Sequencing of selected chromatin remodelling genes reveals increased burden of rare missense variants in ASD patients from the Japanese population

概要

主論文の要旨

Sequencing of selected chromatin remodelling genes
reveals increased burden of rare missense variants in
ASD patients from the Japanese population
クロマチンリモデリング関連遺伝子のシーケンシングにより
日本人ASD患者における稀なミスセンスバリアントの
集積を明らかとした

名古屋大学大学院医学系研究科
脳神経病態制御学講座
（指導：木村 宏之
罗 子尧

総合医学専攻

精神医学分野
准教授）

【Introduction】
Autism spectrum disorder (ASD) and Schizophrenia (SCZ) are both highly heritable mental
disorders considered to stem from disruption of neural development. High heritability indicates
a substantial contribution of genetic factors to the etiology of both disorders. Recent studies
have shown ASD and SCZ may share common pathogenic mechanisms underlain by genetic
overlap. Therefore, it would be valuable to investigate the genetic burdens in a set of genes
that belong to a common biological process involved in both ASD and SCZ. In this study, we
focus on chromatin remodeling. Chromatin remodeling is the process by which chromatin
remodeler regulates transcription by changing accessibility to DNA or histone. Throughout
brain development, chromatin remodeling takes part in diverse aspects. Therefore, it is not
surprising that disturbance in this process was found to be widely associated with psychiatric
disorders including ASD and SCZ, and the genes encoding subunits of chromatin remodeling
complexes, specifically the BRG1/BRM-associated factor (BAF) complex, have been
identified as candidates for ASD and SCZ. Despite accumulating evidence showing the
involvement of the BAF complex in neural development and psychiatric disorders, to our
knowledge, no study has yet focused on rare variants of the BAF genes in ASD and SCZ in the
Japanese population. In the current study, we performed a case-control study based on targeted
deep sequencing on coding regions of four BAF genes: SMARCA2, SMARCA4, SMARCC2, and
ARID1B. Pathogenicity of discovered rare variants was then predicted based on functional
annotations from in-silico tools considering in-silico deleteriousness, constraint, and paralog
conservation. Afterward, an association study was conducted to investigate the burden of rare
variants in BAF genes in ASD and SCZ patients.
【Methods】
Subjects
All subjects in this study are ethnically Japanese. The sample-set used in discovery of rare
variants comprised 437 SCZ cases (mean age=47.1Y; males=56.3%), 187 ASD cases (mean
age=19.9Y; males=75.9%) and 524 healthy controls (mean age=41.3Y; males=59.7%).

ASD

and SCZ patients were recruited upon diagnosis according to DSM-5 criteria. Control subjects
are recruited from the general population and self-reported with no history of mental disorders.
Targeted sequencing and variant annotation
Genomic DNA was extracted from whole peripheral blood or saliva following standard
protocol. After library preparation and sample indexing, templates were prepared and loaded
to the Ion 318™ Chip v2. Sequencing data were collected and analyzed by Ion Torrent suite
(v5.12.1). Variants were called and exonic functions were annotated with Torrent Variant Caller
Plugin (v5.12.0.4) with default parameters.

-1-

Functional annotation of rare missense variants
The functions of rare missense variants were annotated by CADD v1.6, VEST3, PolyPhen2, MPC score, and JalView score. Rare variants with scores over the standard cut-offs were
considered pathogenic (“CADD_del”. “VEST3_del”, “PPh2_del”, “MPC1” or “paralog-conserved”).
(Fig. 1).
Association analysis
Allele counts of rare variants were tested for association using a 2×2 contingency table
between different groups utilizing Fisher’s exact test (one-tailed). Differences were considered
significant with a p value ≤ 0.05.
【Results】
Detection of rare variants
After quality control we obtained targeted sequencing data from 1,134 Japanese individuals
(98.8%, 432 SCZ, 185 ASD, and 517 controls) on coding regions of SMARCA2, SMARCA4,
SMARCC2, and ARID1B genes. 27 rare non-synonymous variants in ASD and SCZ cases,
including 25 missense variants (Table 1), one in-frame deletion (E593del in SMARCA4) in a
SCZ patient and one frame-shift variant (G1044Dfs*17 in SMARCC2) in a pair of ASD
monozygotic twins were identified. All variants were heterozygote. Among them, 15 missense
variants were ultra-rare (minor allele frequency < 0.1%) and 4 were considered novel. With
allele counts of 27 (0.063/sample) in SCZ, 21 (0.114/sample) in ASD and 34 (0.066/sample)
in controls, an enrichment of rare missense variants in BAF genes was seen in ASD compared
to SCZ (p=0.027) and healthy controls (p=0.045) while no enrichment was seen in rare
synonymous variants (Fig. 2a, b).
Association analysis of rare missense variants
Through in silico prediction, we identified 18 CADD_del, 11 VEST3_del, 14 PPh2_del, 12
MPC1 and 8 paralog-conserved variants. We conducted an association analysis to investigate
the burden of these pathogenic-predicted variants in ASD and SCZ compared to healthy
controls. As a result, CADD_del variants were significantly enriched in ASD (p=0.024) but
not in SCZ (p=0.386) (Fig. 2d). VEST3_del variants were significantly enriched in ASD
(p=0.004) and marginally in SCZ (p=0.063) (Fig. 2e). PPh2_del variants were significantly
enriched in ASD (p=0.011) but not in SCZ (p=0.254) (Fig. 2f). No significant enrichment of
MPC1 variants was seen in ASD (AC=10, p=0.147) or SCZ (AC=13, p=0.663) compared to
controls (AC=17) (Fig. 2g). Paralog-conserved variants were found significantly enriched in
ASD (p=0.01) but not in SCZ (p=0.47) (Fig. 2h).
【Discussion】

-2-

Through association analysis, a significantly increased burden of rare missense variants in
BAF genes was found in ASD cases compared to healthy controls (p=0.045) (Fig. 2b), while
not seen in rare synonymous variants (Fig. 2a). This is not surprising since BAF genes are
intolerant of missense variation indicated by their high missense Z-scores (2.59-6.85). Notably,
a similar effect was also seen between ASD and SCZ (p=0.027) (Fig. 2b). Our findings are
consistent with the previous studies suggesting greater burdens of rare genetic variants in ASD
than SCZ. According to functional annotations by in silico tools, more rare missense variants
were predicted pathogenic in both ASD and SCZ than in controls (Fig. 2c). This showcases
rare missense variants in BAF genes discovered in ASD and SCZ cases were more likely
pathogenic. As a result, pathogenic-predicted CADD_del (p=0.024), VEST3_del (p=0.004),
PPh2_del (p=0.011), and paralog-conserved (p=0.01) variants were more significantly
enriched in ASD compared to controls (Fig. 2d-f, h). Interestingly, as predicted by ANCHOR,
the majority of pathogenic-predicted variants discovered in our study (69~92%) were located
in disordered binding regions (DBRs) (Fig. 3). DBRs are regions particularly important in
protein-protein interactions (PPIs). Given the role of protein interactions in core functions of
BAF subunits including complex assembly and transcription factors interaction, pathogenic
variants located in DBRs discovered in our study may exert their impacts by disrupting PPIs.
This may also explain why MPC1 variants were not significantly enriched in ASD, (Fig. 2g),
as PPI disruption was not well identified by MPC scoring as a recent study implicated. There
are several limitations to this study. First, only encoding genes of SMARCA2, SMARCA4,
SMARCC2 and ARID1B were sequenced in this study. Recent studies have found other BAF
subunits involved in the pathogenesis of ASD, therefore future studies should include other
BAF genes to gain a more comprehensive understanding of the involvement of BAF chromatin
remodeling complex in psychiatric disorders. Second, the sample size is relatively limited and
the possibility of bias or artifact could not be ruled out. Larger samples may be required to
examine the reproducibility of this result.
【Conclusion】
This is the first targeted sequencing study focusing on rare variants of the BAF chromatin
remodeling complex genes in ASD and SCZ in the Japanese population. Our case-control study
based on sequencing data reveals a significantly increased burden of rare missense variants in
BAF genes in ASD, and stratification to pathogenic-predicted variants further gains
significance of this effect. Rare pathogenic-predicted variants of BAF genes were positioned
at disordered binding regions and may confer risk for ASD by disrupting PPIs. This study
supports the involvement of rare missense variants of BAF gene in the susceptibility of ASD.

-3-

-4-

Position
(GRCh37)

157405967

157502199

157511174

157517313

157522073

157522274

157525120

157527653

157527680

157528049

157528292

2039561

2060973

2191407

56558236

56558443

56563662

56563987

56566484

56572311

56578708

11094897

11097111

11097224

11098500

Chr6

Chr6

Chr6

Chr6

Chr6

Chr6

Chr6

Chr6

Chr6

Chr6

Chr6

Chr9

Chr9

Chr9

Chr12

Chr12

Chr12

Chr12

Chr12

Chr12

Chr12

Chr19

Chr19

Chr19

Chr19

G

G

A

C

A

T

G

T

C

G

G

G

G

A

C

G

G

A

A

G

G

G

G

A

C

Re fe re nce
alle

A

C

T

G

G

C

T

A

T

A

A

A

A

T

G

A

A

G

G

A

A

A

A

T

T

Alte rnate
alle

Variant information

SMARCA4

SMARCA4

SMARCA4

SMARCA4

SMARCC2

SMARCC2

SMARCC2

SMARCC2

SMARCC2

SMARCC2

SMARCC2

SMARCA2

SMARCA2

SMARCA2

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

ARID1B

Ge ne name

p.A340T

p.G239R

p.Q201L

p.P24A

p.S138P

p.D396G

p.T552N

p.S754C

p.E785K

p.P1071L

p.A1140V

p.R1579H

p.R560K

p.S151C

p.A1993G

p.S1912N

p.R1789Q

p.K1780R

p.N1659S

p.V1503I

p.G1436S

p.E1280K

p.S1218N

p.T1065S

p.P724S

Amino acid
change

SCZ

SCZ

ASD/SCZ

ASD

ASD

SCZ

SCZ

ASD

ASD/SCZ

ASD

SCZ

ASD/SCZ

ASD/SCZ

ASD

SCZ

ASD

ASD/SCZ

SCZ

ASD

ASD/SCZ

ASD

ASD

SCZ

ASD/SCZ

ASD/SCZ

Phenotype

Detailed information of 25 rare missense variants discovered in the current study

C hromosom
e

Table 1

rs371214327

.

rs587778682

tgv61192460

tgv46021374

.

.

rs111975931

tgv46020486

tgv46020167

rs745595139

rs138760068

rs778157524

tgv35226272

.

.

rs751043203

rs753020972

rs140177120

rs567836947

rs141461351

rs752398721

rs574296429

rs771621841

rs534990481

-

-

0.17%

-

0.07%

-

0.02%

0.05%

0.05%

-

0.01%

0.63%

0.03%

0.02%

-

-

0.34%

-

0.08%

0.14%

0.06%

-

0.01%

0.21%

0.11%

-

-

0.13%

-

0.07%

-

0.02%

0.06%

0.07%

-

0.02%

0.71%

0.02%

0.03%

-

-

0.35%

-

0.09%

0.17%

0.06%

-

-

0.21%

0.13%

0.0035%

-

0.0164%

-

0.0004%

-

-

0.0322%

-

-

0.0012%

0.0025%

0.0016%

-

-

-

0.0021%

0.0012%

0.0778%

0.0131%

0.0071%

0.0004%

0.0076%

0.0012%

0.0117%

gnomAD
v2.1.1

Database minor allele frequency

ToMMo 4.7K
TogoVar ID GEM-J WGA
JPN

dbS NP/

19.1

23.4

24.5

23.7

18.5

22.7

10.9

24.5

24

23.8

24.6

23.2

14.3

23.4

22.5

27.2

24.2

4.3

14.1

20.4

23.8

32

20.9

8.2

24.2

CADD

0.102

0.737

0.563

0.512

0.346

0.267

0.123

0.521

0.529

0.515

0.413

0.323

0.136

0.448

0.135

0.154

0.55

0.274

0.043

0.074

0.699

0.876

0.077

0.156

0.52

VEST3

Benign

Benign

Probably
Damaging
Possibly
Damaging
Probably
Damaging

Benign

Benign

Benign

Possibly
Damaging
Probably
Damaging
Possibly
Damaging
Probably
Damaging

8

6

11

11

8

9

8

0

11

11

7

0

9

8
Benign

8
Benign

11

5

7

11

8

11

9

9

8

Probably
Damaging

Possibly
Damaging
Probably
Damaging

Benign

Benign

Benign

Probably
Damaging
Possibly
Damaging

Benign

Benign

Polyphe n-2 JalVie
HDIV
w score
Probably
11
Damaging

In-silico prediction tool

Paternal

Paternal

Paternal

Maternal

Maternal

Paternal

Paternal

Pedigree
analysis

Figure 1. Workflow of the current study

-5-

Figure 2. Association analysis of rare variants
(a) Association analysis of rare synonymous variants. (b) Association analysis of rare missense variants. (c)
Pathogenicity prediction of rare missense variants based on functional annotations from in silico tools. (d)-(h)
Association analysis of pathogenic-predicted variants based on functional annotations by CADD, VEST3, PolyPhen2,
MPC and JalView.
(1) represents the number of rare synonymous variants per sample, (2) represents the number of rare missense variants
per sample, (3) represents the ratio of pathogenic-predicted variants in ASD/SCZ/CON, (4) represents the number of
CADD_del variants per sample, (5) represents the number of VEST3_del variants per sample, (6) represents the number
of PPh2_del variants per sample, (7) represents the number of MPC1 variants per sample, 8) represents the number of
paralog_conserved variants per sample.

-6-

Figure 3. Location of rare missense variants
Location of rare missense variants discovered in (a) SMARCA2 (b) SMARCA4 (c) SMARCC2 and (d) ARID1B. The
protein domain localization is based on the Pfam protein families database and disordered binding regions (DBRs) are
predicted by ANCHOR utilizing the d2p2 database.

-7-