Science - USA (2021-11-12)

contact the SPT5 domain KOWx-4, which
forms a clamp on exiting RNA (Fig. 2B and figs.
S4H, S5C, and S6D). The active center groove of
the INTS11 endonuclease is located between
its MBL andb-CASP domains and is closed in

the unbound structures of the Integrator-

PP2A complex ( 15 ) and the cleavage module

( 19 ). However, theb-CASP domain has rotated

away from the MBL domain by ~17° in our

structure, opening the active center groove

such that the endonuclease may accommo-

date RNA (Fig. 3A and movie S3). Superposi-

tion of the closed INTS11 ( 19 ) onto our structure

shows that theb-CASP domain would clash

with the KOWx-4 domain (fig. S7C). Thus,

binding of Integrator to the PEC apparently

leads to opening of the endonuclease active

center groove, allowing accommodation of

the RNA substrate for cleavage.

The observed open conformation of INTS11

resembles the active conformation of the re-

lated CPSF73 endonuclease when bound to

an RNA substrate ( 26 ). Superposition of the

CPSF73-RNA complex structure ( 26 )ontoour

structure showed that RNA that exits from

Pol II can continue straight into the active

center groove of INTS11, and that a minimum

of ~20 nucleotides of RNA are required to

reach from the Pol II active site to the INTS11

active site (Fig. 3B). This agrees very well with

the observation that Integrator-mediated RNA

cleavage occurs 35 and 21 nt downstream of

the transcription start site in vivo, when the

active site of paused Pol II resides at 55 and

41 nt downstream, respectively ( 9 , 11 ). The

cleavage module also cross-links to the flex-

ible INTS10-13-14 module (figs. S5A and S6E),

which is known to bind RNA ( 21 ) and may

transiently participate in RNA recognition

and cleavage. In summary, interface C may

serve to guide nascent RNA to the Integrator

endonuclease for RNA cleavage.

To test these predictions, we performed RNA

cleavage assays with two different RNA sub-

strates, the hairpin RNA used in our structure

and a single-stranded RNA (Fig. 3C and fig.

S10). We found that Integrator cleaves RNA in

the presence of the PEC but not in its absence,

which is consistent with endonuclease activa-

tion upon PEC binding (Fig. 3C). RNA cleavage

can occur as close as ~20 nt from the RNA 3′

end, consistent with our structure and with in

vivo observations ( 9 , 11 ). Cleavage is more effi-

cient for single-stranded RNA (fig. S10), sug-

gesting that hairpin structures in nascent RNA

interfere with Integrator-mediated RNA cleav-

age. This model is consistent with our previous

finding that hairpins in pre-mRNA transcripts

correlate with Pol II accumulation in the

promoter-proximal region ( 27 ) and with the

observation that 3′processing of snRNAs oc-

curs in the single-stranded region downstream

of an RNA hairpin ( 28 – 30 ).

Interface D is formed between INTS6 and

NELF-B. A loop in INTS6 (residues 403 to 411)

binds to helixa9 of NELF-B and the surround-

ing residues (Fig. 2B and figs. S4F and S5D).

Whereas Integrator regions that form interfaces

A to C are generally conserved over species

(fig. S9), the INTS6 loop is poorly conserved in

invertebrates such asC. elegans, which lack

NELF (fig. S9D), suggesting the absence of in-

terface D in invertebrates. The C-terminal region

of INTS6 is mobile and is known to bind to

884 12 NOVEMBER 2021•VOL 374 ISSUE 6569 science.orgSCIENCE

B

C

Front view

INTS1 N-terminus

NELF

INTS 8

INTS5INTS 5

Phosphatase

Integrator core

PP2A-A

PP2A-C

INTS7

INTS 2

INTS 4

INTS9

INTS 6

Upstream DNA

Back view

Exiting RNA

90°

180°

90°

INTS 11

KOW2-3

KOW5

SPT5 NGN-KOW1

SPT4

DSIF

INTS5INTS 5

Side view INTS^8 Top view

1

200 400 600 800 1000 1200 1400 1600 1800 2000

INTS12 binding

RPB2 binding INTS2 binding

N-terminal region C-terminal region

INTS1

INTS2 Ncap Helical repeats

(^1242) 490 515 1204
33 581608 820 963
INTS4 HEAT repeats HB
-sandwich
INTS7 -sandwich
-barrel
1 537 962
HEAT HB
787
vWA
INTS5
1 290 1019
Helical repeats
INTS8
1 75 995
Helical repeats
INTS6
1 390420 530 600 887
Interface D
Interface A
interface B
Interface B INTS9^1 MBL^272 -CASP^499 MBL^658
INTS11
1 213 449 600
-CASP MBL
Interface C
MBL
Integrator core
Cleavage module
PP2A-A
PP2A-C
1 589
1 327
Helical repeats
A
vWA
INTS1
C-terminus
-barrel Segment 1-4
Phosphatase
Pol II INTS1 N-terminus
Cleavage module
INTS1
C-terminus
KOWx-4
Fig. 1. Structure of PEC-Integrator-PP2A complex.(A) Domain organization of Integrator and PP2A
subunits. Regions that form interfaces with the PEC are indicated. Black solid and dashed lines indicate
regions that are modeled fully or as backbone traces, respectively. Color code is used throughout. HB, helical
bundle. (BandC) Four views of the structure. Pol II is shown as a gray surface, DSIF in green, and
NELF in pink. Nucleic acids are in blue, cyan, and red for DNA template, DNA non-template, and RNA,
respectively. The canonical location of the DNA clamp formed by the SPT4 and SPT5 NGN domain
and the SPT5 KOW1 domain is indicated with a dashed black circle. The active site Mn atoms of PP2A-C
are depicted with red spheres.
RESEARCH | REPORTS