Structure of nucleosome-bound human BAF complex

Abstract

Architecture of human BAF complex
The SWI/SNF family chromatin remodelers regulate chromatin and transcription. The protein complexes BAF and PBAF are mammalian SWI/SNF remodelers that play essential functions in diverse developmental and physiological processes. He et al. determined the structure of the human BAF complex, which contains three modules that bind the nucleosome on the top, bottom, and side, making this nucleosome-recognition pattern distinct from other chromatin remodelers. Mutations in BAF that are frequently associated with human cancer cluster into a nucleosome-interacting region. This structure provides a framework for understanding the BAF-mediated chromatin remodeling mechanism and its dysregulation in cancer.
Science , this issue p. 875

Full text

STRUCTUR AL BIOLOGY
Structure of nucleosome-bound human BAF complex
Shuang He
1,2
*, Zihan Wu
1,2
*, Yuan Tian
1,2
*, Zishuo Yu
1,2
, Jiali Yu
1,2
, Xinxin Wang
1,2
, Jie Li
1
,
Bijun Liu
1
, Yanhui Xu
1,2,3,4
†
Mammalian SWI/SNF family chromatin remodelers, BRG1/BRM-associated factor (BAF) and polybromo-
associated BAF (PBAF), regulate chromatin structure and transcription, and their mutations are linked
to cancers. The 3.7-angstrom-resolution cryo–electron microscopy structure of human BAF bound
to the nucleosome reveals that the nucleosome is sandwiched by the base and the adenosine
triphosphatase (ATPase) modules, which are bridged by the actin-related protein (ARP) module. The
ATPase motor is positioned proximal to nucleosomal DNA and, upon ATP hydrolysis, engages with and
pumps DNA along the nucleosome. The C-terminal ahelix of SMARCB1, enriched in positively charged
residues frequently mutated in cancers, mediates interactions with an acidic patch of the nucleosome.
AT-rich interactive domain-containing protein 1A (ARID1A) and the SWI/SNF complex subunit SMARCC
serve as a structural core and scaffold in the base module organization, respectively. Our study provides
structural insights into subunit organization and nucleosome recognition of human BAF complex.
The adenosine triphosphate (ATP)–dependent
chromatin remodeling complexes (also
known as chromatin remodelers) reg-
ulate the chromatin packing state by
sliding, ejecting, and restructuring the
nucleosome to enable dynamic regulation of
chromatin structure (1,2). As prototype chro-
matin remodelers, the SWI/SNF complexes
demonstrate nucleosome sliding activity and
distinctive ejection activity, by which they
create nucleosome-depleted regions (NDRs)
that are essential for transcriptional regulation
(3–10). Mammalian SWI/SNF (mSWI/SNF)
complexes, BRG1/BRM-associated factor (BAF)
and polybromo-associated BAF (PBAF), consist
of up to 15 subunits encoded by more than
29 genes, generating more than 1400 possible
complexes (6,7,11,12). Up to 20% of malig-
nancies contain mutations of BAF and/or PBAF
subunits, making these complexes among the
most frequently dysregulated targets in human
cancer (4,7,13).
Although the compositions and subunit
functions of SWI/SNF complexes have been
extensively studied (8,11,14–19), the structural
studies of SWI/SNF complexes have been lim-
ited to the low-resolution electron microscopy
(EM) structures of yeast SWI/SNF complexes
(20–23) and structures of isolated domains
(15,24–27). A recent study reported an ~7-
Å-resolution structure of RSC (yeast homolog
of PBAF) bound to the nucleosome with the
substrate recruitment module refined to 3.4-Å
resolution (28). However, the molecular mech-
anisms of subunit organization and nucleo-
some recognition of mammalian BAF complex
remain largely unknown.
Structure determination
We reconstituted human BAF complex con-
sisting of the catalytic subunit SMARCA4
(BRG1) and nine auxiliary subunits (Fig. 1A,
figs. S1 and S2, and table S1). The purified
BAF complexes exhibited nucleosome sliding
activities (fig. S1, D and H to J) and were com-
plexed with nucleosome core particle (NCP) in
the absence of ATP and adenosine diphosphate
(ADP) (fig. S1D). The cryo-EM structure of the
BAF-NCP complex (~1.2 MDa) was determined
at an overall resolution of 3.7 Å, with the map of
thebasemodulelocallyrefinedto3.0-Åres-
olution (Figs. 1, B to F, and 2A; figs. S3 and S4;
table S2; and movies S1 to S5). The structural
models were built by fitting available struc-
turesintothecryo-EMmaps(15,24,26,27,29)
followed by manual model building aided by
cross-linking mass spectrometry (XL-MS) (fig.
S2,AandB,anddataS1andS2).The3.7-
Å-resolution map and corresponding model
were used hereafter unless otherwise specified.
BAF sandwiches the nucleosome
ThehumanBAFcomplexsandwichesthenu-
cleosome, and its nucleosome-binding manner
is distinct from that of other representative
chromatin remodelers, including Chd1, SWR1,
and INO80 (30–32) (Fig. 1, B to D, and fig. S5).
These remodelers primarily bind nucleosomal
DNA and/or histone tails but have fewer con-
tacts with core histones. The sandwiched
nucleosome binding of BAF may provide a
structural basis to support nucleosome ejec-
tion activity (discussed below). The nucleoso-
mal DNA tends to be detached or disordered
at both entry and exit sites (Fig. 1, E and F,
and movie S5), which may allow the DNA
translocation to occur more efficiently owing
to fewer restraints.
The BAF complex consists of three modules:
the adenosine triphosphatase (ATPase) module,
the actin-related protein (ARP) module, and the
base module (Fig. 1A). A large portion of the
catalytic subunit SMARCA4 (residues 521 to
1647) forms the ATPase module, which grasps
the nucleosome with the ATPase motor par-
tially wrapping around the nucleosomal DNA
(Fig. 1, B to D). Within the ARP module, the
helicase-SANT–associated region (HSA; resi-
dues 446 to 520) of SMARCA4 binds the het-
erodimer formed by ACTL6A (BAF53A) and
ACTB (b-actin) (Fig. 1, A and B). The pre-HSA
(residues 350 to 445) of SMARCA4 is anchored
intothebasemodule,inwhichtheSMARCB1
(BAF47,hSNF5,orINI1)packsagainstthe
bottomsurfaceofthenucleosome(Fig.1,A
and B).
The ARP module bridges the ATPase and
base modules
TheARPmoduleisformedbytheACTL6A-
ACTB heterodimer and the long ahelix of the
HSA of SMARCA4 (Fig. 1G), revealing a fold
similar to the yeast HSA
Snf2
-arp7-arp9-Rtt102
structure (27). The ARP module has no direct
contact with the nucleosome, but it associates
with and bridges the ATPase and base mod-
ules(Fig.1,B,C,andG).Acryo-EMmapre-
veals considerable contacts between lobe 1 of
the SMARCA4 ATPase domain and subdo-
main 4 (residues 266 to 305) of ACTL6A (Fig. 1,
B and G). Moreover, the pre-HSA in the base
moduleandpost-HSAintheATPasemodule
aredirectlyconnectedbytheHSAhelix.Thus,
the ARP module maintains a rigid conforma-
tion of the HSA helix and likely couples the
motions of the ATPase and base modules
during chromatin remodeling. This obser-
vation is consistent with the essential role of
the ARP module both in “coupling”the DNA
translocation and ATP hydrolysis and in the
assembly of functional SWI/SNF complexes
(8,18,27).
The ATPase motor engages with
nucleosomal DNA
The cryo-EM density of the ATPase motor is
relatively weak and reveals a tilted and open
conformation (angle of ~90° between the two
ATPase lobes), indicating that the structure
represents a pre-engaged conformation (fig. S6),
which is consistent with the lack of ATP and
ADP in the BAF-NCP structure. The ATPase
motor is positioned at nucleosomal DNA near
superhelical location (SHL) 2.5, and the posi-
tioning is likely guided by the HSA-associated
ARP module and ATPase-associated Snf2 ATP
coupling (SnAC) domain. We observed weak
cryo-EM density stretching across the top
RESEARCH
He et al., Science 367, 875–881 (2020) 21 February 2020 1of7
1
Fudan University Shanghai Cancer Center, Institutes of
Biomedical Sciences, State Key Laboratory of Genetic
Engineering, and Shanghai Key Laboratory of Medical
Epigenetics, Shanghai Medical College of Fudan University,
Shanghai 200032, China.
2
The International Co-laboratory of
Medical Epigenetics and Metabolism, Ministry of Science and
Technology, China, Department of Systems Biology for
Medicine, School of Basic Medical Sciences, Shanghai
Medical College of Fudan University, Shanghai 200032,
China.
3
Human Phenome Institute, Collaborative Innovation
Center of Genetics and Development, School of Life
Sciences, Fudan University, Shanghai 200433, China.
4
CAS
Center for Excellence in Molecular Cell Science, Chinese
Academy of Sciences, Shanghai 200031, China.
*These authors contributed equally to this work.
†Corresponding author. Email: xuyh@fudan.edu.cn
on February 25, 2020 http://science.sciencemag.org/Downloaded from

surface of the nucleosome (Fig. 1D and fig.
S6A). This region is likely derived from the
SnAC and/or bromodomain, which have
been shown to bind DNA and/or histone
(25,33,34).
The structure of BAF-NCP in the presence of
ADP was refined to 10.3-Å resolution, showing
that the ATPase motor adopts a relatively
closed conformation (~70°) and has tight con-
tacts to nucleosomal DNA around SHL 2 (fig.
S6B and movie S6). Thus, ATP or ADP would
promote conformational transition of the BAF
complex from a pre-engaged to an engaged
state, with the ATPase engaging with nu-
cleosomal DNA and, upon ATP hydrolysis,
pumping DNA toward the nucleosome dyad
and generating DNA translocation.
The organization of the base module
The pre-HSA of SMARCA4 is anchored into
the base module, which consists of seven ad-
ditional auxiliary subunits: two BAF-specific
subunits, ARID1A (BAF250A) and DPF2
(BAF45D); and five BAF/PBAF-shared subunits,
SMARCB1, SMARCD1 (BAF60A), SMARCE1
(BAF57), and two copies of SMARCC (8,11,15–19)
(Figs. 1 and 2). These auxiliary subunits exist
exclusively in the base module and account
for ~80% of the total molecular mass of the
BAF complex (fig. S4).
Thebasemodulerevealsacompactfoldand
can be divided into five closely associated sub-
modules: the head, thumb, palm, bridge, and
fingers (Fig. 2). The head and bridge bind the
nucleosome and the ARP module, respectively,
generating intermodular contacts (Fig. 1G).
The thumb is formed by a SANT domain of
SMARCC, the pre-HSA of SMARCA4, and the
C-terminal helices of SMARCD1 (Fig. 2). The
fingers submodule reveals a characteristic
Y-shaped five-helix bundle formed by coiled-
coil (CC) domains of SMARCD1, SMARCE1,
He et al., Science 367, 875–881 (2020) 21 February 2020 2of7
Dyad
1.5
2.5
DNA
entry
3.5
WH
SMARCB1
ACTL6A
lobe 2
lobe 1
ACTB
90°
ATPase
ATPase
DNA
Exit
HSA
HSA
DNA
Exit
DNA
Entry
HSA
HSA
ATPase
ATPase
Thumb
Fingers
Palm
Head
Bridge
Fingers
Palm
Thumb
H2A
H2A
/H2B
H2B
pre-HSA
pre-HSA
SHL 2.5
SnAC?
SnAC?
Dyad
1.5
3.5 4.5
Base Module
ARP Module
ATPase Module
1246
ATPase lobe2
SMARCA4
766
ATPase lobe1
ATPase Module
521
ARP Module
BR
SnAc
RPT1 RPT2
SMARCC2
a
SMARCC2
b
SMARCB1
CC-C2
a
CC-C2
b
CC-D2
CC-E1
DPF2
Req
αC
CC-D1
Base Module
SWIB
H2A/H2B
H2B/H2A
H3/H4H4/H3
SMARCD1
pre-HSA
1647
SANT
SWIRM
SMARCA4 ACTL6A
ACTB
SMARCD1
SMARCE1 SMARCC2
a
SMARCB1
DPF2ARID1A
H2A H2B H4H3 DNA
SMARCC2
b
A
Dyad
-1.5
-2.5
-3.5
-4.5
-5.5
DNA
exit
B
D
EF
lobe 2
lobe 1
ACTL6A
WH
SMARCB1
WH
SMARCB1
ACTL6A
αC
ACTL6A
αC
ATPase
ATPase
HSA
G
RPT1
SHL 2.5
Fingers
Thumb
Palm
Head
Bridge
SMARCC2
a
SMARCC2
b
SMARCE1
ACTL6A
ACTB
ARID1A
ARID1A-insert
subdomain 4
(266-305)
90°
C
180°
ARID1A
WH
higher threshold level NCP
NCP
90°
SANT
SWIRM
Fig. 1. Cryo-EM structure of human BAF bound to NCP. (A) Schematic
architecture of BAF complex organization and domain structure of BAF
subunits. Color scheme for BAF subunits is indicated in (G) and used
throughout all figures. (Bto D) The 3.7-Å-resolution cryo-EM map of
BAF-NCP in the absence of ADP in three different views. The map at a
higher threshold level [(B), right panel] reveals more details, with the
ATPase module fading out. The map at lower threshold level [(B), left
panel] reveals density (likely derived from SnAC) packing across the top
surface of nucleosome. The intermodular contacts and BAF-NCP contacts
are indicated with white circles. (Eand F) Two different views of the
cryo-EM map of the nucleosome showing detached and disordered DNA
at entry (E) and exit (F) sites. The positions of nucleosomal DNA are
labeled with SHL numbers. (G) Cartoon model of BAF-NCP structure
shown in two different views. RPT, repeat domain; Req, Requiem
domain;SWIRM,SWI3,RSC8,andMOIRA;SANT,Swi3,Ada2,N-Cor,
and TFIIIB; ARM, armadillo repeats; CC, coiled coil; HSA, helicase-SANT–
associated; ARP, actin-related protein; BR, bromodomain; WH, winged
helix; SWIB, SWI/SNF complex B/MDM2.
RESEARCH |RESEARCH ARTICLE
on February 25, 2020 http://science.sciencemag.org/Downloaded from

and two SMARCC subunits. The palm con-
nectspre-HSA,SMARCD1,SMARCE1,and
SMARCC, which merge at a four-way junction.
Nucleosome recognition by SMARCB1 in the
head submodule
Theheaddirectlybindsthehistoneoctamer
on the bottom (Fig. 1, B and C, and fig. S7A).
It consists of two repeat (RPT) domains
and a C-terminal a(aC) helix of SMARCB1,
the requiem (Req) domain of DPF2, two SWIRM
domains of SMARCCs, and an insert derived
from ARID1A (ARID1A-insert) (Fig. 3A and
fig. S7B). The two SWIRM domains bind the
RPT1 and RPT2 domains, respectively. The
two RPT-SWIRM complexes adopt a similar
foldandasymmetricallybindeachother,with
the intermolecular interactions buttressed by
the ARID1A-insert (residues 1802 to 1862) and
Req domain (residues 13 to 82) of DPF2 (Fig. 3,
A and B). The head merges with the bridge
and thumb at a three-way junction, and the
interactions are mediated by the interwoven
loops from DPF2, ARID1A, and two SMARCCs
(Fig.3A).TheReqdomainofDPF2isalmost
identical to that of DPF1 and DPF3 in primary
sequence, indicating a similar role of DPF1,
DPF2, and DPF3 in BAF complex organization
(fig. S2D).
SMARCB1 is required for structural integ-
rity and function of the SWI/SNF complex
(16,17,35), and inactivation of SMARCB1
was reported in almost all malignant rhabdoid
tumors (14,36–38). The aC helix, which is
enriched in residues that are frequently mu-
tated in human cancers, packs against the
bottom H2A-H2B heterodimer and serves as a
hinge to connect the base module and nu-
cleosome (Fig. 3C and fig. S7A). The four most
frequently mutated arginine residues (R370,
R373, R374, and R377) together form a pos-
itively charged cluster and pack against an
acidic patch on the bottom of the nucleosome
(Fig. 3D). The four arginine residues are in-
variant across species, from yeast to humans
(Fig. 3E), and were clearly visualized in the
cryo-EM map (Fig. 3F). This observation high-
lights the importance of the aChelixof
SMARCB1 in binding of the nucleosome and
is consistent with recent studies (14,28). Al-
though the aC helix of Sfh1 (yeast homo-
log of SMARCB1) is disordered in the yeast
RSC-NCP complex structure, the deletion of
this helix impairs nucleosome ejection activ-
ity (28).
The RPT1 domain of SMARCB1 is in close
proximity to the a2 helix of histone H2A and
likely causes steric clash between the histone
octamer and DNA (Figs. 1, B and G, and 3C;
and fig. S5A), suggesting that RPT1 serves as a
“wedge”that favors DNA detachment around
the exit site. The N-terminal winged helix (WH)
domain of SMARCB1 binds the ARM domain
of ARID1A and is located more than 40 Å away
from nucleosomal DNA (Fig. 1, C and G), sug-
gesting a role independent of its previously
reported DNA binding ability (26). The WH
domain is in close proximity to the ARP module,
suggesting a role in regulating ARP–base in-
termodular interactions (Fig. 1, B and C).
ARID1A serves as a rigid core to stabilize the
base module
ARID1A is the largest subunit of the BAF
complex. A large portion of ARID1A, includ-
ing the characteristic AT-rich interacting do-
main (ARID), was invisible in the complex
str ucture, possibly owing to intrinsic flexibility
He et al., Science 367, 875–881 (2020) 21 February 2020 3of7
Bridge
Palm
Head
Thumb
Fingers
HSA
HSA
Pillar
Pillar
Req
Req
180°
Bridge
Head
Thumb
Fingers
HSA
HSA
Palm
ARID1A-insert
ARID1A-insert
pre-HSA
pre-HSA
pre-HSA
pre-HSA
Req
Req
αC
SWIB
SWIB
CC-E1
CC-E1
pre-HSA
pre-HSA
HSA
HSA
RPT1
RPT1
RPT2
RPT2
αC
SANT
SANT
a
*
CC-D1
CC-D1
CC-D2
CC-D2
Req
Req
ARM
ARM
Pillar
Pillar
Wedge
Wedge
CC-C2
CC-C2
b
*
SWIRM
SWIRM
a
SWIRM
SWIRM
b
SANT
SANT
b
CC-C2
CC-C2
a
*
A
90°
SMARCA4
ARID1A
SMARCB1
DPF2
SMARCD1
SMARCE1
SMARCC2
a
SMARCC2
b
B
Req
Req
HSA
HSA
RPT1
RPT1
RPT2
RPT2
WH
WH
αC
ARM
ARM
SWIB
SWIB
CC-E1
CC-E1
SWIRM
SWIRM
a
SWIRM
SWIRM
b
CC-C2
CC-C2
a
*
CC-C2
CC-C2
b
*
SANT
SANT
b
4-way junction
4-way junction
Fig. 2. Structure of the base module. (A) The cryo-EM map of the base module locally refined to 3.0-Å resolution. The submodules and elements involved in
complex assembly are indicated. (B) Cartoon model of the base module in two different views. Regions that are involved in nucleosome recognition and complex
formation are indicated. CC-C2
a
* and CC-C2
b
* represent the possibility of two CCs of SMARCC2 because of a disconnected cryo-EM map.
RESEARCH |RESEARCH ARTICLE
on February 25, 2020 http://science.sciencemag.org/Downloaded from

(fig. S2D). The modeled C-terminal ARM do-
main consists of seven ARM repeats arranged
in a superhelical conformation and serves as
a rigid core to bind SMARCA4 and all other
base subunits (Fig. 4A). A zinc finger is formed
with a zinc atom coordinated with residues
H2019 and H2021 on ARM3 and residues H2090
and C2094 on ARM4 (Fig. 4B). The zinc finger
stabilizes two ARM repeats and the associ-
ated loop regions. The ARM domain is lo-
cated in the center of the base module, with
the bottom associating with the fingers, one
endjoiningwiththeheadandthumb,and
theotherendcappedbytheSWIBdomainof
SMARCD1. The N-terminal helical turns of
HSA and the preceding loop pack against the
concave surface of the ARM domain. The sur-
face residues of ARM contacting SMARCA4
and SMARCD1 are highly conserved, suggest-
ing evolutionarily conserved functions in sta-
bilizingthebasemodule.
We found that, in a manner consistent with
the central role of the ARM domain in base
module organization, ARID1A bound to other
base subunits and the purified base subcom-
plex (fig. S8, A to C). ARID1A is one of the most
frequently mutated mSWI/SNF subunits in
human cancer, and frameshift mutations occur
near the C terminus (11,39), supporting the im-
portance of the C-terminal ARM domain in BAF
structure and function. The structure showed
that the overall fold of the BAF complex wo uld
not be properly maintained in the absence of
ARID1A (Fig. 4C), although other BAF subunits
would remain associated with SMARCC1 and
SMARCC2 because the BAF
DARID1A
complex sur-
vived ion-exchange chromatography (fig. S1C).
The sliding activity of BAF
DARID1A
was con-
siderably enhanced by the addition of larger
amounts of purified ARID1A (Fig. 4D and
fig. S8D), indicating that ARID1Ais required
for efficient nucleosome sliding activity.
ARID1A and ARID1B share highly similar
primary sequences and are mutually exclusive
in the BAF complex, suggesting that ARID1B-
containing BAF is assembled in a similar
manner (fig. S2D). As a PBAF-specific subunit,
ARID2 is not a paralog of ARID1A or ARID1B
and shows a distinct domain architecture.
However, the N-terminal helical region (res-
idues 150 to 470) of ARID2 is predicted to form
a seven-repeat ARM domain (40). Thus, the
N-terminal ARM domain of ARID2 and the
C-terminal ARM domain of ARID1A may play
a similar role in organizing PABF and BAF,
respectively.
SMARCCs compose base module scaffold, and
SMARCD1 and SMARCE1 facilitate organization
Asthescaffoldofthebasemodule,thetwo
SMARCC subunits bind all other base sub-
units and thread through the head, thumb,
palm, and fingers submodules (Fig. 4C). The
He et al., Science 367, 875–881 (2020) 21 February 2020 4of7
E
A
R374
R373
R370
R377
R377
R374
R370
R373
H2B
H2B
C
C
C
C
RPT2
RPT2
RPT1
RPT1
DNAexit
H2A
H2A
Wedge
Wedge
Head
R370
R373
R374
R377
R370
R373
R374
R377
D
C
C
F
Y16
Y17 M21
Q23
H25
Y27
N28
L31
R35
R38 F41
Req
Req
SWIRM
SWIRM
b
b
RPT2
RPT2
B
C
Fig. 3. Structure of the head submodule and role of SMARCB1 in
nucleosome recognition. (A) Close-up view of the head submodule and its
interactions with the bridge and thumb submodules. The sandwiched loops
located in a three-way junction are indicated with a dashed blue circle.
(B) Close-up view of interactions between the Req domain of DPF2 shown in
cartoon and RTP2-SWIRM
b
shown in surface representation (red, negative
charge; blue, positive charge). (C) Close-up view of the contacts between
SMARCB1 and the bottom surface of the nucleosome. Four histone-contacting
arginine residues on the aC helix are shown in stick representation.
Single-letter abbreviations for the amino acid residues are as follows: A, Ala;
C, Cys; D, Asp; E, Glu; F, Phe; G, Gly; H, His; I, Ile; K, Lys; L, Leu; M, Met;
N, Asn; P, Pro; Q, Gln; R, Arg; S, Ser; T, Thr; V, Val; W, Trp; and Y, Tyr. (D) Basic
residues of the aC helix pack against the acidic surface of the histone
octamer. The positively charged residues are shown as sticks. (E) Conservation
of the aC helix of SMARCB1. The invariant and highly conserved residues are
highlighted with dark-green and green backgrounds, respectively. (F) The
cryo-EM map around the aC helix is shown in mesh in two views. The side chains
of the abovementioned residues are well covered by the map.
RESEARCH |RESEARCH ARTICLE
on February 25, 2020 http://science.sciencemag.org/Downloaded from

BAF complexes containing SMARCC1 and
SMARCC2 (BAF-CC1/CC2) and two SMARCC2
subunits (BAF-CC2/CC2) generated almost
identical cryo-EM maps (fig. S3). SMARCC1
and SMARCC2 behave similarly in binding
other BAF subunits and subcomplexes (fig. S9,
D to I, and supplementary text). This obser-
vation is consistent with a previous study
showing that SMARCC1 and SMARCC2 are
functionally similar and play critical roles in the
early stage of mSWI/SNF complex assembly (11).
SMARCD1 adopts an elongated conforma-
tion and runs alongside the two SMARCC
subunits (Fig. 4C). The SWIB domain adopts
a compact globular fold and binds the ARM7
of ARID1A and the CC of SMARCC (Fig. 4A).
The a9 helix of SMARCD1 packs against the
a1 helix of SMARCE1 and a loop of ARID1A,
generating a “pillar”that connects to the
bridge (Figs. 2 and 4A). The a10, a11, and
a12 helices of SMARCD1 interact with the
SANT
b
domain and pre-HSA with the a12
helix protruding out of the intersection of the
head,bridge,andthumb(Fig.4,AandC).Thus,
SMARCD1 and SMARCE1 assist SMARCC1 and
SMARCC2 in organizing the base module.
Comparison of BAF-NCP and RSC-NCP
complex structures
The yeast RSC complex is the homolog of the
mammalian PBAF complex, consisting of ARID2
(instead of ARID1A and ARID1B) and a PBAF-
specific subunit, polybromo-1 (PBRM1, or
BAF180). Comparison of our BAF-NCP struc-
ture and a recently published yeast RSC-NPC
structure (28) revealed a similar nucleosome-
binding mode (fig. S10). Unexpectedly, the
nucleosome is mainly bound by the ATPase
domain, and the aC helix of Sfh1 is invisible
in the RSC-NCP structure (28). In contrast,
the nucleosome is sandwiched by BAF with
H2A-H2B dimer stably associated with the aC
helix of SMARCB1. The two complexes may
represent different conformational states, or
this difference might result from different ex-
perimental conditions. Structural comparison
also revealed considerable differences in the
base module organization. ARID1A, SMARCC1,
SMARCC2, SMARCD1, and SMARCE1 exhibit
conformations distinct from their counter-
parts in the RSC complex. The comparison
He et al., Science 367, 875–881 (2020) 21 February 2020 5of7
CC-D1
CC-D1
CC-D2
CC-D2
CC-E1
CC-E1
SWIB
SWIB
SANT
SANT
a
*
CC-C2
CC-C2b*
CC-C2
CC-C2
a
*
CC-D1
CC-D1
CC-D2
CC-D2
CC-E1
CC-E1
SWIB
SWIB
CC-C2
CC-C2b*
CC-C2
CC-C2
a
*
SMARCA4
ARID1A
SMARCB1
DPF2
SMARCD1
SMARCE1
SMARCC2
a
SMARCC2
b
SANT
SANT
a
*
SWIRM
SWIRM
a
SWIRM
SWIRM
b
SANT
SANT
b
Bridge
Fingers
Palm
Head
WH
WH
ARM7
ARM7
HSA
HSA
Bridge
SWIB
SWIB
SWIRM
SWIRM
a
SWIRM
SWIRM
b
Req
Req
ARID1A-insert
ARID1A-insert
Thumb
Fingers
Palm
Head
Bridge
HSA
pre-HSA
SWIB
CC-D2
SWIRM
a
CC-D1
ARM7
Req
SWIRM
b
A
100°
45N45 (80nM)
ATP
Time/min
BAF (120nM)
ARID1A (nM)
1 2 3 4 5 6 7 8
SMARCA4
ARM1
ARM1
ARM2
ARM2
ARM4
ARM4
ARM5
ARM5
ARM6
ARM6
ARM7
ARM7
ARM3
ARM3
SMARCD1
Variable Average Conserved
α9
α1
ARM3
ARM3
ARM4
ARM4
H2019
H2090
C2094
H2021
B
D
C
3-way
3-way
junction
junction
zinc finger
Pillar
α
10
10
α
11
11
α
12
12
Thumb
α9
α1
SANT
b
Th
u
m
Fin
ger
s
g
Pal
m
H
ea
d
Bri
e
Bri
dge
dg
dg
dg
Bri
e
HSA
pre
-HSA
SWI
B
CC-
D2
SWIRM
a
CC-
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
C
D1
ARM7
Req
SWI
R
A
zinc finger
Pillar
α
α
α
α
9
α
1
SANT
b
Fig. 4. Subunit organization in the base module. (A) Cartoon model of
interactions between ARID1A and other subunits is shown in two views. (Left middle
panel) The surface of the ARM domain is shown and colored according to the
conservation scores. (Right middle panel) Seven ARM repeats are shown, with front
helices colored in yellow, back helices colored in green, and ridge helices and loops
colored in blue. (B) Close-up view of the zinc finger connecting ARM3 and ARM4,
with zinc-coordinating residues shown in sticks and zinc atom shown as a gray ball.
The location of the zinc finger is indicated with a blue rectangle in the upper panel of
(A). (C)CartoonofthebasemodulewithARID1AandSMARCA4omitted.Thelackof
ARID1A would lead to a clash of the base module, although other subunits remain
associated by the scaffold subunits, SMARCC1 and SMARCC2. The three-pronged
yellow and blue shape indicates a three-way junction of the head, thumb, and bridge.
(D) In vitro nucleosome sliding assay performed using purified BAF
DARID1A
complex
and increasing amounts of purified ARID1A. Note that BAF
DARID1A
demonstrates
nucleosome sliding activity in higher protein concentration (fig. S8D). 45N45, a
center-positioned nucleosome with two 45–base pair flanking DNA fragments.
RESEARCH |RESEARCH ARTICLE
on February 25, 2020 http://science.sciencemag.org/Downloaded from

also suggested that Rsc7 and Htl1 in RSC are
equivalents of DPF2 and SMARCE1 of BAF,
respectively. BAF has no histone tail–binding
lobe, which may exist exclusively in RSC and
PBAF complexes. Homologs of yeast DNA
interaction subunits, RSC3 and RSC30, do
not exist in mammals, suggesting that PBAF
is also structurally different from RSC.
Model of chromatin remodeling of the
BAF complex
Our structural and biochemical analyses illus-
trated the mechanisms of organization and
nucleosome recognition of the human BAF
complex, the prototype mSWI/SNF remodeler
(Fig. 5A). BAF ejects the nucleosome and cre-
ates and maintains NDRs that are essential
for transcription. Two non–mutually exclusive
models of nucleosome ejection have been
proposed (2). The BAF-NCP structure could
fit both models (Fig. 5B). In the first model,
the BAF-bound histone octamer, and more
possibly the H2A-H2B dimer, could be evicted
because of DNA tension resulting from either
strong DNA translocation (Fig. 5B, model 1a;
independent of the adjacent nucleosome) or
collision with the adjacent nucleosome (Fig. 5B,
model 1b). In the second model, BAF sand-
wiches the nucleosome to ensure that the
ATPase domain stably engages with DNA and
“peels”the DNA off of the adjacent nucleo-
some.Inbothcases,thetwonucleosome-
sandwiching regions, the SnAC domain of
SMARCA4 and the aChelixofSMARCB1,
likely play important roles in nucleosome
ejection, which has been experimentally dem-
onstrated (14,16,28,33). The DNA detachment
may facilitate efficient DNA translocation of
the BAF-associated nucleosome because of
fewer DNA–histone contacts. The chromatin
ejection may also be promoted by the DNA-
interacting ARID domain of ARID1A and/or
acetylated histone tail–binding bromodomain
of SMARCA4, which were not observed in the
cryo-EM map, owing to flexibility.
REFERENCES AND NOTES
1. C. Y. Zhou, S. L. Johnson, N. I. Gamarra, G. J. Narlikar,
Annu. Rev. Biophys. 45, 153–181 (2016).
2. C. R. Clapier, J. Iwasa, B. R. Cairns, C. L. Peterson, Nat. Rev.
Mol. Cell Biol. 18, 407–422 (2017).
3. L. K. Elfring, R. Deuring, C. M. McCallum, C. L. Peterson,
J. W. Tamkun, Mol. Cell. Biol. 14, 2225–2234 (1994).
4. J. Masliah-Planchon, I. Bièche, J. M. Guinebretière, F. Bourdeaut,
O. Delattre, Annu. Rev. Pathol. 10,145–171 (2015).
5. L. Neigeborn, M. Carlson, Genetics 108, 845–858 (1984).
6. W. Wang et al., EMBO J. 15, 5370–5382 (1996).
7. C. Kadoch, G. R. Crabtree, Sci. Adv. 1, e1500447 (2015).
8. C. R. Clapier et al., Mol. Cell 62, 453–461 (2016).
9. M. L. Dechassa et al., Mol. Cell 38, 590–602 (2010).
10. H. Boeger, J. Griesenbeck, J. S. Strattan, R. D. Kornberg,
Mol. Cell 14, 667–673 (2004).
11. N. Mashtalir et al., Cell 175, 1272–1288.e20 (2018).
12. L. Ho, G. R. Crabtree, Nature 463, 474–484 (2010).
13. C. Kadoch et al., Nat. Genet. 45, 592–601 (2013).
14. A. M. Valencia et al., Cell 179, 1342–1356.e23 (2019).
15. L. Yan, S. Xie, Y. Du, C. Qian, J. Mol. Biol. 429, 1650–1660
(2017).
16. P. Sen et al., Cell Rep. 18, 2135–2147 (2017).
17. X. Wang et al., Nat. Genet. 49, 289–295 (2017).
18. H. Szerlong et al., Nat.Struct.Mol.Biol.15, 469–476
(2008).
19. M. L. Phelan, S. Sif, G. J. Narlikar, R. E. Kingston, Mol. Cell 3,
247–253 (1999).
20.M.L.Dechassaet al., Mol. Cell. Biol. 28,6010–6021
(2008).
21. Y. Chaban et al., Nat. Struct. Mol. Biol. 15, 1272–1277
(2008).
22. C. L. Smith, R. Horowitz-Scherer, J. F. Flanagan, C. L. Woodcock,
C. L. Peterson, Nat. Struct. Biol. 10,141–145 (2003).
23. F. J. Asturias, W. H. Chung, R. D. Kornberg, Y. Lorch, Proc. Natl.
Acad. Sci. U.S.A. 99, 13477–13480 (2002).
24. M. Li et al., Nature 567, 409–413 (2019).
25. E. A. Morrison et al., Nat. Commun. 8, 16080 (2017).
26. M. D. Allen, S. M. Freund, G. Zinzalla, M. Bycroft, Structure 23,
1344–1349 (2015).
27. H. L. Schubert et al., Proc. Natl. Acad. Sci. U.S.A. 110,
3345–3350 (2013).
28. Y. Ye et al., Science 366,838–843 (2019).
29. K. Luger, A. W. Mäder, R. K. Richmond, D. F. Sargent,
T. J. Richmond, Nature 389, 251–260 (1997).
30. L. Farnung, S. M. Vos, C. Wigge, P. Cramer, Nature 550,
539–542 (2017).
31. O. Willhoft et al., Science 362, eaat7716 (2018).
32. R. Ayala et al., Nature 556, 391–395 (2018).
He et al., Science 367, 875–881 (2020) 21 February 2020 6of7
ACTB
ACTL6A/B
sliding
?
DNA tension is resolved by
ejection of BAF-bound histone
octamer or dimer
ATPase
post-ATPase
SMARCB1
ARP
Base
ATPase
post-ATPase
SMARCB1
ARP
Base
SMARCB1
ARP
Base
ATPase
post-ATPase
ATPase
post-ATPase
SMARCB1
ARP
Base
Model 2Model 1b
ATPase
post-ATPase
SMARCB1
ARP
Base
Free DNA
ARP
ATPase
post-ATPase
SMARCB1
Base
ARP
strong DNA translocation and
high DNA tension lead to
ejection independent
of adjacent nucleosome
Model 1a
ATPase
SMARCB1
Base
Free DNA
DNA peel-off and ejection of
adjacent histone octamer or dimer
adjacent nucleosome collides
AB
ARP
SMARCB1
SMARCE1
ARID1A/B
DPF1/2/3
SMARCD1/2/3
H2A H2B
H3 H4
SMARCE1
ACTB
ACTL6A/B
SMARCA2/4
H2AH2B
H3
H4
SMARCC1/2
SMARCC1/2
Bromodomain PHD finger domain
DNA binding domain (ARID, HMG, WH)
ATP / A D P
Two CCs of SMARCD
Two CCs of SMARCCOne CC of SMARCE1
Rotation
Free DNA
Fig. 5. Models of chromatin remodeling of the BAF complex. (A) BAF
subunits are shown in cartoon model. The indicated paralogs of BAF subunits are
mutually exclusive in the BAF complex. The nucleosome is sandwiched by
SMARCA2/4 and SMARCB1. The ATPase of SMARCA2/4 grasps nucleosomal
DNA and generates DNA translocation in an ATP-dependent manner. PHD, plant
homeodomain; HMG, high-mobility group. (B) The BAF-NCP structure could
fit two nonexclusive models of nucleosome ejection. The histone octamer or
dimer could be ejected from BAF-bound nucleosome (model 1) or the adjacent
nucleosome (model 2). The ejection could occur independent of (model 1a) or
dependent upon (model 1b) the adjacent nucleosome.
RESEARCH |RESEARCH ARTICLE
on February 25, 2020 http://science.sciencemag.org/Downloaded from

33. P. Sen et al., Mol. Cell. Biol. 33, 360–370 (2013).
34. W. Shen et al., Biochemistry 46, 2100–2110 (2007).
35. R. T. Nakayama et al., Nat. Genet. 49, 1613–1623 (2017).
36. I. Versteege et al., Nature 394, 203–206 (1998).
37. C. Kadoch, G. R. Crabtree, Cell 153,71–85 (2013).
38. M. J. McBride et al., Cancer Cell 33, 1128–1141.e7 (2018).
39. R. Mathur et al., Nat. Genet. 49, 296–302 (2017).
40. L. A. Kelley, S. Mezulis, C. M. Yates, M. N. Wass,
M. J. Sternberg, Nat. Protoc. 10, 845–858 (2015).
ACKNOWL EDGME NTS
We thank the Center of Cryo-Electron Microscopy, Fudan
University; the Center of Cryo-Electron Microscopy, ShanghaiTech
University; the Center for Biological Imaging of Institute of
Biophysics (IBP) of Chinese Academy of Sciences (CAS); and
the National Center for Protein Science Shanghai (NCPSS) for
supporting cryo-EM data collection and data analyses. We also
thank the Biomedical Core Facility, Fudan University, for
supporting mass spectrometry analyses. Funding: This work was
supported by grants from the National Key R&D Program of China
(2016YFA0500700), the National Natural Science Foundation of
China (31830107, 31821002, and 31425008), the National Ten-
Thousand Talent Program (Y.X.), the National Program for Support
of Top-Notch Young Professionals (Y.X.), the Shanghai Municipal
Science and Technology Major Project (2017SHZDZX01), and the
Strategic Priority Research Program of the Chinese Academy of
Sciences (XDB08000000). Author contributions: S.H. prepared
the samples for structural and biochemical analyses with help from
J.Y., J.L., X.W., and B.L. Z.W. and Y.T. performed EM analyses
and model building with help from Z.Y. Y.X. and S.H. wrote the
manuscript. Y.X. supervised the project. Competing interests: The
authors declare no competing interests. Data and materials
availability: Cryo-EM maps have been deposited in the Electron
Microscopy Data Bank (EMDB) under accession numbers
EMD-0968 (base module), EMD-0969 (ARP module), EMD-0970
(ATPase-NCP with aC of SMARCB1), EMD-0972 (ATPase-NCP with
detached DNA), EMD-0974 (BAF-NCP, 3.7 Å), EMD-0971 (BAF-NCP,
6.6 Å), and EMD-0973 (BAF
ADP
-NCP). Atomic coordinates for the
base module and BAF-NCP have been deposited in the Protein Data
Bank under IDs 6LTH and 6LTJ, respectively.
SUPPLEMENTARY MATERIALS
science.sciencemag.org/content/367/6480/875/suppl/DC1
Materials and Methods
Supplementary Text
Figs. S1 to S10
Tables S1 and S2
References (41–61)
Movies S1 to S6
Data S1 and S2
View/request a protocol for this paper from Bio-protocol.
24 October 2019; resubmitted 10 January 2020
Accepted 22 January 2020
Published online 30 January 2020
10.1126/science.aaz9761
He et al., Science 367, 875–881 (2020) 21 February 2020 7of7
RESEARCH |RESEARCH ARTICLE
on February 25, 2020 http://science.sciencemag.org/Downloaded from

Structure of nucleosome-bound human BAF complex
Shuang He, Zihan Wu, Yuan Tian, Zishuo Yu, Jiali Yu, Xinxin Wang, Jie Li, Bijun Liu and Yanhui Xu
originally published online January 30, 2020DOI: 10.1126/science.aaz9761
(6480), 875-881.367Science
, this issue p. 875Science
its dysregulation in cancer.
region. This structure provides a framework for understanding the BAF-mediated chromatin remodeling mechanism and
remodelers. Mutations in BAF that are frequently associated with human cancer cluster into a nucleosome-interacting
nucleosome on the top, bottom, and side, making this nucleosome-recognition pattern distinct from other chromatin
determined the structure of the human BAF complex, which contains three modules that bind theet al.processes. He
PBAF are mammalian SWI/SNF remodelers that play essential functions in diverse developmental and physiological
The SWI/SNF family chromatin remodelers regulate chromatin and transcription. The protein complexes BAF and
Architecture of human BAF complex
ARTICLE TOOLS http://science.sciencemag.org/content/367/6480/875
MATERIALS
SUPPLEMENTARY http://science.sciencemag.org/content/suppl/2020/01/29/science.aaz9761.DC1
REFERENCES http://science.sciencemag.org/content/367/6480/875#BIBL
This article cites 60 articles, 11 of which you can access for free
PERMISSIONS http://www.sciencemag.org/help/reprints-and-permissions
Terms of ServiceUse of this article is subject to the
is a registered trademark of AAAS.ScienceScience, 1200 New York Avenue NW, Washington, DC 20005. The title
(print ISSN 0036-8075; online ISSN 1095-9203) is published by the American Association for the Advancement ofScience
Copyright © 2020, American Association for the Advancement of Science
on February 25, 2020 http://science.sciencemag.org/Downloaded from