TRANSCRIPTION
Structural basis of Integrator-mediated
transcription regulation
Isaac Fianu^1 , Ying Chen^1 , Christian Dienemann^1 , Olexandr Dybkov2,Andreas Linden3,4,
Henning Urlaub3,4, Patrick Cramer^1 *
Integrator and protein phosphatase 2A (PP2A) form a complex that dephosphorylates paused RNA
polymerase II (Pol II), cleaves the nascent RNA, and terminates transcription. We report the structure
of the pretermination complex containing the human Integrator-PP2A complex bound to paused Pol II.
Integrator binds Pol II and the pausing factors DSIF and NELF to exclude binding of the elongation
factors SPT6 and PAF1 complex. Integrator also binds the C-terminal domain of Pol II and positions
PP2A to counteract Pol II phosphorylation and elongation. The Integrator endonuclease docks to the
RNA exit site and opens to cleave nascent RNA about 20 nucleotides from the Pol II active site.
Integrator does not bind the DNA clamps formed by Pol II and DSIF, enabling release of DNA and
transcription termination.
I
ntegrator is a cofactor of RNA polymerase II
(Pol II)–dependent processing of small
nuclear RNAs (snRNAs) ( 1 ) and other non-
coding RNAs ( 2 – 4 ). It enables 3′process-
ing of snRNAs and transcription termination
by interacting with the Pol II–associating fac-
tors DSIF and NELF ( 5 ). Integrator also func-
tions together with DSIF and NELF in the
promoter-proximal region of protein-coding
genes ( 6 , 7 ), where it facilitates premature
termination in a process called transcription
attenuation ( 8 – 13 ). Integrator catalyzes endo-
nucleolytic cleavage of nascent RNA for termi-
nation ( 9 , 10 ). In addition, Integrator counteracts
the transition of paused Pol II to active elongation
by recruiting protein phosphatase 2A (PP2A),
which dephosphorylates Pol II and DSIF ( 14 – 16 ),
and attenuates transcription genome-wide
( 12 ) and also in enhancer regions ( 11 , 12 ).
Integrator has a molecular weight of
~1.5 MDa and contains 14 subunits called
INTS1 to INTS14 ( 17 ). INTS11 is the RNA endo-
nuclease and associates with INTS9 and INTS4
to form the catalytic cleavage module of In-
tegrator ( 18 ). The structure of the cleavage
module ( 19 ) confirmed that INTS11 and INTS9
are closely related to CPSF73 and CPSF100,
respectively, which are subunits of the cleav-
age and polyadenylation specificity factor that
functions in pre-mRNA 3′processing and Pol II
termination at polyadenylation sites at the
3 ′end of protein-coding genes ( 20 ). The struc-
ture of Integrator was recently reported and
revealed nine of its subunits and the asso-
ciated PP2A phosphatase but not the N-terminal
region of INTS1 or subunits INTS3, INTS10,
INTS12, INTS13, and INTS14 ( 15 ). INTS10, INTS13,
and INTS14 form the INTS10-13-14 subcomplex
that has also been analyzed structurally ( 21 ).
To study Integrator-mediated Pol II regula-
tion, we reconstituted the 14-subunit human
Integrator from three recombinant subcom-
plexes, an eight-subunit Integrator core, the
cleavage module, and the INTS10-13-14 module
(Fig. 1A; fig. S1, A to C and H, and materials
and methods). We also prepared recombinant
human PP2A containing subunits PP2A-A and
PP2A-C (fig. S1D). Integrator and PP2A formed
a stable complex with a paused Pol II-DSIF-
NELF elongation complex (PEC), which we
prepared as described previously ( 22 ) (fig. S1,
I and J). The resulting 34-subunit PEC-
Integrator-PP2A complex was subjected to
cryo–electron microscopy (cryo-EM) analysis
(fig. S2 and table S1). We obtained a high-
confidence structure of the complex at 3.6-Å
resolution and used focused refinement to
resolve the PEC at 2.7-Å resolution and regions
of Integrator and PP2A at 2.9- to 3.2-Å reso-
lution (figs. S2 to S4, tables S1 and S2, and
movies S1 and S2). We further used BS3 cross-
linking and mass spectrometry to confirm the
structure and to localize flexible regions (figs.
S5 and S6 and data S1).
The structure shows that Integrator embra-
ces the PEC (Fig. 1 and movie S1). Pol II adopts
a paused conformation and contains a tilted
DNA-RNA hybrid (fig. S4B) that impairs load-
ing of the next nucleoside triphosphate sub-
strate into the active center ( 22 ). NELF binds
to Pol II as in the known PEC structure ( 22 )
and shows only minor alterations in conforma-
tion (fig. S7A). With respect to DSIF, we ob-
served the SPT5 domains KOW2-3, KOWx-4,
and KOW5, whereas other SPT5 domains and
SPT4 are disordered ( 23 ). Comparison of our
structure with the structure of the unbound
Integrator-PP2A complex ( 15 ) shows that sev-eral Integrator elements move upon PEC bind-
ing (fig. S7B and movie S3). As in the reported
structure ( 15 ), Integrator subunits INTS3,
INTS10, INTS12, INTS13, and INTS14 are mo-
bile and therefore not present in our structure.
Superposition of our structure onto recently
reported structures of Pol II preinitiation com-
plexes shows that Integrator is sterically ex-
cluded by bound initiation factors and Mediator
(fig. S7D).
Integrator approaches Pol II from the side
opposite the active center cleft and forms four
contact areas that we refer to as interfaces A to
D (Fig. 2 and movie S4). Integrator contacts
thePolIIsubunitRPB2(interfaceA),thePolII
subunit dimer RPB3-RPB11 (interface B), the
DSIF subunit SPT5 (interface C), and the
NELF subunit NELF-B (interface D). These
contacts with Pol II, DSIF, and NELF explain
how Integrator recognizes the PEC, which is
thought to form mainly in the 5′region of
protein-coding genes and at the 3′end of non-
coding genes.
Interface A is formed by contacts between a
helical part in the N-terminal region of INTS1
(residues 357 to 1316) and RPB2. The INTS1 N-
terminal region contacts the RPB2 external
domains 1 and 2, particularly helicesa15,a16,
anda17 ( 24 ) (Fig. 2A). The N-terminal region
of INTS1 was largely absent in the unbound
Integrator structure ( 15 ) but is observed in our
structure and adopts a defined location upon
PEC binding. Whereas residues ~357 to 948 of
the INTS1 N-terminal region form a helical
domain that approaches RPB2, residues 949
to 1316 bind INTS2 and correspond to a pro-
tein region that could not be assigned previ-
ously ( 15 ). The INTS1 N-terminal region is
connected to the INTS1 C-terminal region by
an ~40-residue mobile linker that runs along
INTS7 (Fig. 1B and fig. S6A). The N-terminal
residues 1 to 300 of INTS1 remain mobile, cross-
link to the Pol II clamp and protrusion ( 24 ),
and form a complex with INTS12 that cross-
links to the NELF and Pol II regions close to
downstream DNA (fig. S6C). An interaction of
INTS1 with INTS12 occurs inDrosophilaIn-
tegrator and is essential for snRNA 3′process-
ing ( 25 ).
Interface B is formed by Integrator subunits
INTS2 and INTS7, which contact RPB11 and
RPB3, respectively (Fig. 2A). Whereas INTS2
makes a minor contact with helixa3 of RPB11
( 24 ), a more intimate contact is formed by the
INTS7 C-terminal helixa40 that is disordered
in the unbound Integrator-PP2A structure ( 15 ).
Helixa40 of INTS7 binds to domain 2 of RPB3
( 24 ) and contacts INTS9 in the Integrator cleav-
age module.
Interface C is formed between the Integrator
cleavage module and the PEC region around
the RNA exit site of Pol II (Fig. 2B and fig. S4H).
INTS9 is located close to the KOW5 domain of
SPT5. The MBL andb-CASP domains of INTS11SCIENCEscience.org 12 NOVEMBER 2021•VOL 374 ISSUE 6569 883
(^1) Department of Molecular Biology, Max Planck Institute for
Biophysical Chemistry, 37077 Göttingen, Germany.
(^2) Department of Cellular Biochemistry, Max Planck Institute
for Biophysical Chemistry, 37077 Göttingen, Germany. 3
Bioanalytical Mass Spectrometry, Max Planck Institute for
Biophysical Chemistry, 37077 Göttingen, Germany.
(^4) Institute of Clinical Chemistry, Bioanalytics Group,
University Medical Center Göttingen, 37075 Göttingen,
Germany.
*Corresponding author. Email: [email protected]
RESEARCH | REPORTS