (i.e., N) was increased. Glutamate is synthesized from glutamine in
393 neurons and released into the synapse, which is converted back into glutamine in glial cells, by
394 the enzyme glutamine-synthetase (Bak et al., 2006; Hertz, 2013). Glutamate to GABA
395 conversion is catalyzed by the neuronal enzyme glutamate decarboxylase. Glutamate and
396 GABA have been found as reliable markers of cortical excitability and inhibition, and thus critical
397 for the mechanisms of neuroplasticity and learning. Moreover, brain excitation and inhibition
398 levels are thought to be critical for triggering the onset of sensitive periods for cognitive skill
399 acquisition by shaping plastic responsiveness of underlying neural systems in response to
400 environmental stimulation (Werker and Hensch, 2015). MRS studies of Glx in
401 neurodevelopmental disorders, including ASD and ADHD, have reported abnormal glutamate or
402 Glx levels in patient populations relative to typically developing subjects (Carrey et al., 2007;
403 Brown et al., 2013; Cochran et al., 2015b). Higher levels of glutamate have been hypothesized
404 to indicate hyperexcitability in these disorders that affect neuronal-network dynamics involved in
405 learning, and memory consolidation (Pugh et al., 2014a). Previous studies using smaller ASD
406 samples (n=7-12, ages 8-17), reported differences in Glx and Glx/Cr in the frontal regions
407 compared to TD groups (Bejjani et al., 2012; Kubas et al., 2012; Joshi et al., 2013). However,
408 no group-level differences were found in Glx or Glx/Cr in our study. Yet, we found a significant
409 positive correlation between Glx/Cr and nonverbal intelligence only in the ASD group. This may
410 suggest that DLPFC Glx in ASD may have a different glutamatergic transmission than that of
411 TD children.
15
413 level cognitive skills, yet the exact mechanisms remain largely unknown (Ross and Sachdev,
414 2004; Jung et al., 2009; Logue and Gould, 2014; Cohen Kadosh et al., 2015; Zacharopoulos et
415 al., 2021). Across developmental disorders such as ASD and attention-related disorders,
416 neurometabolites profiles has been found to vary compared to typically developing age-matched
417 counterparts (Perlov et al., 2009b; Baruth et al., 2013). In a 1H-MRS study using children with
418 reading disabilities (RD), it was shown that RD children had elevated Cho and glutamate
419 relative to TD children. They further indicated that Cho and glutamate levels inversely correlated
420 with reading and related language measures such that increased concentrations were
421 associated with poorer performance (Pugh et al., 2014b). A more recent RD study reported a
422 negative association between Cho and reading ability of children aged between 6 and 8 (Del
423 Tufo et al., 2018). The same study also observed a positive relationship between glutamate
424 concentration and reading performance. Lastly, the fact that in both the ASD and TD groups
425 both Cho and Glx/Cr levels were correlated with attentional switching scores, but not with
426 attentional inhibition, is consistent with the putative role of the prefrontal cortex in attention,
427 since its role for attentional switching, but not inhibition of attention, is well established (Rossi et
428 al., 2009). Attentional switching happens in the twin anticorrelated functional networks involving
429 the DLPFC as a task-positive functional site, demonstrated by Fox and colleagues (Fox et al.,
430 2005). The task-switching aspect and its relationship with neurometabolites in ASD is worth
431 exploring as a biomarker for further investigation.
438 specific biomarkers in ASD. Because ASD is a neurodevelopmental disorder, it is possible that
439 the function of different metabolites is impacted differently due to differing maturation
440 trajectories throughout development. Longitudinal studies of metabolites and their correlation
441 with specific traits at different points in time could help to elucidate this important question.
442 Despite these limitations it should be noted that it is unlikely that our findings can be fully
443 explained by potential confounds, such as differences in voxel tissue composition, age- or IQ.
444 There were no significant between-group differences in voxel brain matter and CSF, metabolite
445 concentrations were tissue corrected, and the results are consistent with prior metabolites
16
446 studies of ASD and of other related disorders. Note that prior studies on neurometabolic
447 changes among ASD and healthy controls is somewhat mixed [Review: (Ford and Crewther,
448 2016)]. For example, within this review, 4 studies employing TD and ASD children (Fayed and
449 Modrego, 2005; Friedman et al., 2006; DeVito et al., 2007; Corrigan et al., 2013) found a group
450 difference in Cho levels in the frontal regions, while 8 studies (Hisaoka et al., 2001; Friedman et
451 al., 2003; Levitt et al., 2003; Endo et al., 2007; Zeegers et al., 2007; Vasconcelos et al., 2008;
452 Fujii et al., 2010; Kubas et al., 2012) did not. The differences across studies could stem from
453 cohort differences, methodological differences, age differences, power differences, etc. Thus,
454 we consider the important finding here to be not the lack of difference in group means, which
455 may be a power issue, but the correlations between metabolite levels and behavioral traits.
456 Furthermore, note that the results of Figure 4A in the ASD group, the correlation between NVIQ
457 and Glx/Cr, were primarily driven by one outlier participant who had not only the lowest NVIQ
458 score in the ASD cohort, but also the lowest VIQ and total IQ scores. This underscores the need
459 to replicate and extend such studies to cohorts of ASD participants with greater IQ ranges, and
460 especially to ASD participants with below average IQ scores.
465 age-, IQ- matched TD children. The results suggests that the extent to which Cho and Glx/Cr
466 levels in the DLPFC mediate VIQ and inhibition of attention, may differ for ASD children relative
467 to TD children. In contrast, for both groups, attention switching scores showed an inverse
468 relationship with Cho and a positive correlation with Glx/Cr, suggesting further that the roles of
469 metabolites for executive function skills, and specifically attentional switching and inhibition of
470 attention, are preserved in ASD. These results are especially interesting in the context of the
471 significant heterogeneity associated with the ASD phenotype (Lenroot and Yeung, 2013), since
472 they suggest possible underlying neural mechanisms for the cognitive heterogeneity in
473 particular, which do not overlap which those expected from typical development. Although the
474 mechanisms connecting these neurometabolites to atypical cognitive functioning alongside
475 typical executive functioning remain to be elucidated, the overall finding that the role of
476 neurometabolites in ASD does not follow a consistent pattern relative to TD children may
477 provide valuable information on potential neurobiological pathways of atypical development, and
478 the heterogeneities associated with cognitive function in ASD in particular. Indeed, a better
479 understanding of the neurobiological/chemical underpinnings of how heterogeneity in cognitive
17
480 and executive function may correlated with metabological underpinnings is critical for advancing
481 our understanding of the neural bases that underlie heterogeneity in ASD.
482
483
484 ACKNOWLEDGMENT
485 We would like to thank our participants and their families. This study was supported by the
486 following: The National Institutes of Health (R01NS048455, MH); the Nancy Lurie Marks
487 Foundation (MH); MIT-MGH Strategic Partnership grant from the Executive Committee on
488 Research (ECOR) at Massachusetts General Hospital (EMR); the National Institute of Child
489 Health and Development (R01HD073254, TK); the National Institute of Mental Health
490 (R01MH117998, TK)
491
500
501
502 REFERENCES
503
510 10.1016/j.pnpbp.2018.09.010.
514 10.1016/j.pnpbp.2018.09.010.
18
517 10.1146/annurev.ne.09.030186.002041.
523 10.5772/20752.
524 Bak, L. K., Schousboe, A., and Waagepetersen, H. S. (2006). The glutamate/GABA-glutamine
525 cycle: Aspects of transport, neurotransmitter homeostasis and ammonia transfer. J
526 Neurochem 98. doi: 10.1111/j.1471-4159.2006.03913.x.
534 Bejjani, A., O’Neill, J., Kim, J. A., Frew, A. J., Yee, V. W., Ly, R., et al. (2012). Elevated
535 glutamatergic compounds in pregenual anterior cingulate in pediatric autism spectrum
536 disorder demonstrated by1H MRS and 1H MRSI. PLoS One 7. doi:
537 10.1371/journal.pone.0038786.
538 BenAmor, L. (2014). 1H-Magnetic resonance spectroscopy study of stimulant medication effect
539 on brain metabolites in French Canadian children with attention deficit hyperactivity
540 disorder. Neuropsychiatr Dis Treat 10. doi: 10.2147/NDT.S52338.
544 10.6064/2012/703675.
550 Buyanova, I. S., and Arsalidou, M. (2021). Cerebral White Matter Myelination and Relations to
551 Age, Gender, and Cognition: A Selective Review. Front Hum Neurosci 15. doi:
552 10.3389/fnhum.2021.662031.
19
562 Cohen Kadosh, K., Krause, B., King, A. J., Near, J., and Cohen Kadosh, R. (2015). Linking
563 GABA and glutamate levels to cognitive skill acquisition during development. Hum Brain
564 Mapp 36. doi: 10.1002/hbm.22921.
565 Cooper, R. (2017). Diagnostic and statistical manual of mental disorders (DSM). Knowledge
566 Organization 44. doi: 10.5771/0943-7444-2017-8-668.
570 doi: 10.1016/j.jad.2017.12.001.
574 doi: 10.1016/j.jad.2017.12.001.
575 Corrigan, N. M., Shaw, D. W. W., Estes, A. M., Richards, T. L., Munson, J., Friedman, S. D., et
576 al. (2013). Atypical developmental patterns of brain chemistry in children with autism
577 spectrum disorder. JAMA Psychiatry 70. doi: 10.1001/jamapsychiatry.2013.1388.
584 Davie, C. A., Feinstein, A., Kartsounis, L. D., Barker, G. J., McHugh, N. J., Walport, M. J., et al.
585 (1995). Proton magnetic resonance spectroscopy of systemic lupus erythematosus
586 involving the central nervous system. J Neurol 242. doi: 10.1007/BF00867424.
590 10.3389/fpsyg.2018.01507.
20
594 DeVito, T. J., Drost, D. J., Neufeld, R. W. J., Rajakumar, N., Pavlosky, W., Williamson, P., et al.
595 (2007). Evidence for Cortical Dysfunction in Autism: A Proton Magnetic Resonance
596 Spectroscopic Imaging Study. Biol Psychiatry 61. doi: 10.1016/j.biopsych.2006.07.022.
597 Doyle-Thomas, K. A. R., Card, D., Soorya, L. V., Ting Wang, A., Fan, J., and Anagnostou, E.
598 (2014). Metabolic mapping of deep brain structures and associations with symptomatology
599 in autism spectrum disorders. Res Autism Spectr Disord 8. doi:
600 10.1016/j.rasd.2013.10.003.
604 Endo, T., Shioiri, T., Kitamura, H., Kimura, T., Endo, S., Masuzawa, N., et al. (2007). Altered
605 Chemical Metabolites in the Amygdala-Hippocampus Region Contribute to Autistic
606 Symptoms of Autism Spectrum Disorders. Biol Psychiatry 62. doi:
607 10.1016/j.biopsych.2007.05.015.
608 Fayed, N., and Modrego, P. J. (2005). Comparative study of cerebral white matter in autism and
609 attention-deficit/hyperactivity disorder by means of magnetic resonance spectroscopy.
610 Acad Radiol 12. doi: 10.1016/j.acra.2005.01.016.
611 Fetit, R., Hillary, R. F., Price, D. J., and Lawrie, S. M. (2021a). The neuropathology of autism: A
614 Fetit, R., Hillary, R. F., Price, D. J., and Lawrie, S. M. (2021b). The neuropathology of autism: A
617 Ford, T. C., and Crewther, D. P. (2016). A comprehensive review of the 1H-MRS metabolite
618 spectrum in autism spectrum disorder. Front Mol Neurosci 9. doi:
619 10.3389/fnmol.2016.00014.
626 Friedman, S. D., Shaw, D. W., Artru, A. A., Richards, T. L., Gardner, J., Dawson, G., et al.
627 (2003). Regional brain chemical alterations in young children with autism spectrum
628 disorder. Neurology 60. doi: 10.1212/WNL.60.1.100.
21
629 Friedman, S. D., Shaw, D. W. W., Artru, A. A., Dawson, G., Petropoulos, H., and Dager, S. R.
630 (2006). Gray and white matter brain chemistry in young children with autism. Arch Gen
631 Psychiatry 63. doi: 10.1001/archpsyc.63.7.786.
632 Fujii, E., Mori, K., Miyazaki, M., Hashimoto, T., Harada, M., and Kagami, S. (2010). Function of
633 the frontal lobe in autistic individuals: A proton magnetic resonance spectroscopic study.
634 Journal of Medical Investigation 57. doi: 10.2152/jmi.57.35.
641 Hertz, L. (2013). The glutamate-glutamine (GABA) cycle: Importance of late postnatal
642 development and potential reciprocal interactions between biosynthesis and degradation.
643 Front Endocrinol (Lausanne) 4. doi: 10.3389/fendo.2013.00059.
644 Hisaoka, S., Harada, M., Nishitani, H., and Mori, K. (2001). Regional magnetic resonance
645 spectroscopy of the brain in autistic individuals. Neuroradiology 43. doi:
646 10.1007/s002340000520.
647 Hodges, H., Fealko, C., and Soares, N. (2020). Autism spectrum disorder: Definition,
648 epidemiology, causes, and clinical evaluation. Transl Pediatr 9. doi:
649 10.21037/tp.2019.09.09.
652 10.1007/s10803-012-1719-1.
653 Ito, H., Mori, K., Harada, M., Hisaoka, S., Toda, Y., Mori, T., et al. (2017). A Proton Magnetic
654 Resonance Spectroscopic Study in Autism Spectrum Disorder Using a 3-Tesla Clinical
655 Magnetic Resonance Imaging (MRI) System: The Anterior Cingulate Cortex and the Left
656 Cerebellum. J Child Neurol 32. doi: 10.1177/0883073817702981.
657 Joshi, G., Biederman, J., Wozniak, J., Goldin, R. L., Crowley, D., Furtak, S., et al. (2013).
658 Magnetic resonance spectroscopy study of the glutamatergic system in adolescent males
659 with high-functioning autistic disorder: A pilot study at 4T. Eur Arch Psychiatry Clin
660 Neurosci 263. doi: 10.1007/s00406-012-0369-9.
664 10.1098/rspb.1999.0790.
665 Jung, R. E., Gasparovic, C., Chavez, R. S., Caprihan, A., Barrow, R., and Yeo, R. A. (2009).
666 Imaging intelligence with proton magnetic resonance spectroscopy. Intelligence 37. doi:
667 10.1016/j.intell.2008.10.009.
22
670 10.1016/j.cortex.2009.07.010.
674 Kubas, B., Kułak, W., Sobaniec, W., Tarasow, E., Łebkowska, U., and Walecki, J. (2012).
675 Metabolite alterations in autistic children: A 1H MR spectroscopy study. Adv Med Sci 57.
676 doi: 10.2478/v10039-012-0014-x.
681 Neuroimage. doi: 10.1016/s1053-8119(96)80249-7.
687 10.3389/fnhum.2013.00733.
688 Levitt, J. G., O’Neill, J., Blanton, R. E., Smalley, S., Fadale, D., McCracken, J. T., et al. (2003).
689 Proton magnetic resonance spectroscopic imaging of the brain in childhood autism. Biol
690 Psychiatry 54. doi: 10.1016/S0006-3223(03)00688-7.
691 Logue, S. F., and Gould, T. J. (2014). The neural and genetic basis of executive function:
692 Attention, cognitive flexibility, and response inhibition. Pharmacol Biochem Behav 123. doi:
693 10.1016/j.pbb.2013.08.007.
699 10.1016/j.jogn.2019.07.005.
701 doi: 10.11919/j.issn.1002-0829.216027.
23
709 10.1016/j.neubiorev.2015.02.009.
716 doi: 10.3389/fnhum.2014.00034.
719 doi: 10.3389/fnhum.2014.00034.
722 8.X5000-1.
723 Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., et al. (2009a).
726 Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., et al. (2009b).
742 Ross, A. J., and Sachdev, P. S. (2004). Magnetic resonance spectroscopy in cognitive
743 research. Brain Res Rev 44. doi: 10.1016/j.brainresrev.2003.11.001.
24
746 008-1642-z.
762 Stroganova, T. A., Kozunov, V. V., Posikera, I. N., Galuta, I. A., Gratchev, V. V., and Orekhova,
763 E. V. (2013). Abnormal Pre-Attentive Arousal in Young Children with Autism Spectrum
764 Disorder Contributes to Their Atypical Auditory Behavior: An ERP Study. PLoS One 8. doi:
765 10.1371/journal.pone.0069100.
768 Rep. doi: 10.1016/j.celrep.2013.10.001.
771 doi: 10.1007/s00401-017-1736-4.
772 Vasconcelos, M. M., Brito, A. R., Domingues, R. C., Da Cruz, L. C. H., Gasparetto, E. L.,
773 Werner, J., et al. (2008). Proton magnetic resonance spectroscopy in school-aged autistic
774 children. Journal of Neuroimaging 18. doi: 10.1111/j.1552-6569.2007.00200.x.
25
782 Yang, Z. Y., Quan, H., Peng, Z. L., Zhong, Y., Tan, Z. J., and Gong, Q. Y. (2015). Proton
783 magnetic resonance spectroscopy revealed differences in the glutamate +
784 glutamine/creatine ratio of the anterior cingulate cortex between healthy and pediatric post-
785 traumatic stress disorder patients diagnosed after 2008 Wenchuan earthquake. Psychiatry
786 Clin Neurosci 69. doi: 10.1111/pcn.12332.
787 Zacharopoulos, G., Sella, F., Kadosh, K. C., Hartwright, C., Emir, U., and Kadosh, R. C. (2021).
788 Predicting learning and achievement using GABA and glutamate concentrations in human
789 development. PLoS Biol 19. doi: 10.1371/journal.pbio.3001325.
790 Zeegers, M., Van Der Grond, J., Van Daalen, E., Buitelaar, J., and Van Engeland, H. (2007).
791 Proton magnetic resonance spectroscopy in developmentally delayed young boys with or
792 without autism. J Neural Transm 114. doi: 10.1007/s00702-006-0501-y.
793
26
Response to Reviewers
Response to reviews
We would like to begin by thanking the reviewers for the valuable comments on this research
article. We are grateful for the constructive suggestions on how to further improve our work and
strengthen our findings. Below, we address the comments point by point. For convenience, the
response to reviewers’ comments is written in blue, and changes to the manuscript have been
highlighted.
Reviewers' comments:
Reviewer #1:
In this cross sectional study of patients with Autism Spectrum Disorder (ASD), authors collected
single-voxel MRS data in DLPFC cortical region of the brain in 14 ASD males and 16 Typically
Developing males who were age matched. Authors used MRS data to estimate Choline (Cho),
Glutamate-Glutamine (Glx), and Creatine (Cr) levels. Authors also collected symptom severity
measures in patients, and NVIQ, IQ, attentional re-orienting, and attentional switching in patients
and healthy. Metabolite levels did not differ between ASD from TD. And metabolite levels did not
correlate with symptom severity. However, Cho levels were inversely associated with NVIQ in ASD
but not in TD participants. Attentional-switching scores were inversely correlated with Cho levels
in the in ASD group; and Glx/Cr levels were positively correlated with attentional-switching scores
the in the ASD and the TD groups.
There are several conceptual problems with the manuscript and analyses
(1) Authors utilized anatomical MRI to estimate gray, white, and CSF fraction in each voxel; and
using these fractions to estimate metabolites levels corrected for partial volume. With MRS data
collected for single voxel, correction for partial volumes is not possible. If partial volume
correction was done, then no details have been provided. Without this, it is impossible to
interpret the findings in the manuscript.
We thank the reviewer for the opportunity to clarify this. In this study, partial volume correction
was performed as now stated in lines 210-217
“All metabolite levels were adjusted for gray matter (GM), white matter (WM), and cerebrospinal fluid
(CSF) contributions as follows: the MEMPRAGE images and the voxel coordinates from the Siemens
RDA files were used in Gannet toolkit version 3.1 (Edden et al., 2014) to generate binary masks of the
voxel location. These masks were then used in SPM 12 (Penny et al., 2007) to calculate the partial
volumes of GM, WM and CSF percentages within the voxel. The segmented tissue fractions were then
used to correct for metabolite concentrations using LC Model (see S1 Appendix). Metabolite levels
are reported relative to the non-suppressed water signal and creatine. The metabolite values are in
institutional units (for absolute concentrations).”
”
(2) The manuscript seems to suggests that they used LCmodel software for absolute
quantification of metabolite levels. If so, then the details for the absolute quantification are
missing. How was absolute quantification done with acquisition of any reference scan?
Absolute quantification was performed using the water unsuppressed reference scan. Information
of this scan is now added to manuscript (line 187):
“In addition, unsuppressed water signal (number of averages = 4) was acquired for absolute
concentration calculation.”
(3) The authors claim to use Glx/Cr ratio to account for the correlations of Glx with age. This
seems strange. The ratio is computed to normalize metabolite levels to Cr levels so that they can
be compared across participants. Even if Glx is correlated with age, Cr has to be similarly
correlated with age to remove the age effects in the ratio; that too, if Glx and Cr depend upon
only age, which is not true. So the justification provided for the use of Glx/Cr ratio is not clear.
Glx was not correlated with age in this study, but a trend was observed in the ASD group
(p=0.052). Cr was also not correlated with age in either group or no trends were observed
(p=0.524 [ASD], p=0.310 [TD]). Because of the Glx trend to correlate with age, we tested whether
the Glx/Cr ratio was correlated with age, following the approach used in several other studies of
Glx in children in general {Benamor 2014; Yang et al., 2015; Perlov et al., 2009}, and children with
ASD {Kubas et al., 2012; Doyle-Thomas et al., 2014; Ito et al., 2017}. This information was added to
the manuscript (lines 233-241).
Kubas, B., Kulak, W., Sobaniec, W., Tarasow, E., Lebkowska, U., and Walecki, J. (2012). Metabolite
alterations in autistic children: a 1H MR spectroscopy study. Adv. Med. Sci. 57, 152–156. Doi:
10.2478/v10039-012-0014-x
Ito, H., Mori, K., Harada, M., Hisaoka, S., Toda, Y., Mori, T., Goji, A., Abe, Y., Miyazaki, M., & Kagami,
S. (2017). A Proton Magnetic Resonance Spectroscopic Study in Autism Spectrum Disorder Using a
3-Tesla Clinical Magnetic Resonance Imaging (MRI) System: The Anterior Cingulate Cortex and the
Left Cerebellum. Journal of child neurology, 32(8), 731–739.
https://doi.org/10.1177/0883073817702981
Doyle-Thomas KA, Card D, Soorya LV, Wang AT, Fan J, Anagnostou E. Metabolic mapping of deep
brain structures and associations with symptomatology in autism spectrum disorders. Res Autism
Spectr Disord. 2014 Jan;8(1):44-51. doi: 10.1016/j.rasd.2013.10.003. PMID: 24459534; PMCID:
PMC3897261.
Benamor L. (2014). (1)H-Magnetic resonance spectroscopy study of stimulant medication effect on

brain metabolites in French Canadian children with attention deficit hyperactivity
disorder. Neuropsychiatric disease and treatment, 10, 47–54. https://doi.org/10.2147/NDT.S52338
Yang, Z. Y., Quan, H., Peng, Z. L., Zhong, Y., Tan, Z. J., & Gong, Q. Y. (2015). Proton magnetic
resonance spectroscopy revealed differences in the glutamate + glutamine/creatine ratio of the
anterior cingulate cortex between healthy and pediatric post-traumatic stress disorder patients
diagnosed after 2008 Wenchuan earthquake. Psychiatry and clinical neurosciences, 69(12), 782–
790. https://doi.org/10.1111/pcn.12332
Perlov, E., Philipsen, A., Matthies, S., Drieling, T., Maier, S., Bubl, E., Hesslinger, B., Buechert, M.,
Henning, J., Ebert, D., & Tebartz Van Elst, L. (2009). Spectroscopic findings in attention-
deficit/hyperactivity disorder: review and meta-analysis. The world journal of biological psychiatry :
the official journal of the World Federation of Societies of Biological Psychiatry, 10(4 Pt 2), 355–365.
https://doi.org/10.1080/15622970802176032
(4) Why was Cho levels used without normalizing it to either background noise levels to account
for gain factor or to Cr levels? If Cho levels were absolute quantities in mmol, than how was the
absolute quantification done?
In this study Cho levels are reported in absolute concentration, referenced to water. This is now
clarified in the methods section.
(5) Partial volume correction was done using anatomical MRI data that were acquired in addition
to the single voxel MRS data. The correction procedure assumes that the patient does not move
between anatomical and MRS scan, which likely is not the case. Patient movement between two
scans will lead to error in locating where the MRS voxel is location in the anatomical space. No
steps have been proposed to minimize motion effects.
A mock scanner was used to prepare (reduced anxiety and motion) the children for actual MRI
scan session. Next, MRS was performed immediately following the anatomical scan. We used QA
(visual inspection) following the scan acquisition to exclude participants with excessive motion or
other artefacts. 3 ASD and 1 TD participant data were excluded due to poor data quality.
(6) Several statements are made in the abstract and discussion that do not really provide any
information and leave a person wondering what is being conveyed. For example, in the abstract
Elucidating the apparently divergent role of neurometabolites in ASD in the absence of significant
group differences in absolute levels is an important step towards understanding and mapping the
neural correlates of ASD.
Why is elucidating divergent role is an important step forward? And why it leads to understanding
and mapping the neural correlates of ASD is metabolites in those regions are neither different
than TD nor correlate with symptom severity? What is being conveyed here by the authors?
We agree that the statement is indeed too vague. We changed it to the following statement:
“Elucidating the role of neurometabolites in ASD also in the absence of significant group differences
in absolute levels is important, as sometimes even in the absence of group differences in means,
differences in the ASD group might still emerge that are lost when only group means are considered.
Understanding how individual neurophysiological markers might correlate with ASD specific traits
similarly to, or differently from, correlations with the same traits in TD populations, and in this
specific case NVIQ, is therefore a step towards understanding and mapping the neural correlates of
ASD.”
(7) Similarly, the next statement in the abstract
These results are also relevant in the context of the significant cognitive function heterogeneity
associated with the ASD phenotype, as they suggest possible underlying neural mechanisms that do
not overlap with those expected from typical development.
What is “cognitive function heterogeneity”? And even if there differences across participants,
findings do not support any bases for heterogeneity. And how do the findings suggest a neural
mechanism for this heterogeneity that does not overlay with TD individuals? The statement as
written does not convey any valid thought at all.
We apologize for this vague statement. We believe that the revised version of the prior statement
is now clear enough as a standalone statement and replaces both of these previous sentences.
Therefore, this statement is now removed.
(8) Authors claim that they focused on only 2 metabolites because of small sample size. How does
focusing on 2 metabolites over the problems of small sample size?
We thank the reviewer for the opportunity to clarify this. Since each added metabolite needs to
be corrected for additional comparisons, thus decreasing power further in this already small
sample, we decided to focus on only two metabolites.
(9) Authors state that the patients had a prior diagnosis of ASD. What does prior diagnosis mean?
Did they have or did not have autism at the scan time?
Clearly this phrasing created confusion. We apologize for this. We meant to say that the
participants had gotten their diagnosis before being recruited. The word “prior” was removed. All
diagnoses were current at the time of recruitment, and further verified as noted.
(10) Authors used multiple and not multivariate linear regression analysis. If they did use
multivariate, then they need to explain what were the dependent variables in the analyses.
The regression equation provided

Y_Behavior= β0 + β1[met1]_group + β2[met2]_group
is not clear. What are the terms in the equation?
We thank the reviewer for pointing this out. We have now changed this to multiple linear
regression. This is now explained in the manuscript (lines 226-230).
“YBehavior= β0 + β1[met1]group + β2[met2]group, where dependent variable Y is the predicted behavior

measure (i.e., NVIQ, ICS-S), and met1/met2 are continuous predictor variables (metabolite levels),
β0 is the estimate of the model intercept and β1/ β2 are the regression coefficients.“
(11) Why did they use Bonferroni correction for 10 comparisons? There seems to be many more
analyses that were run.
We thank the reviewer for the opportunity to clarify this point. As noted in the manuscript, we
actually did not apply any correction for multiple comparisons, following the recent best practices
(Lu and Belitskaya-Levy, 2015; Amrhein et al., 2019; Lowe, 2019), that advise to simply note p
values as they are without corrections, due to the dependencies masked by corrections. Instead,
we indicated using * the significance of the p-value, with ** indicating a p value 10 times more
significant than 0.05. We now removed the sentence “, i.e. p values that would survive a
Bonferroni correction for 10 comparisons”, since it is indeed confusing. The number of
comparisons can be assessed by the reader within each figure. Two groups x 3 metabolites in
figure 2, 2 groups x 4 behavioral measures in figure 3 and in figure 4. Note also that a Bonferroni
correction would be too strict, as some of these measures, e.g. the behavioral measures, are not
actually completely independent. Hence the practice of not correcting, allowing the reader to
assess significance using uncorrected values,
(12) Lines 387-388, Discussion Section
This may suggest that DLPFC Glx in ASD may have a different glutamatergic transmission than that
of TD children.
What does different glutamatergic transmission means? And why is it different because of
different correlations with NVIQ?
The absence of a group difference in the concentrations of Glx or Glx/Cr indicate that Glx levels
themselves are not altered in the autistic DLPFC. However, Glx/Cr predicted the dynamics of NVIQ
only in ASD. This study is clearly not equipped to answer these important questions (what does it
mean and why), but we know this is not the first time such differences have been observed
(Coplan et al., 2018). ASD and anxiety can impact development, and we know neurotransmission
is impacted by development. How and why exactly the changes occur this way in this and other
studies seem to indicate will hopefully be explored in future studies.
(13) Line 415, discussion section
Therefore, longitudinal MRS studies are clearly necessary.
Why are longitudinal MRS studies clearly needed?
This sentence is now replaced with: “Because ASD is a neurodevelopmental disorder, it is possible
that the function of different metabolites is impacted differently due to differing maturation
trajectories throughout development. Longitudinal studies of metabolites and their correlation with
specific traits at different points in time could help to elucidate this important question.” (line 436-
439)
There were no significant between-group differences in voxel brain matter and CSF, metabolite
concentrations were tissue corrected, and the results are consistent with prior metabolites studies of
ASD and of other related disorders.
Prior studies have reported significant metabolite differences in ASD. So how are the reported
findings of no difference in this manuscript consistent with those?
Prior studies show that the neurometabolic changes among ASD and healthy controls is
somewhat mixed (Review: Ford TC and Crewther DP 2016). For example, within this review, 4
studies employing TD and ASD children (Corrigan et al., 2013; Friedman et al., 2006; Fayed and
Modrego 2005; DeVito et al., 2007) found a group difference in Cho levels in the frontal regions,
while 8 studies (Friedman et al., 2003; Levitt et al., 2003; Zeegers et al., 2007; Hisaoka et al., 2001;
Kubas et al., 2012; Fujii et al., 2010; Vasconcelos et al., 2008; Endo et al., 2007) did not. The
differences across studies could stem from cohort differences, methodological differences, age
differences, power differences, etc. Thus, we consider the important finding here to be not the
lack of difference in group means, which may be a power issue, but the correlations between
metabolite levels and behavioral traits. This has now been added to the manuscript (line 444).
This suggests that Cho and Glx/Cr may have different neurometabolic roles in ASD compared to TD
children in overall cognitive function.
Different neurometabolic role? Really? Cho and Glx are mediating different molecular processes in
ASD than in TD? If so, then at is the evidence for making such a claim?
We agree that this sentence was written poorly. It is now revised to read:
“The results suggests that the extent to which Cho and Glx/Cr levels in the DLPFC mediate VIQ and
inhibition of attention, may differ for ASD children relative to TD children.” (line 463)
Reviewer #2:
In this work, Weerasekera et al investigated choline and glutamatergic compounds in the left
dorsolateral prefrontal cortex of 14 ASD and 16 TD children. While there were no significant
between group metabolite differences, correlations between metabolite levels and intelligence
and cognitive scores were observed. NVIQ was negatively correlated with Cho in the ASD group
but positively correlated in the TD group, whereas Glx/Cr was positively correlated with NVIQ in
ASD but not correlated in the TD group. On the other hand, attention switching showed similar
patterns of correlations across groups, it was negatively correlated with NVIQ and positively
correlated with Glx/Cr. Hence the authors conclude that Cho and Glx/Cr have different roles in
NVIQ modulation in ASD, while their role in attention switching seems preserved in ASD, and
state that these results are relevant in the context of cognitive heterogeneity as they suggest
possible neural mechanisms in ASD that do not overlap with the expected from typical
development.
I have the following comments and suggestions:
Abstract:
1. The authors state that no significant differences were observed for absolute concentrations but
there were both absolute and relative measures reported. Please correct.
We thank the Reviewer for point this out, this now changed as; “While no significant group
differences were observed in concentrations (absolute or creatine referenced)” (line 44)
2. The findings of the submitted work to not speak to the cognitive heterogeneity within ASD,
only to a different relationship between metabolites and cognitive profile between ASD and non-
ASD groups. The statement regarding heterogeneity should be removed as it is not supported by
the data.
We thank the reviewer for the suggestion. We have now removed this statement from the
abstract.
Introduction:
3. Please revise the references throughout the manuscript, a few examples in the introduction are:
Line 66: APA DSM 5 ref missing and the others are not adequate.
Line 68: ref missing
Line 69: Ref Howlin 2021 does not cover neuropathology of ASD, hence it’s not adequate.
Line 102: Ref Lam et al 2006 is out of context
We have now revised the references as suggested by the reviewer.
Methods and materials
4. What was the medication status in the ASD group?
Seven of the 14 boys in the ASD group used psychotropic medication (Ritalin, Citalopram, Focalin,
Adderall, Sertraline, Concerta and Risperidone). None of the participants in the TD group were on
medication. This information is now added to the methods section (line 166).
5. What do the ICS-I and ICS-S scores mean, do higher scores mean worse/better performance? It
would be helpful to briefly explain this assessment.
We thank the reviewer for the opportunity to clarify this.
-The NEPSY-II Inhibition Contrast Scaled Score (ICS-I): measures the ability to voluntarily inhibit
attention.
- NEPSY-II Switching Contrast Scaled Score (ICS-S): measures the ability for switching attention
between competing stimuli.
The higher the ICS-I and ICS-S scores, the better the performance. This information is now
included in Figure (3-4) legends.
6. There’s no a priori sample size calculation and, as noted by the authors, the study sample is
small. Since an a priori sample size was not calculated, the authors should at least discuss their
finds compared to other similar studies regarding their sample sizes.
We thank the reviewer for the suggestion. Now we have added:
“Previous studies using smaller ASD samples (n=7-12, ages 8-17), reported differences in Glx and
Glx/Cr in the frontal regions compared to TD groups (Bejjani et al., 2012; Kubas et al., 2012; Joshi et
al., 2013). However, no group-level differences were found in Glx or Glx/Cr in our study. Yet, we
found a significant positive correlation between Glx/Cr and nonverbal intelligence only in the ASD
group.” (line 403).
7. How were the metabolite absolute concentrations estimated? It seems that no water reference
was acquired for that purpose. If that’s the case, ideally all values reported should be relative, for
instance, to creatine. Please, report the analysis using Cho/Cr in addition to Cho.
We apologize for not including this information. The absolute concentrations are estimated using
an unsuppressed water signal. This information is now added to the methods section (line 187).
8. While it is stated that metabolites levels were adjusted for grey matter, white matter and CSF,
later on it is stated that CSF content was used (line 222). Please clarify which correction was made
and provide the formula used.
We have now added S1 Appendix which contains the detailed information of this process. This is
referenced in the methods section.
We used LC Model (LCM) to estimate the corrected metabolite concentrations. LCM uses the
following formula:
Cmet = Ratioarea × (2/N1Hmet) × (ATTH2O/attmet) × WCONC
Cmet = corrected metabolite concentration in mmol/l

Ratioarea= ratio of area of metabolite/area of unsuppressed water
N1Hmet = number of equivalent protons contributing to the resonance
ATTH2O & attmet are the factors (< 1) by which the water and metabolite resonance areas are
attenuated by relaxation. Default; ATTH2O = 0.7, attmet = 1
WCONC = water concentration (in mM) in the voxel.
WCONC = (43300fgm + 35880fwm + 55556fcsf) / (1 – fcsf), fgm, fwm and fcsf are the volume fractions
segmented using the T1-mprage.
The product: [Ratioarea × (2/N1Hmet) × (ATTH2O/attmet)] is the default output of metabolite levels
from LCM and by multiplying this product with WCONC, we get the corrected metabolite
concentration.
9. State that values are in institutional units (for absolute concentrations) as no other corrections
were made (e.g. relaxation times).
This is now added to the methods section (line 217).
10. Line 243 – number 5 should be removed here as there were only 4 behavioural measures for
TD.
We have now removed number 5 from the sentence.
11. Some MRSinMRS reporting standards (Lin et al 2021, NMR in Biomedicine) have been
followed (e.g. visualisation of all spectra), but please provide as much recommended details as
possible for acquisition, quantification, and quality assurance, for instance, which water
suppression method was used, the basis set used, a table with SNR, FWHM and CRLB mean values
and range for each metabolite (could be supp material), etc.
This information is now added to the methods section and a table with quality parameters is in
the supplementary materials.
Parameter TD ASD p-value

SNR 17.2±4.1 17.1±3.3 0.940
FWHM 6.8±2.2 7.9±2.2 0.203
ChoCRLB 2.9±0.6 2.8±0.5 0.697
NAACRLB 2.9±0.5 3.0±0.9 0.666
CrCRLB 2.5±0.5 2.4±0.5 0.551
GlxCRLB 7.1±1.6 6.9±0.9 0.706
Results:
12. What is the test value for the slope comparison? Only p value is reported.
This information is now added to the manuscript (lines: 292, 321 and figure 3-4 legends)
13. Line 305 – Combining the two groups decreased the p value, but the correlation coefficient
was smaller, so it did not strengthen, this last part of the sentence should be removed.
We have now revised this sentence as follows: combining the two groups slightly decreased the
correlation coefficient but increased the statistical significance of the correlation (r=-0.56, p=0.002).
[line: 312].
14. What does the multivariate analysis add to the work? This has not been addressed.
We thank the reviewer for pointing this out and now below statement is added to the discussion
(line 416). The multivariate analysis was added to test whether combining measures improved
predictions. While r and p values for predicting NVIQ and ICS-S were somewhat improved, the
improvement was not statistically significant. This is now added in the results (lines 325).
Discussion
15. Line 348: Again, the correlation is not stronger as the r value is smaller. Please remove.
Now the sentence reads:

“Glx/Cr ratio also showed a positive association with the ICS-S score for attention switching for both
ASD and TD groups, and the combination of the two groups strengthened the statistical significance
since the power (i.e., N) was increased.” (line 389).
16. In the second paragraph the authors discuss the findings in light of the ASD group
correlations, but the Cho findings in the TD group don’t seem to fit here. The discussion is mixing
findings in clinical and non-clinical populations. Please rewrite this part of the discussion in the
light of both ASD and TD findings and previous literature regarding Cho to be clearer on what is
the point of the paragraph.
This paragraph is now re-written as requested by the reviewer (line 369-385).
“Choline containing compounds are key components of cell membranes and the myeline sheets and
associated with white matter density and speculated to reflect excessive neuronal connectivity or
abnormal myelination (Laycock et al., 2008). Our observation of a significant association between
low Cho levels and higher NVIQ in the ASD group is similar to the findings of a previous study
where an inverse association with the left DLPFC Cho/Cr and full-scale IQ (FSIQ) was reported in
generalized anxiety disorder (Coplan et al., 2018b). During disease states, elevated Cho is associated
with membrane breakdown and inflammation (Davie et al., 1995), processes which are likely to
impair executive functioning. However, we also observed a significant positive association between
DLPFC Cho and NVIQ in the control group, where Coplan et al reported no significant association.
Another previous study reported an inverse relationship between Cho and performance IQ in
healthy adults (Jung et al., 1999). A major difference between our study and aforementioned
previous studies is that those included adult populations. Progressive myelination of white matter
pathways during typical development of children and adolescents has significant changes in
cognitive abilities, due to more rapid neural communication (Buyanova and Arsalidou, 2021).
Therefore, dynamic myelin turnover, which is indicative of active Cho metabolism in typically
developing children may explain the positive association with cognitive performance. “
17. Line 388-389: “It also suggests a possible neural mechanism within the DLPFC for the
heterogeneity observed in cognitive function in ASD.” – This statement is not supported by the
data and should be removed.
This statement is now removed.
18. The first paragraph on page 13 (starting on line 368) is missing references for the glutamine-
glutamate-GABA cycle.
19. Paragraph starting on line 390:

First sentence is missing references. The authors discuss literature reporting opposite findings
regarding glutamate correlations and reading performance, but it’s not clear how this discussion
relates to the current manuscript findings, please clarify.
20. In the limitations sections the authors mention that future studies with larger sample sizes are
needed, aren’t there previous studies with larger sample sizes published already, at least for
metabolite between group comparisons? Current findings should be discussed in the light of
previous works.
We have re-worded this sentence as follows:
Because ASD is a neurodevelopmental disorder, it is possible that the function of different

metabolites is impacted differently due to differing maturation trajectories throughout development.
Longitudinal studies of metabolites and their correlation with specific traits at different points in
time could help to elucidate this important question. (lines 436-439).
Reviewer #3:
The authors brought novel ideas of examining neurometabolic profiles and their relatedness with
behavioral measures in autism. It was also clearer that the authors worked hard and put a great deal
of effort in data collection, and data analyses sections. However, the study would become more
interpretable if it was conceptualized better, such that, inclusion of right frontal cortex, wider range of
executive function measures, and full-scale IQ.
We thank the reviewer for the encouraging comments. We acknowledge Reviewer’s concern regarding
the conceptualization of the study. The data for this study was collected a while ago, as a part of a
multimodal study, therefore due to time limitation MRS data was limited to a single location.

Pone D 22 34955 - R1

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Pone D 22 34955 - R1

Uploaded by

Copyright:

Available Formats

PLOS ONE

Dorsolateral Prefrontal Cortex Metabolic Profiles in Autism Spectrum Disorder

Manuscript Number: PONE-D-22-34955R1

Article Type: Research Article

Corresponding Author: Akila S Weerasekera

Keywords: MRS; choline; glutamate; Glx; Executive function

Order of Authors: Akila S Weerasekera

Benamor L. (2014). (1)H-Magnetic resonance spectroscopy study of stimulant

Elucidating the apparently divergent role of neurometabolites in ASD in the absence of

“Elucidating the role of neurometabolites in ASD also in the absence of significant

(7) Similarly, the next statement in the abstract

What is “cognitive function heterogeneity”? And even if there differences across

The regression equation provided

is not clear. What are the terms in the equation?

“YBehavior= β0 + β1[met1]group + β2[met2]group, where dependent variable Y is the

(12) Lines 387-388, Discussion Section

(13) Line 415, discussion section

Therefore, longitudinal MRS studies are clearly necessary.

Why are longitudinal MRS studies clearly needed?

This sentence is now replaced with: “Because ASD is a neurodevelopmental disorder,

(14) Line 418 and 419, discussion section

(15) Line 430 and 431, discussion section

In this work, Weerasekera et al investigated choline and glutamatergic compounds in

I have the following comments and suggestions:

We have now revised the references as suggested by the reviewer.

Methods and materials

4. What was the medication status in the ASD group?

We thank the reviewer for the opportunity to clarify this.

We thank the reviewer for the suggestion. Now we have added:

We used LC Model (LCM) to estimate the corrected metabolite concentrations. LCM

Cmet = Ratioarea × (2/N1Hmet) × (ATTH2O/attmet) × WCONC

Cmet = corrected metabolite concentration in mmol/l

The product: [Ratioarea × (2/N1Hmet) × (ATTH2O/attmet)] is the default output of

This is now added to the methods section (line 217).

We have now removed number 5 from the sentence.

Now the sentence reads:

This paragraph is now re-written as requested by the reviewer (line 369-385).

This statement is now removed.

19. Paragraph starting on line 390:

Appropriate references are now added.

We have re-worded this sentence as follows:

Because ASD is a neurodevelopmental disorder, it is possible that the function of

This statement is required for submission

This statement is required for submission

View published research articles from

Enter: The authors have declared that no

Authors with competing interests

Enter competing interest details beginning

I have read the journal's policy and the

Write "N/A" if the submission does not

General guidance is provided below.

Human Subject Research (involving human

Animal Research (involving vertebrate

animals, embryos or tissues)

Include the following details if this study

involves the collection of plant, animal, or

other materials from a natural setting:

Authors are required to make all data

Important: Stating ‘data available on request

Do the authors confirm that all data

• If the data are held or will be held in a

Data cannot be shared publicly because

The data underlying the results