Received: 11 September 2023 Accepted: 18 January 2024

DOI: 10.1002/aur.3101

RESEARCH ARTICLE

Evaluating the regulatory function of non-coding autism-associated

single nucleotide polymorphisms on gene expression in human brain
tissue

Kealan Pugsley | Atefeh Namipashaki | Mark A. Bellgrove | Ziarih Hawi

Turner Institute for Brain and Mental Health Abstract

and School of Psychological Sciences, Monash
University, Melbourne, Victoria, Australia
Common variants account for most of the estimated heritability associated with
autism spectrum disorder (autism). Although several replicable single nucleotide
Correspondence polymorphisms (SNPs) for the condition have been detected using genome-wide
Ziarih Hawi, Turner Institute for Brain and association study (GWAS) methodologies, their pathophysiological relevance
Mental Health, Monash University,
18 Innovation Walk, Clayton, Victoria 3800,
remains elusive. Examining this is complicated, however, as all detected loci are
Australia. situated within non-coding regions of the genome. It is therefore likely that they
Email: [email protected] possess roles of regulatory function as opposed to directly affecting gene coding
sequences. To bridge the gap between SNP discovery and mechanistic insight, we
Funding information applied a comprehensive bioinformatic pipeline to functionally annotate autism-
National Health and Medical Research Council
of Australia, Grant/Award Numbers: associated polymorphisms and their non-coding linkage disequilibrium (i.e., non-
APP1146644, APP1002458 randomly associated) partners. We identified 82 DNA variants of probable regu-
latory function that may contribute to autism pathogenesis. To validate these pre-
dictions, we measured the impact of 11 high-confidence candidates and their
GWAS linkage disequilibrium partners on gene expression in human brain tissue
from Autistic and non-Autistic donors. Although a small number of the surveyed
variants exhibited measurable influence on gene expression as determined via
quantitative polymerase chain reaction, these did not survive correction for multi-
ple comparisons. Additionally, no significant genotype-by-diagnosis effects were
observed for any of the SNP-gene associations. We contend that this may reflect
an inability to effectively capture the modest, neurodevelopmental-specific impact
of individual variants on biological dysregulation in available post-mortem tissue
samples, as well as limitations in the existing autism GWAS data.

Lay Summary
Although researchers have identified genetic variants associated with autism, it is
not well understood how exactly they contribute to the onset of the condition.
This study examined whether a set of likely functional variants impacted on gene
expression in human brain tissue from Autistic and non-Autistic donors. We
found that none of the variants were associated with changes in the amount of
gene product expressed. This may reflect problems with the way researchers regu-
larly predict and measure the biological function of genetic mutations.

KEYWORDS
autism, gene regulation, GWAS, post-mortem, SNP

This is an open access article under the terms of the Creative Commons Attribution-NonCommercial-NoDerivs License, which permits use and distribution in any
medium, provided the original work is properly cited, the use is non-commercial and no modifications or adaptations are made.
© 2024 The Authors. Autism Research published by International Society for Autism Research and Wiley Periodicals LLC.

Autism Research. 2024;1–15. wileyonlinelibrary.com/journal/aur 1

 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2 PUGSLEY ET AL.

INTRODUCTION specific polymorphisms and phenotypes, systematic co-

inheritance of alleles at independent loci (i.e., linkage dis-
Autism spectrum disorder (autism) is a complex neurode- equilibrium [LD]) creates difficulty in discerning the
velopmental condition marked by extensive clinical and causal pathogenic locus (Uffelmann & Posthuma, 2021).
etiological heterogeneity. Autism is broadly characterized It may therefore be the case that a variant surpassing the
by difficulties in social communication and restricted threshold of significance may simply tag a SNP of genu-
and/or repetitive behaviors or interests (American Psychi- ine effect and be functionally neutral. Secondly, GWAS
atric Association, 2022). The clinical phenotype results in associations alone provide no indication as to the biologi-
varying degrees of support needs for Autistic individuals cal relevance of the genome-wide significant SNP(s) to
across the lifespan. Although several heritable and non- their associated condition or trait, necessitating further
heritable factors are implicated in the etiology of autism, investigation. These limitations require resolution on two
the exact cause of the condition remains unknown. This fronts: (1) a process of functional annotation and prioriti-
has severely limited the design and implementation of zation to isolate variants likely to influence the target
pre-manifest diagnostic practices, and hindered the estab- phenotype from their non-functional LD partners, and
lishment of therapeutics capable of directly addressing (2) validation of the biological mechanisms through
the core challenges experienced by members of the Autis- which the prioritized SNPs may confer disorder liability.
tic community (Vorstman et al., 2017). The latter is of Consistent with hypotheses for the pathogenic relevance
particular concern, as interindividual heterogeneity in of common variants to health phenotypes (Schork
diagnostic presentations has given rise to an increasing et al., 2009), all autism-GWAS SNPs were situated within
trend of trial-and-error drug prescription and polyphar- non-coding regions of the genome. Unlike exonic variants,
macy (Bowden et al., 2020; Vohra et al., 2016). these SNPs do not directly disrupt the genic sequences
It is widely acknowledged that genetic factors play a encoding gene products. Instead, they contribute to patho-
role in influencing the development of autism, with heri- genesis through subtle regulatory effects on gene expression
tability estimates in the range of 60%–90% (Castelbaum (The ENCODE Project Consortium, 2012). Though a rec-
et al., 2019; Tick et al., 2016). The genetic architecture of ognized source of genetic burden for etiologically complex
the condition includes contributions from variants of dif- conditions including autism (e.g., Alex et al., 2019;
fering penetrance and expressivity (Weiner et al., 2017). Golovina et al., 2021), the distinct mechanisms driving
Common inherited variants (i.e., single nucleotide poly- these associations remain difficult to investigate. As such,
morphisms [SNPs]) predominate the overall genetic lia- we sought to identify non-coding GWAS and/or LD SNPs
bility of autism (
12%–52%; Gaugler et al., 2014). The of regulatory capacity by applying a modified in silico
small effect sizes conferred by these variants necessitate pipeline previously described by Tong et al. (2016). This
exceptionally large samples to detect genuine genotype– pipeline has been specifically designed to facilitate the pri-
phenotype relationships (Schork et al., 2009), historically oritization of non-coding variants for functional analysis,
hampering the feasibility of SNP discovery analyses. and comprises several bioinformatic tools which annotate
However, international genomic working groups the likely effects of individual polymorphisms on gene
(e.g., Psychiatric Genetic Consortium [PGC]) have now expression. To validate these predictions, we examined the
amassed sufficiently large samples to permit the identifi- impact of a subset of the bioinformatically selected SNPs
cation of genome-wide significant variants for a range of on putative gene(s) expression in post-mortem human brain
traits and conditions, including autism. tissue using quantitative polymerase chain reaction
A recent meta-analysis of genome-wide association (qPCR). Samples were obtained from Autistic and non-
study (GWAS) data of 18,381 Autistic and 27,969 Autistic donors, permitting detection of the independent
non-Autistic individuals, in combination with multi-trait regulatory effects of the prioritized SNPs, as well as synergy
analysis of GWAS (MTAG) data from three genetically between these and individuals’ genetic liability for the con-
overlapping phenotypes (i.e., depression, schizophrenia, dition. We anticipated that allelic variation in the SNPs
educational attainment), identified 12 significant risk loci would impact overall expression of the target gene(s), pro-
for autism (Grove et al., 2019). These SNPs represent the viding preliminary evidence of their functionality.
first replicable genome-wide significant common variants
(p ⩽ 5  108) for the condition, providing novel clues for
biological discovery. In addition, gene-based and gene-set MATERIALS AND METHODS
analyses of the GWAS summary statistics (via multi-
marker analysis of genoMic annotation [MAGMA]) iden- Prioritization of functional SNPs
tified 15 genetic associations with autism, seven of which
existed within loci not implicated in the primary GWASs. Identification of non-coding autism-associated
Delineating the functional role of genome-wide signif- variants
icant SNPs in human health is challenging for a number
of reasons. Firstly, though the GWAS discovery Genome-wide significant SNPs (i.e., autism-GWAS and
approach provides insights into associations between MTAG) and leading candidates detected via MAGMA,
 TABLE 1 Index SNPs and Their Prioritized LD SNP Partners selected for Experimental Validation via qPCR.
SNP GRCh37 Index SNP partner Distance (bp) LD (r 2) Analysis GWAS p GWAS OR A1/A2 MAF Class Gene(s)

Index variants (all)

PUGSLEY ET AL.

rs910805 20:21248116 - - - Autism GWAS 2.04  109 0.909 A/G 0.263 NF KIZ, XRN2, NKX2–2, NKX2–4

rs10099100 8:10576775 - - - Autism GWAS 1.07  108 1.088 C/G 0.355 NF C8orf74, SOX7, PINX1
8
rs201910565 1:96561801 - - - Comb. Autism GWAS 2.48  10 0.927 A/T 0.337 NF LINC02790, PTBP2

rs71190156 20:14836243 - - - Autism GWAS 2.75  108 0.925 GTTTTTT/G 0.482 NF MACROD2

rs111931861 7:104744219 - - - Comb. Autism 3.53  108 0.805 A/G 0.045 NF KMT2E, SRPK2
12
rs2388334 6:98591622 - - - MTAG (Autism-Edu) 3.34  10 0.935 A/G 0.495 NF MMS22L, POU3F2

rs325506 5:104012303 - - - MTAG (Autism-MDD) 3.26  1011 1.074 C/G 0.484 NF NUD12
9
rs11787216 8:142615222 - - - MTAG (Autism-Edu) 1.99  10 0.933 T/C 0.353 NF MROH5

rs1452075 3:62481063 - - - MTAG (Autism-Edu) 3.17  109 1.084 T/C 0.279 NF CADPS

rs1620977 1:72729142 - - - MTAG (Autism-MDD) 6.66  109 1.064 A/G 0.292 NF NEGR1
9
rs10149470 14:104017953 - - - MTAG (Autism-MDD) 8.52  10 0.946 A/G 0.442 NF MARK3, CKB, TRMT61, BAG5,
APOPT1, KLC1, XRCC3
6
rs16854048 4:42123728 - - - MTAG (Autism-MDD) 3.04  10 1.086 A/C 0.124 NF SLC30A9, BEND4,
TMEM33, DCAF4L1

rs13188074 5:113801423 - - - MAGMA 3.04  106 1.068 A/G 0.342 NF KCNN2

rs142920272 17:44301840 - - - MAGMA 2.89  107 0.913 T/C 0.221 NF KANSL1

rs146122400 17:44782225 - - - MAGMA 2.26  106 0.918 /GT 0.205 NF WNT3

7
rs141455452 17:44019083 - - - MAGMA 8.94  10 1.082 T/G 0.413 NF MAPT

rs17701675 8:8850633 - - - MAGMA 3.79  106 0.936 G/A 0.384 NF MFHAS1

rs6992403 8:10790758 - - - MAGMA 1.28  107 0.925 C/G 0.447 NF XKR6

5
rs7831557 8:10280228 - - - MAGMA 1.75  10 1.062 G/A 0.429 NF MSRA

17:43965129 - - - MAGMA 1.23  106 1.086 T/ 0.245 NF CRHR1

rs529507 11:131773383 - - - MAGMA 5.76  106 0.913 A/G 0.145 NF NTM

6
11:102751102 - - - MAGMA 3.88  10 0.851 C/ 0.040 NF MMP12

rs2736342 8:11347289 - - - MAGMA 3.33  106 0.937 T/G 0.403 NF BLK

Prioritized Variants Selected for Experimental Validation

rs2668627 17:44353693 rs142920272 51,853 0.913 - 4.27  106 0.923 T/C 0.232 I ARL17B

rs17763086 17:43905481 chr17:43965129a 59,648 0.903 - 5.94  106 0.925 T/G 0.226 I CRHR1, LINC02210-CRHR1
a 6
rs17574361 17:44108202 rs142920272, chr17:43965129 193,638, 143,073 0.941,0.903 - 8.39  10 0.926 A/G 0.226 II KANSL1

rs1918788 17:44267617 rs142920272 34,223 0.912 - 3.86  106 0.924 A/G 0.232 I KANSL1

rs366858 17:43726588 chr17:43965129a 238,541 0.903 - 9.32  106 1.079 T/C 0.226 I LINC02210-CRHR1
a 6
rs1052594 17:44102689 rs142920272, chr17:43965129 199,151, 137,560 0.926,0.889 - 9.63  10 1.079 C/G 0.229 II MAPT

rs111972148b 17:43973121 chr17:43965129a 7992 0.903 - 8.26  106 1.080 C/G 0.226 I MAPT

(Continues)
3

19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
4 PUGSLEY ET AL.

henceforth referred to as index SNPs/variants, are listed

(autism or combined [comb.]-autism), MTAG (Autism-Educational Attainment [Edu]/ Autism-Major Depressive Disorder [MDD]) or MAGMA analysis in which the SNP surpassed the threshold of significant association or was a lead candidate for a significant gene/
gene-set association. The variants’ significance value ( p) and odds ratio (OR) are reported as per the PGC GWAS/multi-trait analysis of GWAS. All MAGMA and prioritized variants’ p and OR reported as per the PGC autism GWAS. Genes presented in bold were
Note: Prioritized SNPs with >1 index listed were found to meet LD threshold parameters for multiple autism GWAS/ lead candidate index variants. Prioritized variants with negative distance values are upstream of the index SNP. Analysis refers to the PGC GWAS
in Table 1. SNPs in LD with these were identified via the
LDProxy web tool (LDLink Project; Machiela &
Chanock, 2015). Variants within ±250 kb of the GWAS
index SNP(s) with a correlation coefficient estimate of
r 2 ≥ 0.3 in the European 1000 Genomes project popula-
MFHAS1

SLC30A9
Gene(s)

MSRA

tions were considered in LD.

XKR6
NSF

The statistical association of each index and LD SNP

partner with autism was determined according to the
Class

autism-GWAS data as analyzed by the PGC (Grove

II
I

et al., 2019). Variants achieving a significance value of

p ≤ 5  103 were retained for prioritization, and their
MAF

0.411

0.445

0.213

0.242

0.482

genic classification was determined using the Annotate

Abbreviations: A1/A2, autism-associated allele/non-associated allele; Class I, intronic/50 UTR TF binding SNPs; Class II, 30 UTR microRNA binding SNPs; MAF, minor allele frequency; NF, non-functional.

Variation (ANNOVAR) software (Wang et al., 2010).

A/G

T/G
C/G

T/C

T/C
A1/A2

Classification of pathogenicity

Two complementary machine learning algorithms were

GWAS OR

utilized to identify SNPs of functional potential: the

0.953

0.953

1.074

1.056

1.044

Combined Annotation Dependent Depletion (CADD,

mapped to a significant GWAS hit and were identified via MAGMA as a significantly associated gene/gene-set. GWAS and associated data derived from Grove et al. (2019).

Version GRCh37-v1.6; Kircher et al., 2014) and the

6.12  104

4.92  104

4.35  105
4

1.97  103

Genome-Wide Annotation of Variants (GWAVA,

8.68  10
GWAS p

Ensembl Variation Database Release 70; Ritchie

et al., 2014). Variants predicted to be pathogenic were
denoted by a CADD combined C score ≥10 and/or a
GWAVA Transcription Start Site score (TSS) ≥ 0.5. Var-
iants captured at this point formed the pool of prioritized
variants/SNPs.
Excluded from analysis as all ABN and 77/80 successfully genotyped ABBN samples possessed the dominant homozygote genotype.

Each SNPs’ expression quantitative trait loci (eQTLs)

Analysis

were manually curated according to the Genotype-Tissue

-

Expression (GTEx, Release 7) project (Lonsdale

et al., 2013). SNPs with predicted regulatory function
within the central nervous system (CNS) datasets were
LD (r 2)

0.451

0.706

0.833

0.417

0.302

retained for annotation.

Functional annotation of prioritized SNPs

Distance (bp)

120,872

90,188

34,551

148,515
15,695

To determine the likely biological consequences of the

prioritized variants, the SNPs were screened in silico for
predicted regulatory effects within three broad functional
classes: (I) 50 flanking and intronic variants affecting tran-
scription factor (TF) binding and/or transcription
Index SNP partner

enhancers/suppressors; (II) 30 untranslated region (UTR)

Excluded from analyses due to unsuccessful genotyping.
rs146122400

SNPs affecting messenger RNA (mRNA) stability/trans-

rs17701675

rs16854048

rs6992403a
rs7831557

lation; and (III) intronic variants likely to affect splicing.

Class I
8:8729761

8:10190040

17:44797919

4:42089177

8:10939273
(Continued)

To identify SNPs likely to influence TF binding, priori-

tized variants mapped to the 50 UTR/50 flanking or intron
GRCh37

of a protein-coding gene were assessed via RegulomeDB

(Version 2; Boyle et al., 2012). RegulomeDB incorpo-
rates information from a range of databases such as the
TABLE 1

rs11991118
rs7832708
rs907183

rs199456

ENCODE Project Consortium and NCBI Sequence

rs11051

Read Archive to predict regulatory function. Variants

SNP

classified as “likely to affect binding” (1a-2c) were then

b
a
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PUGSLEY ET AL. 5

annotated in HaploReg (Version 4; Ward & Kellis, 2016) of Autistic donors was confirmed by the ABN via consul-
to determine their specific impact on TF binding motifs. tation with the donor family and review of available med-
Further, the SNPs were manually annotated via Ensembl ical, auditory/speech-language, and neuropsychological
(Genome Reference Consortium Human Build records inclusive of the Autism Diagnostic Interview—
37 [GRCh37]; Yates et al., 2020) and the University of Revised (Kim & Lord, 2012; Lord et al., 1994) and/or
California–Santa Cruz (UCSC) Genome Browser (Kent other equivalent assessments. Autistic and non-Autistic
et al., 2002) to establish whether they overlapped signifi- donors were predominantly male (Autistic donors:
cant regulatory regions. Specifically, UCSC database n = 28, 87.5%; non-Autistic donors: n = 22, 70.9%), and
tracks EPDnew human version 006 (Dreos et al., 2013), were an average of 32.1 years and 31.4 years of age at
Open Regulatory Annotation (Lesurf et al., 2016), and time of tissue acquisition, respectively. Individuals with
GeneHancer (Fishilevich et al., 2017) were examined. known genetic syndromes associated with secondary
autism presentation (e.g., Fragile X syndrome, Rett’s
Class II syndrome, Neurofibromatosis) were excluded from the
SNPs at the 30 UTR can modulate post-transcriptional current sample. Additional non-Autistic IFG samples
and/or translational processes of gene expression via dis- were acquired from 87 Caucasian donors from the
ruption of microRNA-based targeting. As such, priori- Australian Brain Bank Network (ABBN). The sample
tized variants mapped to the 30 UTR were analyzed for consisted of 64 male donors (73.6%), with an overall
their predicted consequences on microRNA binding sites mean donor age of 51.59 years. Ethical approval for the
using two independent measures: TargetScan Context+ use of human post-mortem tissue was obtained from
scores derived from the polymorphism in microRNAs the Monash University Human Research Ethics Commit-
and their Target Sites (polymiRTS) database (Agarwal tee (Project ID: 25786).
et al., 2015) and miRanda-mirSVR scores (Bino
et al., 2004). SNPs scoring ≤ 0.5 on either measure were
considered functional. Nucleotide extraction and SNP genotyping
Class III RNA and DNA were extracted using TRIzol
SNP variation in splice junctions represents an important (Invitrogen), and RNA was purified using the RNase-
determinant of gene expression. To assess for predicted Free DNase Set and RNAeasy Mini Kit (Qiagen) accord-
effects on gene splicing, prioritized variants mapped to ing to manufacturer instruction. DNA samples were
intronic regions of protein-coding genes were evaluated commercially genotyped via a custom Agena Bioscience
using the Ensembl Variant Effect Predictor web-tool MassARRAY on a MA4 Mass Spectrometer at the
(VEP, https://ensembl.org/Tools/VEP; McLaren Australian Genome Research Facility (Queensland,
et al., 2016). SNPs were analyzed for overlap with canon- Australia). Four SNPs did not meet quality control met-
ical splicing regions and the creation of de novo acceptor rics for genotyping analysis (chr17:43965129, rs6992403,
or donor splice sites via the SpliceAI (Jaganathan rs111972148, rs142920272), necessitating application of
et al., 2019) and Maximum Entropy Scan (Yeo & alternate genotyping methods. rs111972148 genotypes
Burge, 2004) VEP plugins. were determined using a TaqMan probe assay (Applied
Biosystems) performed on a LightCycler-480 thermocy-
cler (Roche). Genotyping of rs142920272 was determined
Gene mapping using a restriction fragment length polymorphism
approach. The genomic region flanking rs142920272 was
Positional mapping was performed according to ANNO- amplified (forward primer sequence: 50 -GATCCTCGTT-
VAR annotations. These describe the genic context of a CAAGTCCTG-30 ; reverse primer sequence: 50 -
given SNP in relation to known coding and non-coding GTGAGCGGTAAGGAAAGTG-30 ) and resulting
genes. amplicons restricted with FauI (New England Biolabs) at
55 C for 2 h. Three fragments (58, 159, and 244 bp) were
fractionated on a 3% agarose gel, permitting delineation
Validation of SNP variant regulatory function of the allelic variants. Suitable experimental genotyping
alternatives were not available for either of
Post-mortem tissue acquisition and chr17:43965129 or rs6992403 due to high false primer
characteristics potential and high local similarity, respectively. These
variants were excluded from analyses.
Evidence supports the role of the inferior frontal gyrus
(IFG) as a region of interest in autism neuropathology
(Ikeda et al., 2018; Li et al., 2020; Peng et al., 2020). Gene expression quantification
Post-mortem IFG samples from 32 donors with and
31 donors without a diagnosis of autism were obtained Complementary DNA (cDNA) was prepared using the
from the Autism BrainNet (ABN). The clincial diagnosis High Capacity cDNA Reverse Transcription Kit
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
6 PUGSLEY ET AL.

(Applied Biosystems). qPCR experiments were per- unequal distribution of the autism-associated (A1) and
formed as recommended by the manufacturer on a non-associated (A2) alleles was observed. In these
LightCycler-480 thermocycler using SYBR Green I Mas- instances, a recessive (r, A1A1 + A1A2 vs. A2A2) or
ter Mix (Roche). For data normalization purposes, dominant model (D, A2A2 + A1A2 vs. A1A1) was used
β-2-Macroglobulin (B2M) and β-actin (ACTB) genes in lieu of an additive model to improve the sample sizes
were included as internal references in each reaction. for subsequent statistical analyses. Model comparisons
qPCR primers were designed using the freely available were conducted with the “anova” function of the R
RealTime qPCR Assay design tool (IDT, https://sg. “stats” package (R Core Team, 2020). To account for
idtdna.com/scitools/Applications/RealTimePCR/; multiple comparisons, the significance threshold for dis-
Table S1). Each sample was plated, amplified, and ana- covery was adjusted using the spectral decomposition
lyzed in triplicate. Gene expression was measured accord- method (matSpD, https://research.qut.edu.au/sgel/
ing to the samples’ cycles threshold (Ct) value as software/matspd-local-version/; Nyholt, 2004).
calculated via relative quantification (Wong &
Medrano, 2005).
The authors acknowledge that qPCR cannot be used RESULTS
to assess the expression of alternate gene isoforms associ-
ated with Class III SNP regulation. This method was Identification and functional annotation of non-
deemed suitable for the current study after in silico vari- coding autism-associated SNPs
ant annotation revealed there to be no Class III variants
likely to impact the splicing of any putative brain- Application of the threshold for allelic co-inheritance
expressed genes. identified 7706 SNPs in LD with one or more of the
23 autism-associated index variants. Of these, 86%
(n = 6647) were associated with autism, suggesting that
Statistical analysis application of a lowered LD threshold permits detection
of many SNPs likely to exhibit genuine correlation with
Data analyses were performed in RStudio (22.12.0 the target phenotype. Six-thousand five-hundred and
+ 353) using parametric tests or non-parametric equiva- eighty-nine of the associated variants (99.13%) were
lents when indicated by assumption violation. retained following non-coding genic classification,
The influence of increasing copies of the autism- 62.13% (n = 4094) of which were observed to be mapped
associated allele (i.e., 0 vs. 1 vs. 2 copies) on gene expres- to introns, 1.58% (n = 104) to the 30 UTR, and 0.17%
sion was evaluated in a series of multiple linear regression (n = 11) to the 50 UTR/flank of protein-coding genes. The
analyses (MLRAs) using the “lm” function of the R remaining SNPs were located in intergenic and intronic
“stats” package (R Core Team, 2020). To account for non-coding RNA sequences.
potential non-genotypic effects on expression, we first To determine the functional capabilities of these vari-
constructed a covariate model (Model 1) consisting of ants, their CADD C, scores and GWAVA TSS scores were
demographic characteristics (age, sex). In the second curated. In total, 583 unique SNPs exceeded C ≥ 10
model (Model 2), diagnostic status (Autistic vs. non- (n = 404) and/or TSS score ≥ 0.5 (n = 237). Fifty-eight
Autistic) was included as an additional predictor, allow- (9.94%) of these were classified as pathogenic by CADD
ing for group level differences in gene expression to be and GWAVA. As both tools’ models and annotations were
captured. Allelic dosage of the variant of interest was independently validated at time of publication (Kircher
subsequently added in the third model (Model 3), where et al., 2014; Ritchie et al., 2014), the predictive utility of
additional variance in gene expression would be CADD and GWAVA at the individual level were consid-
explained by SNP genotype. It is widely accepted that the ered appropriate for the prioritization of functional SNPs.
pathogenicity conferred by common variants represents Three-hundred and fifty-four of the functionally
the aggregate of many SNPs of low penetrance (Reich & classified variants (60.7%) were known eQTLs for
Lander, 2001). As such, it may be the case that synergism protein-coding genes within CNS tissue. These variants
between the elevated genetic burden among Autistic demonstrated regulatory functions across all neural tissue
donors and individual SNPs may uniquely augment gene types within the GTEx database, with several SNPs iden-
expression. To account for this, an interaction term for tified as influencing gene regulation across multiple dis-
diagnostic status and allelic dosage was included in the tinct cortical areas. Regions with known associations
final model (Model 4). with autism and autism-like behaviors such as the cere-
Assumption testing was performed prior to applica- bellum (n = 314, 88.7%) (e.g., D’Mello et al., 2015) and
tion of the regression analyses. Where normality was vio- frontal cortex (n = 299, 84.4%) (e.g., Dickstein
lated, appropriate transformations were performed using et al., 2013) were consistently represented among the
either the generic “log()” or “sqrt()” R functions (R Core eQTLs. This suggests that the prioritized SNPs may con-
Team, 2020), and all analyses were conducted on tribute to gene dysregulation in cortices underpinning
transformed data. For several variants of interest, key behavioral domains of the condition.
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PUGSLEY ET AL. 7

Functional variants show high localisation Dehydrogenase (TDH). TDH encodes a non-functional
within 50 flanking and intronic regions of brain truncated protein, the ancestor of which is predicted to
expressed genes have been involved in the conversion of L-threonine
to glycine (Edgar, 2002). The biological relevance of
In silico functional annotation identified 82 candidate TDH to autism is not apparent, particularly given the
variants meeting the pre-defined criteria for (I) TF bind- gene product lacks known functionality in human cells.
ing, (II) microRNA binding, and/or (III) splicing func-
tionality (Table S2). Most candidates were classified
within Class I (n = 76, 92.7%). Manual annotation of Experimental validation of variant regulatory
these variants revealed that 64.5% (n = 49) overlapped functionality
salient regulatory features such as promoters (n = 6) and
promoter flanking regions (n = 8), TF binding sites SNPs selected for qPCR validation are listed in
(n = 32), and enhancers (n = 22), supporting their patho- Table 1. SNPs mapped to protein-coding genes were
genic potential. considered for follow-up. Class I variants were decided
Notably, Ensembl and UCSC annotations converged by (1) RegulomeDB rank scores, and (2) consensus pre-
for rs1918788, a SNP mapped to a promoter overlapping dictions of function according to HaploReg/Ensembl/
intron 2 of the KAT8 Regulatory NSL Complex Subunit UCSC annotation. Additionally, SNPs overlapping
1 (KANSL1). Heterogeneous microdeletions of KANSL1 gene promoter regions were included for validation. All
are causative of Koolen de Vries syndrome: a multi- variants meeting miRanda-mirSVR and TargetScan
system disorder prominently characterized by mild- Context+ score cut-offs were retained for analysis. The
to-moderate intellectual disability (Koolen et al., 2016). donors’ genotype carrier status for the selected SNPs is
The gene expands over 198 kilobases, making it very presented in Table S3. Due to significant differences in
large and transcriptionally complex. Internal promoters demographic characteristics of the ABN and ABBN
likely provide greater endogenous control over KANSL1 samples (age, t(123) = 9.06, p < 0.001), MLRAs for
expression (Singer et al., 2008). As such, disruption of each were conducted separately. As all donors within
this promoter and, by proxy, KANSL1 dysregulation the ABBN sample were of the same diagnostic status
may contribute to the presentation of related neurodeve- (i.e., non-Autistic), MLRAs for these data consisted of
lopmental phenotypes including autism, making the covariate (Model 1) and allelic dosage (Model 3)
rs1918788 a strong candidate for functional validation. models only.
Analyses revealed that neither diagnosis nor variation
in allelic dosage of the bioinformatically prioritized vari-
Prioritized variants may influence processes of ants or their index LD partners accounted for discernible
translation regulation and isoform expression changes in the expression of their respective target
gene(s) in the ABN tissue (see Table 2). A trend to signifi-
Interrogation of 30 UTR variants led to the identification cance was observed for the model of the SLC30A9-
of five SNPs (6.6%) likely to influence mRNA translation mapped prioritized variant rs11051 (Model 3, p = 0.006),
through microRNA binding (Class II). This included where changes in gene expression were not meaningfully
rs1052594, of which the autism-associated allele was pre- predicted by variation in donor demographics (Model
dicted to disrupt a conserved microRNA binding site 1, p = 0.085) or diagnosis alone (Model 2, p = 0.462).
(hsa-miR-6767-5p, hsa-miR-4781-5p, hsa-miR-486-3p, However, this did not survive adjustment for multiple
hsa-miR-5090, hsa-miR-6775-5p, hsa-miR-744-5p) in the comparisons (adjusted α = 0.005). Though less pro-
30 UTR of the Microtubule Associated Protein Tau nounced, this was mirrored for the rs11051 LD index
(MAPT) gene. Tau proteins exert influence on processes SNP genotype model (rs16854048, Model 3, p = 0.045),
such as neuronal migration (Sapir et al., 2012) and neur- where inclusion of allelic dosage as a predictor captured
ite elongation (Leugers & Lee, 2010) in the developing additional variance in SLC30A9 expression above and
cortex. In addition, partial and/or complete ablation of beyond that of age and sex (Model 1, p = 0.085)
MAPT has been shown to inhibit autism-associated neu- and diagnostic group (Model 2, p = 0.481). These trends
rological phenotypes (e.g., megalencephaly) and behav- were not observed in the ABBN analyses (see Table S4)
iors (e.g., repetitive/stereotyped movements, social despite comparable allele frequencies across samples for
preference) in murine models of the condition (Tai the prioritized (Fisher’s Exact Test, two-sided p = 0.107)
et al., 2020). Variability in rs1052594 genotype may and index (Fisher’s Exact Test, two-sided p = 0.353)
therefore contribute to autism pathogenesis via MAPT variants.
dysregulation. Inclusion of an interaction effect between diagnosis
Only one candidate SNP, rs2736277, was classified as and SNP genotype did not yield significant models for
likely to affect splicing (Class III). Both VEP and Max- any of the analyzed gene-variant pairs in the ABN data-
EntScan annotations predicted rs2736277 to correspond set. A trend was observed for those inclusive of a donor
to creation of a splice site in pseudogene L-Threonine group*allelic dosage predictor for both the index
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
8 PUGSLEY ET AL.

TABLE 2 Summary of MLRA results for qPCR gene expression results in ABN Autistic and non-Autistic samples.

Model and predictor(s) R2 Adj. R 2 ΔR 2 ΔF Pr(>ΔF) df

Model Log(ARL17B), rs2668627p

1 Age, sex 0.025 0.010 0.025 0.705 0.499 (2, 55)
2 Model 1 + group 0.057 0.005 0.032 1.988 0.165 (1, 54)
3 Model 2 + rs2668627 genotypeD 0.085 0.016 0.028 1.709 0.197 (1, 53)
4 Model 3 + group*rs2668627 genotypeD 0.157 0.076 0.072 4.426 0.040 (1, 52)
i
Model Log(ARL17B), rs142920272
1 Age, sex 0.032 0.000 0.032 0.999 0.375 (2, 60)
2 Model 1 + group 0.068 0.020 0.035 2.342 0.131 (1, 59)
3 Model 2 + rs142920272 genotypeD 0.078 0.014 0.010 0.685 0.411 (1, 58)
4 Model 3 + group*rs142920272 genotypeD 0.138 0.063 0.060 3.986 0.051 (1, 57)
Model Log(CRHR1), rs17763086p,a
1 Age, sex 0.018 0.016 0.018 0.518 0.598 (2, 58)
2 Model 1 + group 0.028 0.023 0.010 0.592 0.445 (1, 57)
3 Model 2 + rs17763086 genotypeD 0.036 0.033 0.008 0.458 0.502 (1, 56)
4 Model 3 + group*rs17763086 genotypeD 0.039 0.048 0.003 0.197 0.659 (1, 55)
Model Log(KANSL1), rs17574361p,b
1 Age, sex 0.003 0.031 0.003 0.080 0.923 (2, 60)
2 Model 1 + group 0.029 0.021 0.026 1.584 0.213 (1, 59)
3 Model 2 + rs17574361 genotypeD 0.032 0.034 0.004 0.216 0.644 (1, 58)
4 Model 3 + group*rs17574361 genotype D
0.057 0.026 0.025 1.484 0.228 (1, 58)
Model Log(KANSL1), rs142920272i
1 Age, sex 0.003 0.031 0.003 0.080 0.923 (2, 60)
2 Model 1 + group 0.029 0.021 0.026 1.569 0.216 (1, 59)
3 Model 2 + rs142920272 genotypeD 0.030 0.037 0.001 0.048 0.827 (1, 58)
4 Model 3 + group*rs142920272 genotype D
0.048 0.035 0.018 1.099 0.299 (1, 57)
Model Log(LINC02210-CRHR1), rs366858p,a
1 Age, sex 0.051 0.017 0.051 1.515 0.229 (2, 56)
2 Model 1 + group 0.171 0.126 0.120 7.682 0.008 (1, 55)
3 Model 2 + rs17763086 genotypeD 0.172 0.111 0.001 0.047 0.830 (1, 54)
4 Model 3 + group*rs17763086 genotype D
0.174 0.096 0.002 0.122 0.728 (1, 53)
Model MAPT, rs1052594p
1 Age, sex 0.137 0.108 0.137 4.754 0.012 (2, 60)
2 Model 1 + group 0.142 0.098 0.005 0.354 0.554 (1, 59)
3 Model 2 + rs1052594 genotyper 0.161 0.103 0.019 1.299 0.259 (1, 58)
4 Model 3 + group*rs1052594 genotyper 0.168 0.095 0.007 0.456 0.502 (1, 57)
i
Model MAPT, rs142920272
1 Age, sex 0.137 0.108 0.137 4.754 0.012 (2, 60)
2 Model 1 + group 0.142 0.098 0.005 0.348 0.558 (1, 59)
3 Model 2 + rs142920272 genotypeD 0.150 0.091 0.008 0.516 0.475 (1, 58)
4 Model 3 + group*rs142920272 genotypeD 0.153 0.079 0.004 0.237 0.628 (1, 57)
p
Model Log(MFHAS1), rs907183
1 Age, sex 0.002 0.031 0.002 0.073 0.930 (2, 59)
2 Model 1 + group 0.003 0.049 0.000 0.024 0.876 (1, 58)
3 Model 2 + rs907183 genotype 0.032 0.036 0.030 1.739 0.193 (1, 57)
4 Model 3 + group*rs907183 genotype 0.049 0.036 0.017 0.987 0.325 (1, 56)
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PUGSLEY ET AL. 9

TABLE 2 (Continued)

Model and predictor(s) R2 Adj. R 2 ΔR 2 ΔF Pr(>ΔF) df

Model Log(MFHAS1), rs17701675i

1 Age, sex 0.002 0.031 0.002 0.073 0.930 (2, 59)
2 Model 1 + group 0.003 0.049 0.000 0.024 0.879 (1, 58)
3 Model 2 + rs17701675 genotype 0.005 0.065 0.002 0.116 0.735 (1, 57)
4 Model 3 + group*rs17701675 genotype 0.012 0.076 0.007 0.423 0.518 (1, 56)
Model MSRA, rs7832708p
1 Age, sex 0.044 0.010 0.044 1.279 0.286 (2, 56)
2 Model 1 + group 0.044 0.008 0.000 0.005 0.947 (1, 55)
3 Model 2 + rs7832708 genotype 0.044 0.027 0.000 0.013 0.911 (1, 54)
4 Model 3 + group*rs7832708 genotype 0.044 0.046 0.000 0.010 0.923 (1, 53)
Model MSRA, rs7831557i
1 Age, sex 0.044 0.010 0.044 1.313 0.277 (2, 57)
2 Model 1 + group 0.044 0.007 0.000 0.003 0.957 (1, 56)
3 Model 2 + rs7831557 genotype 0.044 0.025 0.000 0.003 0.954 (1, 55)
4 Model 3 + group*rs7831557 genotype 0.044 0.044 0.000 0.001 0.975 (1, 54)
Model NSF, rs199456p
1 Age, sex 0.034 0.002 0.034 1.061 0.352 (2, 60)
2 Model 1 + group 0.049 0.001 0.015 0.890 0.350 (1, 59)
3 Model 2 + rs199456 genotyper 0.049 0.016 0.000 0.016 0.900 (1, 58)
4 Model 3 + group*rs199456 genotype r
0.051 0.032 0.002 0.129 0.721 (1, 57)
Model NSF, rs146122400i
1 Age, sex 0.052 0.019 0.052 1.570 0.217 (2, 57)
2 Model 1 + group 0.068 0.018 0.016 0.930 0.339 (1, 56)
3 Model 2 + rs146122400 genotypeD 0.071 0.004 0.004 0.219 0.642 (1, 55)
4 Model 3 + group*rs146122400 genotype D
0.099 0.016 0.028 1.674 0.201 (1, 54)
Model Sqrt(SLC30A9), rs11051p
1 Age, sex 0.080 0.049 0.080 2.565 0.085 (2, 59)
2 Model 1 + group 0.088 0.040 0.008 0.548 0.462 (1, 58)
3 Model 2 + rs11051 genotype 0.202 0.146 0.114 8.252 0.006 (1, 57)
4 Model 3 + group*rs11051 genotype 0.227 0.158 0.025 1.814 0.184 (1, 56)
Model Sqrt(SLC30A9), rs16854048i
1 Age, sex 0.080 0.049 0.080 2.565 0.085 (2, 59)
2 Model 1 + group 0.088 0.040 0.008 0.504 0.481 (1, 58)
3 Model 2 + rs16854048 genotypeD 0.151 0.091 0.063 4.194 0.045 (1, 57)
4 Model 3 + group*rs16854048 genotypeD 0.159 0.084 0.009 0.595 0.444 (1, 56)
p,a
Model XKR6, rs11991118
1 Age, sex 0.006 0.030 0.006 0.157 0.855 (2, 56)
2 Model 1 + group 0.008 0.046 0.002 0.123 0.728 (1, 55)
3 Model 2 + rs11991118 genotype 0.042 0.029 0.034 1.894 0.175 (1, 54)
4 Model 3 + group*rs11991118 genotype 0.050 0.040 0.008 0.450 0.505 (1, 53)
Abbreviations: Adj. R , adjusted R ; D, dominant model (A2A2 + A1A2 vs. A1A1) applied due to unequal distribution of the autism-associated/ non-associated alleles;
2 2

df, degrees of freedom; i, index variant; Log, analysis performed on natural logarithmic transformed qPCR data; p, prioritized variant; Pr(>F), significance value of the
ΔF; r, recessive model (A1A1 + A1A2 vs. A2A2) applied due to unequal distribution of the autism-associated/ non-associated alleles; Sqrt, analysis performed on square-
root transformed qPCR data; ΔF, change in F; ΔR 2, change in R 2.
a
Index/prioritized LD SNP excluded from analysis due to unsuccessful genotyping.
b
KANSL1 associated variants rs17574361 and rs1918788 and LINC02210-CRHR1 associated variants rs17763086 and rs366858 demonstrated perfect co-inheritance in
the current sample. As such, analyses for these SNP-genes pairs were only conducted for rs17574361 and rs366858 KANSL1 and LINC02210-CRHR1 SNPs only.
*Adjusted p < 0.005.
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
10 PUGSLEY ET AL.

(rs14292027, Model 4, p = 0.040) and prioritized combination of predictive and experimental data. These
(rs2668627, Model 4, p = 0.051) SNPs mapped to collectively afforded strong in silico evidence of genuine
ARL17B, though these did not surpass the adjusted regulatory functionality and, by proxy, a potential
threshold for statistical significance. influence on autism etiology. Notably, none of the
autism-GWAS SNPs identified by Grove et al. (2019)
were predicted to possess regulatory functions by any of
DISCUSSION the applied bioinformatic resources. This aligns with the
supposition that—rather than themselves representing
Common allelic variation predominates the genetic liabil- markers of risk—GWAS variants may simply tag co-
ity of autism (Gaugler et al., 2014). However, the inabil- inherited loci of pathogenicity. Such results exemplify the
ity to distinguish causal from co-inherited SNPs limits need to interrogate LD partners of genome-wide signifi-
the mechanistic insights derived from GWAS findings. cant SNPs for functional potential, substantiating our
Translation of these statistical signals into etiological examination of these within our prioritization process.
knowledge is essential not only to enhance the genetic lit- Eight of the 12 experimentally interrogated variants
eracy of the condition, but also spearhead development (75%) were mapped to either or both index SNPs
of novel therapeutics. Overcoming this barrier is compli- rs142920272 and chr17:43965129, situating them within
cated, as blind spots regarding the role of non-coding a
1 Mb region of cytoband 17q21.31. Many of the
variants in autism pathogenesis hinders biological inter- selected SNPs shared LD estimates exceeding the thresh-
pretation of their genic effects. Here, we sought to sys- old for co-inheritance (r 2 ≥ 0.88), suggesting that this
tematically characterize the likely influence of the region is a hotspot for genetic variability relevant to
recently identified autism-GWAS SNPs and/or their non- autism. Comprising several genes of known CNS func-
coding LD partners on gene expression in human brain tioning, copy number variations to 17q21.31 frequently
tissue. Several SNPs of pathogenic potential were identi- coincide deviations from typical neurodevelopmental tra-
fied through in silico prioritization. Although a small jectories. For example, aforementioned microdeletions in
number showed nominally significant or trending the region encompassing KANSL1 have been causally
SNP-gene relations (e.g., rs11051-SLC30A9), these did linked to presentation of Koolen de Vries syndrome
not survive correction for multiple testing. In line with (Koolen et al., 2016), resulting in a diverse array of
the additive polygenic nature of autism, we investigated motor, intellectual, and congenital symptoms of varying
whether a genotype-by-diagnosis effect may lead to dif- clinical severity. In addition to these, a consistent attri-
ferential impacts of allelic dosage on gene expression out- bute of the condition is elevated amiability, with individ-
comes. However, this did not yield significant results. uals exhibiting stereotypically friendly social behaviors.
Of the examined variants, those proximally mapped This distinctly contrasts with 17q21.31 microduplication
to SLC30A9 (Solute Carrier Family 30 Member 9) dem- syndrome, wherein developmental and intellectual fea-
onstrated the most promising evidence of gene expression tures persist, however interpersonal tendencies are more
regulation functionality. SLC30A9 is predicted to encode akin to those traditionally associated with autism
a nuclear receptor co-activator important in zinc ion (e.g., reduced social reciprocity) (Gregor et al., 2012;
homeostasis and transport. The gene’s protein product Grisart et al., 2009; Kitsiou-Tzeli et al., 2012; Martorell
has been shown to be ubiquitously expressed in the devel- et al., 2022; Mc Cormack et al., 2014; Natacci
oping human cortex (Yang & Shcheglovitov, 2020). As et al., 2016). Carriers of clinically salient 17q21.31 micro-
such, SLC30A9 likely impacts an array of neurodevelop- duplications additionally report concurrent psychomotor
mental processes which, if disrupted, may influence and language delays/difficulties (Gregor et al., 2012;
autism onset. Loss-of-function mutations in SLC30A9 Grisart et al., 2009; Martorell et al., 2022; Mc Cormack
have been associated with a rare recessive cerebro-renal et al., 2014). This closely mirrors the secondary chal-
syndrome featuring intellectual disability, motor difficul- lenges experienced by many members of the Autistic
ties, and oculomotor disturbances (Kleyner et al., 2022; community (Lord et al., 2020), suggestive of a shared bio-
Perez et al., 2017), each of which are known to co-occur logical antecedent for the two presentations.
with autism (Bhat, 2020; Etyemez et al., 2022; Johnson Our inability to experimentally establish functionality
et al., 2016). Although the leading prioritized SNP of or the prioritized SNPs is surprising. Approximately
(i.e., rs11051) and the associated index variant 40% of genome-wide significant polymorphisms localized
(i.e., rs16854048) did not surpass the corrected signifi- to non-coding genic regulators exhibit some effect over
cance level, this phenotypic overlap suggests SLC30A9 to their proximal protein coding genes (Fadason et al.,
be of clinical relevance to autism. As such, it should be 2018). Further, our previous work (Dark et al., 2020;
considered a gene of interest for future investigation. Hawi et al., 2013) and similar experimental approaches
Our prioritization pipeline integrated several estab- by others (Law et al., 2007; Zhang et al., 2019) have dem-
lished bioinformatic resources designed to annotate our onstrated that the regulatory effects of high-confidence
trait-associated variants for experimental validation. The non-coding variants can be effectively captured using the
shortlisted SNPs were selected on the grounds of a methods applied for the current study. It may be the case
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PUGSLEY ET AL. 11

that our sample sizes for the qPCR analyses were under- time, there have been changes in the top-ranking GWAS
powered to detect the subtle influence of the selected vari- SNPs from iteration to iteration (e.g., attention-deficit/
ants on gene expression. This is difficult to overcome hyperactivity disorder; Demontis et al., 2023, 2019). This
given the paucity of available post-mortem brain tissue suggests that the PGC autism-GWAS index SNPs may
from Autistic individuals. Additionally, both the ABN not necessarily reflect the most etiologically relevant
and ABBN tissue collections have been aggregated across common genetic variants to the autism phenotype, hence
multiple time points and research sites, meaning that our prioritization efforts may have failed to capture LD
inter- and intra-sample variations in tissue handling are SNPs of clinical importance.
likely to impact the preservation and integrity of RNA Given the many challenges associated with variant
post-mortem. This may have contributed to the non- modeling, as well as the inherent difficulties in identifying
significant qPCR results observed at the group level, as SNPs of functional relevance as reported here, focussing
nucleotide degradation could theoretically mask potential on the role of individual polymorphisms on genetic liabil-
differences in gene expression attributable to diagnostic ity may be inefficient in bridging the gap between candi-
status. date discovery and biological annotation. This is not to
Although secondary to the SNP-gene relationships say that understanding the influence of these variants is
central to the current study, it is worth elaborating on the unimportant, but rather that existing barriers to effec-
absence of group-by-group differences in gene expression tively denoting their impact may hinder timely transla-
between Autistic and non-Autistic samples of the ABN. tion of GWAS findings into community health outcomes.
The effects of etiological factors contributory to autism It is well understood that trait-associated SNPs do not
onset are most pronounced when exposure occurs during act individually but rather in aggregate to augment phe-
prenatal and early post-natal periods (Pugsley notypic likelihood (Reich & Lander, 2001). As such, an
et al., 2022). From a genetics perspective, genes of known approach that considers the cumulative effects of multi-
or suspected pathophysiological relevance to autism are ple allelic variants may provide greater insight into the
focally expressed in cell types present in early phases of genes and biological pathways key to the pathophysiol-
cortical development (e.g., neural progenitor cells, imma- ogy of autism. Within our own dataset, all but one of the
ture neuron phenotypes) (Yang & Shcheglovitov, 2020). prioritized variants shortlisted in Table 1 and Table S2
Although gene function may be sustained across the life- were in LD with a leading candidate from the MAGMA
span, it is reasonable to conclude that dysregulation of as opposed to the GWAS and MTAG analyses. This sug-
these during this critical phase would profoundly influ- gests that a multi-variant approach may be more fruitful
ence spatiotemporally regulated processes of neurogen- in detecting loci of genuine pathogenicity than the current
esis. Failure to include samples representative of this methods.
period may have precluded detection of differences in
putative gene expression between the sample groups, cul-
minating into the present null findings. This is an impor- CONCLUSION
tant caveat in the utilization of adult post-mortem tissues
when studying the biological underpinnings of early-onset Although functional annotation of autism-associated
neurodevelopmental conditions. To circumvent this, the SNPs was possible using the described in silico resources,
generation and application of novel experimental models failure to experimentally corroborate these predictions
should be considered. For example, patient-derived suggests our method to be an imperfect solution to sup-
induced pluripotent stem cells afford an alternate means port the transition of GWAS signals to clinical applica-
to examine research questions relating to the genetics of tion. This may be attributed, in part, to the small sample
neurodevelopment (Yang & Shcheglovitov, 2020). These of donor tissue used to derive the current research find-
tools and their derivatives assist to overcome limitations ings. However, ours represents one of largest clinical
of paucity and inaccessibility of living tissue, and are samples of post-mortem brain tissue applied to facilitate
capable of recapitulating processes of neurogenesis pres- an investigation of this nature, indicative of the need to
ently absent in donor sample collections. collate greater resources to support experimental efforts.
It is important to note that existent limitations in the Achieving the current aims amid the rapid variant
available autism GWAS data pose additional challenges identification afforded by GWAS and other discovery
to the current aims. Although a significant development methodologies therefore requires significant research
in the autism genetics space, the genome-wide significant investment. Most immediately, validation of the existent
markers identified by Grove et al. (2019) have not been autism-GWAS SNPs is needed to ensure that all subse-
consistently replicated (e.g., Matoba et al., 2020). This is quent analyses are founded on the most etiologically
essential to verify the reliability of the primary GWAS salient candidates for the condition. As evidenced by the
findings (Kraft et al., 2009), ensuring that the present results, it is essential that prioritization efforts
genotype-to-phenotype correlations are accurate for the additionally attend to the regulatory potential of SNPs
target disease/disorder. As sample sizes and representa- beyond those surpassing genome-wide significance.
tion of clinical cases have increased in consortia data over Expansion of existent collections of donor tissue would
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
12 PUGSLEY ET AL.

greatly assist such experimental endeavors. However, this Alex, A. M., Saradalekshmi, K. R., Shilen, N., Suresh, P. A., &
comes with myriad challenges associated with acquisition Banerjee, M. (2019). Genetic association of DNMT variants can
play a critical role in defining the methylation patterns in autism.
and handling, and is unlikely to circumvent practical IUBMB Life, 71(7), 901–907.
issues relating to the use of adult tissue in studies of child- American Psychiatric Association. (2022). Diagnostic and statistical
hood conditions. It may be the case that the application manual of mental disorders (5th ed., text rev.). American Psychiat-
of alternative models should be implemented, where ric Association Publishing. https://doi.org/10.1176/appi.books.
available, to facilitate future experimental works and bet- 9780890425787
Bhat, A. N. (2020). Is motor impairment in autism spectrum disorder
ter contextualize research outcomes to relevant periods of distinct from developmental coordination disorder? A report from
neurodevelopment. Finally, interrogation of the con- the SPARK study. Physical Therapy, 100(4), 633–644. https://doi.
certed influence of multiple variants as opposed to single org/10.1093/ptj/pzz190
polymorphisms on the condition’s pathophysiology Bino, J., Enright, A. J., Aravin, A., Tuschl, T., Sander, C., &
should be considered, as this may permit more timely Marks, D. S. (2004). Human microRNA targets. PLoS Biology,
2(11), e363. https://doi.org/10.1371/journal.pbio.0020363
translation of statistical association to molecular Bowden, N., Thabrew, H., Kokaua, J., & Braund, R. (2020). National
discovery. prescribing rates and polypharmacy for children and young people
in New Zealand with and without autism spectrum disorder.
A CK NO W L E D G M E N T S Research in Autism Spectrum Disorders, 78, 101642. https://doi.
Autistic and non-Autistic donor tissue samples and data org/10.1016/j.rasd.2020.101642
Boyle, A. P., Hong, E. L., Hariharan, M., Cheng, Y., Schaub, M. A.,
were obtained from Autism BrainNet, a resource of the Kasowski, M., Karczewski, K. J., Park, J., Hitz, B. C., Weng, S.,
Simons Foundation Autism Research Initiative (SFARI). Cherry, J. M., & Snyder, M. (2012). Annotation of functional vari-
Autism BrainNet also manages the Autism Tissue Pro- ation in personal genomes using RegulomeDB. Genome Research,
gram (ATP) collection, previously funded by Autism 22(9), 1790–1797. https://doi.org/10.1101/gr.137323.112
Speaks. We are grateful and indebted to the families who Castelbaum, L., Sylvester, C. M., Zhang, Y., Yu, Q., &
Constantino, J. N. (2019). On the nature of monozygotic twin con-
donated tissue for research purposes to Autism BrainNet cordance and discordance for autistic trait severity: A quantitative
and the ATP. Australian Brain Bank tissue samples were analysis. Behavior Genetics, 50, 263–272. https://doi.org/10.1007/
received from the Sydney Brain Bank which is supported s10519-019-09987-2
by Neuroscience Research Australia. The authors would Dark, C., Williams, C., Bellgrove, M. A., Hawi, Z., & Bryson-
like to acknowledge the support of participating families. Richardson, R. J. (2020). Functional validation of CHMP7 as an
ADHD risk gene. Translational Psychiatry, 10(1), 385. https://doi.
KP and AN are supported by an Australian Government org/10.1038/s41398-020-01077-w
Research Training Program (RTP) Scholarship. MAB is Demontis, D., Walters, G. B., Athanasiadis, G., Walters, R.,
supported by a Senior Research Fellowship from the Therrien, K., Nielsen, T. T., Farajzadeh, L., Voloudakis, G.,
National Health and Medical Research Council Bendl, J., Zeng, B., Zhang, W., Grove, J., Als, T. D., Duan, J.,
(NHMRC) of Australia. ZH is supported by a research Satterstrom, F. K., Bybjerg-Grauholm, J., Bækved-Hansen, M.,
Gudmundsson, O. O., Magnusson, S. H., … iPSYCH-Broad Con-
grant funded by the National Health and Medical sortium. (2023). Genome-wide analyses of ADHD identify 27 risk
Research Council of Australia (APP1146644 & loci, refine the genetic architecture and implicate several cognitive
APP1002458). Open access publishing facilitated by domains. Nature Genetics, 55(2), 198–208. https://doi.org/10.1038/
Monash University, as part of the Wiley - Monash Uni- s41588-022-01285-8
versity agreement via the Council of Australian Univer- Demontis, D., Walters, R. K., Martin, J., Mattheisen, M., Als, T. D.,
Agerbo, E., Baldursson, G., Belliveau, R., Bybjerg-Grauholm, J.,
sity Librarians. Bækvad-Hansen, M., Cerrato, F., Chambert, K.,
Churchhouse, C., Dumont, A., Eriksson, N., Gandal, M.,
D A T A A V A I L AB I L I T Y S T A T E M E N T Goldstein, J. I., Grasby, K. L., Grove, J., … Neale, B. M. (2019).
The data that support the findings of this study are avail- Discovery of the first genome-wide significant risk loci for atten-
able on request from the corresponding author. The data tion deficit/hyperactivity disorder. Nature Genetics, 51(1), 63–75.
https://doi.org/10.1038/s41588-018-0269-7
are not publicly available due to privacy or ethical Dickstein, D. P., Pescosolido, M. F., Reidy, B. L., Galvan, T.,
restrictions. Kim, K. L., Seymour, K. E., Laird, A. R., Di Martino, A., &
Barrett, R. P. (2013). Developmental meta-analysis of the func-
ORCID tional neural correlates of autism spectrum disorders. Journal of
Kealan Pugsley https://orcid.org/0000-0001-7507-9532 the American Academy of Child & Adolescent Psychiatry, 52(3),
279–289.e216. https://doi.org/10.1016/j.jaac.2012.12.012
Atefeh Namipashaki https://orcid.org/0000-0001-9350- D’Mello, A. M., Crocetti, D., Mostofsky, S. H., & Stoodley, C. J.
6155 (2015). Cerebellar gray matter and lobular volumes correlate with
Mark A. Bellgrove https://orcid.org/0000-0003-0186- core autism symptoms. NeuroImage: Clinical, 7, 631–639. https://
8349 doi.org/10.1016/j.nicl.2015.02.007
Ziarih Hawi https://orcid.org/0000-0002-0814-9112 Dreos, R., Ambrosini, G., Cavin Périer, R., & Bucher, P. (2013). EPD
and EPDnew, high-quality promoter resources in the next-
generation sequencing era. Nucleic Acids Research, 41(D1), D157–
R E FE RE NC E S D164. https://doi.org/10.1093/nar/gks1233
Agarwal, V., Bell, G. W., Nam, J.-W., & Bartel, D. P. (2015). Predict- Edgar, A. J. (2002). The human L-threonine 3-dehydrogenase gene is an
ing effective microRNA target sites in mammalian mRNAs. eLife, expressed pseudogene. BMC Genetics, 3(1), 18. https://doi.org/10.
4, e05005. https://doi.org/10.7554/eLife.05005 1186/1471-2156-3-18
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PUGSLEY ET AL. 13

Etyemez, S., Esler, A., Kini, A., Tsai, P.-C., DiRienzo, M., Biobehavioral Reviews, 69, 260–279. https://doi.org/10.1016/j.
Maenner, M., & Lee, L.-C. (2022). The role of intellectual disabil- neubiorev.2016.08.007
ity with autism spectrum disorder and the documented cooccurring Kent, W. J., Sugnet, C. W., Furey, T. S., Roskin, K. M., Pringle, T. H.,
conditions: A population-based study. Autism Research, 15(12), Zahler, A. M., & Haussler, D. (2002). The human genome browser
2399–2408. at UCSC. Genome Research, 12(6), 996–1006. https://doi.org/10.
Fadason, T., Schierding, W., Lumley, T., & O’Sullivan, J. M. (2018). 1101/gr.229102
Chromatin interactions and expression quantitative trait loci Kim, S. H., & Lord, C. (2012). New autism diagnostic interview-revised
reveal genetic drivers of multimorbidities. Nature Communications, algorithms for toddlers and young preschoolers from 12 to
9(1), 5198. https://doi.org/10.1038/s41467-018-07692-y 47 months of age. Journal of Autism and Developmental Disorders,
Fishilevich, S., Nudel, R., Rappaport, N., Hadar, R., Plaschkes, I., Iny 42, 82–93. https://doi.org/10.1007/s10803-011-1213-1
Stein, T., Rosen, N., Kohn, A., Twik, M., Safran, M., Kircher, M., Witten, D. M., Jain, P., O’Roak, B. J., Cooper, G. M., &
Lancet, D., & Cohen, D. (2017). GeneHancer: Genome-wide inte- Shendure, J. (2014). A general framework for estimating the rela-
gration of enhancers and target genes in GeneCards. Database: tive pathogenicity of human genetic variants. Nature Genetics,
The Journal of Biological Databases and Curation, 2017, bax028. 46(3), 310–315. https://doi.org/10.1038/ng.2892
https://doi.org/10.1093/database/bax028 Kitsiou-Tzeli, S., Frysira, H., Giannikou, K., Syrmou, A., Kosma, K.,
Gaugler, T., Klei, L., Sanders, S. J., Bodea, C. A., Goldberg, A. P., Kakourou, G., Leze, E., Sofocleous, C., Kanavakis, E., &
Lee, A. B., Mahajan, M., Manaa, D., Pawitan, Y., Reichert, J., Tzetis, M. (2012). Microdeletion and microduplication 17q21.31
Ripke, S., Sandin, S., Sklar, P., Svantesson, O., Reichenberg, A., plus an additional CNV, in patients with intellectual disability,
Hultman, C. M., Devlin, B., Roeder, K., & Buxbaum, J. D. identified by array-CGH. Gene, 492(1), 319–324. https://doi.org/
(2014). Most genetic risk for autism resides with common varia- 10.1016/j.gene.2011.10.023
tion. Nature Genetics, 46(8), 881–885. https://doi.org/10.1038/ng. Kleyner, R., Arif, M., Marchi, E., Horowitz, N., Haworth, A.,
3039 King, B., Gavin, M., Amble, K., Velinov, M., & Lyon, G. J.
Golovina, E., Fadason, T., Lints, T. J., Walker, C., Vickers, M. H., & (2022). Autosomal recessive SLC30A9 variants in a proband with
O’Sullivan, J. M. (2021). Understanding the impact of SNPs asso- a cerebrorenal syndrome and no parental consanguinity. Cold
ciated with autism spectrum disorder on biological pathways in Spring Harbor Molecular Case Studies, 8(2), a006137. https://doi.
the human fetal and adult cortex. Scientific Reports, 11(1), 15867. org/10.1101/mcs.a006137
https://doi.org/10.1038/s41598-021-95447-z Koolen, D. A., Pfundt, R., Linda, K., Beunders, G., Veenstra-
Gregor, A., Krumbiegel, M., Kraus, C., Reis, A., & Zweier, C. (2012). Knol, H. E., Conta, J. H., Fortuna, A. M., Gillessen-
De novo triplication of the MAPT gene from the recurrent Kaesbach, G., Dugan, S., Halbach, S., Abdul-Rahman, O. A.,
17q21.31 microdeletion region in a patient with moderate intellec- Winesett, H. M., Chung, W. K., Dalton, M., Dimova, P. S.,
tual disability and various minor anomalies. American Journal of Mattina, T., Prescott, K., Zhang, H. Z., Saal, H. M., … de
Medical Genetics Part A, 158A(7), 1765–1770. https://doi.org/10. Vries, B. B. A. (2016). The Koolen-de Vries syndrome: A pheno-
1002/ajmg.a.35427 typic comparison of patients with a 17q21.31 microdeletion versus
Grisart, B., Willatt, L., Destrée, A., Fryns, J. P., Rack, K., de a KANSL1 sequence variant. European Journal of Human Genet-
Ravel, T., Rosenfeld, J., Vermeesch, J. R., Verellen- ics, 24(5), 652–659. https://doi.org/10.1038/ejhg.2015.178
Dumoulin, C., & Sandford, R. (2009). 17q21.31 microduplication Kraft, P., Zeggini, E., & Ioannidis, J. P. (2009). Replication in genome-
patients are characterised by behavioural problems and poor social wide association studies. Statistical Science, 24(4), 561–573.
interaction. Journal of Medical Genetics, 46(8), 524–530. https:// https://doi.org/10.1214/09-sts290
doi.org/10.1136/jmg.2008.065367 Law, A. J., Kleinman, J. E., Weinberger, D. R., & Weickert, C. S.
Grove, J., Ripke, S., Als, T. D., Mattheisen, M., Walters, R. K., (2007). Disease-associated intronic variants in the ErbB4 gene are
Won, H., Pallesen, J., Agerbo, E., Andreassen, O. A., Anney, R., related to altered ErbB4 splice-variant expression in the brain in
Awashti, S., Belliveau, R., Bettella, F., Buxbaum, J. D., Bybjerg- schizophrenia. Human Molecular Genetics, 16(2), 129–141. https://
Grauholm, J., Bækvad-Hansen, M., Cerrato, F., Chambert, K., doi.org/10.1093/hmg/ddl449
Christensen, J. H., … Børglum, A. D. (2019). Identification of com- Lesurf, R., Cotto, K. C., Wang, G., Griffith, M., Kasaian, K.,
mon genetic risk variants for autism spectrum disorder. Nature Jones, S. J. M., Montgomery, S. B., Griffith, O. L., & Open Regu-
Genetics, 51(3), 431–444. https://doi.org/10.1038/s41588-019-0344-8 latory Annotation Consortium. (2016). ORegAnno 3.0: A
Hawi, Z., Matthews, N., Wagner, J., Wallace, R. H., Butler, T. J., community-driven resource for curated regulatory annotation.
Vance, A., Kent, L., Gill, M., & Bellgrove, M. A. (2013). DNA Nucleic Acids Research, 44(D1), D126–D132. https://doi.org/10.
variation in the SNAP25 gene confers risk to ADHD and is associ- 1093/nar/gkv1203
ated with reduced expression in prefrontal cortex. PLoS One, 8(4), Leugers, C. J., & Lee, G. (2010). Tau potentiates nerve growth factor-
e60274. https://doi.org/10.1371/journal.pone.0060274 induced mitogen-activated protein kinase signaling and neurite ini-
Ikeda, T., Tokuda, T., Monden, Y., Hirai, M., Mizushima, S. G., tiation without a requirement for microtubule binding. Journal of
Nagashima, M., Kyutoku, Y., Taniguchi, T., Shimoizumi, H., Biological Chemistry, 285(25), 19125–19134. https://doi.org/10.
Dan, I., & Yamagata, T. (2018). Hypoactivation of the right pre- 1074/jbc.M110.105387
frontal cortex underlying motor-related inhibitory deficits in chil- Li, Y., Zhu, Y., Nguchu, B. A., Wang, Y., Wang, H., Qiu, B., &
dren with autism spectrum disorder: A functional near-infrared Wang, X. (2020). Dynamic functional connectivity reveals abnor-
spectroscopy study. Japanese Psychological Research, 60(4), 251– mal variability and hyper-connected pattern in autism spectrum
264. https://doi.org/10.1111/jpr.12204 disorder. Autism Research, 13(2), 230–243. https://doi.org/10.1002/
Jaganathan, K., Kyriazopoulou Panagiotopoulou, S., McRae, J. F., aur.2212
Darbandi, S. F., Knowles, D., Li, Y. I., Kosmicki, J. A., Lonsdale, J., Thomas, J., Salvatore, M., Phillips, R., Lo, E., Shad, S.,
Arbelaez, J., Cui, W., Schwartz, G. B., Chow, E. D., Hasz, R., Walters, G., Garcia, F., Young, N., Foster, B.,
Kanterakis, E., Gao, H., Kia, A., Batzoglou, S., Sanders, S. J., & Moser, M., Karasik, E., Gillard, B., Ramsey, K., Sullivan, S.,
Farh, K. K.-H. (2019). Predicting splicing from primary sequence Bridge, J., Magazine, H., Syron, J., … Moore, H. F. (2013). The
with deep learning. Cell, 176(3), 535–548.e524. https://doi.org/10. genotype-tissue expression (GTEx) project. Nature Genetics, 45(6),
1016/j.cell.2018.12.015 580–585. https://doi.org/10.1038/ng.2653
Johnson, B. P., Lum, J. A., Rinehart, N. J., & Fielding, J. (2016). Ocu- Lord, C., Brugha, T. S., Charman, T., Cusack, J., Dumas, G.,
lar motor disturbances in autism spectrum disorders: Systematic Frazier, T., Jones, E. J. H., Jones, R. M., Pickles, A.,
review and comprehensive meta-analysis. Neuroscience & State, M. W., Taylor, J. L., & Veenstra-VanderWeele, J. (2020).
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
14 PUGSLEY ET AL.

Autism spectrum disorder. Nature Reviews Disease Primers, 6(1), Sapir, T., Frotscher, M., Levy, T., Mandelkow, E.-M., & Reiner, O.
5. https://doi.org/10.1038/s41572-019-0138-4 (2012). Tau’s role in the developing brain: Implications for intellec-
Lord, C., Rutter, M., & Le Couteur, A. (1994). Autism diagnostic tual disability. Human Molecular Genetics, 21(8), 1681–1692.
interview-revised: A revised version of a diagnostic interview for https://doi.org/10.1093/hmg/ddr603
caregivers of individuals with possible pervasive developmental Schork, N. J., Murray, S. S., Frazer, K. A., & Topol, E. J. (2009). Com-
disorders. Journal of Autism and Developemtal Disorders, 24, 659– mon vs. rare allele hypotheses for complex diseases. Current Opin-
685. https://doi.org/10.1007/BF02172145 ion in Genetics & Development, 19(3), 212–219. https://doi.org/10.
Machiela, M. J., & Chanock, S. J. (2015). LDlink: A web-based applica- 1016/j.gde.2009.04.010
tion for exploring population-specific haplotype structure and link- Singer, G. A. C., Wu, J., Yan, P., Plass, C., Huang, T. H. M., &
ing correlated alleles of possible functional variants. Bioinformatics, Davuluri, R. V. (2008). Genome-wide analysis of alternative pro-
31(21), 3555–3557. https://doi.org/10.1093/bioinformatics/btv402 moters of human genes using a custom promoter tiling array.
Martorell, L., Yubero, D., Capdevila, E. C., Fern andez Isern, G., BMC Genomics, 9(1), 349. https://doi.org/10.1186/1471-2164-9-349
Salinas, D., Mari Vico, R., Rebollo, M., Muchart, J., Tai, C., Chang, C.-W., Yu, G.-Q., Lopez, I., Yu, X., Wang, X.,
Armstrong, J., & Ortigoza-Escobar, J. D. (2022). The diagnosis of Guo, W., & Mucke, L. (2020). Tau reduction prevents key features
the first-documented intragenic KANSL1 microduplication patient of autism in mouse models. Neuron, 106(3), 421–437.e411. https://
broadens the genetic spectrum of Koolen de Vries syndrome. Clini- doi.org/10.1016/j.neuron.2020.01.038
cal Genetics, 101(5–6), 575–576. https://doi.org/10.1111/cge.14124 The ENCODE Project Consortium. (2012). An integrated encyclopedia
Matoba, N., Liang, D., Sun, H., Aygün, N., McAfee, J. C., of DNA elements in the human genome. Nature, 489(7414), 57–
Davis, J. E., Raffield, L. M., Qian, H., Piven, J., Li, Y., 74. https://doi.org/10.1038/nature11247
Kosuri, S., Won, H., & Stein, J. L. (2020). Common genetic risk Tick, B., Bolton, P., Happé, F., Rutter, M., & Rijsdijk, F. (2016). Heri-
variants identified in the SPARK cohort support DDHD2 as a tability of autism spectrum disorders: A meta-analysis of twin
candidate risk gene for autism. Translational Psychiatry, 10(1), studies. Journal of Child Psychology and Psychiatry, 57(5), 585–
265. https://doi.org/10.1038/s41398-020-00953-9 595. https://doi.org/10.1111/jcpp.12499
Mc Cormack, A., Taylor, J., Te Weehi, L., Love, D. R., & Tong, J. H. S., Hawi, Z., Dark, C., Cummins, T. D. R., Johnson, B. P.,
George, A. M. (2014). A case of 17q21.31 microduplication and Newman, D. P., Lau, R., Vance, A., Heussler, H. S.,
7q31.33 microdeletion, associated with developmental delay, Matthews, N., Bellgrove, M. A., & Pang, K. C. (2016). Separating
microcephaly, and mild dysmorphic features. Case Reports in the wheat from the chaff: Systematic identification of functionally
Genetics, 2014, 658570. https://doi.org/10.1155/2014/658570 relevant noncoding variants in ADHD. Molecular Psychiatry,
McLaren, W., Gil, L., Hunt, S. E., Riat, H. S., Ritchie, G. R. S., 21(11), 1589–1598. https://doi.org/10.1038/mp.2016.2
Thormann, A., Flicek, P., & Cunningham, F. (2016). The Ensembl Uffelmann, E., & Posthuma, D. (2021). Emerging methods and
variant effect predictor. Genome Biology, 17(1), 122. https://doi. resources for biological interrogation of neuropsychiatric poly-
org/10.1186/s13059-016-0974-4 genic signal. Biological Psychiatry, 89(1), 41–53. https://doi.org/10.
Natacci, F., Alfei, E., Tararà, L., D’Arrigo, S., Zuffardi, O., 1016/j.biopsych.2020.05.022
Gentilin, B., & Pantaleoni, C. (2016). Chromosome 17q21.31 dupli- Vohra, R., Madhavan, S., Sambamoorthi, U., StPeter, C., Poe, S.,
cation syndrome: Description of a new familiar case and further Dwibedi, N., & Ajmera, M. (2016). Prescription drug use and
delineation of the clinical spectrum. European Journal of Paediatric polypharmacy among medicaid-enrolled adults with autism: A ret-
Neurology, 20(1), 183–187. https://doi.org/10.1016/j.ejpn.2015.09.010 rospective cross-sectional analysis. Drugs Real World Outcomes,
Nyholt, D. R. (2004). A simple correction for multiple testing for single- 3(4), 409–425. https://doi.org/10.1007/s40801-016-0096-z
nucleotide polymorphisms in linkage disequilibrium with each Vorstman, J. A. S., Parr, J. R., Moreno-De-Luca, D., Anney, R. J. L.,
other. American Journal of Human Genetics, 74(4), 765–769. Nurnberger, J. I., Jr., & Hallmayer, J. F. (2017). Autism genetics:
https://doi.org/10.1086/383251 Opportunities and challenges for clinical translation. Nature
Peng, Z., Chen, J., Jin, L., Han, H., Dong, C., Guo, Y., Kong, X., Reviews Genetics, 18(6), 362–376. https://doi.org/10.1038/nrg.
Wan, G., & Wei, Z. (2020). Social brain dysfunctionality in indi- 2017.4
viduals with autism spectrum disorder and their first-degree rela- Wang, K., Li, M., & Hakonarson, H. (2010). ANNOVAR: Functional
tives: An activation likelihood estimation meta-analysis. annotation of genetic variants from high-throughput sequencing
Psychiatry Research: Neuroimaging, 298, 111063. https://doi.org/ data. Nucleic Acids Research, 38(16), e164. https://doi.org/10.1093/
10.1016/j.pscychresns.2020.111063 nar/gkq603
Perez, Y., Shorer, Z., Liani-Leibson, K., Chabosseau, P., Kadir, R., Ward, L. D., & Kellis, M. (2016). HaploReg v4: Systematic mining of
Volodarsky, M., Halperin, D., Barber-Zucker, S., Shalev, H., putative causal variants, cell types, regulators and target genes for
Schreiber, R., Gradstein, L., Gurevich, E., Zarivach, R., human complex traits and disease. Nucleic Acids Research,
Rutter, G. A., Landau, D., & Birk, O. S. (2017). SLC30A9 muta- 44(D1), D877–D881. https://doi.org/10.1093/nar/gkv1340
tion affecting intracellular zinc homeostasis causes a novel Weiner, D. J., Wigdor, E. M., Ripke, S., Walters, R. K.,
cerebro-renal syndrome. Brain, 140(4), 928–939. https://doi.org/10. Kosmicki, J. A., Grove, J., Samocha, K. E., Goldstein, J. I.,
1093/brain/awx013 Okbay, A., Bybjerg-Grauholm, J., Werge, T., Hougaard, D. M.,
Pugsley, K., Scherer, S. W., Bellgrove, M. A., & Hawi, Z. (2022). Envi- Taylor, J., Bækvad-Hansen, M., Dumont, A., Hansen, C.,
ronmental exposures associated with elevated risk for autism spec- Hansen, T. F., Howrigan, D., Mattheisen, M., … Robinson, E. B.
trum disorder may augment the burden of deleterious de novo (2017). Polygenic transmission disequilibrium confirms that com-
mutations among probands. Molecular Psychiatry, 27(1), 710–730. mon and rare variation act additively to create risk for autism
https://doi.org/10.1038/s41380-021-01142-w spectrum disorders. Nature Genetics, 49(7), 978–985. https://doi.
R Core Team. (2020). R: A language and environment for statistical org/10.1038/ng.3863
computing. R Foundation for Statistical Computing. https://www. Wong, M. L., & Medrano, J. F. (2005). Real-time PCR for mRNA
R-project.org quantitation. BioTechniques, 39(1), 75–85. https://doi.org/10.2144/
Reich, D. E., & Lander, E. S. (2001). On the allelic spectrum of human 05391rv01
disease. Trends in Genetics, 17(9), 502–510. https://doi.org/10.1016/ Yang, G., & Shcheglovitov, A. (2020). Probing disrupted neurodevelop-
s0168-9525(01)02410-6 ment in autism using human stem cell-derived neurons and orga-
Ritchie, G. R. S., Dunham, I., Zeggini, E., & Flicek, P. (2014). Func- noids: An outlook into future diagnostics and drug development.
tional annotation of noncoding sequence variants. Nature Developmental Dynamics, 249(1), 6–33. https://doi.org/10.1002/
Methods, 11(3), 294–296. https://doi.org/10.1038/nmeth.2832 dvdy.100
 19393806, 0, Downloaded from https://onlinelibrary.wiley.com/doi/10.1002/aur.3101 by Cochrane Luxembourg, Wiley Online Library on [09/03/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
PUGSLEY ET AL. 15

Yates, A. D., Achuthan, P., Akanni, W., Allen, J., Allen, J., SU P P O R T I N G I N FO R M A T I O N
Alvarez-Jarreta, J., Amode, M. R., Armean, I. M., Additional supporting information can be found online
Azov, A. G., Bennett, R., Bhai, J., Billis, K., Boddu, S.,
Marugan, J. C., Cummins, C., Davidson, C., Dodiya, K.,
in the Supporting Information section at the end of this
Fatima, R., Gall, A., … Flicek, P. (2020). Ensembl 2020. article.
Nucleic Acids Research, 48(D1), D682–D688. https://doi.org/
10.1093/nar/gkz966
Yeo, G., & Burge, C. B. (2004). Maximum entropy modeling of short How to cite this article: Pugsley, K., Namipashaki,
sequence motifs with applications to RNA splicing signals. Journal A., Bellgrove, M. A., & Hawi, Z. (2024).
of Computational Biology, 11(2–3), 377–394. https://doi.org/10.
Evaluating the regulatory function of non-coding
1089/1066527041410418
Zhang, L., Qin, Y., Gong, X., Peng, R., Cai, C., Zheng, Y., Du, Y., & autism-associated single nucleotide polymorphisms
Wang, H. (2019). A promoter variant in ZNF804A decreasing its on gene expression in human brain tissue. Autism
expression increases the risk of autism spectrum disorder in the Research, 1–15. https://doi.org/10.1002/aur.3101
Han Chinese population. Translational Psychiatry, 9(1), 31.
https://doi.org/10.1038/s41398-019-0369-x