RESEARCH ARTICLES
◥MOLECULAR BIOLOGY
Pol IV and RDR2: A two-RNA-polymerase machine
that produces double-stranded RNA
Kun Huang1,2†, Xiao-Xian Wu^1 †, Cheng-Li Fang1,2†, Zhou-Geng Xu2,3†, Hong-Wei Zhang1,2, Jian Gao2,3,
Chuan-Miao Zhou^3 , Lin-Lin You1,2, Zhan-Xi Gu1,2, Wen-Hui Mu^4 , Yu Feng^5 , Jia-Wei Wang^3 , Yu Zhang^1 *
DNA methylation affects gene expression and maintains genome integrity. The DNA-dependent RNA
polymerase IV (Pol IV), together with the RNA-dependent RNA polymerase RDR2, produces double-
stranded small interfering RNA precursors essential for establishing and maintaining DNA methylation in
plants. We determined the cryo–electron microscopy structures of the Pol IV–RDR2 holoenzyme and
the backtracked transcription elongation complex. These structures reveal that Pol IV and RDR2 form a
complex with their active sites connected by an interpolymerase channel, through which the Pol IV–
generated transcript is handed over to the RDR2 active site after being backtracked, where it is used
as the template for double-stranded RNA (dsRNA) synthesis. Our results describe a 'backtracking-
triggered RNA channeling' mechanism underlying dsRNA synthesis and also shed light on the
evolutionary trajectory of eukaryotic RNA polymerases.
T
o date, five multiple-subunit DNA-dependent
RNA polymerases (msDdRPs) have been
identified in eukaryotes. Polymerase (Pol)
I, Pol II, and Pol III are conserved in all
eukaryotic organisms ( 1 ), whereas Pol IV
and Pol V are specific to land plants ( 2 – 5 ). Dec-
ades of studies have demonstrated that Pol IV
and Pol V—together with an RNA-dependent
RNA polymerase (RDR2)—are key members
of the plant RNA-directed DNA methylation
(RdDM) pathway. Pol IV, in association with
RDR2, generates double-stranded RNAs (dsRNAs)
at the RdDM locus of genomic DNA ( 6 – 8 ). Sub-
sequently, the dsRNA precursors are processed
into 24-nucleotide (nt) small interfering RNAs
(siRNAs) that pair with Pol V–generated scaf-
fold RNAs to recruit the DNA methyltransferase
DRM2 for DNA methylation ( 9 – 13 ). There is
increasing evidence to suggest that the RdDM
pathway is essential for repressing transpos-
able elements, establishing genome imprinting,
and maintaining genome integrity ( 13 – 15 ).
Although it has been proposed that Pol IV is
derived from Pol II ( 16 , 17 ), the catalytic activ-
ity ( 18 – 20 ), regulatory transcription factors
( 21 – 25 ), and targeting of genomic loci of Pol IV
differ radically from those of Pol II ( 22 , 23 , 26 , 27 ).
Precisely how Pol IV transcription initiates,
elongates, and terminates, and how Pol IV part-
ners with RDR2 to produce dsRNA precursors
at specific genomic loci, remain elusive. To ad-
dress these questions, we solved cryo–electron
microscopy (cryo-EM) structures of theArabidopsis
Pol IV–RDR2 holoenzyme and Pol IV–RDR2
backtracked transcription elongation com-
plex (bTEC).A cryo-EM structure of theArabidopsis
Pol IVÐRDR2 holoenzyme
Direct purification of low-abundance, multiple-
subunit protein complexes for structural study
has not yet been achieved in plants. To cir-
cumvent this problem, we purified endogenous
Pol IV from an engineeredArabidopsis thaliana
cell line (T87) stably expressing 3xFLAG-tagged
NRPD1, the largest subunit of Pol IV (fig. S1, A
to C) ( 28 ). Liquid chromatography–mass spec-
trometry (LC-MS/MS) and SDS–polyacrylamide
gel electrophoresis (PAGE) analysis of the pu-
rified complex shows the presence of Pol IV
subunits and RDR2, suggesting that we ob-
tained a stable Pol IV–RDR2 complex (fig. S1D
and table S1), in agreement with previous re-
portsthatRDR2isassociatedwithPolIVinvivo
( 18 , 25 ). The purified Pol IV–RDR2 complex
is catalytically active in extending a Pol IV
RNA primer with a bipartite DNA scaffold
(fig. S1E) ( 18 ).
The structure of the Pol IV–RDR2 holoenzyme
was determined by using the single-particle
cryo-EM method at a nominal resolution of
3.86 Å (fig. S2 and movie S1). Pol IV resembles
a crab claw, the characteristic shape for all
msDdRPs (Fig. 1) ( 29 ). Structure superimpo-
sition of Pol IV with other msDdRPs confirmeda closer relationship between Pol IV and Pol II
than that between Pol I and Pol III (Fig. 2B and
fig. S4, A to C) ( 16 , 17 ). The structure of RDR2
is superimposable on the crystal structure
of the catalytic domain of a cellular-encoded
RNA-dependent RNA polymerase (cRdRP)
fromNeurospora crassa[root mean square
deviation (RMSD) 1.3 Å for all Caatoms; fig.
S4D] ( 30 ). Pol IV and RDR2 interact with each
other through a large otherwise solvent-exposed
surface of 2265 Å^2 , where the secondary chan-
nel of Pol IV approaches the active center of
RDR2 (Fig. 1C and movie S2).Pol IV has a characteristic active center cleft
Pol IV resembles Pol II in overall shape but dif-
fers in the conformations of key motifs in its
active center cleft (Fig. 2, A to C, and fig. S4,
E to G) ( 31 ). No large domain rearrangement
of the 10 subunits of Pol IV occurs compared
with Pol II, with the exception of the zinc rib-
bondomainofPolIVNRPB9subunit(NRPB9
zinc ribbon), which relocates to a new posi-
tion ~40 Å from where the RPB9 zinc ribbon
is located in Pol II and establishes extensive
interactions with NRPD1 funnel helices (NRPD1
FH) at the new site (Fig. 2F). The Pol IV–specific
DeCL-like domain (Fig. 1A) is disordered in our
structure. Similar to Pol II, Pol IV contains a
main cleft for accommodating DNA and RNA
at the deep bottom of the two lobes (Fig. 2A).
Protein sequence alignment reveals that sev-
eralkeymotifsinsidethemaincleftofPolIV—
the rudder loop, the lid loop, the bridge helix,
and the trigger loop—exhibit consensus se-
quences distinct from those in Pol II (fig. S6
and table S3) ( 31 ). In our Pol IV–RDR2 structure,
these key structural elements either become dis-
ordered or adopt distinct conformations com-
pared with those in the Pol II structure (Fig.
2Candfig.S4,EtoG,andI).
In the main cleft of Pol IV, the rudder and
lid loops exhibit disorder, rendering the entire
active center cleft wide open and indicating
that the two loops are flexible and exist as an
ensemble of multiple conformations (Fig. 2C
and fig. S9B). In the active site of Pol IV, the
bridge helix bends in the middle (Fig. 2C and
fig. S9C). The two-residue deletion at the
central region compared with Pol II accounts
for the bent bridge helix of Pol IV, and also
likely hinders a bent-to-straight conforma-
tional transition (necessary for the nucleotide
addition cycle of Pol II) (fig. S6). The trigger
loop of Pol IV adopts an open conformation in
our structure (Fig. 2C and fig. S9D). It bears a
six-residue deletion and sequence variation at
a key position for catalysis; i.e., Pol IV NRPD1
contains S898 at the position of the conserved
histidine (H1085 inSaccharomyces cerevisiae
Pol II–RPB1 subunit) in Pol II (fig. S6) ( 32 ). In
summary, our structures reveal substantial
conformational differences in key structural
motifs at the Pol IV active site (i.e., the rudderRESEARCHSCIENCEscience.org 24 DECEMBER 2021•VOL 374 ISSUE 6575 1579
(^1) Key Laboratory of Synthetic Biology, CAS Center for Excellence
in Molecular Plant Sciences, Shanghai Institute of Plant
Physiology and Ecology, Chinese Academy of Sciences,
Shanghai 200032, China.^2 University of Chinese Academy of
Sciences, Beijing 100049, China.^3 National Key Laboratory
of Plant Molecular Genetics, CAS Center for Excellence in
Molecular Plant Sciences, Shanghai Institute of Plant
Physiology and Ecology, Chinese Academy of Sciences,
Shanghai 200032, China.^4 Key Laboratory of Plant Stress
Biology, State Key Laboratory of Cotton Biology, School of
Life Sciences, Henan University, Kaifeng 475004, China.
(^5) Department of Biophysics, and Department of Pathology of
Sir Run Run Shaw Hospital, Zhejiang University School of
Medicine, Hangzhou 310058, China.
*Corresponding authors. Email: [email protected] (Y.F.);
[email protected] (J.-W.W.); [email protected] (Y.Z.)
These authors contributed equally to this work.