02 - Genomics, Neuroscience and ASD
02 - Genomics, Neuroscience and ASD
02 - Genomics, Neuroscience and ASD
Autism spectrum disorder (ASD) constitutes a group of genetic contribution3–20. In fact, over the past 15 years,
aetiologically and symptomatically heterogeneous, despite an extraordinary degree of aetiological and clini
1
Department of Psychiatry
early-onset developmental disorders that are defined by cal heterogeneity, the pursuit of specific genetic contri
and Behavioral Sciences,
UCSF Weill Institute for
core deficits in social communication and the presence butors to ASD has been remarkably successful. Similarly
Neurosciences, University of repetitive, stereotyped behaviours1. Although these to that of other relatively common disorders, the
of California, San Francisco, features are the diagnostic sine quo non, individuals genetic architecture of ASD is now confirmed to include
San Francisco, CA, USA. come to clinical attention representing a very wide range a wide range of variation with regard to type, frequency
2
Quantitative Biosciences of severity and often with co-occurring signs and symp- and effect size. However, the greatest promise for insight
Institute, University of
toms that can include (but are not limited to) intellectual into the pathobiology of ASD presently is the result of
California, San Francisco,
San Francisco, CA, USA.
disability, anxiety, depression, sleep disturbance, epi- striking progress in gene discovery based on the identi
3
Langley Porter Psychiatric
lepsy, specialized talents (also known as savant skills), fication of large-effect, rare (often de novo) damaging
Institute, University of delayed motor development and neurological soft signs. variants in the coding portion of the genome21–34.
California, San Francisco, The inherent breadth of the presentation has led to an The success in characterizing many rare large-effect
San Francisco, CA, USA. often-stated clinical pearl that “if you have seen one mutations and concomitant gene discovery would reaso
4
These authors contributed person with ASD, you have seen one person with ASD”. nably be expected to set the stage for the rapid identifi-
equally: Helen Rankin Indeed, the notion of an autism spectrum is a central cation of key biological features and potential treatment
Willsey, A. Jeremy Willsey.
feature of the diagnostic nosology. targets. However, the extensive locus heterogeneity and
✉e-mail: jeremy.willsey@
However, despite the wide variety of clinical mani biological pleiotropy of ASD risk genes, the develop-
ucsf.edu; matthew.state@
ucsf.edu festations and the absence of reliable diagnostic mental dynamism of the human brain along with its
https://doi.org/10.1038/ biomarkers2, ASD persistently ranks at the very top of the structural and functional complexity, the wide range
s41583-022-00576-7 list of neuropsychiatric disorders with regard to relative of clinical outcomes associated with even large-effect
0123456789();:
Reviews
www.nature.com/nrn
0123456789();:
Reviews
Box 2 | Major methods to identify AsD risk genes and loci track record of success relying on homozygosity map-
ping of consanguineous pedigrees in a wide range of
rare variant-based approaches (allele frequency <1%) severe neurological and neurodevelopmental disorders
Whole-exome sequencing (WES). This approach aims to identify rare, especially de novo, (NDDs), including brain malformation syndromes,
coding variants recurring in the same gene in unrelated patients more often than severe intellectual disability and epilepsy67–69, it has
expected by chance (based on theoretical mutation rates, unaffected siblings or
been somewhat surprising that the yield of novel genes
population frequency). These approaches have historically focused on ‘simplex’ families
(unaffected parents with one child with autism), which tend to be enriched for de novo from similar efforts have been relatively limited for the
variants. However, rare inherited variants and case–control data from simplex, multiplex ASD phenotype, although there have been some notable
or unknown family types have increasingly contributed to gene discovery, especially exceptions70–74. Additionally, several more recent stud-
as databases accumulate information on population variant frequency and the ability ies have demonstrated an overall modest contribution
to stratify the most deleterious variants improves58–60. As these databases grow, the of recessive inheritance, including compound hetero
definition of rare is beginning to shift to <0.1% or even <0.01%. Genes identified by zygous damaging mutations in non-consanguineous
this approach tend to carry large effect sizes (Fig. 1). families74–77.
Whole-genome sequencing (WGS). Conceptually similar to WES, these analyses expand Studies have also supported the contribution of rare
to include non-coding variants. Identifying non-coding loci requires very large cohort somatic variation to ASD, suggesting that these vari-
sizes and is challenging due to incomplete characterization of regulatory loci (that is, ants may contribute to risk in approximately 5–22% of
functional boundaries, target gene), the relative inability to prioritize deleterious simplex cases56,57,78–83. However, gene discovery based
variants (especially compared with coding variants) and the tendency for these effects on somatic variants alone has been underpowered.
to be restricted to particular (yet currently unknown) developmental epochs, tissues Additionally, because these studies generally use data
and/or cell types. Accordingly, specific non-coding loci have not yet been identified
derived from the DNA of clinically accessible tissues
at genome-wide significant thresholds.
such as peripheral blood and saliva57,79–82, brain-specific
Copy number variant (CNV) detection. CNVs are deletions or duplications of DNA (also somatic mutations will, by their nature, be missed.
known as structural variants). The approach to detecting CNVs is conceptually similar Identifying somatic variants within brain tissue from
to WES and WGS: identify rare and/or de novo CNVs from genotyping or sequencing
individuals with ASD is an active area of investigation56,
data, and then search for recurrent variants overlapping the same locus in multiple
unaffected individuals. Associated loci tend to overlap genes and carry large effects
but the obstacles to this as a systematic approach to
(Fig. 1). When combined with WES and WGS, small coding CNVs, especially deletions, identifying novel ASD risk genes remain formidable.
can help identify individual risk genes24. Collectively, rare coding variants identified from
WES not only have been a particularly valuable resource
common variant-based approach (allele frequency >1%)
for biological studies but also have been shown to contri
Genome-wide association. This approach relies on common variants (also known as
single-nucleotide polymorphisms (SNPs)), occurring at a frequency greater than 1% bute to a substantial minority of individuals presenting
in the population. Variants are generally identified with genotyping arrays. Cases and for clinical evaluation. The estimates of clinical con-
controls are compared to identify SNPs with an allele that occurs more frequently in tribution range from a low of about 10% to a high of
cases. Very large cohorts are required to identify risk loci as they carry small effect sizes approximately one half of the clinical ASD population,
(Fig. 1). Only five genome-wide significant loci for autism spectrum disorder (ASD) have depending in part on the ascertainment criteria. For
been identified to date92 but this number is expected to grow as sample sizes increase. example, increasing parental age, low IQ relative to
mid-parental expectation, dysmorphology, epilepsy,
congenital heart disease, female sex and simplex status
The initial progress in ASD gene discovery relied — along with a larger number of unaffected siblings —
heavily on the identification of rare heterozygous are all known to lead to higher positive yields in genetic
de novo mutations leading to putative loss of function of testing21,22,24,28–31,33,44,64,84–89.
the encoded protein (also referred to as protein-truncating
Pleiotropy
A phenomenon in which
variants). However, missense de novo mutations30 as well Whole-genome sequencing. To date, the contribution of
a single gene contributes as rare inherited and somatic mutations (also known as whole-genome sequencing (WGS) studies to identify-
to multiple processes mosaic mutations) have now also been demonstrated to ing risk genes and loci for ASD has been limited com-
or phenotypic traits. carry risk23,24,29,53–57. In addition, the dramatic expansion pared with the approaches noted above. Capturing the
of databases of control exome and genome sequences full complement of variation across the entire genome,
Convergent neuroscience
Studies that address the has enabled the identification of ultra-rare genetic vari including both sequence and structural mutations,
overlap, or intersection, ation, the characterization of evolutionarily constrained clearly represents the future of genomic analyses for ASD
of genetic risk for a psychiatric genes (that is, those likely intolerant to variation) and and, indeed, all of medicine. However, several factors
disorder with respect to the prioritization of missense variants most likely to be limit its current utility: broadly, that the analysis and
molecular-level, cellular-level
and circuit-level function
deleterious (for example, as determined by the MPC or storage of WGS data require extensive computational
as well as across multiple ‘missense badness, PolyPhen-2 and constraint’ score)58–62, infrastructure and confer substantial costs; the ‘search
dimensions of analysis, all of which have demonstrated utility in parsing risk for space’ for WGS is 100-fold greater than for WES, yet is
including anatomical localization ASD and, consequently, the potential to improve gene less well characterized; the ability to interpret the conse-
and developmental timing.
discovery23,29,55,60,63–65. Indeed, characterizing ultra-rare quences of non-coding variation is far more challenging
Gene Ontology terms variants among highly conserved genes in case–control than assigning functional significance to canonical cod-
A hierarchical set of terms that data alone has emerged as an efficient, viable alternative ing mutations; and, as a group, rare non-coding variation
aim to define the universe of to family-based studies designed to identify rare de novo is estimated to carry relatively more modest risks for
possible descriptors a gene can mutations (for example, Fig. 1b; also ref.66). ASD as compared to rare damaging coding mutations90.
have, including properties such
as molecular function, cellular
Likewise, there have been long-standing efforts to Consequently, the cohort sizes needed to identify asso
component and biological identify rare recessive variants and compound hetero ciated loci with confidence are estimated to be an order
process. zygous variants contributing to ASD. Based on the strong of magnitude or more larger than for contemporary
0123456789();:
Reviews
a Relative risk versus allele frequency for ASD genetic variants b Relative risk versus allele frequency for schizophrenia genetic variants
300 300
De novo damaging variant De novo damaging variant
CNV CNV
50 50 (case–control)
20 20
10 10
5 5
2 2
1 1
1 × 10–5 1 × 10–4 0.001 0.01 0.1 0.5 1 × 10–5 1 × 10–4 0.001 0.01 0.1 0.5
Allele frequency in population Allele frequency in population
PTEN FOXP1
ADNP SCN2A CUL1
22q11.21(del)
SYNGAP1
SLC6A1
SHANK3
SUV420H1 GRIA3 RB1CC1
KDM5B
DYRK1A
CHD8 SP4 SETD1A
DEAF1
TLK2 7q11.23(dup)
ARID1B GRIN2A 16p11.2, distal(del)
DNMT3A
1q21.1(dup) XPO7
CTNNB1 POGZ 15q13.3(del)
GRIN2B 3q29(del) 3q29(del) 2p16.3 (NRXN1)(del)
GIGYF1 15q12(dup)
CHD2 ANK2 7q11.23(dup)
KDM6B ASH1L TRIO
MED13L DSCAM 15q13.3(del) 16p11.2, proximal(dup)
ANKRD11 2p16.3(del, NXRN1)
1q21.1(del/dup)
16p11.2(del) CACNA1G
22q11.21(del, dup)
22q13.33(del, SHANK3) 16p11.2(dup) HERC1
c Relative risk for de novo damaging variants in ASD versus d Relative risk versus false discovery rate for de novo damaging
schizophrenia genes variants in high-confidence ASD risk genes (FDR <0.01)
PTEN
FOXP1
150
150 SCN2A SLC6A1
AP2S1
ASD relative risk
ADNP
Relative risk
SUV420H1
100
100 DYRK1A
SYNGAP1
SHANK3
ARID1B
50 CHD8
50
POGZ GRIN2B
0
0
ASD Schizophrenia 1 × 10–12 1 × 10–9 1 × 10–6 1 × 10–3
FDR
www.nature.com/nrn
0123456789();:
Reviews
◀ Fig. 1 | relationship between effect size and allele frequency for loci discovered power91. However, more recent studies, involving more
in autism spectrum disorder or schizophrenia. a,b | Systematic studies of rare and than 18,000 cases and almost 28,000 controls, have suc-
common variants have identified risk loci for autism spectrum disorder (ASD) and for cessfully identified five common risk alleles of modest
schizophrenia, but the trajectory of discovery differs between these conditions. In ASD, effect (Box 2; Fig. 1a), meeting rigorous statistical criteria
the majority of loci discovered to date are genes and have been identified by exome-wide
for association92. Undoubtedly, this number will con-
sequencing studies of rare, generally de novo, coding variants (P < 2.5 × 10–6; red dots;
for ASD, 26 genes23; for schizophrenia, 10 genes66), whereas in schizophrenia the majority
tinue to grow as case–control cohorts further increase
are single-nucleotide polymorphisms (SNPs) and have been discovered based on genome- in size91. Moreover, despite the relatively small number
wide association studies (GWAS) of common variants (P < 5 × 10–8; teal dots; for ASD, of loci implicated to date, polygenic risk scores, which are
5 SNPs92; for schizophrenia, 270 SNPs130, of which only the 132 SNPs with relative a means to quantify cumulative common genetic risks,
risk >1 are shown). None of these genes (red) or SNPs (teal) overlap between ASD and have already demonstrated a modest ability to diffe
schizophrenia. Both disorders have a similar number of loci associated based on rare, rentiate cases versus controls92. Additionally, polygenic
generally de novo, copy number variants (CNVs; blue dots; for ASD, permutation- risk scores appear to differ by sex93,94, with a higher
based false discovery rate (FDR) ≤ 0.02, 10 loci24; for schizophrenia, Benjamini–Hochberg genetic load observed in females, providing evidence
FDR ≤ 0.02, 8 loci222). In contrast to rare coding variants and common variants, most of for a ‘female protective effect’ (FPE; discussed below)
the top CNVs overlap between ASD and schizophrenia. In general, rare variants carry
— a phenomenon that has also been widely observed
substantially higher relative risks than common variants. c | De novo damaging variants
in the top 26 ASD risk genes tend to carry higher relative risk than de novo damaging
with respect to rare variants22,24,27,28,30,32,33,95. There has
variants within the top 7 schizophrenia risk genes (de novo damaging variants have been also been some success in identifying shared common
identified in only 7 out of the 10 schizophrenia risk genes), underscoring the particularly genetic risks that cross a wide range of apparently
large contribution of rare variants to ASD. d | Utilizing a conservative FDR threshold disparate psychiatric diagnoses96.
instead of an exome-wide significant P value (FDR ≤ 0.01 versus P < 2.5 × 10–6) highlights There is also evidence that among individuals carry-
47 high-confidence ASD risk genes23, with varying relative risks. In general, the most ing large-effect rare risk variants, common alleles and
strongly associated genes carry the largest relative risks. See Supplementary Table 1 for polygenic inheritance likely contribute to idiopathic
a complete list of gene names, FDRs and effect sizes; see ref.223 for a broader list of ASD ASD97,98. Indeed, the question of the extent to which
risk genes as determined by FDR. Note that we defined relative risk as the ratio of the polygenic ‘background’ may shape the impact of rare
frequency of a given variant in cases versus unaffected controls. Dot size is proportional
mutations and help define an individual’s clinical pre
to relative risk. Shaded area represents 95% confidence interval of the locally weighted
least squares regression. Location on x axis in parts a,b is based on frequency in
sentation and natural history has been long-standing99.
unaffected controls. As there are no associations in ASD for alleles with frequencies The potential for very large, well-powered cohort stud-
between 0.001 and 0.1, no curve is shown in this interval in part a. Damaging variants ies in ASD makes this an increasingly tractable area of
consist of protein-truncating variants (frameshift, canonical splice-acceptor, canonical investigation93,100. Similarly, as the yield of SNPs from
splice-donor and nonsense variants) and missense 3 (Mis3) variants (those predicted ASD GWAS grows, important questions regarding
to be probably damaging by PolyPhen-2 (ref.62)). To estimate the per gene relative risk the genetic architecture underlying various groups of
for de novo damaging variants, we defined the frequency in cases as the total number individuals with autism, including those with so-called
of de novo damaging variants observed in a given gene23,66 and the frequency in controls high-functioning autism, may well be clarified. Finally,
based on estimates of the number of mutations expected per generation23. All 26 genes the discovery of a much larger set of genome-wide
identified in ASD have multiple de novo damaging variants in probands23. However,
significant common alleles will undoubtedly empower
only seven of ten genes identified in schizophrenia have de novo variants (the rest
were identified by rare damaging variants in case–control data only)66. Therefore, we
studies of convergence similar to those currently
estimated relative risk for de novo damaging variants when possible (red dots) and for successfully leveraging rare, large-effect mutations.
rare damaging variants when necessary (orange dots). To estimate relative risk for rare
damaging variants (case–control, schizophrenia only), we defined the frequency of each The number of ASD-associated genes. Contemporary
allele in cases based on reported frequency for schizophrenia66 and the frequency in association analyses of common, genetically complex
controls based on the frequency observed in the non-psychiatric subset of the gnomAD and heterogeneous phenotypes such as ASD rely heavily
v2 data set58. To estimate relative risk for rare CNVs, we defined the frequency in cases on the application of rigorous statistical methods that
based on reported frequencies for ASD24 and for schizophrenia222 and the frequency in effectively protect against false positive findings, which
controls based on frequency observed in the Database of Genomic Variants (DGV)224 have historically bedevilled psychiatric genetic studies.
(ASD) or directly from controls (schizophrenia)222. We estimated the 3q29(del) frequency
Over the past decade, two broad approaches to the
in schizophrenia as 0.00005 because no 3q29(dels) were observed in 20,227 controls.
The 15q13.3(del) is referred to as 15q13.2-13.3 in ref.24 and we estimated its frequency as analysis of rare variants, using either a Bayesian or a
0.000092 based on structural variants in gnomAD225 because of inconsistent frequencies frequentist framework, have emerged, providing the
reported in smaller studies in DGV. We estimated relative risk for common variants based field with robust and consistent results. For WES stud-
on reported odds ratios for ASD92 or for schizophrenia130. Part b adapted from ref.66, ies specifically, it has become commonplace to require
CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). a discovery P value of less than or equal to 2.5 × 10–6
— effectively correcting for the multiple comparisons
WES studies, and, currently, no published investiga- among the approximately 20,000 genes comprising
tions have yet realized sufficient numbers to identify, the human exome. Widely used alternative methods
definitively, individual non-coding risk variants (Box 2). estimate false discovery rates (FDRs) per gene using
a Bayesian approach, assigning significant associa-
Common variant-based discovery. Although most of the tion typically to genes showing FDR ≤ 0.1 and reserv-
ASD risk loci confirmed to date have come from stud- ing a ‘highest-confidence’ designation to those with
ies of rare coding variants of large effect, there is strong FDR ≤ 0.01. By contrast, GWAS arrived relatively quickly
Penetrance evidence that common variants of small effect carry at a single consensus standard and threshold, namely a
The probability that an
individual with a given
the majority of population risk4,5. Despite this, initial discovery P value of less than 5 × 10–8 followed by the
genotype will exhibit the ASD genome-wide association studies (GWAS) yielded finding of nominal significance in a replication cohort101.
associated phenotype. disappointing results owing to insufficient statistical Importantly, the fact that there remain alternative
0123456789();:
Reviews
Resilience
accepted statistical approaches to WES studies of rare as noted above, the yield of rare risk mutations from
The capacity of an individual variants contributes to some degree of variability in gene high-throughput sequencing studies of ASD cohorts has
to have a ‘better than expected ‘accounting’. For example, the recent omnibus analysis by consistently been found to increase as IQ declines. Not
outcome’ (for example, being the Autism Sequencing Consortium23 is commonly cited surprisingly, these kinds of observations have fuelled
unaffected despite having
as identifying 102 ASD risk genes, based on the Bayesian debates over whether these diagnoses are truly sepa-
multiple autism spectrum
disorder-associated genetic FDR threshold of 0.1 noted above. Using the alterna- rable at the genetic level6,92,108–112. For CNVs, this vari-
variants). tive accepted exome-wide significant P value threshold ability in phenotypic outcomes has been particularly
(2.5 × 10–6) yields 26 risk genes (Fig. 1a) and restricting dramatic. From over a decade of progress in studies of
Protein-truncating variants to the highest-confidence genes based on FDR ≤ 0.01 ASD, schizophrenia, epilepsy and intellectual disability,
(Also referred to as loss-
of-function or likely gene
identifies 47 risk genes (Fig. 1d). it has become increasingly difficult to identify any struc-
disrupting variants). Sequence Additional definitional issues contribute to a range tural variant associated with one of these diagnoses that
variants that are predicted to of views on what constitutes the definitive list of bona is not also associated with some risk for one or more of
shorten the protein-coding fide risk genes. For example, some widely used databases the others — and the list of additional clinical outcomes
sequence of a gene — usually
(for example, SFARI Gene102,103) include genes discovered showing overlap continues to grow113–116. Although there
to the point of resulting in a
non-functional protein or no in studies of syndromic forms of ASD. Such approaches is some evidence for differing relative representations of
protein at all due to nonsense- have the advantage of being more complete than those clinical manifestations for individual CNVs, including
mediated decay — including relying solely on large-scale association studies restricted for duplications versus deletions of the same genomic
nonsense, frameshift and
to ‘idiopathic’ clinical samples (Box 1). However, consen- interval24, overall, the data have consistently pointed to
essential splice site variants.
sus approaches also introduce a degree of subjectivity the absence of diagnostic specificity6,108,110–112 (Fig. 1a,b).
Somatic mutations and potential ascertainment bias that may be largely These findings, not surprisingly, led to concern that
(Also known as mosaic avoided by focusing only on genes that exceed predeter- the pursuit of biological mechanisms related to any given
mutations). Variants that are mined statistical thresholds23,24. The task of building a CNV would lack the ability to inform development of
present in fewer than 100%
singular high-confidence ASD risk gene list is similarly ASD therapeutics, as well as to speculation that indivi
of the cells of an individual
(for example, present in brain made more challenging by studies that cross diagnostic dual genes of large effect would pose similar challenges.
cells but not present in categories. For instance, although approaches that com- A related concern was, and remains, that the wide range
germ-line cells), generally bine ASD and other NDD cohorts increase the yield of of potential clinical outcomes from any given variant
because the mutation
gene discovery, they also complicate our understanding would pose serious obstacles to the design and execu-
occurred after fertilization.
of the relative risk that each identified gene carries for tion of early intervention trials, including those targeting
Compound heterozygous ASD versus other NDDs. Further, results from studies individuals with large-effect genetic risks. For example,
variants that rely on targeted sequencing can be difficult to it is not obvious how one would assess efficacy in a group
Variants similar to recessive integrate with findings from WES or WGS, given the of high genetic-risk individuals when the clinical course
variants, in that both alleles of
a gene are mutated; however,
challenges of arriving at a single applicable statistical might evolve over weeks, months or years and include
in this case, the two alleles are approach or threshold. symptoms that vary dramatically from one person to the
different. next, irrespective of treatment.
ASD versus other NDDs. With recent success in gene As both rare and common variant studies have
Neurodevelopmental identification across multiple NDDs and psychiatric progressed, a more nuanced view of the genetic archi-
disorders
(NDDs). A group of conditions
disorders, the question of overlap of genetic risks among tecture and loci contributing to ASD versus other neuro
(including autism spectrum apparently distinct clinical syndromes has emerged as a developmental syndromes is beginning to emerge.
disorder, attention deficit key area of inquiry, for both conceptual and pragmatic For example, in comparison with schizophrenia,
hyperactivity disorder, tic reasons. Some of the areas currently of greatest interest much smaller cohorts have been required to identify
disorders and intellectual
disability) characterized
in this regard include the degree to which ASD risks are large-effect risk genes in ASD. Conversely, much larger
by developmental onset, distinguishable from the risks for schizophrenia or from sample sizes have been required to identify the first
atypical brain development intellectual disability. Both questions have historical rele genome-wide significant common variants in ASD ver-
and resultant impairments vance as well: autism was initially conceptualized clini sus in schizophrenia92,117–119. Consistent with these obser-
in cognition, communication,
cally as a form of childhood-onset schizophrenia104,105, vations, rare variants show, on average, a larger relative
behaviour and/or motor skills.
but conventional wisdom subsequently shifted and risk in ASD than in schizophrenia (Fig. 1a–c). Moreover,
Polygenic risk scores the notion was formally rejected after the second edi- in a recent study of schizophrenia focused on identifying
Estimates of an individual’s tion of the Diagnostic and Statistical Manual of Mental ultra-rare coding variants, none of the 10 genes meet-
genetic predisposition for a Disorders106. Similarly, it is now clear that early diagnostic ing the most rigorous cut-off for statistical significance
disorder or disease based
criteria for autism tended to enrich for individuals with (P ≤ 2.5 × 10–6) overlaps with the 26 genes identified at
on the collective effects of
many genetic variants. co-existing intellectual disability, resulting in a long- the same threshold in prior studies of ASD23,66 (Fig. 1a,b).
standing overestimate of the prevalence of intellectual Moreover, an examination of a broader set of genes
SFARI Gene disability in the population with ASD107. from this recent study that fell just below a threshold of
An evolving database compiled
The question of whether and to what degree there is FDR ≤ 0.1 for schizophrenia risk demonstrates that only
by the Simons Foundation
Autism Research Initiative
specificity to ASD genetic risk versus other developmen- 3 out of these 34 probable schizophrenia risk genes
(SFARI) research program tal phenotypes remains a controversial topic. Successive overlap with the 102 high-confidence ASD risk genes
that includes a list of genes generations of systematic CNV studies have shown that identified based on the same statistical cut-off23,66. In a
and copy number variants identical structural variations can contribute to multiple similar vein, recent detailed investigations of diverse
(CNVs) associated with autism
distinct diagnostic outcomes, including, but not limited mutations in a high-confidence ASD risk gene have
spectrum disorder curated
from human studies and to, epilepsy, schizophrenia, bipolar disorder, intellec- shown a stronger relationship between individual alleles
animal models. tual disability, attention deficit hyperactivity disorder, and specific phenotypic outcomes than would be antici
Tourette disorder and ASD (see also Fig. 1a,b). Moreover, pated based on the findings from genotype–phenotype
www.nature.com/nrn
0123456789();:
Reviews
studies of high-confidence CNVs120. Moreover, given conclusions. Moreover, the relationships between diag-
strong evidence that rare and common alleles combine nosis, variant class (for example, protein-truncating vari
to contribute to both disorders93,100 and the current state ants versus missense variants123), direction of effect (for
of locus discovery overall — with a paucity of common example, duplication versus deletion and loss of function
alleles identified in ASD and rare alleles in schizophrenia versus gain of function120) and patient sex are likely to
— definitive conclusions about the presence or absence provide key insights into the mediators and moderators
of specificity between ASD and schizophrenia genetic of whatever specificity is found to be present. Finally,
risks are likely premature. although there is understandable consternation regard-
There has also been long-standing interest in the ing the possibility that genetic risks for ASD, intellectual
overlap, both clinically and aetiologically, of ASD and disability, epilepsy and other NDDs are inseparable, the
intellectual disability. As noted, early diagnostic crite- high degree of comorbidity that is also observed, includ-
ria for autism were biased towards the ascertainment of ing for individuals with large-effect rare mutations in
individuals with concomitant intellectual impairment. the highest-confidence ASD genes, may turn out to
However, as the wider spectrum of ASD clinical manifes- facilitate early therapeutic trials. For example, pheno-
tations has been increasingly appreciated, the percentage types apart from social functioning, such as seizures,
of individuals with ASD and without intellectual disabi motor development and overall intellectual function-
lity has increased markedly and now reflects the majority ing, are presently more amenable to quantification and
of those who carry an ASD diagnosis. Nonetheless, the to reliable longitudinal assessment, particularly early in
frequent co-occurrence clinically of intellectual disability development.
and ASD has prompted an ongoing debate over the sepa-
rability of social versus intellectual functioning and has Future work in ASD genetics. For those who recall the early
raised key questions as to whether recent gene discovery days of gene discovery efforts in psychiatric disorders
efforts in ASD cohorts have the potential to offer insight writ large, and in ASD in particular, the past 15 years
into the core biology of social disability110–112. There has of progress has been nothing short of spectacular.
been a long-standing hypothesis that social ability is con- However, considerable opportunities remain: the meth-
tinuously distributed in the population and that higher ods that have been most successful in elaborating the
IQs buffer behaviourally against autism symptoms121,122. growing list of risk genes introduce some biases owing,
One contemporary interpretation of this idea is that the in part, to differential detection of types of variants. For
recent spate of large-effect mutations identified in WES example, the sex chromosomes have been largely unex-
studies primarily lead to lower IQ, which, in turn, limits plored to date, owing to complexities in accurate variant
an individual’s ability to compensate for social impair- calling at scale. In addition, their distinctive inheritance
ment mediated by common alleles. Along these lines, patterns and selection pressures preclude the use of
multiple analyses have noted that the rate of rare de novo current metrics (for example, the probability of being
mutations in ASD cohorts increases as IQ declines33,44,64,88 loss-of-function intolerant) that have accelerated gene
and some have concluded that rare de novo sequence discovery on the autosomes. It is also likely that the
mutations are over-represented only in individuals with relative sparsity of female samples in existing research
IQ below the mean64. cohorts, especially from simplex families, has limited the
By contrast, multiple studies have also shown that ability to detect sex-specific or sex-biased risk associated
rare sequence variants and CNVs contributing large with specific large-effect variants. Similarly, missense
amounts of risk are found across the full range of IQ23,24,33. mutations, gain-of-function mutations, non-canonical
Moreover, one of these studies evaluated ASD genes with splice site mutations, somatic mutations and the universe
regard to whether they showed evidence for being either of non-coding variants remain far more difficult to inter-
‘ASD-predominant’ or ‘ASD/NDD-predominant’ and pret at scale than canonical loss-of-function mutations.
found evidence for differing distributions of phenotypes Recessive variants, inherited variants and variants with
corresponding to specific genes23. Although the authors low probability of being loss-of-function intolerant also
of the study did not conclude that any genes exclusively remain difficult to identify, although network-based
carried risk for ASD alone, their analysis was consistent approaches have begun to fill in these gaps, by identi-
with numerous observations suggesting that the rare fying putative novel risk genes based on proximity to
variant genetics of human social and intellectual func- strongly associated risk genes54,124,125.
tioning are separable to some degree. For example, one Similarly, efforts to identify genotype–phenotype
Missense variants
Single base-pair coding
study showed that duplications versus deletions of the relationships have so far been met with limited
variants that result in altered identical Williams syndrome region on chromosome success24,126 but this is a highly promising area for future
amino acids, for which several 7q11.23 leads, on average, to similar levels of intellectual investigation. Hypotheses regarding the utility of endo-
metrics have been developed impairment but also, typically, to strikingly divergent phenotypes to increase the power of gene discovery
to predict the functional
social phenotypes33. efforts have not panned out127. Thankfully, however, the
consequence, including the
PolyPhen-2 (Polymorphism As progress in both common and rare variant studies well-known limitations of categorical psychiatric diag-
Phenotyping v2) and MPC paints a more complete picture of the allelic architecture noses combined with the extreme clinical heterogeneity
(missense badness, PolyPhen-2 of ASD, intellectual disability, schizophrenia and other of ASD have not thwarted the identification of dozens of
and constraint) metrics, for psychiatric disorders, multiple related questions will large-effect genes and the first set of common risk alleles.
which a categorization of
PolyPhen-2 missense 3 (Mis3)
become increasingly tractable. Studies examining the The combination of increasingly large cohorts combined
or an MPC score ≥2 reflects a correlation of specific genes with differing distributions with the ability to work from a well-established list of
probably damaging variant. of diagnoses will have increased power to draw definitive definitive genes and mutations promises to provide
0123456789();:
Reviews
insight into questions, such as the relative specificity of a mutation in a human ASD risk gene in, for example,
genetic risks, that have long been vexing to investigators. rodents, has become routine, as has the ability to chara
As foundational network data characterizing the rela- cterize a very wide range of neurobiological sequelae,
tionship between diverse risk genes and the proteins they including in awake behaving animals. However, the
encode are generated, so-called ‘network–phenotype’ observation of even the most plausibly ASD-relevant
analyses may increase power to extract meaningful phenotype in a construct-valid mutant does not, on its
relationships between risk variants co-localizing in net- own, provide reassurance that the observation is point-
work space and patient phenotypes37,124,128. That being ing to causal mechanisms in the human syndrome — a
said, despite the abundance of genes identified based on problem that is compounded by the well-established
rare and de novo mutations, there is a strong rationale to challenges of relying on face-valid behavioural pheno-
continue to pursue common variant analyses. Although types in evolutionarily distant species to illuminate psy-
studying individual small-effect non-coding alleles may chiatric syndromes35. Similarly, even for a single gene,
not be as direct a path to exploring core biology, they limitations in the ability to assay molecular and cellu-
account for a substantial proportion of population risk lar phenotypes in a hypothesis-free manner across the
and the interplay of common and rare variant risk is entire brain at high resolution in more complex models
now a tractable area of inquiry that could have profound introduces inherent analytical biases. In this regard, hav-
importance124,129,130. The ability to evaluate the combined ing strong evidence regarding when and where to look
contributions of variation across the frequency and for ASD-related pathobiology is potentially invaluable
effect spectrum promises to yield critical insights into in sorting through an immensely complex and dynamic
natural history and treatment response, and potentially biology. These types of considerations have motivated
to contribute to diagnostic strategies especially aimed at convergent neuroscience approaches over the past sev-
very early detection. eral years. These studies assess risk genes in parallel, in a
hypothesis-free manner, to identify and confirm shared
Progress in convergent neuroscience developmental epochs, brain regions, specific cell types
The progress made to date in gene and locus discovery and biological pathways associated with human vulner-
in ASD has set the stage for understanding the role of ability to social impairment as a prelude to resolving the
these genes and loci in human brain development and pleiotropy of ASD risk genes37.
Gene Ontology enrichment
function, and for clarifying how disruptions can lead
analysis
A statistical approach that to the emergence of social impairments. However, the Incomplete gene ontologies. Convergent neuroscience
assesses whether specific extensive biological and locus heterogeneity revealed approaches are based on the hypothesis that subsets of
Gene Ontology terms are by contemporary genomic analyses has, simultaneously, risk genes for a given disorder, such as ASD, will point
statistically over-represented led to concern that any individual gene or mutation, and to shared vulnerability somewhere along the path from
in a (generally large) set of
genes or proteins (for example,
consequently any related therapy, might be relevant only gene to complex behaviour. This logic is reflected in the
among autism spectrum for a fleetingly small number of individuals with autism. earliest sequencing studies in ASD as successful gene
disorder risk genes). At the same time, from the earliest successful studies, discovery was immediately followed by a search for func-
there has also been evidence that a larger number of tional overlap. Many of the first genes associated with
Gene set enrichment
apparently functionally diverse ASD risk genes tend to ASD based on targeted sequencing studies were known
analysis
Like Gene Ontology converge on a smaller number of biological pathways, to encode proteins localized to the synapse (for example,
enrichment analysis, developmental stages, brain regions and cell types. NRXN1, SHANK3, NLGN3 and NLGN4)46,131,132. Not
this statistical approach Indeed, steady progress has been made in specifying and surprisingly, as a result, the field focused on the struc-
assesses whether specific defining these points of biological overlap through reli- ture and function of the synapse as a convergent point
Gene Ontology terms are
statistically over-represented
ance on relatively hypothesis-free approaches and on the of ASD-relevant biology41,133,134. As systematic gene and
in a large set of genes or expanding armamentarium of developmentally contextu- locus discovery proceeded, prima facie evidence for con-
proteins; however, unlike Gene alized neurobiological data resources, spanning levels of vergence was also found for genes involved in chroma-
Ontology enrichment analysis, analysis, construct-valid models and human brain tissue. tin modification and transcriptional regulation27,28,30,33,43.
gene set enrichment analysis
As risk gene lists have grown and the vast degree of
also incorporates the rank of
genes within the set into the The problem of pleiotropy. A major challenge in trans- allelic and locus heterogeneity has been revealed, enrich-
statistical test (for example, lating successful gene discovery in ASD into a reliable ment analyses (for example, Gene Ontology enrichment
differentially expressed genes understanding of pathobiological mechanisms — and analysis and gene set enrichment analysis ) have been
ranked based on the fold- the identification of related treatment targets — has invoked to identify statistical over-representation of
change between cases and
controls).
been the pleiotropy of genes, that is, their tendency to molecular functions, cellular components, biological
be involved in different roles throughout brain deve processes and other gene lists from various pathway
Pathway databases lopment and function, potentially dependent on, for databases43,135–140. These analyses have repeatedly high-
Databases that annotate example, cell type, brain region and/or developmental lighted gene targets of FMRP and the β-catenin path-
genes with an ontological
stage35,37,38. Consequently, although an identified risk ways28,34,42 and confirmed significant enrichment of Gene
approach to capture functional
relationships, including gene may have a well-studied and well-characterized role Ontology terms associated with chromatin modifica-
molecular interactions, in a biologically plausible process, it does not necessarily tion and synaptic function23,24,29,41,43,133,134. However, it is
regulation and phenotype follow that either this is the only function that the gene important to note that the underlying data that support
associations; common pathway plays or it is necessarily the role that contributes to the Gene Ontology terms and other annotations may offer
databases include Gene
Ontologies (GO) and Kyoto
emergence of ASD. an incomplete and biased view of gene function and
Encyclopedia of Genes and The challenge of pleiotropy is, of course, also highly that these approaches are generally unable to account
Genomes (KEGG). relevant in the study of animal models. Recapitulating for pleiotropy and/or missing data37,141. The annotations
www.nature.com/nrn
0123456789();:
Reviews
underlying these terms are often created by aggregating develop spatio-temporally defined co-expression net-
existing data, a process that can skew towards catego- works and then examined their relationship to a set of
ries of function and, potentially, cell types and develop- ASD genes meeting specified statistical thresholds for
mental time points that are more commonly or readily association. The observation of a statistically signifi-
studied, while minimizing or omitting others142,143. For cant over-representation of risk genes mapping with
example, given several decades of effort aimed at chara co-expression networks or of greater than expected
cterizing neurons and a particular emphasis on illumi- expression correlation between ASD genes within net-
nating the structure and function of the synapse, it is not works was considered evidence for convergent vulner-
surprising that ontologies tend to be relatively enriched ability. Three groups published the first examples of
for related annotation terms. These types of inherent, these approaches leveraging somewhat different gene
and often cryptic, ascertainment biases can cascade into lists and approaches to network generation. Strikingly,
self-reinforcing, but potentially incomplete, biological despite these differences and the sparse lists of associated
interpretations. genes at the time, all three groups identified evidence for
Indeed, recent analyses of some of these same genes convergence in prenatal development25,146,147, with two
annotated at the synapse have shown experimental evi- resolving this further to mid-gestational prefrontal cor-
dence of functional convergence that precedes synapse tex (PFC) and developing excitatory neurons25,146. More
formation — for example, in neurogenesis40,144. Of course, recent studies — using similar or divergent methods,
these genes probably have multiple roles during brain rapidly expanding gene lists and more comprehensive
development with their observed function dependent, foundational transcriptomic resources — have consis
in part, on spatiotemporal parameters. In addition, it is tently replicated these findings and, additionally, pro-
possible that both processes are relevant to ASD and vided evidence for involvement of the cerebellum and
interrelated, with synapse dysfunction following from striatum126,148–153 (Fig. 2a).
earlier developmental derangements. To address the Based on the aforementioned reproducible asso
inherent limitations of Gene Ontology analyses, recent ciation of the developing mid-gestational frontal cortex,
work has increasingly focused on the development of analyses of the more fine-grained BrainSpan prenatal
highly parallelized, systematic, experimental character- laser micro-dissected data set154 have similarly been
izations of relevant genes and/or proteins across deve used to identify points of laminar convergence for ASD
lopmental stages, tissues and cell types. Similarly, data risk genes in the developing cortex, through the analysis
sets developed through the manipulation of known ASD of the preservation of network connections by layer25,
genes within intact biological systems also offer a basis the over-representation of associated gene lists146 or the
for convergence analysis that mitigates some of the limi degree of connectivity of ASD-associated interaction
tations of ontology-based approaches. These efforts are networks40 (Fig. 2b). Early analyses implicated the inner
described in more detail below. cortical plate (iCP)25,146, and a recent analysis using more
than 100 ASD genes and a comprehensive database of
Examining neurotypical brains. Several groups have molecular interactions confirmed this finding and fur-
taken advantage of the growing list of reproducibly asso- ther identified the inner subventricular zone (iSVZ) and
ciated risk genes to identify anatomical regions, cell types subplate40. Developmentally, a key process occurring in
and/or developmental stages of the typically developing mid-gestational cortex is neurogenesis, with neural
brain of particular relevance to ASD. Conceptually, these progenitors actively undergoing cell division and diffe
studies are based on two underlying hypotheses: first, rentiation in the subventricular zone, leading to the gene
genes expressed in a highly coordinated spatiotemporal ration of the postmitotic excitatory neurons of the cortex
fashion probably share functions; and, second, identifying (migrating through the subplate and into the cortical
points of significant overlap among a group of disparate plate) (Fig. 2b). The findings from the biological analyses
Neurogenesis ASD risk genes in the context of typical development is of these systems are also consistent with findings from
The process by which new a means of triangulating key parameters associated with orthogonal enrichment analyses23,25,29,146,148, as well as
neurons are generated, in
core pathobiology. These hypotheses highlight the impor- emerging evidence from model system studies, as noted
humans most active during
gestational weeks 10–25. tance of the nature of the input data: these analyses are in later sections.
dependent on gene lists derived in a systematic, unbiased Recent advances in single-cell RNA sequencing
BrainSpan Atlas of the fashion. For example, the potential for circular reasoning (scRNA-seq) and single-nucleus RNA sequencing tech-
Developing Human Brain introduced by investigating convergence among a set of nology have enabled the generation of high-resolution
A foundational resource that
includes bulk transcriptome
‘candidate’ genes identified based, in part or in whole, expression data at the level of single cells or nuclei
profiling of up to 16 cortical on biological plausibility is obvious. For this reason, from the typically developing prenatal cortex151,155–158.
and subcortical structures the emergence of widely accepted statistical thresholds Analyses of scRNA-seq data tend to first identify the cell
across human brain in genome-wide and exome-wide studies of idiopathic types present based on clustering patterns of individual
development (prenatal
ASD cohorts to identify gene lists has been a key to the cells and expression of cell type-specific marker genes,
to adult).
development of meaningful convergence analyses. and then assess each cell type for concurrent (that is,
Neurotypical Some of the earliest efforts in this vein used network- co-expression), high or specific expression of ASD risk
Description of individuals with based approaches to analyse data from the BrainSpan genes. These studies have again implicated excitatory
apparently typical intellectual Atlas of the Developing Human Brain, which generated progenitor cells and maturing excitatory cortical neu-
and cognitive development
(for example, not affected
transcriptomic data from early gestation to late adult rons, and have additionally implicated inhibitory pro-
by a neurodevelopmental stages across multiple anatomical regions of ‘neurotypical’ genitor cells and maturing inhibitory neurons of the
or psychiatric disorder). human brains145. Several studies leveraged these data to cortex23,54,151,157,159 (Fig. 2c, left).
0123456789();:
Reviews
a Studies on neurotypical controls indicate points of convergence of ASD genetic risk (bulk tissue data)
Early gestation Early mid-gestation Late mid-gestation Late gestation
Parietal Parietal
Temporal Occipital cortex cortex
Ventral cortex cortex
forebrain
Parietal DFC
cortex Occipital
PFC cortex
OFC
Frontal URL
cortex OFC
Ventral CBC ITC
Dorsal forebrain Ventral
forebrain forebrain
b ASD genetic risk may disrupt neurogenesis in midgestational cortex (dissected tissue data)
Cortical plate
Excitatory
neuron
Subplate
Migrating
neuron Intermediate zone
oSVZ
Intermediate
progenitor iSVZ Significant ASD risk gene
network connectivity
in layers iSVZ, subplate and
Radial glia iCP of neurotypical controls
Ventricular zone
c ASD risk genes are differentially enriched in cell types of the human brain (single-cell data)
Neurotypical developing human cortex Post-mortem ASD cortex
Inhibitory progenitors
30 and neurons 30
20 20
Increased
10 Excitatory 10 representation of
progenitors
t-SNE2
t-SNE2
excitatory neurons
0 and neurons 0
–10 –10
–20 –20
Increased representation
–30 –30 of microglia
d Post-mortem human brains studies suggest differences in ASD (bulk tissue data)
www.nature.com/nrn
0123456789();:
Reviews
◀ Fig. 2 | emerging patterns of AsD convergence identified from gene expression similar enrichment in excitatory neurons, although the
data. Numerous studies investigating human gene expression data sets for convergence signal was stronger for more superficial layers146. Two
related to autism spectrum disorder (ASD) risk have been conducted, spanning so-called additional studies in mice generated and analysed a data
‘neurotypical’ (control) brain samples (parts a–c) as well as post-mortem brain samples set with gene expression profiles for 25 distinct CNS cell
from individuals with ASD (parts c,d). These data sets encompass bulk tissue from major
types and observed enrichment of ASD-associated genes
regions of the brain (parts a,d), bulk tissue from more finely dissected regions of the
developing cortex (part b) and single cells from the cortex (part c). Neurotypical data
in cortical excitatory neurons, as well as inhibitory neu-
tend to include prenatal and postnatal samples across the entire spectrum of human rons of the cortex and striatum126,148,161. A recent study
brain development whereas patient-derived data include postnatal samples only (ASD using Xenopus tropicalis observed that ten ASD risk
diagnosis requires behavioural assessment). Generally, approaches utilizing neurotypical genes chosen solely based on the strength of statistical
expression data seek to identify convergence by searching for ‘enrichment’ of ASD risk evidence for association are co-expressed in the deve
genes in particular developmental epochs, brain regions and/or cell types. In this case, loping telencephalon at time points that correspond to
enrichment is indicated by the ‘strength’ of co-expression of ASD risk genes or by ‘high’ human mid-gestational periods in the frontal cortex40.
or ‘specific’ expression of ASD risk genes. By contrast, approaches utilizing patient These non-human studies suggest that expression con-
expression data generally aim to identify convergence by characterizing systematic vergence among ASD genes is at least partially conserved
differences between cases and matched controls (that is, genes consistently differentially
through evolution, although this needs to be examined
expressed in brain tissue from individuals with ASD). Although data and approaches
differ substantially across studies, the developing frontal cortex and excitatory neurons in more detail and across additional models.
are recurring points of enrichment. a | Studies investigating spatiotemporal convergence
of ASD genetic risk in gene expression data from bulk samples of neurotypical brains have Comparative post-mortem studies. An alternative
consistently highlighted mid-gestational frontal and parietal cortices23,25,126,129,146,148,149,153. approach to identifying convergence of ASD risk involves
We summarize the extent of replication by color intensity, with the strongest replication comparing the molecular profiles of post-mortem brain
indicated by dark red and the weakest by light red. These studies varied widely in tissue from individuals with ASD and from neuro
granularity of developmental periods and brain regions studied, and therefore we typical controls. One of the first published studies of this
considered studies that examined two or fewer regional (for example, grouping all type profiled the transcriptomes of ASD and typically
cortical regions, studying only one brain region) or temporal (for example, prenatal developing brains across the PFC, temporal cortex and
versus postnatal) contexts to provide ‘limited’ evidence. Strong replication denotes a
cerebellum, and observed that regional expression diffe
significant finding in at least three ‘non-limited’ studies, moderate replication indicates
a significant finding in two non-limited studies and weak replication corresponds to a rences identified in control brains were attenuated in the
significant finding in one non-limited study and at least one limited study. We define frontal and temporal lobes of individuals with an ASD
early gestation as 8–10 post-conception weeks (PCW), early mid-gestation as 10–16 PCW, diagnosis162 (Fig. 2d). This observation was replicated
late mid-gestation as 16–24 PCW and late gestation as 24–38 PCW. b | Based on findings in a subsequent PsychENCODE study with larger cohort
summarized in part a, several studies have examined laminar convergence of ASD sizes163. These studies and others have also identified
genetic risk in neurotypical early mid-gestational frontal cortex. An initial study assessing downregulation of neuron and synapse-related genes
preservation of spatiotemporal co-expression networks by layer implicated the inner and upregulation of microglia and immune-related
cortical plate25 whereas a contemporaneous study assessing specificity of expression genes in the brains of individuals with ASD compared
of ASD gene lists by layer indicated a relative lack of layer-specific enrichment146. A later with controls162–166 (Fig. 2d).
study incorporating orthogonal gene interaction data and assessing ASD gene network
scRNA-seq has provided some additional granularity
connectivity by layer replicated the inner cortical plate finding and newly implicated
the inner subventricular zone and subplate40. c | Analyses of single-cell gene expression
to the observed differences between brains of individ-
data from neurotypical controls and individuals with ASD have consistently highlighted uals with ASD and neurotypical brains (Fig. 2c, right).
excitatory neurons. In neurotypical mid-gestational cortex, ASD risk genes tend to be Experiments leveraging data from scRNA-seq to decon-
highly and/or specifically expressed in developing excitatory and inhibitory neurons volve bulk RNA-sequencing (RNA-seq) expression data
(progenitors and maturing neurons)23,54,151,157,159. Analyses of single-cell gene expression have shown that the majority of expression variation
data from (postnatal) cortex of individuals with ASD have similarly identified consistent across brain tissue samples is attributable to varying
differences between cases and controls in excitatory neurons, as well as potential proportions of basic cell types, and that that the cortex
differences in microglia and oligodendrocytes166,167. d | Analyses of bulk samples of brain of individuals with ASD may have higher fractions of
tissue from individuals with ASD compared with those from matched neurotypical excitatory neurons, microglia and astrocytes, and lower
controls support the major findings from parts a–c. These studies implicate frontal cortex,
fractions of oligodendrocytes167. This suggests that prior
excitatory and inhibitory neurons, and microglia, and additionally highlight differences in
genes associated with astrocytes and in expression modules related to synaptic function
observations, from comparative bulk RNA-seq studies,
and inflammation162–165,226. CBC, cerebellar cortex; DFC, dorsal frontal cortex; iCP, inner of alterations in synapse-related and immune-related
cortical plate; iSVZ, inner subventricular zone; ITC, inferior temporal cortex; OFC, orbital genes may derive from changes in cell representation
frontal cortex; oSVZ, outer subventricular zone; PFC, prefrontal cortex; URL, upper rhombic rather than from large alterations in gene expression
lip. Some of the brain images in parts a,b and d are adapted with permission from ref.199. within given cell types162–164.
Overall, studies that compare the transcriptome of
ASD and neurotypical brains have repeatedly noted that
PsychENCODE Finally, these observations of convergent expression there may be altered cortical patterning and changes in
A consortium-based project of ASD risk genes in excitatory and inhibitory neurons relative abundance and/or function of excitatory neurons
that aims to produce
and the developing cortex have also been reproduced and microglia (Fig. 2d). As noted above, these findings are
multidimensional genomic
data from human post-mortem
in analyses of expression data from wild type model largely congruent with data from studies searching for
brain tissue from neurotypical organisms. In one of the earliest studies, analysis of layer- convergence of ASD risk gene expression in neurotypical
and patient donors to begin specific expression data from developing mouse brains brain tissue, in that both implicate the frontal cortex and
to functionally characterize supported the idea that ASD-associated genetic risk excitatory neurons in ASD risk. However, they differ in
risk variants in model systems,
with an initial focus on autism
converges in developing deep layer cortical excitatory that neurotypical data implicate inhibitory neurons but
spectrum disorder, bipolar neurons25. A contemporaneous study leveraging gene not microglia23,25,54,126,129,146,148,151,153,157,159 (Fig. 2c). There
disorder and schizophrenia. expression data from adult primate cortex160 observed are multiple potential explanations for these divergent
0123456789();:
Reviews
findings, including that they may reflect some of the phenotypes (such as macrocephaly), the diverse genetic
inherent practical limitations of studies involving com- backgrounds of probands and controls, and the vari
parisons of affected versus unaffected post-mortem ability in derivation of iPSC lines and downstream
tissue. For example, brain bank resources from indi- cell types, it is not yet clear whether these findings will
viduals with ASD remain quite limited and, therefore, replicate in larger, better-powered analyses.
most studies involve very small sample sizes and have Transcriptomic and epigenetic profiling of
attendant challenges with statistical power. Additionally, iPSC-derived telencephalic organoids has identified
those samples that are available are often highly hetero- ASD-associated co- regulated gene modules related
geneous in characteristics such as age and sex, as well as to synaptic function that overlap with those pre
with regard to important covariates such as medication viously identified in studies using post-mortem brain
history, post-mortem interval, agonal events and/or time tissue146,163,178. Recently, organoid models of developing
of collection165,168–170. Moreover, as the diagnosis of ASD forebrain generated from individuals with ASD and
is based on behaviour, these studies involve postnatal from neurotypical controls were profiled for chromatin
samples only, generally from patients 2 years of age or accessibility and gene expression over time179. Overall,
older, raising the critical issue of how to relate postna- ASD genetic risk mapped to glial progenitor cells,
tal changes in cases versus controls to pathobiology of mid-stage excitatory neurons and late-stage excitatory
a developmental disorder with strong evidence for a and inhibitory neurons. These results are therefore
nexus of prenatal risk, such as ASD. Given the natural largely consistent with the aforementioned findings
history, with symptoms emerging typically in the first from human brain. It is important to note, however,
1–2 years of life, as well as the evidence for the relevance that numerous distinctions exist between organoids and
of prenatal events in syndrome aetiology, it is plausible primary tissue, including limited spatial organization,
and, indeed, likely that observations in postnatal tissue impaired cell-type maturation, the absence of cell types
may capture events downstream of primary patho found in primary tissue and the fact that organoids may
biology and may be offering insights into ‘effect’ rather be significantly impacted by factors related to oxygen
than ‘cause’. Although it would be ideal to perform dif- diffusion and stress180–182. As organoid technology con-
ferential expression analyses in prenatal affected versus tinues to develop (for example, see refs183–185), it will be
neurotypical tissue, these types of experiments pose an increasingly promising avenue for a more granular
formidable logistical challenges, including as a result of understanding of the molecular-level and cell type-level
ASD being a clinical and behavioural diagnosis that can changes taking place during early neurodevelopment,
only formally be made postnatally. Although one might including with respect to ASD.
consider experiments leveraging prenatal post-mortem
tissue carrying large-effect ASD-associated mutations Parallelized modelling of genetic risk. The efforts to iden-
without relying on clinical diagnosis, the observation of tify where and when ASD pathology arises have already
the wide range of potential diagnostic and phenotypic provided important insights into underlying biology
outcomes from even the largest-effect risk genes would and now offer opportunities for parallel investigations of
place limits on the ability to interpret findings with ASD risk genes that constrain or define key experimen-
respect to ASD pathology specifically. tal parameters including cell types and developmental
Induced pluripotent
stem cell stages of interest. In addition, in recent years, advances
(iPSC). A pluripotent cell Patient-derived model systems. Functional and tran- in CRISPR-based technology have facilitated the paral-
obtained by reprogramming scriptional profiling has also been conducted on human lelized study of multiple ASD-associated genes in vitro
somatic cells through ectopic induced pluripotent stem cell (iPSC)-derived monolayer and in vivo to identify points of functional convergence.
expression of defined
pluripotency factors and/or
models of cell types from the developing brain as well as Initial in vitro studies have characterized the effect of
treatment with small in 3D models of human brain development (also known perturbing roughly a dozen ASD-associated risk genes
molecules. as organoids and spheroids)171. Over the past few years, in neurons derived from human iPSCs or embryonic
there have been several studies comparing neurons neuronal precursor cells186,187. These initial studies sug-
Organoids
generated from iPSCs derived from donors with ASD gest that perturbation of ASD genes results in impaired
(Also known as spheroids)
3D structures, derived versus those from neurotypical individuals. Some recur- neuronal differentiation and reduced spontaneous syn-
in vitro from primary tissue, rent findings in iPSC-derived neurons from patients aptic activity, and highlight the feasibility of assessing
embryonic stem cells or with ASD and macrocephaly, compared with controls, the functional consequence of multiple ASD genes in
induced pluripotent stem cells include increased cellular proliferation, accelerated parallel. Going forward, advances in pooled CRISPR
(iPSCs), that self-organize and
recapitulate aspects of organ
neuronal differentiation, impaired synapse development screening will enable hypothesis-n aive functional
development, anatomy, cellular and decreased spontaneous and synchronous neuronal screening of larger numbers of genes with a wide range
composition, physiology and activity172–174. Interestingly, a recent study showed that of read-outs, including proliferation, differentiation
function. iPSCs derived from individuals with ASD but without and cellular function as well as gene expression and
macrocephaly exhibited impairments in neurogenesis chromatin accessibility at single-cell resolution188–195.
Gene modules
Typically, a set of genes with compared with those from neurotypical individuals175. Patient-derived iPSC model systems still tend to suffer
similar expression profiles that Thus far, systematic profiling of iPSCs derived from indi- from small sample sizes in addition to variability among
are inferred to be functionally viduals with idiopathic autism have been limited by small iPSC lines and heterogeneity in the derived cell popu-
related and co-regulated, and sample sizes (on the order of ten individuals carrying lations196. In this regard, systems with isogenic or com-
that can be identified through
various module-detection
an ASD diagnosis) and varied timing and methods of paratively homogeneous genetic backgrounds are an
methods (for example, gene profiling cellular phenotypes172–178. Given the aetiological important addition to the experimental armamentarium,
co-expression). heterogeneity of the syndrome, the presence of comorbid as they moderate some of the challenges of interpreting
www.nature.com/nrn
0123456789();:
Reviews
subtle biological effects across varied genetic back- predominance for individuals with autism with normal
grounds39,40,197 — although the generalizability of findings to high IQ tend to be even higher. Similarly, although the
in isogenic backgrounds needs further investigation, as findings have been roughly consistent across studies, it is
does the effect of sex (see below). also clear that these estimates are vulnerable to several
Studying ASD risk genes in vivo in intact biological potential confounds beyond IQ, including ascertainment
systems is critical to model dynamic neurodevelopmental methods and potential diagnostic bias200–202. Still, studies
contexts, circuit-level function and higher-order pheno- that have endeavoured to account for these issues have,
types. However, generating animal models of individual nonetheless, found consistent evidence for male bias in
ASD genes is labour and time-intensive. Consequently, ASD risk95,202,203. One putative avenue of explanation for
historically, studying multiple ASD genes in parallel these findings is the FPE, evidence for which derives
to identify points of functional convergence has been mostly from genetic studies that show enrichment of
challenging. Increasingly, these challenges can be miti- risk in females versus males — that is, female probands
gated through the use of pooled screening approaches tend to carry a relatively increased burden of rare dam-
or using model organisms that allow for scalable arrayed aging de novo and transmitted mutations, including
screening. In a recent study, an in vivo pooled CRISPR– CNVs and sequence variants23,24,29,31,33,44, as well as rare
scRNA-seq system (in vivo Perturb-seq) that combines complete knockouts76. Additionally, females — including
genetic editing in the developing mouse brain with post- unaffected mothers of children with autism — also
natal scRNA-seq was used to identify the cell type-specific appear to be enriched for common variant risk93,94.
functional consequences of perturbing 35 ASD risk genes Similarly, unaffected female siblings of children diag-
during brain development39. The authors observed over- nosed with ASD appear to carry an increased rate of
lapping gene expression changes linked to perturbation rare variants22–24,27,28,30,32,33,95 and appear to inherit more
of small subsets of ASD genes (up to four) in different cell polygenic risk93,94, as compared with male siblings. These
types, hinting at cell type-specific functional convergence. observations support the hypothesis that females are
Importantly, this study highlights the feasibility of pooled more resilient to genetic risks, resulting in the observation
in vivo CRISPR screens with single-cell profiling to of higher thresholds for females to become symptomatic
interrogate many risk genes for functional convergence and meet diagnostic criteria. Interestingly, to date, there
across multiple cell types. However, there were no clear has been no evidence of sex-biased risk for individual
convergent findings involving large numbers of genes. autosomal genes24 and rare genetic risks mapping to the
Larger-scale experiments with greater numbers of cells sex chromosomes have been relatively unexplored owing
and multiple time points across brain development may to technical and methodological barriers.
be better powered to identify convergent gene modules In contrast to the types of data noted above, family-
within specific cell types. based and twin-based studies of recurrence rates have
Our group recently leveraged CRISPR alongside provided less clear evidence for the FPE. Younger
the X. tropicalis model system to study the effects of siblings of females diagnosed with ASD may be more
loss-of-function mutations in ten ASD risk genes in likely to receive an ASD diagnosis than younger sib-
parallel in an arrayed format40. Strikingly, perturbation lings of males diagnosed with ASD94,204; however, there
of all ten ASD risk genes altered telencephalic size and are conflicting data in this regard and these analyses are
impaired neurogenesis. This convergent neurogenesis constrained by limited statistical power owing to the
phenotype, spanning ASD genes with apparently dis- relatively small female cohort sizes95,153,203.
parate cellular functions, was conserved in human 2D There is also the possibility of male-specific ‘risk’,
and 3D in vitro models of brain development. Overall, which is not mutually exclusive with the FPE model.
these observations are in line with prior findings that For example, males with ASD have an increased rate of
ASD risk converges in mid-gestational telencephalic hemizygous rare loss-of-function variants on chromo-
excitatory neurons, and provide strong evidence for some X76. However, the sex chromosomes, in general,
impaired neurogenesis as a point of vulnerability, as has appear to be a minor factor in determining risk versus
been previously hypothesized198. These findings under- resilience31,205,206, although they also tend to be omitted
score the power of Xenopus for productive functional from large-scale genetic studies owing to technical diffi-
Functional convergence genomics studies, in particular highlighting the ability of culties and complexities in accurate variant calling due to
A situation in which seemingly medium-throughput parallelized in vivo studies to iden- confounding sequence homology between sex chromo
disparate entities are tify convergent mechanisms that can then be prioritized somes and differences in ploidy by sex. Therefore, it is
associated with an overlapping
function, which can occur at
for validation and extension in human systems199. possible that there are as yet undiscovered factors on the
various levels of investigation, sex chromosomes.
including molecular, cell Sex differences and genetic risk These findings suggest that sex differences are a critical
taxonomic, morphological, Although, in general, the field is enjoying success in aspect of the biology of ASD and a potentially promising
neural circuit or phenotype
elaborating points of functional and spatiotemporal avenue for the development of broadly applicable thera-
level.
convergence among a rapidly expanding list of auto- peutics, especially in light of the possibility that unravel-
Sex differences somal ASD genetic risks, far less is understood about ling underlying mechanisms may illuminate resilience as
Differences between individuals the nature of the contribution of sex to ASD aetiology. well as risk factors. The lack of mechanistic understand-
of different sex in the same A strong male sex bias has consistently been observed ing of sex bias is probably a consequence of the relative
species, often including
secondary sex characteristics,
for ASD, but this remains largely unexplained200. The paucity of knowledge regarding differential molecular
size and behavioural or male to female ratio has been consistently estimated and cellular mechanisms in male and female brain devel-
cognitive traits. to be around 4:1 (ref.201), although estimates of male opment, inherent limitations in commonly employed
0123456789();:
Reviews
Table 1 | Available human developmental brain expression data sets for studying sex differences
resource Brain regions Developmental sample size Technology Number Year
age (female, male) of cells
Colantuoni et al.218 PFC 12–18 PCW, 269 (92, 177) Microarray NA 2011
birth–78 years
BrainSpan 16 (12 cortical) 5 PCW–82 years 57 (26, 31) Microarray NA 2011
microarray145
BrainSpan LMD154 ~300 cortical regions 15–21 PCW 4 (3, 1) Microarray NA 2014
(LMD of 6 layers)
BrainSpan RNA-seqa 16 (12 cortical) 8 PCW–40 years 42 (19, 23) Bulk RNA-seq NA 2013
O’Brien et al.219 Entire brain hemisphere 12–19 PCW 120 (50, 70) Bulk RNA-seq NA 2018
Li et al. 16 regions (11 cortical) 8 PCW–40 years 41 (18, 23) Bulk RNA-seq NA 2018
(PsychENCODE)151
Walker et al.152 Unspecified cortex 14–21 PCW 201 (84, 117) Bulk RNA-seq NA 2019
Werling et al. 1 region (DPFC) 6 PCW–20 years 176 (72, 104) Bulk RNA-seq NA 2020
(BrainVar)153
Pollen et al.155 Unspecified cortex 14–16 PCW 3b (3, 0) scRNA-seq 393 2015
(MD of 2 layers)
Nowakowski et al.156 2 cortical regions 5–37 PCW 48b (21, 27) scRNA-seq 4,261 2017
(MD of 2 layers), MGE
Zhong et al.220 PFC 6–24 PCW 12b (9, 3) scRNA-seq 2,309 2018
Fan et al.159 22 regions (20 cortical) 22–23 PCW 3 (2, 1) scRNA-seq 4,213 2018
Li et al. Cortex (fronto-parietal) 5–20 PCW 9 (3, 6) scRNA-seq 1,195 2018
(PsychENCODE)151
Polioudakis et al.157 Unspecified cortex 15–16 PCW 4 (3, 1) scRNA-seq 33,976 2019
(MD of 2 groups of
cortical layers)
Bhaduri et al.182 7 regions (all cortical) 4–20 PCW 5b (4, 1) scRNA-seq 189,409 2020
Eze et al.158 18 regions (3 cortical) 4–8 PCW 10b (7 , 3) scRNA-seq 289,000 2021
Trevino et al.221 Unspecified cortex 16–24 PCW 4b (3, 1) scRNA-seq 57 ,868 2021
DPFC, dorsolateral prefrontal cortex; LMD, laser micro-dissection; MD, micro-dissection; MGE, medial ganglionic eminence;
NA, not applicable; PCW, post-conception weeks; PFC, prefrontal cortex; RNA-seq, RNA sequencing; scRNA-seq, single-cell RNA
sequencing. aBrainSpan RNA-seq data can be accessed at http://brainspan.org/static/download.html. bSex inferred by X-inactive
specific transcript expression.
model systems for interrogating these differences and characterizing developmental differences in sex hor-
the lack of any immediately obvious explanation for mone levels in the brain and whether they result in sex
sex bias in otherwise highly successful gene discovery differences in brain development and function. With the
efforts. Identifying robust differences in ‘neurotypical’ recent discovery that oestrogen inhibits sonic hedgehog
male and female brain development will probably facili- (SHH) signalling endogenously in the developing
tate understanding of risk for many psychiatric disorders brain40, critical future areas of research will include
beyond ASD, as many show some degree of sex bias207. determining whether SHH signalling is sexually dimor-
Rich developmental human brain omics data sets from phic during prenatal development and whether it is a
both males and females (see Table 1 for a list of currently convergent pathway in ASD risk. SHH has key roles
available gene expression data) will be a critical resource throughout brain development, spanning patterning,
in efforts to identify these differences. proliferation, arealization and differentiation 213–215,
In considering points of intersection between ASD suggesting that there may be many points of potential
risk factors and biological sex differences, there have contribution to sexually dimorphic outcomes just within
been some recent data pointing to the potential contri- this single pathway.
bution of the sex hormones, testosterone and oestrogen,
to ASD sex bias95,208–210. Indeed, two independent in vivo Perspective
resilience screens identified oestrogen as a protective We are at a highly opportune moment in the effort to
factor in the context of modelled ASD genetic risks40,45, identify core pathobiological processes underlying
implying that oestrogen could be an endogenous pro- ASD. There is a long and growing list of reliably asso-
tective factor if it is active at higher levels in females ciated large-effect genes accompanied by an expanding
than in males. However, testosterone excess has also armamentarium of multidimensional omics data sets,
been hypothesized to be a risk factor for ASD209,211,212. higher-throughput model systems and CRISPR-based
Further research into the precise manner by which sex genetic tools that are empowering studies focused on
hormones intersect with ASD risk is needed, especially convergence. Although Gene Ontology enrichment
www.nature.com/nrn
0123456789();:
Reviews
analyses highlighting chromatin modification and syn- forms of syndromic ASD, these initiatives are gain-
apse function have been a readily accessible resource for ing momentum and clinical trials are on the horizon.
developing evidence of shared biology, newer hypothesis- Similar efforts for the most highly penetrant high-
free system biological approaches are important comple- confidence ASD genes are unlikely to be far behind.
ments to these types of studies and have already provided The pursuit of convergence neuroscience approaches
strong evidence for a nexus of ASD risk, particularly in can contribute to these and related efforts, in part, by
developing excitatory neurons in the mid-gestational clarifying when and where vulnerability to ASD risk
frontal cortex. emerges, and therefore can help answer key remaining
Of course, critical challenges remain. First, it is not questions regarding when interventions addressing
clear how to span the gap from molecular and cellu- ASD would need to be delivered and where in the
lar mechanisms of convergence to differences in the brain37,38. At the same time, points of convergence,
clinical presentation of ASD. Animal models provide with regard to risk and resilience, have the potential to
an opportunity to observe and manipulate behaviours, support more traditional therapeutics development by
as well as the ability to identify underlying circuit-level identifying biological mechanisms and targets that are
changes. However, the limitations of anthropomorphi ‘downstream’ of individual genes and applicable to more
zing and relying on face-valid behaviours for complex than one gene at a time. These points of convergence
neuropsychiatric phenotypes are well known35. Second, may also have diagnostic utility. Also, there are critical
despite the long-observed male predominance in ASD, practical and ethical issues that still confront the design
the field still lacks a detailed understanding of how and execution of clinical trials in ASD, particularly
this phenomenon is generated and therefore, going early in development. Nonetheless, progress in other
forward, it will be critical to investigate sexual dimor- areas of medicine, including with regard to early-onset
phism in brain development and to intersect this with neurodegenerative disorders216,217, combined with the
genetic risk. progress in multiple areas noted above points to several
Despite these challenges, there are already ongoing viable paths towards more effective, personalized and
efforts to exploit successful gene discovery directly to rational treatments for ASD, particularly for individuals
develop novel therapies35. These include gene replace- with severe autism.
ment, gene editing and antisense oligonucleotide strat-
egies. Particularly with regard to the most common Published online xx xx xxxx
1. Lord, C. et al. Autism spectrum disorder. Nat. Rev. Dis. analysis of twin studies. J. Child. Psychol. Psychiatry pathogenesis of autism. Cell 155, 997–1007
Prim. 6, 5 (2020). 57, 585–595 (2016). (2013).
2. Díaz-Caneja, C. M. et al. A white paper on a 16. Ritvo, E. R., Freeman, B. J., Mason-Brothers, A., Mo, A. One of the first studies to assess spatiotemporal
neurodevelopmental framework for drug discovery & Ritvo, A. M. Concordance for the syndrome of autism convergence of ASD risk genes, this study performs
in autism and other neurodevelopmental disorders. in 40 pairs of afflicted twins. Am. J. Psychiatry 142, a co-expression network analysis that integrates
Eur. Neuropsychopharmacol. 48, 49–88 74–77 (1985). BrainSpan developmental expression data with
(2021). 17. Taniai, H., Nishiyama, T., Miyachi, T., Imaeda, M. genetic data from simplex families with ASD to
3. Folstein, S. & Rutter, M. Infantile autism: a genetic & Sumi, S. Genetic influences on the broad spectrum identify deep layer cortical projection neurons
study of 21 twin pairs. J. Child. Psychol. Psychiatry of autism: study of proband-ascertained twins. in prefrontal and primary motor-somatosensory
18, 297–321 (1977). Am. J. Med. Genet. B Neuropsychiatr. Genet. 147B, cortical regions during mid-gestational
4. Gaugler, T. et al. Most genetic risk for autism resides 844–849 (2008). development as a critical nexus of ASD risk.
with common variation. Nat. Genet. 46, 881–885 18. Lichtenstein, P., Carlström, E., Råstam, M., Gillberg, C. 26. O’Roak, B. J. et al. Exome sequencing in sporadic
(2014). & Anckarsäter, H. The genetics of autism spectrum autism spectrum disorders identifies severe de novo
5. Klei, L. et al. Common genetic variants, acting disorders and related neuropsychiatric disorders mutations. Nat. Genet. 43, 585–589 (2011).
additively, are a major source of risk for autism. in childhood. Am. J. Psychiatry 167, 1357–1363 One of the first WES studies of people with ASD
Mol. Autism 3, 9 (2012). (2010). and their families.
6. Cross-Disorder Group of the Psychiatric Genomics 19. Nordenbæk, C., Jørgensen, M., Kyvik, K. O. & 27. Neale, B. M. et al. Patterns and rates of exonic
Consortium. et al. Genetic relationship between five Bilenberg, N. A Danish population-based twin study de novo mutations in autism spectrum disorders.
psychiatric disorders estimated from genome-wide on autism spectrum disorders. Eur. Child. Adolesc. Nature 485, 242–245 (2012).
SNPs. Nat. Genet. 45, 984–994 (2013). Psychiatry 23, 35–43 (2014). Together with refs. 28, 30 and 34, this paper
7. Bailey, A. et al. Autism as a strongly genetic disorder: 20. Colvert, E. et al. Heritability of autism spectrum cemented the contribution of de novo sequence
evidence from a British twin study. Psychol. Med. 25, disorder in a UK population-based twin sample. variants to ASD risk.
63–77 (1995). JAMA Psychiatry 72, 415–423 (2015). 28. Iossifov, I. et al. De novo gene disruptions in children
8. Devlin, B. & Scherer, S. W. Genetic architecture in 21. Sebat, J. et al. Strong association of de novo on the autistic spectrum. Neuron 74, 285–299 (2012).
autism spectrum disorder. Curr. Opin. Genet. Dev. 22, copy number mutations with autism. Science 316, 29. De Rubeis, S. et al. Synaptic, transcriptional and
229–237 (2012). 445–449 (2007). chromatin genes disrupted in autism. Nature 515,
9. Hallmayer, J. et al. Genetic heritability and shared This study conducts comparative genomic 209–215 (2014).
environmental factors among twin pairs with autism. hybridization in simplex and multiplex families This large ASD WES study, contemporaneous with
Arch. Gen. Psychiatry 68, 1095–1102 (2011). with ASD and establishes that de novo CNVs ref. 44, confirms that de novo loss-of-function
10. Steffenburg, S. et al. A twin study of autism in significantly contribute to ASD risk. mutations contribute significantly to ASD, identifies
Denmark, Finland, Iceland, Norway and Sweden. 22. Levy, D. et al. Rare de novo and transmitted 33 ASD risk genes and implicates ASD risk genes
J. Child. Psychol. Psychiatry 30, 405–416 copy-number variation in autistic spectrum in chromatin remodelling, transcription and splicing,
(1989). disorders. Neuron 70, 886–897 (2011). and synaptic function.
11. Ozonoff, S. et al. Recurrence risk for autism spectrum 23. Satterstrom, F. K. et al. Large-scale exome sequencing 30. Sanders, S. J. et al. De novo mutations revealed
disorders: a Baby Siblings Research Consortium study. study implicates both developmental and functional by whole-exome sequencing are strongly associated
Pediatrics 128, 488–495 (2011). changes in the neurobiology of autism. Cell 180, with autism. Nature 485, 237–241 (2012).
12. Constantino, J. N., Zhang, Y., Frazier, T., 568–584 (2020). As well as contributing as noted above, this
Abbacchi, A. M. & Law, P. Sibling recurrence This work presents the largest published ASD WES paper outlined a key paradigm for the statistical
and the genetic epidemiology of autism. Am. J. study to date, which identifies 102 high-confidence assessment of a gene’s association to ASD, based
Psychiatry 167, 1349–1356 (2010). ASD risk genes. on the number and type of recurrent de novo
13. Rosenberg, R. E. et al. Characteristics and 24. Sanders, S. J. et al. Insights into autism spectrum variants.
concordance of autism spectrum disorders among disorder genomic architecture and biology from 31. Jacquemont, S. et al. A higher mutational burden
277 twin pairs. Arch. Pediatr. Adolesc. Med. 163, 71 risk loci. Neuron 87, 1215–1233 (2015). in females supports a ‘female protective model’ in
907–914 (2009). This large study combines WES data from neurodevelopmental disorders. Am. J. Hum. Genet.
14. Le Couteur, A. et al. A broader phenotype of autism: De Rubeis et al. (2014) and Iossifov et al. (2014) 94, 415–425 (2014).
the clinical spectrum in twins. J. Child. Psychol. as well as de novo CNV data to identify 65 ASD 32. Dong, S. et al. De novo insertions and deletions of
Psychiatry 37, 785–801 (1996). risk genes and 6 ASD risk loci. predominantly paternal origin are associated with
15. Tick, B., Bolton, P., Happé, F., Rutter, M. & Rijsdijk, F. 25. Willsey, A. J. et al. Coexpression networks implicate autism spectrum disorder. Cell Rep. 9, 16–23
Heritability of autism spectrum disorders: a meta- human midfetal deep cortical projection neurons in the (2014).
0123456789();:
Reviews
33. Sanders, S. J. et al. Multiple recurrent de novo 45. Hoffman, E. J. et al. Estrogens suppress a behavioral 61. Huang, N., Lee, I., Marcotte, E. M. & Hurles, M. E.
CNVs, including duplications of the 7q11.23 Williams phenotype in zebrafish mutants of the autism risk Characterising and predicting haploinsufficiency
syndrome region, are strongly associated with autism. gene, CNTNAP2. Neuron 89, 725–733 (2016). in the human genome. PLoS Genet. 6, e1001154
Neuron 70, 863–885 (2011). This is one of the first studies to identify a (2010).
One of several early studies associating de novo functional interaction between oestrogen and This work develops a paradigm to prioritize
CNVs with ASD risk, this paper also generated an ASD risk gene. genes based on likelihood of haploinsufficiency
a statistical framework for quantifying the 46. Jamain, S. et al. Mutations of the X-linked genes (i.e. likelihood that a heterozygous loss of
significance of recurrent de novo CNVs. This encoding neuroligins NLGN3 and NLGN4 are associated function variant results in a clinical phenotype).
framework was later adapted to early studies with autism. Nat. Genet. 34, 27–29 (2003). Haploinsufficient genes tend to carry higher
of de novo sequence variants. This work uses linkage analysis and targeted risk for ASD and other human disorders, and is
34. O’Roak, B. J. et al. Sporadic autism exomes reveal DNA sequencing to identify two ASD-associated therefore a key metric for prioritizing functional
a highly interconnected protein network of de novo genes, NLGN3 and NLGN4, which are among the mutations.
mutations. Nature 485, 246–250 (2012). earliest individual genes with robust evidence of 62. Adzhubei, I. A. et al. A method and server
As well as contributing as noted above, this paper association with idiopathic (non-syndromic) ASD. for predicting damaging missense mutations.
was one of the first to demonstrate that genes with 47. Laumonnier, F. et al. X-Linked mental retardation and Nat. Methods 7, 248–249 (2010).
damaging mutations in individuals with ASD are autism are associated with a mutation in the NLGN4 A longstanding and widely used resource for
highly connected in protein–protein interaction gene, a member of the neuroligin family. Am. J. Hum. predicting the severity of missense mutations.
modules, suggesting that they may converge on Genet. 74, 552–557 (2004). 63. Kosmicki, J. A. et al. Refining the role of de novo
similar biological functions. 48. Kumar, R. A. et al. Recurrent 16p11.2 microdeletions protein-truncating variants in neurodevelopmental
35. Sestan, N. & State, M. W. Lost in translation: in autism. Hum. Mol. Genet. 17, 628–638 (2008). disorders by using population reference samples.
traversing the complex path from genomics to 49. Marshall, C. R. et al. Structural variation of Nat. Genet. 49, 504–510 (2017).
therapeutics in autism spectrum disorder. Neuron chromosomes in autism spectrum disorder. 64. Samocha, K. E. et al. A framework for the interpretation
100, 406–423 (2018). Am. J. Hum. Genet. 82, 477–488 (2008). of de novo mutation in human disease. Nat. Genet. 46,
36. State, M. W. & Sestan, N. The emerging biology of 50. Weiss, L. A. et al. Association between microdeletion 944–950 (2014).
autism spectrum disorders. Science 337, 1301–1303 and microduplication at 16p11.2 and autism. N. Engl. This study presents a frequentist framework to
(2012). J. Med. 358, 667–675 (2008). evaluate excesses of de novo mutations at the level
This review describes the rationale for and 51. Pinto, D. et al. Functional impact of global rare of individual genes, for which simple case–control
approaches to leveraging convergence among ASD copy number variation in autism spectrum disorders. comparisons cannot achieve meaningful levels of
genes to identify not only relevant molecules and Nature 466, 368–372 (2010). significance due to the rarity of de novo events;
pathways but also when and where they act in the 52. SPARK Consortium. SPARK: a US cohort of 50,000 when applied to ASD WES data, this method
developing human brain. families to accelerate autism research. Neuron 97, highlights several risk genes at a genome-wide
37. Willsey, A. J. et al. The psychiatric cell map initiative: 488–493 (2018). level of statistical significance.
a convergent systems biological approach to illuminating 53. He, X. et al. Integrated model of de novo and inherited 65. Samocha, K. E. et al. Regional missense constraint
key molecular pathways in neuropsychiatric disorders. genetic variants yields greater power to identify risk improves variant deleteriousness prediction.
Cell 174, 505–520 (2018). genes. PLoS Genet. 9, e1003671 (2013). Preprint at bioRxiv https://doi.org/10.1101/148353
This review describes convergent system This work develops a Bayesian model (transmission (2017).
biological approaches to study genes implicated and de novo association (TADA)) that integrates 66. Singh, T. et al. Rare coding variants in ten genes confer
in neuropsychiatric disorders such as ASD, information from multiple types of genetic variation substantial risk for schizophrenia. Nature https://
emphasizing the importance of parallel from large-scale human gene sequencing studies doi.org/10.1038/s41586-022-04556-w (2022).
investigations of multiple genes and mutations, to improve the power to detect novel risk genes. The largest WES study published on schizophrenia
in as unbiased a fashion as possible, to identify A variation of this model is still used to discover to date, this paper identifies 10 exome-wide
points of convergence and reveal high-order ASD risk genes. This model has also been applied significant schizophrenia risk genes of large
(molecular, cellular and circuit-level) phenotypes. to other psychiatric disorders, such as Tourette effect size.
38. Willsey, A. J. & State, M. W. Autism spectrum syndrome. 67. Bilgüvar, K. et al. Whole-exome sequencing identifies
disorders: from genes to neurobiology. Curr. Opin. 54. Ruzzo, E. K. et al. Inherited and de novo genetic recessive WDR62 mutations in severe brain
Neurobiol. 30, 92–99 (2015). risk for autism impacts shared networks. Cell 178, malformations. Nature 467, 207–210 (2010).
39. Jin, X. et al. In vivo Perturb-seq reveals neuronal and 850–866.e26 (2019). 68. Gilmore, E. C. & Walsh, C. A. Genetic causes of
glial abnormalities associated with autism risk genes. 55. Satterstrom, F. K. et al. Autism spectrum disorder microcephaly and lessons for neuronal development.
Science 370, eaaz6063 (2020). and attention deficit hyperactivity disorder have a Wiley Interdiscip. Rev. Dev. Biol. 2, 461–478 (2013).
This paper conducted in vivo transcriptional similar burden of rare protein-truncating variants. 69. Dias, C. M. et al. Homozygous missense variants in
screening of a large number of ASD risk genes Nat. Neurosci. 22, 1961–1965 (2019). NTNG2, encoding a presynaptic Netrin-G2 adhesion
using single-cell transcriptomics and CRISPR 56. Rodin, R. E. et al. The landscape of somatic mutation protein, lead to a distinct neurodevelopmental
technology. in cerebral cortex of autistic and neurotypical disorder. Am. J. Hum. Genet. 105, 1048–1056
40. Willsey, H. R. et al. Parallel in vivo analysis of large- individuals revealed by ultra-deep whole-genome (2019).
effect autism genes implicates cortical neurogenesis sequencing. Nat. Neurosci. 24, 176–185 (2021). 70. Strauss, K. A. et al. Recessive symptomatic focal
and estrogen in risk and resilience. Neuron 109, 1409 This is one of the largest studies to date assessing epilepsy and mutant contactin-associated protein-like 2.
(2021). the frequency of somatic (mosaic) variants in brain N. Engl. J. Med. 354, 1370–1377 (2006).
This study conducts parallel in vivo analysis of the tissue from people with ASD and controls. 71. Novarino, G. et al. Mutations in BCKD-kinase lead
ten highest-confidence ASD risk genes in Xenopus, 57. Sherman, M. A. et al. Large mosaic copy number to a potentially treatable form of autism with epilepsy.
finding that the functional consequences of variations confer autism risk. Nat. Neurosci. 24, Science 338, 394–397 (2012).
mutations in these genes converge upon impaired 197–203 (2021). 72. Morrow, E. M. et al. Identifying autism loci and
neurogenesis in the developing forebrain — a 58. Karczewski, K. J. et al. The mutational constraint genes by tracing recent shared ancestry. Science 321,
phenotype that could be rescued by exogenous spectrum quantified from variation in 141,456 218–223 (2008).
oestrogen, suggesting oestrogen is a resilience humans. Nature 581, 434–443 (2020). 73. Manzini, M. C. et al. CC2D1A regulates human
factor that may mitigate various ASD genetic risks. 59. Lek, M. et al. Analysis of protein-coding genetic intellectual and social function as well as NF-κB
41. Zoghbi, H. Y. Postnatal neurodevelopmental disorders: variation in 60,706 humans. Nature 536, 285–291 signaling homeostasis. Cell Rep. 8, 647–655
meeting at the synapse? Science 302, 826–830 (2016). (2014).
(2003). This study aggregates and analyses high-quality 74. Yu, T. W. et al. Using whole-exome sequencing to
This early commentary proposes that ASD genetic WES data for 60,706 individuals of diverse identify inherited causes of autism. Neuron 77,
risk may mechanistically converge at the synapse. ancestries generated as part of the Exome 259–273 (2013).
42. O’Roak, B. J. et al. Multiplex targeted sequencing Aggregation Consortium (ExAC), providing a key 75. Doan, R. N. et al. Recessive gene disruptions in autism
identifies recurrently mutated genes in autism foundation for the field (which has since grown to spectrum disorder. Nat. Genet. 51, 1092–1098
spectrum disorders. Science 338, 1619–1622 (2012). include more than 100,000 individuals). Together (2019).
43. Gilman, S. R. et al. Rare de novo variants associated with several other studies, this work develops 76. Lim, E. T. et al. Rare complete knockouts in humans:
with autism implicate a large functional network of a paradigm to prioritize genes on the basis of population distribution and significant role in autism
genes involved in formation and function of synapses. ‘tolerance’ to genetic variation. Intolerant genes spectrum disorders. Neuron 77, 235–242 (2013).
Neuron 70, 898–907 (2011). tend to carry higher risk for ASD and other human This paper characterizes the rate of homozygous
This study develops a framework (NETBAG) that disorders, so tolerance is a key metric for knockout mutations in ASD and is one of the first
integrates multiple levels of molecular data to prioritizing functional mutations. The associated studies to characterize the contribution of rare
demonstrate that genes affected by CNVs in resource also facilitates the identification of ultra- mutations on chromosome X to ASD risk.
ASD are functionally interconnected. rare genetic variants — a similarly critical metric 77. Schmitz-Abe, K. et al. Homozygous deletions implicate
44. Iossifov, I. et al. The contribution of de novo coding for prioritizing mutations. non-coding epigenetic marks in autism spectrum
mutations to autism spectrum disorder. Nature 515, 60. Petrovski, S., Wang, Q., Heinzen, E. L., Allen, A. S. disorder. Sci. Rep. 10, 14045 (2020).
216–221 (2014). & Goldstein, D. B. Genic intolerance to functional 78. D’Gama, A. M. et al. Targeted DNA sequencing from
This large ASD WES study, contemporaneous with variation and the interpretation of personal genomes. autism spectrum disorder brains implicates multiple
ref. 29, confirms that de novo missense and likely PLoS Genet. 9, e1003709 (2013). genetic mechanisms. Neuron 88, 910–917 (2015).
gene-disrupting mutations contribute significantly Together with several other studies, this work 79. Freed, D. & Pevsner, J. The contribution of mosaic
to ASD and identifies 27 ASD risk genes. It also develops a paradigm to prioritize genes based on variants to autism spectrum disorder. PLoS Genet. 12,
carefully quantifies the contributions of different “tolerance” to genetic variation. Intolerant genes e1006245 (2016).
types of rare variants to ASD risk by comparing tend to carry higher risk for ASD and other human 80. Krupp, D. R. et al. Exonic mosaic mutations contribute
normalized rates in probands versus unaffected disorders, and is therefore a key metric for risk for autism spectrum disorder. Am. J. Hum. Genet.
sibling controls. prioritizing functional mutations. 101, 369–390 (2017).
www.nature.com/nrn
0123456789();:
Reviews
81. Lim, E. T. et al. Rates, distribution and implications 103. Abrahams, B. S. et al. SFARI Gene 2.0: a community- genetic homogeneity? Biol. Psychiatry 77, 775–784
of postzygotic mosaic mutations in autism spectrum driven knowledgebase for the autism spectrum (2015).
disorder. Nat. Neurosci. 20, 1217–1224 (2017). disorders (ASDs). Mol. Autism 4, 36 (2013). 128. Huang, J. K. et al. Systematic evaluation of molecular
82. Dou, Y. et al. Postzygotic single-nucleotide mosaicisms 104. Bleuler, E. in Dementia Praecox or the Group of networks for discovery of disease genes. Cell Syst. 6,
contribute to the etiology of autism spectrum disorder Schizophrenias (ed. Zinkin, J.) 548 (International 484–495.e5 (2018).
and autistic traits and the origin of mutations. Universities Press, 1950). 129. Ben-David, E. & Shifman, S. Networks of neuronal
Hum. Mutat. 38, 1002–1013 (2017). 105. Evans, B. How autism became autism: the radical genes affected by common and rare variants in autism
83. D’Gama, A. M. & Walsh, C. A. Somatic mosaicism transformation of a central concept of child spectrum disorders. PLoS Genet. 8, e1002556
and neurodevelopmental disease. Nat. Neurosci. 21, development in Britain. Hist. Hum. Sci. 26, 3–31 (2012).
1504–1514 (2018). (2013). This is one of the first studies to assess
84. Kong, A. et al. Rate of de novo mutations and the 106. Spitzer, R. L., Williams, J. B. & Skodol, A. E. DSM-III: spatiotemporal convergence of ASD risk genes,
importance of father’s age to disease risk. Nature the major achievements and an overview. Am. J. highlighting the prenatal period as well as
488, 471–475 (2012). Psychiatry 137, 151–164 (1980). chromatin regulation as putative factors in ASD
85. Leppa, V. M. et al. Rare inherited and de novo CNVs 107. Thurm, A., Farmer, C., Salzman, E., Lord, C. & pathogenesis.
reveal complex contributions to ASD risk in multiplex Bishop, S. State of the field: differentiating intellectual 130. Trubetskoy, V. et al. Mapping genomic loci implicates
families. Am. J. Hum. Genet. 99, 540–554 (2016). disability from autism spectrum disorder. Front. genes and synaptic biology in schizophrenia. Nature
86. Jin, S. C. et al. Contribution of rare inherited and Psychiatry 10, 526 (2019). https://doi.org/10.1038/s41586-022-04434-5
de novo variants in 2,871 congenital heart disease 108. Brainstorm Consortium. et al. Analysis of shared (2022).
probands. Nat. Genet. 49, 1593–1601 (2017). heritability in common disorders of the brain. Science 131. Durand, C. M. et al. Mutations in the gene encoding
87. Capkova, Z. et al. Differences in the importance 360, eaap8757 (2018). the synaptic scaffolding protein SHANK3 are associated
of microcephaly, dysmorphism, and epilepsy in the 109. Malhotra, D. & Sebat, J. CNVs: harbingers of a rare with autism spectrum disorders. Nat. Genet. 39,
detection of pathogenic CNVs in ID and ASD patients. variant revolution in psychiatric genetics. Cell 148, 25–27 (2007).
PeerJ 7, e7979 (2019). 1223–1241 (2012). 132. Kim, H.-G. et al. Disruption of neurexin 1 associated
88. Robinson, E. B. et al. Autism spectrum disorder 110. Myers, S. M. et al. Insufficient evidence for ‘autism- with autism spectrum disorder. Am. J. Hum. Genet.
severity reflects the average contribution of de novo specific’ genes. Am. J. Hum. Genet. 106, 587–595 82, 199–207 (2008).
and familial influences. Proc. Natl Acad. Sci. USA 111, (2020). 133. Bourgeron, T. A synaptic trek to autism. Curr. Opin.
15161–15165 (2014). 111. Buxbaum, J. D. et al. Not all autism genes are created Neurobiol. 19, 231–234 (2009).
89. Guo, H. et al. Genome sequencing identifies multiple equal: a response to Myers et al. Am. J. Hum. Genet. 134. Walsh, C. A., Morrow, E. M. & Rubenstein, J. L. R.
deleterious variants in autism patients with more 107, 1000–1003 (2020). Autism and brain development. Cell 135, 396–400
severe phenotypes. Genet. Med. 21, 1611–1620 112. Myers, S. M., Challman, T. D., Martin, C. L. (2008).
(2019). & Ledbetter, D. H. Response to Buxbaum et al. 135. Subramanian, A. et al. Gene set enrichment analysis:
90. Werling, D. M. et al. An analytical framework for Am. J. Hum. Genet. 107, 1004 (2020). a knowledge-based approach for interpreting genome-
whole-genome sequence association studies and its 113. Gudmundsson, O. O. et al. Attention-deficit wide expression profiles. Proc. Natl Acad. Sci. USA
implications for autism spectrum disorder. Nat. Genet. hyperactivity disorder shares copy number variant 102, 15545–15550 (2005).
50, 727–736 (2018). risk with schizophrenia and autism spectrum disorder. 136. Ashburner, M. et al. Gene Ontology: tool for the
This study develops an analytical framework for Transl. Psychiatry 9, 258 (2019). unification of biology. Nat. Genet. 25, 25–29
WGS termed a category-wide association study, 114. Martin, J. et al. Biological overlap of attention-deficit/ (2000).
which mirrors the statistical rigour of GWAS with hyperactivity disorder and autism spectrum disorder: A marker paper for the Gene Ontology Consortium,
annotation categories in place of SNPs. evidence from copy number variants. J. Am. Acad. this manuscript outlines their goal to produce a
91. Sullivan, P. F. et al. Psychiatric genomics: an update Child. Adolesc. Psychiatry 53, 761–70.e26 (2014). structured, precisely defined, common, controlled
and an agenda. Am. J. Psychiatry 175, 15–27 115. Zarrei, M. et al. A large data resource of genomic vocabulary for describing the roles of genes and
(2018). copy number variation across neurodevelopmental gene products in any organism. These ‘GO terms’
92. Grove, J. et al. Identification of common genetic risk disorders. NPJ Genom. Med. 4, 26 (2019). have formed the foundation of many systems
variants for autism spectrum disorder. Nat. Genet. 51, 116. Williams, N. M. et al. Rare chromosomal deletions and biological analyses conducting GO enrichment
431–444 (2019). duplications in attention-deficit hyperactivity disorder: or gene set enrichment analysis.
This work is the largest ASD GWAS published a genome-wide analysis. Lancet 376, 1401–1408 137. Gene Ontology Consortium. The Gene Ontology
to date, including ~18,000 cases and ~28,000 (2010). resource: enriching a GOld mine. Nucleic Acids Res.
controls, and identifies 5 genome-wide significant 117. Shi, J. et al. Common variants on chromosome 49, D325–D334 (2021).
loci. 6p22.1 are associated with schizophrenia. Nature 138. Mi, H., Muruganujan, A., Ebert, D., Huang, X. &
93. Antaki, D. et al. A phenotypic spectrum of autism is 460, 753–757 (2009). Thomas, P. D. PANTHER version 14: more genomes,
attributable to the combined effects of rare variants, 118. Stefansson, H. et al. Common variants conferring risk a new PANTHER GO-slim and improvements in
polygenic risk and sex. Preprint at medRxiv https:// of schizophrenia. Nature 460, 744–747 (2009). enrichment analysis tools. Nucleic Acids Res. 47,
doi.org/10.1101/2021.03.30.21254657 (2021). 119. International Schizophrenia Consortium. et al. D419–D426 (2019).
94. Wigdor, E. M. et al. The female protective effect Common polygenic variation contributes to risk 139. Chen, J., Bardes, E. E., Aronow, B. J. & Jegga, A. G.
against autism spectrum disorder. Preprint at of schizophrenia and bipolar disorder. Nature 460, ToppGene Suite for gene list enrichment analysis and
medRxiv https://doi.org/10.1101/2021.03.29. 748–752 (2009). candidate gene prioritization. Nucleic Acids Res. 37,
21253866 (2021). 120. Ben-Shalom, R. et al. Opposing effects on Na1.2 W305–W311 (2009).
95. Werling, D. M. The role of sex-differential biology in function underlie differences between SCN2A variants 140. Chen, E. Y. et al. Enrichr: interactive and collaborative
risk for autism spectrum disorder. Biol. Sex. Differ. 7, observed in individuals with autism spectrum disorder HTML5 gene list enrichment analysis tool. BMC
58 (2016). or infantile seizures. Biol. Psychiatry 82, 224–232 Bioinforma. 14, 128 (2013).
This review explores the magnitude of male bias in (2017). 141. Dutkowski, J. et al. A Gene Ontology inferred from
ASD prevalence, describes the FPE and highlights This study identifies a link between putative loss- molecular networks. Nat. Biotechnol. 31, 38–45
sex-differential pathways that may underlie sex of-function and gain-of-function mutations in ASD (2013).
bias in ASD. and epilepsy, respectively, providing one of the 142. Schnoes, A. M., Ream, D. C., Thorman, A. W.,
96. Cross-Disorder Group of the Psychiatric Genomics first successful examples of understanding how Babbitt, P. C. & Friedberg, I. Biases in the experimental
Consortium. Identification of risk loci with shared a single gene (SCN2A) contributes risk to multiple annotations of protein function and their effect on our
effects on five major psychiatric disorders: a genome- disorders. understanding of protein function space. PLoS Comput.
wide analysis. Lancet 381, 1371–1379 (2013). 121. Skuse, D. H. Rethinking the nature of genetic Biol. 9, e1003063 (2013).
A critical study that assesses the common vulnerability to autistic spectrum disorders. 143. Haynes, W. A., Tomczak, A. & Khatri, P. Gene
variant risk shared across ASD, attention deficit- Trends Genet. 23, 387–395 (2007). annotation bias impedes biomedical research.
hyperactivity disorder, bipolar disorder, major 122. Skuse, D. H. et al. Social communication competence Sci. Rep. 8, 1362 (2018).
depressive disorder and schizophrenia. and functional adaptation in a general population of This paper analyses multiple annotation databases
97. Weiner, D. J. et al. Polygenic transmission disequilibrium children: preliminary evidence for sex-by-verbal IQ and finds significant inequalities across genes that
confirms that common and rare variation act additively differential risk. J. Am. Acad. Child. Adolesc. Psychiatry have become more prominent over time, highlighting
to create risk for autism spectrum disorders. Nat. Genet. 48, 128–137 (2009). a self-perpetuating cycle that may be driven by the
49, 978–985 (2017). 123. Rees, E. et al. Schizophrenia, autism spectrum tendency of researchers to focus their efforts on
98. Iakoucheva, L. M., Muotri, A. R. & Sebat, J. Getting disorders and developmental disorders share specific richly annotated genes rather than those with the
to the cores of autism. Cell 178, 1287–1298 (2019). disruptive coding mutations. Nat. Commun. 12, 5353 strongest molecular data.
99. State, M. W. & Levitt, P. The conundrums of (2021). 144. Zhao, H. et al. Altered neurogenesis and disrupted
understanding genetic risks for autism spectrum 124. Rosenthal, S. B. et al. A convergent molecular expression of synaptic proteins in prefrontal cortex
disorders. Nat. Neurosci. 14, 1499–1506 (2011). network underlying autism and congenital heart of SHANK3-deficient non-human primate. Cell Res.
100. Davies, R. W. et al. Using common genetic variation disease. Cell Syst. 12, 1094–1107 (2021). 27, 1293–1297 (2017).
to examine phenotypic expression and risk prediction 125. Liu, L. et al. DAWN: a framework to identify autism 145. Kang, H. J. et al. Spatio-temporal transcriptome
in 22q11.2 deletion syndrome. Nat. Med. 26, genes and subnetworks using gene expression and of the human brain. Nature 478, 483–489 (2011).
1912–1918 (2020). genetics. Mol. Autism 5, 22 (2014). This work presents a large gene expression
101. Jannot, A.-S., Ehret, G. & Perneger, T. P < 5 × 10–8 has 126. Chang, J., Gilman, S. R., Chiang, A. H., Sanders, S. J. database that includes human brain samples from
emerged as a standard of statistical significance for & Vitkup, D. Genotype to phenotype relationships before birth to late adulthood in 16 brain regions,
genome-wide association studies. J. Clin. Epidemiol. in autism spectrum disorders. Nat. Neurosci. 18, which allows spatially and temporally informed
68, 460–465 (2015). 191–198 (2015). analyses. This key resource has been widely used
102. Banerjee-Basu, S. & Packer, A. SFARI Gene: an 127. Chaste, P. et al. A genome-wide association study to generate critical insights into convergence of
evolving database for the autism research community. of autism using the Simons Simplex Collection: does genetic risk for ASD and other disorders of the
Dis. Model. Mech. 3, 133–135 (2010). reducing phenotypic heterogeneity in autism increase human brain.
0123456789();:
Reviews
146. Parikshak, N. N. et al. Integrative functional genomic 167. Wang, D. et al. Comprehensive functional genomic unfolded protein response. Cell 167, 1867–1882.e21
analyses implicate specific molecular pathways and resource and integrative model for the human brain. (2016).
circuits in autism. Cell 155, 1008–1021 (2013). Science 362, eaat8464 (2018). 190. Jaitin, D. A. et al. Dissecting immune circuits by linking
One of the first studies to assess spatiotemporal This study describes molecular data resources CRISPR-pooled screens with single-cell RNA-seq. Cell
convergence of ASD risk genes, this study conducts generated by the PsychENCODE Consortium 167, 1883–1896.e15 (2016).
a weighted gene co-expression network analysis from adult human brain tissue, and conducts 191. Datlinger, P. et al. Pooled CRISPR screening with
using BrainSpan developmental expression data a deconvolution analysis of bulk and single-cell single-cell transcriptome readout. Nat. Methods 14,
and assesses modules for enrichment of a broad expression data that finds that the majority 297–301 (2017).
range of ASD-associated genes, implicating the of expression variation across bulk brain tissue 192. Rubin, A. J. et al. Coupled single-cell CRISPR
early and late foetal periods and upper layer samples is attributable to varying proportions screening and epigenomic profiling reveals causal
glutamatergic neurons in ASD. of basic cell types. gene regulatory networks. Cell 176, 361–376.e17
147. Ben-David, E. & Shifman, S. Combined analysis of 168. Jaffe, A. E. et al. qSVA framework for RNA quality (2019).
exome sequencing points toward a major role for correction in differential expression analysis. 193. Tian, R. et al. CRISPR interference-based platform
transcription regulation during brain development Proc. Natl Acad. Sci. USA 114, 7130–7135 for multimodal genetic screens in human iPSC-derived
in autism. Mol. Psychiatry 18, 1054–1056 (2013). (2017). neurons. Neuron 104, 239–255.e12 (2019).
148. Xu, X., Wells, A. B., O’Brien, D. R., Nehorai, A. 169. Zhu, Y., Wang, L., Yin, Y. & Yang, E. Systematic 194. Kampmann, M. CRISPR-based functional genomics for
& Dougherty, J. D. Cell type-specific expression analysis of gene expression patterns associated with neurological disease. Nat. Rev. Neurol. 16, 465–480
analysis to identify putative cellular mechanisms for postmortem interval in human tissues. Sci. Rep. 7, (2020).
neurogenetic disorders. J. Neurosci. 34, 1420–1431 5435 (2017). 195. Tian, R. et al. Genome-wide CRISPRi/a screens in
(2014). 170. Li, J. Z. et al. Systematic changes in gene expression human neurons link lysosomal failure to ferroptosis.
149. Uddin, M. et al. Brain-expressed exons under purifying in postmortem human brains associated with tissue Nat. Neurosci. 24, 1020–1034 (2021).
selection are enriched for de novo mutations in autism pH and terminal medical conditions. Hum. Mol. Genet. 196. Volpato, V. & Webber, C. Addressing variability in
spectrum disorder. Nat. Genet. 46, 742–747 (2014). 13, 609–616 (2004). iPSC-derived models of human disease: guidelines
150. Lin, G. N. et al. Spatiotemporal 16p11.2 protein 171. Chiaradia, I. & Lancaster, M. A. Brain organoids for to promote reproducibility. Dis. Model. Mech. 13,
network implicates cortical late mid-fetal brain the study of human neurobiology at the interface of dmm042317 (2020).
development and KCTD13–Cul3–RhoA pathway in in vitro and in vivo. Nat. Neurosci. 23, 1496–1508 197. Genç, Ö. et al. Homeostatic plasticity fails at the
psychiatric diseases. Neuron 85, 742–754 (2015). (2020). intersection of autism-gene mutations and a novel
151. Li, M. et al. Integrative functional genomic analysis of 172. Mariani, J. et al. FOXG1-dependent dysregulation class of common genetic modifiers. eLife 9, e55775
human brain development and neuropsychiatric risks. of GABA/glutamate neuron differentiation in autism (2020).
Science 362, eaat7615 (2018). spectrum disorders. Cell 162, 375–390 (2015). 198. Packer, A. Neocortical neurogenesis and the etiology
152. Walker, R. L. et al. Genetic control of expression and 173. Schafer, S. T. et al. Pathological priming causes of autism spectrum disorder. Neurosci. Biobehav. Rev.
splicing in developing human brain informs disease developmental gene network heterochronicity in 64, 185–195 (2016).
mechanisms. Cell 179, 750–771.e22 (2019). autistic subject-derived neurons. Nat. Neurosci. 22, This comprehensive review details the potential
153. Werling, D. M. et al. Whole-genome and RNA 243–255 (2019). convergence of ASD genetic risk around
sequencing reveal variation and transcriptomic 174. Marchetto, M. C. et al. Altered proliferation and neurogenesis.
coordination in the developing human prefrontal networks in neural cells derived from idiopathic 199. Exner, C. R. T. & Willsey, H. R. Xenopus leads the way:
cortex. Cell Rep. 31, 107489 (2020). autistic individuals. Mol. Psychiatry 22, 820–835 frogs as a pioneering model to understand the human
154. Miller, J. A. et al. Transcriptional landscape of the (2017). brain. Genesis 59, e23405 (2021).
prenatal human brain. Nature 508, 199–206 One of the first studies comparing iPSC-derived 200. Werling, D. M. & Geschwind, D. H. Sex differences
(2014). neural cells from neurotypical controls to those in autism spectrum disorders. Curr. Opin. Neurol. 26,
This study presents gene expression data from from people with ASD and macrocephaly. 146–153 (2013).
laser-microdissected human mid-gestational brain 175. Adhya, D. et al. Atypical neurogenesis in induced 201. Loomes, R., Hull, L. & Mandy, W. P. L. What is the
tissue. These data have been central to increasing pluripotent stem cells from autistic individuals. male-to-female ratio in autism spectrum disorder?
the resolution of systems biological analyses of Biol. Psychiatry 89, 486–496 (2021). A systematic review and meta-analysis. J. Am.
ASD genetic risk. 176. DeRosa, B. A. et al. Convergent pathways in idiopathic Acad. Child. Adolesc. Psychiatry 56, 466–474
155. Pollen, A. A. et al. Molecular identity of human outer autism revealed by time course transcriptomic analysis (2017).
radial glia during cortical development. Cell 163, of patient-derived neurons. Sci. Rep. 8, 8423 (2018). 202. Kim, Y. S. et al. Prevalence of autism spectrum
55–67 (2015). 177. Griesi-Oliveira, K. et al. Transcriptome of iPSC-derived disorders in a total population sample. Am. J.
156. Nowakowski, T. J. et al. Spatiotemporal gene neuronal cells reveals a module of co-expressed genes Psychiatry 168, 904–912 (2011).
expression trajectories reveal developmental consistently associated with autism spectrum disorder. 203. Bai, D. et al. Inherited risk for autism through
hierarchies of the human cortex. Science 358, Mol. Psychiatry 26, 1589–1605 (2021). maternal and paternal lineage. Biol. Psychiatry 88,
1318–1323 (2017). 178. Amiri, A. et al. Transcriptome and epigenome 480–487 (2020).
157. Polioudakis, D. et al. A single-cell transcriptomic atlas landscape of human cortical development modeled 204. Palmer, N. et al. Association of sex with recurrence
of human neocortical development during mid- in organoids. Science 362, eaat6720 (2018). of autism spectrum disorder among siblings. JAMA
gestation. Neuron 103, 785–801.e8 (2019). 179. Trevino, A. E. et al. Chromatin accessibility dynamics Pediatr. 171, 1107–1112 (2017).
158. Eze, U. C., Bhaduri, A., Haeussler, M., Nowakowski, T. J. in a model of human forebrain development. Science 205. Gockley, J. et al. The female protective effect in autism
& Kriegstein, A. R. Single-cell atlas of early human 367, eaay1645 (2020). spectrum disorder is not mediated by a single genetic
brain development highlights heterogeneity of human 180. Pollen, A. A. et al. Establishing cerebral organoids as locus. Mol. Autism 6, 25 (2015).
neuroepithelial cells and early radial glia. Nat. Neurosci. models of human-specific brain evolution. Cell 176, 206. Martin, H. C. et al. The contribution of X-linked
24, 584–594 (2021). 743–756.e17 (2019). coding variation to severe developmental disorders.
159. Fan, X. et al. Spatial transcriptomic survey of human 181. Andrews, M. G. & Nowakowski, T. J. Human brain Nat. Commun. 12, 627 (2021).
embryonic cerebral cortex by single-cell RNA-seq development through the lens of cerebral organoid 207. McCarthy, M. M. Multifaceted origins of sex
analysis. Cell Res. 28, 730–745 (2018). models. Brain Res. 1725, 146470 (2019). differences in the brain. Philos. Trans. R. Soc.
160. Bernard, A. et al. Transcriptional architecture of 182. Bhaduri, A. et al. Cell stress in cortical organoids Lond. B Biol. Sci. 371, 20150106 (2016).
the primate neocortex. Neuron 73, 1083–1099 impairs molecular subtype specification. Nature 578, 208. Baron-Cohen, S. The extreme male brain theory
(2012). 142–148 (2020). of autism. Trends Cogn. Sci. 6, 248–254 (2002).
161. Doyle, J. P. et al. Application of a translational profiling 183. Andersen, J. et al. Generation of functional human 3D 209. Baron-Cohen, S. et al. Elevated fetal steroidogenic
approach for the comparative analysis of CNS cell cortico-motor assembloids. Cell 183, 1913–1929.e26 activity in autism. Mol. Psychiatry 20, 369–376
types. Cell 135, 749–762 (2008). (2020). (2015).
162. Voineagu, I. et al. Transcriptomic analysis of autistic 184. Miura, Y. et al. Generation of human striatal 210. Manoli, D. S. & Tollkuhn, J. Gene regulatory
brain reveals convergent molecular pathology. Nature organoids and cortico-striatal assembloids from mechanisms underlying sex differences in brain
474, 380–384 (2011). human pluripotent stem cells. Nat. Biotechnol. 38, development and psychiatric disease. Ann. N. Y.
One of the first papers comparing gene expression 1421–1430 (2020). Acad. Sci. 1420, 26–45 (2018).
profiles in postmortem brain tissue of neurotypical 185. Birey, F. et al. Assembly of functionally integrated 211. Ingudomnukul, E., Baron-Cohen, S., Wheelwright, S.
controls versus ASD patients, this study identifies human forebrain spheroids. Nature 545, 54–59 & Knickmeyer, R. Elevated rates of testosterone-
several molecular signatures that have now been (2017). related disorders in women with autism spectrum
well replicated (e.g. downregulation of neuron- 186. Deneault, E. et al. Complete disruption of autism- conditions. Horm. Behav. 51, 597–604 (2007).
and synapse-related genes and upregulation of susceptibility genes by gene editing predominantly 212. Auyeung, B. et al. Fetal testosterone and autistic
microglia-related genes). reduces functional connectivity of isogenic human traits. Br. J. Psychol. 100, 1–22 (2009).
163. Parikshak, N. N. et al. Genome-wide changes in neurons. Stem Cell Rep. 11, 1211–1225 (2018). 213. Komada, M. et al. Hedgehog signaling is involved
lncRNA, splicing, and regional gene expression 187. Lalli, M. A., Avey, D., Dougherty, J. D., Milbrandt, J. in development of the neocortex. Development 135,
patterns in autism. Nature 540, 423–427 (2016). & Mitra, R. D. High-throughput single-cell functional 2717–2727 (2008).
164. Gupta, S. et al. Transcriptome analysis reveals elucidation of neurodevelopmental disease-associated 214. Britto, J., Tannahill, D. & Keynes, R. A critical role for
dysregulation of innate immune response genes genes reveals convergent mechanisms altering sonic hedgehog signaling in the early expansion of the
and neuronal activity-dependent genes in autism. neuronal differentiation. Genome Res. 30, 1317–1331 developing brain. Nat. Neurosci. 5, 103–110 (2002).
Nat. Commun. 5, 5748 (2014). (2020). 215. Echelard, Y. et al. Sonic hedgehog, a member of a
165. Gandal, M. J. et al. Transcriptome-wide isoform-level 188. Dixit, A. et al. Perturb-seq: dissecting molecular family of putative signaling molecules, is implicated
dysregulation in ASD, schizophrenia, and bipolar circuits with scalable single-cell RNA profiling of in the regulation of CNS polarity. Cell 75, 1417–1430
disorder. Science 362, eaat8127 (2018). pooled genetic screens. Cell 167, 1853–1866.e17 (1993).
166. Velmeshev, D. et al. Single-cell genomics identifies cell (2016). 216. Finkel, R. S. et al. Nusinersen versus sham control
type-specific molecular changes in autism. Science 189. Adamson, B. et al. A multiplexed single-cell CRISPR in infantile-onset spinal muscular atrophy. N. Engl.
364, 685–689 (2019). screening platform enables systematic dissection of the J. Med. 377, 1723–1732 (2017).
www.nature.com/nrn
0123456789();:
Reviews
217. Mendell, J. R. et al. Single-dose gene-replacement 233. D’Esposito, M. et al. Isolation, physical mapping, 250. Guo, H. et al. Disruptive mutations in TANC2 define
therapy for spinal muscular atrophy. N. Engl. J. Med. and northern analysis of the X-linked human gene a neurodevelopmental syndrome associated with
377, 1713–1722 (2017). encoding methyl CpG-binding protein, MECP2. psychiatric disorders. Nat. Commun. 10, 4679
218. Colantuoni, C. et al. Temporal dynamics and genetic Mamm. Genome 7, 533–535 (1996). (2019).
control of transcription in the human prefrontal 234. Kremer, E. J. et al. Mapping of DNA instability at the 251. Van Dijck, A. et al. Clinical presentation of a complex
cortex. Nature 478, 519–523 (2011). fragile X to a trinucleotide repeat sequence p(CCG)n. neurodevelopmental disorder caused by mutations
219. O’Brien, H. E. et al. Expression quantitative trait loci in Science 252, 1711–1714 (1991). in ADNP. Biol. Psychiatry 85, 287–297 (2019).
the developing human brain and their enrichment 235. Verkerk, A. J. et al. Identification of a gene (FMR-1) 252. Blackburn, A. T. M. et al. DYRK1A-related intellectual
in neuropsychiatric disorders. Genome Biol. 19, 194 containing a CGG repeat coincident with a breakpoint disability: a syndrome associated with congenital
(2018). cluster region exhibiting length variation in fragile anomalies of the kidney and urinary tract. Genet.
220. Zhong, S. et al. A single-cell RNA-seq survey of the X syndrome. Cell 65, 905–914 (1991). Med. 21, 2755–2764 (2019).
developmental landscape of the human prefrontal 236. Pieretti, M. et al. Absence of expression of the FMR-1 253. Ji, J. et al. DYRK1A haploinsufficiency causes a new
cortex. Nature 555, 524–528 (2018). gene in fragile X syndrome. Cell 66, 817–822 (1991). recognizable syndrome with microcephaly, intellectual
221. Trevino, A. E. et al. Chromatin and gene-regulatory 237. Li, J. et al. PTEN, a putative protein tyrosine disability, speech impairment, and distinct facies.
dynamics of the developing human cerebral cortex at phosphatase gene mutated in human brain, breast, Eur. J. Hum. Genet. 23, 1473–1481 (2015).
single-cell resolution. Cell 184, 5053–5069.e23 (2021). and prostate cancer. Science 275, 1943–1947 254. Bernier, R. et al. Disruptive CHD8 mutations define
222. Marshall, C. R. et al. Contribution of copy number (1997). a subtype of autism early in development. Cell 158,
variants to schizophrenia from a genome-wide study 238. Li, D. M. & Sun, H. TEP1, encoded by a candidate 263–276 (2014).
of 41,321 subjects. Nat. Genet. 49, 27–35 (2017). tumor suppressor locus, is a novel protein tyrosine This is one of the first studies to recognize
223. Searles Quick, V. B., Wang, B. & State, M. W. phosphatase regulated by transforming growth distinctive physical features and co-morbidities
Leveraging large genomic datasets to illuminate factor β. Cancer Res. 57, 2124–2129 (1997). shared across patients with rare mutations in
the pathobiology of autism spectrum disorders. 239. Steck, P. A. et al. Identification of a candidate tumour the same ‘idiopathic’ ASD risk gene — opening the
Neuropsychopharmacology 46, 55–69 (2021). suppressor gene, MMAC1, at chromosome 10q23.3 door to the presence of previously unappreciated
224. MacDonald, J. R., Ziman, R., Yuen, R. K. C., Feuk, L. that is mutated in multiple advanced cancers. syndromes within idiopathic ASD.
& Scherer, S. W. The Database of Genomic Variants: Nat. Genet. 15, 356–362 (1997).
a curated collection of structural variation in the 240. van Slegtenhorst, M. et al. Identification of the Acknowledgements
human genome. Nucleic Acids Res. 42, D986–D992 tuberous sclerosis gene TSC1 on chromosome The authors thank S. Wang for data analysis and graphical
(2014). 9q34. Science 277, 805–808 (1997). representation for Fig. 1 and S. Pyle for graphic design. This
225. Collins, R. L. et al. A structural variation reference 241. European Chromosome 16 Tuberous Sclerosis work was supported by a gift from the Overlook International
for medical and population genetics. Nature 581, Consortium. Identification and characterization of the Foundation and grant support from the National Institutes of
444–451 (2020). tuberous sclerosis gene on chromosome 16. Cell 75, Mental Health (NIMH) (U01MH116487, R01MH115963,
226. Purcell, A. E., Jeon, O. H., Zimmerman, A. W., 1305–1315 (1993). 1U01MH115747, R25MH06048) and the Psychiatric Cell
Blue, M. E. & Pevsner, J. Postmortem brain 242. Marchuk, D. A. et al. cDNA cloning of the type 1 Map Initiative (pcmi.ucsf.edu).
abnormalities of the glutamate neurotransmitter neurofibromatosis gene: complete sequence of the
system in autism. Neurology 57, 1618–1628 (2001). NF1 gene product. Genomics 11, 931–940 (1991). Author contributions
227. Sutcliffe, J. S. et al. The E6-Ap ubiquitin-protein ligase 243. Cawthon, R. M. et al. A major segment of the The authors all contributed to all aspects of preparing the
(UBE3A) gene is localized within a narrowed Angelman neurofibromatosis type 1 gene: cDNA sequence, article.
syndrome critical region. Genome Res. 7, 368–377 genomic structure, and point mutations. Cell 62,
(1997). 193–201 (1990). Competing interests
228. Greenberg, F. & Ledbetter, D. H. Deletions of proximal 244. Viskochil, D. et al. Deletions and a translocation The authors declare no competing interests.
15q without Prader–Willi syndrome. Am. J. Med. Genet. interrupt a cloned gene at the neurofibromatosis
28, 813–820 (1987). type 1 locus. Cell 62, 187–192 (1990). Peer review information
229. Kaplan, L. C. et al. Clinical heterogeneity associated 245. Wallace, M. R. et al. Type 1 neurofibromatosis gene: Nature Reviews Neuroscience thanks R. Anney, S. Shifman and
with deletions in the long arm of chromosome 15: identification of a large transcript disrupted in three F. Volkmar for their contribution to the peer review of this work.
report of 3 new cases and their possible genetic NF1 patients. Science 249, 181–186 (1990).
significance. Am. J. Med. Genet. 28, 45–53 (1987). 246. Rouleau, G. A. et al. Alteration in a new gene encoding Publisher’s note
230. Magenis, R. E., Brown, M. G., Lacy, D. A., Budden, S. a putative membrane-organizing protein causes Springer Nature remains neutral with regard to jurisdictional
& LaFranchi, S. Is Angelman syndrome an alternate neuro-fibromatosis type 2. Nature 363, 515–521 claims in published maps and institutional affiliations.
result of del(15)(q11q13)? Am. J. Med. Genet. 28, (1993).
829–838 (1987). 247. Trofatter, J. A. et al. A novel moesin-, ezrin-, radixin- Supplementary information
231. Williams, C. A., Gray, B. A., Hendrickson, J. E., like gene is a candidate for the neurofibromatosis 2 The online version contains supplementary material available
Stone, J. W. & Cantú, E. S. Incidence of 15q deletions tumor suppressor. Cell 72, 791–800 (1993). at https://doi.org/10.1038/s41583-022-00576-7.
in the Angelman syndrome: a survey of twelve affected 248. Guo, H. et al. NCKAP1 disruptive variants lead to a
persons. Am. J. Med. Genet. 32, 339–345 (1989). neurodevelopmental disorder with core features of
232. Vilain, A., Apiou, F., Vogt, N., Dutrillaux, B. & Malfoy, B. autism. Am. J. Hum. Genet. 107, 963–976 (2020).
Related links
Brainspan Atlas of the Developing Human Brain: https://
Assignment of the gene for methyl-CpG-binding 249. Mirzaa, G. M. et al. De novo and inherited variants in
brainspan.org
protein 2 (MECP2) to human chromosome band Xq28 ZNF292 underlie a neurodevelopmental disorder with
by in situ hybridization. Cytogenet. Cell Genet. 74, features of autism spectrum disorder. Genet. Med. 22,
293–294 (1996). 538–546 (2020). © Springer Nature Limited 2022
0123456789();: