Needham
et al.
Supplement
1
Plasma and Fecal Metabolite Prof
iles in Autism Spectrum Disorder
Supplement 1
SUPPLEMENTAL METHODS
Extended Participant Information:
Briefly, the GISS consists of 7 sections, each section asking
a series of 1 to 6 questions to determine if the pa
rticipant met the criteria for diarrhea, constipation
or irritable bowel syndrome (IBS)- like symptoms, a
nd if stooling/symptoms
have been consistent
for the last 6 months. Any individual with an
tibiotic use within 3 months was excluded, but
multivitamins and over the counter supplements
were permitted. Based on their responses,
participants were placed in one of four groups
, ASD+GI, ASD-GI, TD+GI or TD-GI. Due to the
low incidence of GI issues in typically developi
ng children enrolled in th
e study, we only have 9
individuals in the TD+GI groups. Our original stat
istical analysis resulted in many seemingly
intriguing metabolites with differential leve
ls between the ASD+GI and TD+GI samples.
However, upon closer inspection, most
of these differences were driv
en by the fact that the TD+GI
group (with a small n) was very di
fferent from all other samples
rather than the ASD+GI being
unique. There were occasional differences that were indeed convincing between the ASD+GI and
TD+GI, but these differences also arose in the compar
ison of -GI samples, so they were not specific
to the GI phenotype and presenting them in an
ASD+GI compared to TD+GI context would have
been misleading. Thus, the TD+GI group was removed
from the analysis and is
not presented here.
This study was approved by institutional review
boards for the State of California and the
University of California, Davis. Informed cons
ent is obtained from a legal guardian for all study
participants prior to data
collection in accordance with the UC Davis IRB protocol.
Needham
et al.
Supplement
2
Blood and Stool Collection
: Peripheral blood was collected fro
m non-fasting partic
ipants in acid-
citrate dextrose Vacutainers (BD Biosciences;
San Jose, Ca) before 12PM in a blinded fashion
from ASD and TD groups. Blood was centrifuged at 2100 rpm for 10 minutes followed by the
collection of plasma into cryovials. Plasma was stor
ed at -80 °C until analysis. In addition to blood,
stool samples were also obtaine
d, parents were given
collection containers with RNA later to
collect stool samples at home and asked to stor
e the samples in the freeze
r and brought back frozen
to the clinic within 24 hrs.
Metabolite Analysis:
All metabolite analysis, identification, quality control were performed by
standard procedures at Me
tabolon Inc. as follows.
Sample Preparation:
All samples were maintained at -80
o
C until processed. Samples were
prepared using the automated MicroLab ST
AR® system from Hamilt
on Company. Several
recovery standards were added prior to the first st
ep in the extraction process for QC purposes. To
remove protein, dissociate small mo
lecules bound to protein or trap
ped in the precipitated protein
matrix, and to recover chemically diverse metabolit
es, proteins were precipitated with methanol
under vigorous shaking for 2 min (Glen Mills Ge
noGrinder 2000) followed by centrifugation. The
resulting extract was divided into five fractions
: two for analysis by two
separate reverse phase
(RP)/UPLC-MS/MS methods with
positive ion mode electrospray ionization (ESI), one for
analysis by RP/UPLC-MS/MS with negative ion
mode ESI, one for analysis by HILIC/UPLC-
MS/MS with negative ion mode ESI, and one sa
mple was reserved for backup. Samples were
placed briefly on a TurboVap® (Zymar
k) to remove the organic solvent. The sample extracts were
stored overnight under ni
trogen before preparation for analysis.
Needham
et al.
Supplement
3
QA/QC:
Several types of controls were
analyzed in concert with
the experimental samples: a
pooled matrix sample generated
by taking a small volume of
each experimental sample (or
alternatively, use of a pool of
well-characterized human plasma)
served as a technical replicate
throughout the data set; extracted water samples serv
ed as process blanks; and a cocktail of QC
standards that were carefully chosen not to
interfere with the m
easurement of endogenous
compounds were spiked into ever
y analyzed sample, allowed inst
rument performance monitoring
and aided chromatographic alignment. Instrument
variability was deter
mined by calculating the
median relative standard
deviation (RSD) for the standards that
were added to e
ach sample prior
to injection into the mass spectro
meters. Overall process variability was determined by calculating
the median RSD for all endogenous metabolites (i
.e., non-instrument standards) present in 100%
of the pooled matrix samples. Experimental
samples were randomized
across the platform run
with QC samples spaced ev
enly among the injections.
Ultrahigh Performance Liquid Chromatogr
aphy-Tandem Mass Spectroscopy (UPLC-
MS/MS):
All methods utilized a Waters ACQUITY
ultra-performance liquid chromatography
(UPLC) and a Thermo Scientific Q-Exactiv
e high resolution/accurate mass spectrometer
interfaced with a heated electrospray ioniza
tion (HESI-II) source and Orbitrap mass analyzer
operated at 35,000 mass resolu
tion. The sample extr
act was dried then reconstituted in solvents
compatible to each of the four methods. Each r
econstitution solvent contained a series of standards
at fixed concentrations to ensu
re injection and chromatographi
c consistency. One aliquot was
analyzed using acidic positive ion conditions, ch
romatographically optimi
zed for more hydrophilic
compounds. In this method, the extract was grad
ient eluted from a C18
column (Waters UPLC
BEH C18-2.1x100 mm, 1.7 μm) using
water and methanol, containi
ng 0.05% perfluoropentanoic
acid (PFPA) and 0.1% formic acid (F
A). Another aliquot was also
analyzed using acidic positive
Needham
et al.
Supplement
4
ion conditions, however it was
chromatographically
optimized for more hydrophobic compounds.
In this method, the extract was gradient eluted
from the same afore mentioned C18 column using
methanol, acetonitrile, water, 0.05% PFPA and 0.01%
FA and was operated at an overall higher
organic content. Another aliquot was analyzed
using basic negative
ion optimized conditions
using a separate dedicated C18 column. The basic
extracts were gradient eluted from the column
using methanol and water, however with 6.5mM A
mmonium Bicarbonate at pH 8. The fourth
aliquot was analyzed via negative ionization
following elution from a HILIC column (Waters
UPLC BEH Amide 2.1x150 mm, 1.7 μm) using a gradie
nt consisting of water and acetonitrile
with 10mM Ammonium Formate, pH 10.8. The MS
analysis alternated
between MS and data-
dependent MS
n
scans using dynamic exclusion. The scan
range varied slight
ed between methods
but covered 70-1000 m/z. Raw da
ta files are archived and ex
tracted as described below.
Bioinformatics:
The informatics system consisted of
four major component
s, the Laboratory
Information Management System
(LIMS), the data extraction and
peak-identification software,
data processing tools for QC and compound id
entification, and a colle
ction of information
interpretation and visualization tools for use by
data analysts. The hardware and software
foundations for these informatic
s components were the LAN bac
kbone, and a database server
running Oracle 10.2.0.1 Enterprise Edition.
LIMS:
The purpose of the Metabolon LIMS system
was to enable fully auditable laboratory
automation through a secure, easy
to use, and highly specialized
system. The scope of the
Metabolon LIMS system encompasses sample acce
ssioning, sample prepar
ation and instrumental
analysis and reporting and advanced data analysis
. All of the subsequent software systems are
grounded in the LIMS data structures. It has been
modified to leverage
and interface with the in-
Needham
et al.
Supplement
5
house information extraction and data visualization
systems, as well as thir
d party instrumentation
and data analysis software.
Data Extraction and Compound Identification:
Raw data was extracted, peak-identified and
QC processed using Metabolon’s hard
ware and software. These syst
ems are built on a web-service
platform utilizing Microsoft’
s .NET technologies, which run
on high-performance application
servers and fiber-channel storage arrays in cluste
rs to provide active failo
ver and load-balancing.
Compounds were identified by comparison to library
entries of purified st
andards or recurrent
unknown entities. Metabolon maintain
s a library based on authentica
ted standards that contains
the retention time/index (RI), mass to charge ratio (
m/z)
, and chromatographic data (including
MS/MS spectral data) on all molecules present
in the library. Furthermore, biochemical
identifications are based on three
criteria: retention index within
a narrow RI window of the
proposed identification, accurate
mass match to the library +/-
10 ppm, and the MS/MS forward
and reverse scores between the experimental data
and authentic standards. The MS/MS scores are
based on a comparison of the ions pr
esent in the experimental spectr
um to the ions present in the
library spectrum. While there ma
y be similarities between these
molecules based on one of these
factors, the use of all three data points can be utilized to distinguish and differentiate biochemicals.
More than 3300 commercially av
ailable purified standard co
mpounds have been acquired and
registered into LIMS for analysis on all platforms for determination of their analytical
characteristics. Additional ma
ss spectral entries have
been created for
structurally unnamed
biochemicals, which have been iden
tified by virtue of their recurre
nt nature (both chromatographic
and mass spectral). These compoun
ds have the potential to be id
entified by future
acquisition of
a matching purified standard or by
classical structural analysis.
Needham
et al.
Supplement
6
Complex Lipids Platform:
Lipids were extracted from samples in methanol:dichloromethane in
the presence of internal sta
ndards. The ex
tracts were concentrat
ed under nitrogen and
reconstituted in 0.25mL of 10mM ammonium acetat
e dichloromethane:methanol (50:50). The
extracts were transferred to inserts and placed
in vials for infusion-MS analysis, performed on a
Shimazdu LC with nano PEEK tubing and the Sciex SelexIon-5500 QTRAP. The samples were
analyzed via both positive and ne
gative mode electrospray. Th
e 5500 QTRAP scan was performed
in MRM mode with the total of
more than 1,100 MRMs. Individua
l lipid species were quantified
by taking the peak area ratios of target compounds
and their assigned internal standards, then
multiplying by the concentration of internal
standard added to the sample. Lipid class
concentrations were calculated from the sum of a
ll molecular species within a class, and fatty acid
compositions were determined by calculating th
e proportion of each class comprised by individual
fatty acids.
Peak Quantification, Normalizat
ion and Statistical Analysis:
Peaks were quantified using area-
under-the-curve. The present dataset compri
ses a total of 1611 and 814 compounds of known
identity for the human plasma a
nd fecal samples, respectively. The mouse dataset comprises a total
of 746 known biochemicals. Following log transfor
mation and imputation of
missing values, if
any, with the minimum observed value for each
compound, ANOVA contrast
s using ArrayStudio
(Qiagen) were used to identify
biochemicals that differed signi
ficantly between experimental
groups. Metabolites with greater th
an 30% missing values or inconsis
tent levels of missing values
across sample groups were met with
extreme prejudice for inclusion
in the results section and are
presented along with box plots in order to visua
lly show disparities be
tween detection levels.
Analysis by two-way ANOVA identi
fied biochemicals exhibiting sign
ificant interaction and main
effects for experimental parameters of disease a
nd GI symptoms. For mouse samples, in addition
Needham
et al.
Supplement
7
to two-way ANOVA contrasts, in
order to account for
non-independence (match
to donor), we ran
a linear mixed effect
model where the donor match is a rand
om effect. The analysis was done in
R.gui version 3.5.1. Correlation an
alysis was performed compar
ing the relationship between
observed changes in metabolite levels and measures of severity. An estimate of the false discovery
rate (q-value) is calculated to take into account
the multiple comparison
s that normally occur in
metabolomic-based studies. In all graphs of i
ndividual metabolites, the si
gnificance noted in the
graph is based on p-value from the ANOVA and
can be found in the corresponding Supplemental
Tables along with the q-values, wh
ich were set to a cutoff of 0.1.
Principle Components Analysis (PCA):
All PCA plots were generated using the ClustVis web
tool with ellipses marking the 95% confidence
interval. In cases where ASD samples were
subdivided by severity, samples were separated
and grouped by behavioral score as close to
quartiles as possible without spli
tting samples with th
e same score into separate groups. The
groupings were as follows: for human feces sc
ores for lower and higher severity groups,
respectively: ADOS scores 4-5 an
d 8-10; nonverbal scores 1-4 a
nd 7-10, social sc
ores 10-14 and
20-27, verbal scores 1-3 and 5. Groupings for human
plasma scores for lowe
r and higher severity
groups, respectively: ADOS scores 4-5 and 9-10;
nonverbal scores 1-4 and
8-11, social scores 10-
14 and 22-27, verbal scores 1-3 and 5.
Random Forest Analysis:
Random Forest Analysis was perf
ormed as described previously
1
. A
random subset of the data with
identifying true class information
was selected to
build the tree
(“bootstrap sample” or “training
set”), and then the remaining
data, the “out-of-bag” (OOB)
variables, were passed down the tree to obtain
a class prediction for each sample. This process
was repeated thousands of times to produce th
e forest (total of 50,
000 trees). The final
classification of each sample was determined by
computing the class pred
iction frequency for the
Needham
et al.
Supplement
8
OOB variables over the whole forest
. When the full forest is gr
own, the class predictions were
compared to the true classes, generating the “OOB
error rate” as a measure
of prediction accuracy.
The mean decrease in
accuracy was determined
by randomly permuting
a variable, running the
observed values through the trees, and then reassessing the prediction accuracy. Thus, the Random
Forest analysis provides an “importance” rank ordering of biochemicals.
Details on the “top 30” analysis
are as follows. Running a random
forest and then re-running it
again with just the top 30 will give
a biased estimator of the accuracy
(too optimistic). Because of
this, leave-one-out cross-valida
tion (LOO-CV) was performed as fo
llows: fit the random forest
for all observations ex
cept subject j. Then re-fit the random
forest using only the top 30 variables,
and this forest is fitted using al
l observations except subj
ect j. Then predict subject j from this
random forest. Repeat for all j.
The areas under the receiver operator curves (AUC
s) and their respective 95% confidence intervals
based on the DeLong method
are as follows: Plasma: ASD(
all)/TD: AUC = 0.7716; 95% CI:
0.7082-0.835; ASD/TD, top 30 panel: AUC = 0.7773;
95% CI: 0.7148-0.8398; ASD-GI/TD, top
30 panel: AUC = 0.7948; 95% CI: 0.7287-0.8609. Feces
: ASD(all)/TD: AUC = 0.7358; 95% CI:
0.6276-0.8441; ASD(all)/TD top 30: AUC = 0.7953;
95% CI: 0.6943-0.8964; ASD-GI/TD top 30
panel: AUC = 0.6647; 95% CI = 0.5276-0.8018.
Pathway Enrichment Analysis:
For each individual pair-wise co
mparison, pathway enrichment
determined the number of
statistically sign
ificantly different
compounds relative
to all detected
compounds in a sub-pathway, compared to the tota
l number of statisticall
y significantly different
compounds relative to all detect
ed compounds in the study. A path
way enrichment value greater
than one indicates that
the pathway contains more
significantly changed co
mpounds relative to the
study overall, suggesting that the pathway may be a
target of interest fo
r further investigation.
Needham
et al.
Supplement
9
Enrichment Value = (# of significan
t metabolites in pathway(k) / total # of detected metabolites in
pathway(m) ) / (total # of significant metabo
lites(n) / total # of de
tected metabolites(N) )
(k/m)/(n/N).
Animal Husbandry and Sample Collection:
All mouse housing and experiments were approved
by the California Institute of T
echnology IACUUC. Fecal samples were thawed, 300mls of 1.5%
bicarbonate in PBS was added, sample
s were vortexed, and then allo
wed to settle. 150uL of fecal
sample was then delivered by oral gavage to 5-we
ek-old germ free mice that were then maintained
in sterile, microisolator cages for 3 weeks. After
three weeks, fecal sample
s were collected for 16s
sequencing. Blood was collected at the same
time for all mice by cardiac puncture following
euthanasia by CO
2
, plasma was isolated on ice using EDTA
-treated collection tubes (Thermo), and
samples were stored at -80C until analysis. One
replicate GP sample was thawed prematurely and
thus removed from further analysis.
Fecal Analysis for 16s Sequencing:
Fecal samples collected and
immediately put into empty
sterile tubes, flash frozen, and
maintained at -80C until proces
sing. Total DNA was isolated with
Qiagen DNeasy powersoil extrac
tion kit (Qiagen) following
manufacturer instructions.
Hypervariable V4 region of the 16s gene wa
s amplified by PCR usi
ng 5Xprime master mix
(Prime). Barcoded 806 reverse primers and uni
que forward 515 primer (IDT) were used as
previously described. The amplif
ication was confirmed by elec
trophoresis and
the amplified
products were purified with Quiaqu
ick PCR purification kit (Qiagen).
Samples were sent to MGH NGS Core facility to
be sequenced on the Illumina MiSeq instrument
using MiSeq v2 500-cycle sequencing kit, resulti
ng in approximately 25 million paired-end 250
bp reads covering amplicon region
s. Data were analyzed QIIM
E2 software package at the
Bioinformatic core at MGH. Low
quality score sequencing reads (a
verage Q < 25) were truncated
Needham
et al.
Supplement
10
to 240bp followed by filtering using
deblur
algorithm with default setting
and the
high-quality
reads were aligned to the reference library using
mafft
. The aligned reads were masked to remove
highly variable positions, and a
phylogenetic tree was generated fro
m the masked a
lignment using
the FastTree method. Alpha and beta diversity me
trics and Principal Com
ponent Analysis plots
based on Jaccard distance were generated using default QIIME2 plugins. Taxonomy assignment
was performed using
feature-classifier
method and naïve Bayes cl
assifier trained on the
SILVA138 database operational taxo
nomic units (OTUs). Linear disc
riminant analysis Effect Size
(LEfSe) was performed as
described previously
2
using the Galaxy web application.
Supplemental References
1. Breiman, L. Random Forests.
Machine Learning
45
, 5–32 (2001).
2. Segata, N.
et al.
Metagenomic biomarker di
scovery and explanation.
Genome Biology
12
, R60
(2011).
Needham
et al.
Supplement
11
ASD
GP
0
50
100
150
200
Age (mos)
*
All Samples
A
B
Supplementary figure 1.
Plasma
Sex
GI
Symptoms
Age range
(mos)
Mean
Age
(mos)
Total
N
Group
Male Female +GI -GI ND
TD
89
12
0 90
2
39-151
80
101
ASD
110
20
40 86
4
37-153
88
130
Feces
Sex
GI
Symptoms
Age range
(mos)
Mean
Age
(mos)
Total
N
Group Male Female ND +GI -GI ND
TD
36
3
1
0 33
1
39-151
86
40
ASD
48
7
2 22 31
4
37-149
91
57
D
ASD+GI ASD-GI
0
50
100
150
200
Age (mos)
ns
G
ASD
GP
0
50
100
150
Age (mos)
ns
132 mos (11yrs) and under
GP
ASD
0
50
100
150
200
Age (mos)
ns
ASD+GI ASD-GI
0
50
100
150
200
Age (mos)
ns
All Samples
All Samples
All Samples
ASD+GI
ASD-GI
0
5
10
15
ADOS SS
*
F
0
50
100
150
200
0
5
10
15
Age
Sev Score
Corr=-0.039 p=0.66
0
50
100
150
200
0
2
4
6
Age
Verbal
Corr=-0.120 p=0.18
0
50
100
150
200
0
5
10
15
Age
Nonverbal
Corr=-0.160 p=0.07
0
50
100
150
200
0
10
20
30
Age
Social
Corr=-0.092 p=0.30
E
TD
TD
TD
ASD
TD
ASD
TD
C
0.1
1
10
100
10
100
1000
10000
CM PF #1 (relative)
CMPF #2 (actual, ng/m L)
CMPF
r
2
= 0.98
0.01
0.1
1
10
10
100
1000
10000
pCS #1 (relative)
pCS #2 (actual, ng/mL)
p-cresol
sulfate
r
2
= 0.98
0.1
1
10
100
10
100
1000
HHA #1 (relative)
H HA #2 (actual, ng/mL)
3-hydroxyhippurate
r
2
= 0.96
0.01
0.1
1
10
100
1000
0.1
1
10
100
1000
4-EPS #1 (relative)
4-EPS #2 (actual, ng/mL)
4EPS
Below limit of quantification
r
2
= 0.97
0.1
1
10
10
100
1000
10000
3-IS #1 (relative)
3-IS #2 (actual, ng/m L)
3-indoxylsulfate
r
2
= 0.98
0.0
0.5
1.0
1.5
2.0
0
50
100
150
200
N-acetylserine #1 (relative)
N-acetylserine #2 (actual, ng/m L)
N-acetylserine
r
2
= 0.63
0.01
0.1
1
10
100
10
100
1000
pCG #1 (relative)
pCG #2 (actual, ng/mL)
p-cresol
glucuronide
r
2
= 0.96
H
Needham
et al.
Supplement
12
Figure S1. Group characteristics.
(
A) Numbers of plasma samples: sex, GI symptoms, age range, and total n. (B) Numbers of fecal
samples: sex, GI symptoms, age range, and total n.
(C) PCA plot of plasma (left) and feces (right)
with all metabolites as input. (D
) Ages of donors of all plasma
samples, all plasma samples 11
years and younger, and according to
GI status. (E) Ages of donors
of all fecal samp
les and samples
according to GI status. (F) ADOS-SS graphed accordi
ng to GI status within
the ASD sample group.
(G) Correlations between age (in
months) ASD plasma sample indi
viduals and behavior metrics
are shown, with linear regression line, spearman co
rrelation value, and p-value. (H) Correlation
plots comparing the relative abundance measured
in the untargeted analysis and the follow-up
quantitated values for a subset
of 139 plasma samples. Correlati
ons for 4-ethylphenyl sulfate, p-
cresol sulfate, 3-carboxy-4-methyl-5-propyl
-2-furanpropanoate (CM
PF), 3-hydroxyhippurate, 3-
indoxylsulfate, N-acetylserine, and
p-cresol glucuronide are shown.
(D-F) Data are represented as
mean ± SEM analyzed by Wel
ch’s t-test: *p-value<0.05.
Needham
et al.
Supplement
13
E
Supplementary figure 2.
D
Verbal
Social
Nonverbal
ADOS SS
PC Ester
111
DAG E s ter
101
TAG E s ter
11
MF A
101
Acyl Carn
101
Primary Bile
01
1
Pentose
11
1
Hemoglobin
111
Food/Plant
11
Polyamine
10
1
Glu Met
11
0
Me t, C y s , S A M
100
#
Glutathione
100
Pyrimidine
100
ɣ
-glut A A
100
Ceramides
11
S phingolipid
11
Second. Bile
100
MA G
111
Gly, Ser, Thr
0
11
Negative corr:
Amino Acid
Carbohydrate
Cofactors and Vit.
Complex Lipids
Energy
Lipid
Nucleotide
Peptide
Xenobiotics
Super Pathway
p<0.05
Positive corr:
p<0.05
C
02468
Pentose Metabolism
Phosphatidylcholine (PC)
Lysophospholipid
Phospholipid Metabolism
Monoacylglycerol
DAG Ester
Met, Cys, SAM and Taur Met.
Hexosylceramide
CE Ester
Acyl Carnitine (long)
Saturated FFA
Androgenic Steroids
LPC Ester
Pregnenolone Steroids
Acyl Carnitine (polyusat.)
Enrichment Value
1.7E-5
6.7E-4
3.4E-13
3.7E-7
9.1E-4
0.03
2.8E-4
0.03
8.9E-3
0.02
2.9E-7
2.4E-3
3.8E-5
0.02
0.02
Plasma
Feces
02468
Androgenic Steroids
Phosphatidylcholine (PC)
Monoacylglycerol
Diacylglycerol
PC Ester
Food Component/Plant
PE Plasmalogen
Chemical
PE Ether
Acyl Carnitine (long)
CE Ester
LPC Ester
Phospholipid Metabolism
Acyl Carnitine (medium)
Androgenic Steroids
Pregnenolone Steroids
Enrichment Value
6.1E-4
3.4E-10
0.01
7.0E-3
1.4E-6
9.5E-7
0.03
2.7E-3
1.2E-4
8.4E-3
5.8E-9
7.9E-4
0.02
8.4E-3
0.03
0.01
Plasma
Feces
A
B
ASD+GI compared to TD controls
ASD-GI compared to TD controls
Needham
et al.
Supplement
14
Figure S2. Extended Enriched Pathway
and Behavioral Correlation Results
(A-B) All pathways significantly altered in th
e comparison between ASD+GI and ASD-GI human
plasma samples compared to TD, with enrichment
value plotted and p-value
to the right of each
bar. Metabolites within each pathway could be obser
ved at either higher or lower levels, as this
plot only indicates significant changes. (C) Spear
man correlations between behavior scores of
ASD children in the ADIR diagnostic test (Verbal,
Social, and Nonverbal
metrics) and the ADOS-
SS and metabolite pathways in AS
D samples. Directionality of
correlation is indicated in the
legend at bottom. Colors of pathwa
ys are defined at the top left of
the chart. A split box means that
both positive and negative correl
ations occur with metabolites within that pathway. (D)
Representative correlation plot of levels of a
phosphatidylcholine and verbal
behavior scores. (E)
Representative correlation plots of FFA le
vels and social behavior scores. HExCer,
Hexosylceramide; PC, phosphatidylcholine; DAG, di
acylglycerol; TAG, tr
iacylglycerol; MFA,
monounsaturated fatty acid; PE plas, phos
phatidylethanolamine plasmalogens; Endocann,
endocannabinoid; Met, methionine; Cys, cysteine
; SAM, s-adenosyl methionine; Ala, alanine;
Asp, aspartate;
ɣ
-glutAA, gamma-glutamyl amino acids;
Second. Bile, secondary bile; MAG,
monoacylglycerol; Gly, glycine;
Ser, serine; thr, threonine
Needham
et al.
Supplement
15
Supplementary Figure 3.
0
5
10
15
20
25
30
35
40
45
taurine
**
ASD
TD
P
U
0.0
0.5
1.0
1.5
2.0
2.0
2.5
3.0
taurine
ASD
TD
C
GP
ASD
0
1
2
3
4
6
12
argininate
*
GP
ASD
0
1
2
3
5
10
15
norvaline
*
GP
ASD
0
1
2
3
4
5
8.65
urea
**
Dimethylarginine
(SDMA+ADMA)
homocitrulline
citrulline
arginine
argininosucc
urea
homoarginine
ornithine
*
Trans-4-hydroxyproline,
N-methylproline
proline
*
D
AB
Lowered in ASD:
p<0.05, q<0.1
p<0.05, q>0.1
Elevated in ASD:
p<0.05, q<0.1
p<0.05, q>0.1
No sig. change:
Not detected:
* = Only -GI group
TD
TD
TD
(q value >0.1)
(q value >0.1)
***
(p value 0.06,
q value >0.1)
(q value >0.1)