Article

Arabidopsis Serrate Coordinates Histone

Methyltransferases ATXR5/6 and RNA Processing
Factor RDR6 to Regulate Transposon Expression
Graphical Abstract Authors
Zeyang Ma, Claudia Castillo-González,
Zhiye Wang, ..., Xuefeng Shen,
Magdalena E. Potok, Xiuren Zhang

Correspondence
xiuren.zhang@tamu.edu

In Brief
Serrate is a key factor in RNA metabolism.
Here, Ma et al. uncover its new roles in
regulating TE expression: SE promotes
ATXR5/6-mediated H3K27me1 mark to
repress TE transcription while protecting
TE transcripts from the RDR6 silencing
activity. Thus, Serrate coordinates the
TGS and PTGS machineries to fine-tune
TE expression.

Highlights
d Serrate interacts with H3K27me1 methyltransferase ATXR5/6

d Serrate positively regulates ATXR5/6 function to repress

transposable element (TE) expression

d Serrate stabilizes TE transcript via counteracting the

silencing function of RDR6

d Serrate coordinates the TGS and PTGS machineries to

control the level of TE transcripts

Ma et al., 2018, Developmental Cell 45, 769–784

June 18, 2018 ª 2018 Elsevier Inc.
https://doi.org/10.1016/j.devcel.2018.05.023
Developmental Cell
Article
Arabidopsis Serrate Coordinates Histone
Methyltransferases ATXR5/6 and RNA Processing
Factor RDR6 to Regulate Transposon Expression
Zeyang Ma,1,2 Claudia Castillo-González,1,2 Zhiye Wang,1,2 Di Sun,1,2 Xiaomei Hu,1,2 Xuefeng Shen,1,2,3
Magdalena E. Potok,4 and Xiuren Zhang1,2,5,*
1Department of Biochemistry and Biophysics, Texas A&M University, College Station, TX 77843, USA
2Institutefor Plant Genomics and Biotechnology, Texas A&M University, College Station, TX 77843, USA
3College of Agriculture, South China Agricultural University, Guangzhou 510642, China
4Department of Molecular, Cell and Developmental Biology, University of California at Los Angeles, Los Angeles, CA 90095, USA
5Lead Contact
*Correspondence: xiuren.zhang@tamu.edu
https://doi.org/10.1016/j.devcel.2018.05.023
SUMMARY
Serrate (SE) is a key component in RNA metabolism.
Little is known about whether and how it can regulate
epigenetic silencing. Here, we report histone methyl-
transferases ATXR5 and ATXR6 (ATXR5/6) as novel
partners of SE. ATXR5/6 deposit histone 3 lysine 27
monomethylation (H3K27me1) to promote hetero-
chromatin formation, repress transposable elements
(TEs), and control genome stability in Arabidopsis. SE
binds to ATXR5/6-regulated TE loci and promotes
H3K27me1 accumulation in these regions. Further-
more, SE directly enhances ATXR5 enzymatic activity
in vitro. Unexpectedly, se mutation suppresses the
TE reactivation and DNA re-replication phenotypes
in the atxr5 atxr6 mutant. The suppression of TE
expression results from triggering RNA-dependent
RNA polymerase 6 (RDR6)-dependent RNA silen-
cing in the se atxr5 atxr6 mutant. We propose that
SE facilitates ATXR5/6-mediated deposition of the
H3K27me1 mark while inhibiting RDR6-mediated
RNA silencing to protect TE transcripts. Hence, SE
coordinates epigenetic silencing and RNA process-
ing machineries to fine-tune the TE expression.
INTRODUCTION
TEs make up a substantial portion of most eukaryotic genomes.
Due to potentially deleterious effects, TEs usually undergo tran-
scriptional gene silencing (TGS) (Matzke and Mosher, 2014;
Sienski et al., 2012). However, TEs can also contribute to the
regulatory variation of endogenous gene expression and impact
the organism’s fitness in certain circumstances, like environ-
mental stresses (Iwasaki and Paszkowski, 2014; Lisch, 2013;
Makarevitch et al., 2015). Hence, the reactivation and
expression of TEs need to be tightly regulated to meet various
functional needs. One mechanism to control TE expression is
through chromatin modifications, such as histone methylation
Developmental Cell 45, 769–784, June 18, 2018 ª 2018 Elsevier Inc. 769
(Tan et al., 2011). H3K27 methylations, present in mono-, di-,
and trimethylated states (H3K27me1/2/3), are repressive marks
(Deal and Henikoff, 2011). In animals, the polycomb repressive
complex 2 (PRC2) accounts for H3K27me3 but also affects all
forms of H3K27 methylation (Jiao and Liu, 2015). In Arabidopsis,
H3K27me2/3 are primarily deposited by PRC2 (Liu et al., 2010),
whereas H3K27me1 is predominantly catalyzed by ATXR5/6
(Jacob et al., 2009). H3K27me2/3 are repressive marks in the
facultative heterochromatin and are responsible for the repres-
sion of many coding genes (Margueron and Reinberg, 2011). In
contrast, H3K27me1 is mainly enriched at heterochromatin
where TEs and repetitive sequences are located. Loss of
ATXR5/6 causes a reduction in H3K27me1, transcriptional
activation of TEs, and a genomic instability defect termed DNA
re-replication (Jacob et al., 2009, 2010).
Regulation of endogenous loci also involves cooperation be-
tween the TGS and posttranscriptional gene silencing (PTGS)
pathways. For examples, Dicer-like 1 (DCL1), a core enzyme in
the plant microprocessor (Zhu et al., 2013), participates in
silencing of a subset of transposons through positive regulation
of DNA methylation (Laubinger et al., 2010). Second, DCL1-
dependent microRNAs (miRNAs) can target TE transcripts,
which are canonically repressed by the TGS pathway, to pro-
duce epigenetically activated small interfering RNAs (easiRNAs)
via RNA-dependent RNA polymerase 6 (RDR6) and DCL4
(Creasey et al., 2014). The easiRNAs not only function through
RNA-induced silencing complexes (RISC) to clear transcripts
of reactivated TEs but also affect the DNA methylation status
at the corresponding loci. Third, AGO2 and DCL2, among other
PTGS components, can direct DNA methylation by priming an
RNA Polymerase II–RDR6-engaged RNA-directed DNA methyl-
ation (RdDM) pathway (Matzke et al., 2015). In addition, cap-
binding complex (CBC), best known for its roles in RNA process-
ing, interacts with H3K4me3 and H3K36me3 methyltransferases
complexes. The CBC and histone methyltransferases function
interdependently to regulate the activation of gene expression
and efficient pre-mRNA splicing (Li et al., 2016).
SE, an ortholog of the mammalian Ars2, is known as an essen-
tial component of the plant microprocessor, acting with DCL1
and HYL1 to produce miRNAs (Gruber et al., 2009; Sabin
et al., 2009; Yang et al., 2006). SE also participates in splicing
A B
35S: ATXR5-CFP 35S: YFP-SE Merged
PHD PIP NLS SET PHD PIP NLS SET
ATXR5 379 ATXR6 349

ATXR5ΔC 133 ATXR6ΔC 103

ATXR5ΔN 134 379 ATXR6ΔN 104 349

35S: ATXR6-CFP 35S: YFP-SE Merged

C
AD BD SD-TL SD-TLH AD BD SD-TL SD-TLH
Vector SE Vector SE
ATXR5 Vector ATXR6 Vector
ATXR5ΔC Vector ATXR6ΔC Vector
35S: SE-CFP 35S: DCL1-YFP Merged ATXR5ΔN Vector ATXR6ΔN Vector
ATXR5 SE ATXR6 SE
ATXR5ΔC SE ATXR6ΔC SE
ATXR5ΔN SE ATXR6ΔN SE
Positive control Positive control
Negtive control Negtive control

D Input α-Myc IP Input α-Myc IP E Input α-Flag IP

35S-3HA-SE 35S-3HA-SE 35S-3HA-SE 35S-3HA-SE RNaseA RNaseA
- +
-ATXR5ΔPIP

-ATXR5ΔPIP

PATXR5-FM-ATXR5

PATXR5-FM-ATXR5

PATXR5-FM-ATXR5
35S-FM

35S-FM
35S-FM

35S-FM
35S-FM
-ATXR5

35S-FM
-ATXR5

-SGS3

-SGS3
Empty
Vector
Empty
Vector

kDa kDa

Col-0

Col-0

Col-0
100 100 SGS3
75 75 kDa
ATXR5
ATXR5
50 ATXR5
50 ΔPIP 50
100 100
SE SE
75 75
50 50 100 SE
ACTIN ACTIN

F G H
ZnF CR AD BD SD-TL SD-TLH
MBP Pull down SE 702
Vector SE-a
SE-a 240
His-Sumo-SE Vector SE-b
SE-b 239 470
Vector SE-c
MBP-ATXR5ΔPIP

SE-c 469 702

SE-d 543 ATXR5 Vector
469
MBP-ATXR5

MBP-ATXR6

SE-e 544 641 ATXR5 SE-a

SE-f 642 702 ATXR5 SE-b
SE-g 469Δ(642-694) 702 ATXR5 SE-c
Mock

MBP

His-Sumo-SE
Western

kDa SE-h 469 695 Positive Control

blot

AD BD SD-TL SD-TLH AD BD SD-TL SD-TLH

100
Vector SE-d Vector SE-f
Vector SE-e Vector SE-g
Vector SE-f Vector SE-h
70 ATXR5 Vector ATXR5 Vector
ATXR5 SE-d ATXR5 SE-f
ATXR5 SE-e ATXR5 SE-g
ATXR5 SE-f ATXR5 SE-h
50 Positive Control Positive Control

Figure 1. ATXR5 Is a Bona Fide Partner of SE in Arabidopsis

(A) Confocal microscopy image shows co-localization of 35S-YFP-SE with 35S-ATXR5/6-CFP and 35S-DCL1-YFP with 35S-SE-CFP in N. benthamiana.
Bar, 5 mM.
(B) Schematic diagram of the full-length and truncated ATXR5/6 used in the Y2H assay. NLS, nuclear localization signal; SET, su(var) E(Z) trithorax domain.
(C–F) Validation of SE and ATXR5/6 interaction by Y2H assays (C), co-IP in N. benthamiana (D) and in Arabidopsis (E), and in vitro pull-down assays (F). In (C), AD,
GAL4 activation; BD, DNA-binding domain. SD, synthetic dropout medium.
TLH, minus Trp, Leu, and His. Positive control, pGADT7-bC1 and pGBKT7-NtRFP.
(legend continued on next page)

770 Developmental Cell 45, 769–784, June 18, 2018

of pre-mRNA, especially in the processing of the first introns
(Laubinger et al., 2008; Raczynska et al., 2014). This function is
fulfilled presumably through the interaction with the CBC (Hallais
et al., 2013; Laubinger et al., 2008; Raczynska et al., 2014).
Previous studies show that the expression of some TEs and
long intergenic noncoding RNAs (lncRNAs) are also affected in
se mutants (Laubinger et al., 2010; Liu et al., 2012), raising the
possibility that SE might function in transcriptional regulation.
Indeed, Ars2 has recently been reported as a new transcription
factor in mammals, directly participating in the transcriptional
regulation of the pluripotency factor Sox2 (Andreu-Agullo et al.,
2012). However, whether and how SE regulates epigenetic
silencing globally has not been explored.
In this study, we identified ATXR5/6 as components of SE in
plants. We found that SE is essential for the maintenance of
ATXR5/6-catalyzed H3K27me1 in vivo and this role is indepen-
dent of its canonical function in RNA metabolism. SE protein as-
sociates with the ATXR5/6-regulated TE loci in vivo and directly
enhances ATXR5 enzymatic activity in vitro. Surprisingly,
although SE is a positive regulator of H3K27me1 repressive
mark, the se mutation does not generally activate the expression
of ATXR5/6-regulated TE loci and DNA re-replication. Rather, the
se mutation suppresses the transcriptional activation of TEs and
DNA re-replication phenotypes in the atxr5 atxr6 mutant. We also
found that the suppression of TE expression in the se-2 atxr5
atxr6 mutant results from triggering the RDR6-dependent RNA
silencing. Thus, SE plays dual roles in regulating TE expression
at both the TGS and PTGS levels. Whereas SE promotes
ATXR5/6-mediated H3K27me1 deposition, this protein also pro-
tects the reactivated TE transcripts through repressing the
RDR6-dependent PTGS pathway. This study provides new
insight into the synchronization between the epigenetic and
RNA silencing machineries to tightly control the TE expression
in the eukaryotes.
RESULTS
ATXR5 and ATXR6 Are Bona Fide Partners of SE
To study the function of SE, we proposed to identify its cofactors
in vivo. To this end, we generated 35S-Flag-4Myc(FM)-SE trans-
genic plants and isolated the SE complexes by two-step immu-
noprecipitation (Wang et al., 2018). Mass spectrometry (MS)
analysis of the immunoprecipitate recovered ATXR5 from the
FM-SE transgenic plants but not from that of the control Col-0
(Figure S1A). ATXR5 has a genetic paralog, ATXR6, with
62.3% amino acid (aa) sequence similarity and 52.6% identity
(Figure S1B). Both proteins are H3K27me1 methyltransferases
and function in epigenetic silencing of TEs (Jacob et al., 2009).
ATXR6 was not detected by our MS analysis, likely because
ATXR5 transcript is much more abundant than that of ATXR6
in apex and leaf tissues (Jacob et al., 2009). Moreover, a single
mutant of atxr5, but not atxr6, exhibits a defect in rRNA gene
expression, which is proposed to be regulated by H3K27me1,
while the atxr5 atxr6 double mutant shows synergistic effect
(Pontvianne et al., 2012), suggesting that ATXR5 predominantly
contributes to certain physiological conditions. Therefore,
ATXR5 methyltransferase is the focus in our study with some ex-
periments including ATXR6.
To verify the proteomics result, we first evaluated SE and
ATXR5/6 localization in vivo by confocal microscopy. SE and
ATXR5/6 tagged with yellow fluorescent protein/cyan fluores-
cent protein (YFP/CFP) are localized at discrete nuclear foci,
and they also showed dispersed nuclear signals when transiently
expressed under the 35S promoter in Nicotiana (N) benthamiana
(Figure S1C). When co-expressed, YFP-SE and ATXR5/6-CFP
were co-localized at nuclear speckles (Figure 1A). Notably, these
speckles of ATXR5/6-CFP were different from Dicer (D) bodies,
which are subcellular compartments for the plant micropro-
cessor (Fang and Spector, 2007), suggesting unidentified func-
tions of SE (Figure S1C). We then conducted yeast-two-hybrid
(Y2H) experiments and observed SE interaction with both
ATXR5 and ATXR6 (Figures 1B, 1C, and S1D), but not with
PRC2 components (Figure S1D). We performed coimmunopre-
cipitation (co-IP) assays through transiently expressing the
tested components in N. benthamiana, and also from PATXR5-
FM-ATXR5 Arabidopsis transgenic lines. SE was coimmunopre-
cipitated with ATXR5 but not with the control proteins (Figure 1D).
Importantly, the presence of SE in the ATXR5 IP was not abol-
ished by RNase A treatment, suggesting that their interaction is
RNA-independent in vivo (Figure 1E). Furthermore, in vitro pull-
down assays showed that MBP-ATXR5 and MBP-ATXR6, but
not MBP protein alone, directly interacted with SE (Figure 1F).
Taken together, these results indicate that ATXR5/6 are bona
fide partners of SE.
Next, we roughly mapped the interaction interface between
ATXR5/6 and SE. ATXR5/6 harbor plant homeodomain (PHD)
and PCNA-interacting protein (PIP) domains at their N-termini
and methyltransferase catalytic SET domain at the C-termini.
Y2H assays showed that SE interacts with the SET domain-
containing C-termini of the ATXR5/6 proteins (Figures 1B and
1C). Consistent with the Y2H assay, the ATXR5 isoform (ATXR5D
PIP) that lacks the PIP motif due to alternative splicing was also
associated with SE in the co-IP assay, suggesting that PIP
domain is dispensable for the SE-ATXR5 interaction (Figure 1D).
On the other hand, SE (702 aa) protein comprises a core domain
(194–543 aa) containing zinc-finger (ZnF) motif with a walking
man-like topology, flanked by two large unstructured regions
at the N- and C-termini. The C-terminus harbors a highly

Negative control, pGAD/pGBK vectors. In (D), constructs harboring 35S-FM-ATXR5, 35S-FM-ATXR5DPIP, and 35S-FM-SGS3 or empty vector were co-
infiltrated with 35S-3HA-SE. IP was conducted by an anti-Myc antibody. Western blot analyses were done using anti-Myc, -HA, or -Actin antibodies to detect
the indicated proteins in the Input and IP products. ACTIN and FM-SGS3 served as negative controls. In (E), co-IP was conducted with protein extracts from
PATXR5-FM-ATXR5 transgenic seedlings and Col-0 plants using an anti-Flag antibody. Coimmunoprecipitated protein was detected with the SE-specific antibody
(Figure S1E). In (F), MBP-tagged bait proteins and His-Sumo-SE were mixed and pulled down using amylose resin. The recovered MBP-tagged bait proteins were
monitored by Coomassie blue staining. The output of the His-tagged prey proteins was analyzed by western blot analysis using a monoclonal anti-His antibody.
(G) Schematic diagram of the full-length and truncated versions of SE used in the Y2H assay. ZnF, zinc finger; CR, conserved region.
(H) Y2H assays show that ATXR5 interacts with the C-terminus of SE protein. Positive control, pGADT7-ATXR5 and pGBKT7-SE full length.
See also Figure S1.

Developmental Cell 45, 769–784, June 18, 2018 771

A
Actin TA3 AT1TE22850 AT1TE57220
GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks

Col-0 Col-0 Col-0 Col-0

atxr5 atxr6 atxr5 atxr6 atxr5 atxr6 atxr5 atxr6
se-2 se-2 se-2 se-2
se-2 atxr5 atxr6 se-2 atxr5 atxr6 se-2 atxr5 atxr6 se-2 atxr5 atxr6

AT2TE35840 AT4TE15820 AT4TE28870 AT5TE77610

GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks

Col-0 Col-0 Col-0 Col-0

atxr5 atxr6 atxr5 atxr6 atxr5 atxr6 atxr5 atxr6
se-2 se-2 se-2 se-2
se-2 atxr5 atxr6 se-2 atxr5 atxr6 se-2 atxr5 atxr6 se-2 atxr5 atxr6

B C
H3K27me1 hypomethylated H3K27me1 hypomethylated H3K27me1 plot for hypomethylated
TEs in indicated mutants genes in indicated mutants TEs in atxr5 atxr6

2.5
atxr5 atxr6 atxr5 atxr6 Col-0
atxr5 atxr6
se-2 66 se-2 se-2
87 8
Log2(H3K27me1 vs H3)
517 se-2 atxr5 atxr6
316
91
2.0

1132 135
1461 143
259 22
1.5

778
95
1.0

se2 atxr5 atxr6 se2 atxr5 atxr6 -1000 5’ end TEs body 3’ end 1000

D TEs families of H3K27me1 E

(%)
hypomeylated TEs H3K27me1 ChIP
100 LTR/Copia
DNA/En-Spm 1.0
DNA/Mariner Col-0
ChIP signal (%Input)

DNA/MuDR
80 DNA/Pogo atxr5 atxr6
LINE/L1 0.8 se-2
DNA/Tc1
60 DNA se-2 atxr5 atxr6
SINE
DNA/HAT 0.5
DNA/Harbinger
40 LTR/Gypsy
RC/Helitron
Unassigned 0.3
RathE2_cons
20 RathE3_cons
RathE1_cons
null
LINE
0.0
0
lin 40 N1 I
All Tub G314 AtS Ta3 TS 5390 2850 7220 5840 5820 8870 7610
6

E4 1TE2 1TE5 2TE3 4TE1 4TE2 5TE7

-2
6

5 2
xr
xr

xr e-

1 T
se

at

T 1
at

AT AT AT AT AT AT AT
at s

A
5
xr
at

Figure 2. SE Promotes H3K27me1 Mark in TEs

(A) ChIP-seq shows that normalized H3K27me1 counts were significantly reduced on exemplified TE loci. TE names and chromosome coordinates are shown on
top of each panel. H3K27me1 peaks, which are identified by published data (GSE54894, orange) and recovered in Col-0 and indicated mutants, are shown in
blue. Actin serves as a negative control.
(B) Venn diagrams show the overlap of H3K27me1 hypomethylated transposons in the indicated mutants.

(legend continued on next page)

772 Developmental Cell 45, 769–784, June 18, 2018
conserved and functionally important but yet uncharacterized
sequence (Figure 1G) (Wang et al., 2018). When the C-terminal
unstructured domain (642–702 aa) was deleted, the SE-ATXR5
interaction was completely abolished. More precisely, the
conserved region (642–695 aa) was sufficient to mediate the
interaction between SE and ATXR5 in yeast (Figure 1H). Thus,
we conclude that the distal C-terminus of SE interacts with the
SET domain containing part of ATXR5.
SE Promotes H3K27me1 Deposition on Transposons
We initially hypothesized that ATXR5/6 affects the functions of
SE-mediated RNA metabolism. However, this possibility was
excluded as atxr5 atxr6 lacked the se mutant-characteristic
defects in miRNA biogenesis and alternative splicing (Laubinger
et al., 2008) (Figure S2; Tables S1 and S2). Next, we hypo-
thesized that SE might participate in ATXR5/6-regulated
epigenetic silencing. To test this, we performed the genome-
wide H3K27me1 chromatin immunoprecipitation sequencing
(ChIP-seq) analysis. We recovered 7,651 genomic regions where
H3K27me1 occupancy was significantly decreased in the atxr5
atxr6 mutant. Of those, 5,884 were overlapped with 3,197
TEs, whereas 837 covered 468 genes (Figures 2A and S2G,
Table S3). Notably, among the H3K27me1 hypomethylated
regions, 6843 (89%) were also recovered by an independent
H3K27me1 ChIP-seq analysis (Willing et al., 2015). Our ChIP-
seq results were also consistent with previous reports (Jacob
et al., 2009, 2010), and thus, the datasets are reliable and phys-
iologically relevant.
Interestingly, we recovered a large number of H3K27me1
hypomethylated regions in se-2 and se-2 atxr5 atxr6 triple mu-
tants (Figure S2G). These included 1,678 TEs and 256 genes in
the se-2 mutant and 3,514 TEs and 511 genes in the se-2 atxr5
atxr6 mutant, respectively (Table S3). Comparative analyses re-
vealed that 73% and 74% of the H3K27me1 hypomethylated
TEs in se-2 and se-2 atxr5 atxr6 mutants overlapped with those in
the atxr5 atxr6 mutant, respectively (Figure 2B, Table S3). On the
other hand, among 3,197 hypomethylated TEs in the atxr5 axtr6
mutant, 38% and 81% overlapped with those in se-2 and
se-2 atxr5 axtr6 mutants, respectively. The overlap between
the hypomethylated TEs in atxr5 axtr6 and se-2 mutants is statis-
tically significant (1,219 overlapped TEs; Log[p] =
2,190.541;
hypergeometric test). In total, 1,132 H3K27me1 hypomethylated
TEs were shared by se-2, atxr5 atxr6, and se-2 atxr5 atxr6 mu-
tants (Figure 2B). In addition to the TEs, significant overlapping
of H3K27me1 hypomethylated regions was also found for the
protein-coding genes among these mutants (Figure 2B). Next,
we focused on the 3,197 hypomethylated TEs in the atxr5 atxr6
mutant, and plotted H3K27me1 signal for TE bodies with 1-kilo-
base pairs (kb) upstream and downstream flanking regions.
Remarkably, the H3K27me1 signal was also reduced for the
ATXR5/6-regulated loci in the se-2 mutant, but to a lesser extent
when compared with that in the atxr5 atxr6 mutant. No synergis-
(C) Plotting of normalized ChIP-seq signals of H3K27me1 in the indicated mutants along TE bodies, including 1 kb of flanking upstream and downstream regions;
3,197 TEs that were hypomethylated in the atxr5 atxr6 mutant were used.
(D) Classification of the hypomethylated TEs in the indicated mutants. The y axis represents the percentage of individual categories.
(E) ChIP-qPCR confirmation of H3K27me1 hypomethylated TEs in the indicated mutants and Col-0. Tubulin and AT1G31440 are negative controls. Error bars
indicate standard deviations of three replicates.
See also Figure S2 and Tables S1, S2 and S3.
tic effect of the H3K27me1 depletion was observed in the se-2
atxr5 atxr6 mutant (Figures 2C and S2G). Together, these results
indicate a substantial overlap between ATXR5/6 and SE-regu-
lated H3K27me1 loci.
TEs can be further divided into different classes. Among them,
LTR/Gypsy type TEs are predominantly regulated by ATXR5/6,
whereas LTR/Copia and LINE/L1 are regulated through
H3K9me2 and non-CG DNA methylation pathways that involve
the histone methyltransferases Su(var)3–9 homolog (SUVH)4,
SUVH5, SUVH6, and CHROMOMETHYLASE3 (CMT3) (Stroud
et al., 2012). In our experiments, the hypomethylated TEs in
se-2 and se-2 atxr5 atxr6 mutants were over-represented in the
LTR/Gypsy family, as also observed in the atxr5 atxr6 mutant
(Figure 2D). These results suggest that SE affects the same
type of TE targets that are regulated by ATXR5/6 but not by
the non-CG DNA methylation or H3K9me2 pathways.
To validate the H3K27me1 hypomethylation on target TEs, we
conducted ChIP-qPCR assays with 10 TE loci, including some
previously reported ATXR5/6-regulated ones (Jacob et al.,
2010) and several newly identified hypomethylated TEs in the
atxr5 atxr6 mutant from the ChIP-seq assay (Figure 2E). Consis-
tently, the H3K27me1 levels of the tested TEs were decreased
in the atxr5 atxr6 mutant. Notably, although some of the loci
were not recovered as hypomethylation TEs in the se-2 mutant
from our ChIP-seq data, the H3K27me1 levels of these TEs
were decreased in the se-2 mutant when tested by the more
sensitive ChIP-qPCR assay. These results suggest that the ef-
fect on the number of H3K27me1 hypomethylation loci by SE
loss-of-function was likely undervalued in our H3K27me1
ChIP-seq results due to the relatively low sequencing depth.
If so, SE probably impacts H3K27me1 methylation of most
of the ATXR5/6-regulated loci. Together, our results demon-
strate that SE is a positive regulator of ATXR5/6-catalyzed
H3K27me1 mark in vivo.
SE Binds to ATXR5/6-Regulated TE Loci
To test whether SE protein bound to the target TE loci in vivo, we
performed ChIP-qPCR using the se-2; PSE-FM-SE complemen-
tation lines (Figure S3A). We initially tested a few TE loci that are
hypomethylated in both atxr5 atxr6 and se-2 mutants. ChIP-
qPCR assays showed that SE was significantly enriched at the
tested TE targets but not at the negative control (Figures S3B
and S3C). These results indicate that SE protein is indeed en-
riched at the tested TE loci.
To investigate how SE associates with chromatin globally, we
conducted ChIP-seq experiments. Whole genome ChIP-seq
resulted in 24 million reads uniquely mapped to the reference
genome and revealed 5,400 significantly enriched peaks for SE
protein compared with a control ChIP-seq with Col-0 plants in
the nuclear genome (Table S4). These peaks ranged from
hundreds to several kilobases (Figure S3D) throughout hetero-
chromatic and euchromatic regions (Figure 3A). Among the
Developmental Cell 45, 769–784, June 18, 2018 773
A B
FM-SE ChIP binding genes FM-SE ChIP binding TEs
1.2 0.12
FM-SE FM-SE

Ch
5

ChIP signal (%Input)

hr

r1
C

Col-0 Col-0
0.9 0.09

FM-SE 0.6 0.06

peaks
4
Chr

0.3 0.03
Ch
r2

0.0 0.00

AtSN1
AT1G31440

AT1TE22850

AT4TE28870
AT1TE45390

AT1TE57220

AT3TE72850

AT4TE15820
Ta3

AT2TE35840
Tubulin

AT5G06320
AT1G76170

AT2G23118

AT4G32020
AT1G31440

AT3G15210
Chr3

C H3K27me1 H3K27me1 E
hypomethylated TEs hypomethylated genes
ChIP-seq plot for TEs ChIP-seq plot for genes
343

0.6
0.6

Log2(H3K27me1/H3)
FM-SE FM-SE

Log2(FM-SE/Col-0)
Log2(FM-SE/Col-0)

RPM(H3K4me3)
125 H3K4me3

1.8
H3K27me1

3
0.4
binding genes

0.4
binding TEs

1.6
FM-SE

FM-SE

0.2
0.2

2
2523 674 424 4933

0.0
0.0

1.4

1
-0.2
-0.2

1.2

-0.4
-0.4

0
-2000 5’end 3’end 2000 -2000 5’end 3’end 2000
TEs body genes body

D
AT1TE57220 AT2TE35840 AT4TE15820 AT4TE28870

GSE54894 H3K27me1 GSE54894 H3K27me1 GSE54894 H3K27me1 GSE54894 H3K27me1

H3K27me1 H3K27me1 H3K27me1 H3K27me1

FM-SE FM-SE FM-SE FM-SE

AT2G23118 AT3G15210 AT4G32020 AT5G06320

GSE54894 H3K27me1 GSE54894 H3K27me1 GSE54894 H3K27me1 GSE54894 H3K27me1

H3K27me1 H3K27me1 H3K27me1 H3K27me1

FM-SE FM-SE FM-SE FM-SE

Figure 3. SE Binds to Targeted TE Loci

(A) Circos plot shows the genome-wide distribution of FM-SE binding peaks. Outermost to innermost tracks depict Arabidopsis chromosome (Chr) 1–5
(ideogram), transposon density (black, histograms), and FM-SE binding peaks (blue, histograms), respectively.

(legend continued on next page)

774 Developmental Cell 45, 769–784, June 18, 2018
SE-bound peaks, 5,058, 1,098, and 159 loci reside in genes,
TEs, and intergenic regions, respectively (Table S4). We
selected 14 loci to perform ChIP-qPCR assays. Of those, 13
displayed SE enrichment, indicating the reliability of our ChIP-
seq data and suggesting that the identified loci were bona
fide binding sites of SE in vivo (Figure 3B). Next, we compared
the SE ChIP peaks of TEs and ATXR5/6-regulated H3K27me1
TE loci. Approximately 61% SE-bound TE loci (674/1,098) over-
lapped with H3K27me1 hypomethylated TEs detected in the
atxr5 atxr6 mutant. This result represents a significant overlap
between SE-bound TE peaks and ATXR5/6-regulated TE loci
(Log[p] =
915.5; hypergeometric test). The consistency of
SE-bound peaks and H3K27me1 mark in the TE regions is
also evident in the Integrative Genomics Viewer (IGV) plot (Fig-
ures 3C and 3D).
Interestingly, ATXR5/6-regulated hypomethylated genes also
tended to be bound by SE (Log[p] =
14.5; hypergeometric
test) (Figure 3C). However, only 125/5,058 of SE-bound
protein-coding genes coincided with ATXR5/6-catalyzed
H3K27me1 marks. The association of SE with the non-TE loci
implied that SE might play a critical role in regulating transcription
at these loci. However, the lack of overlap between the SE-bound
peaks and H3K27me1 signals in the non-TE loci suggests that
the SE-regulatory function at protein-coding gene regions was
probably not through the H3K27me1 mark (Figures 3C and 3D).
To further characterize the binding profile of SE in TE and
gene bodies, we plotted the SE enrichment signal alongside
TEs and protein-coding genes. We observed that the SE bind-
ing pattern resembles the distribution profile of H3K27me1 for
TEs (Figure 3E), revealing a strong correlation of SE occupancy
with H3K27me1 modification at the TE regions in vivo. For the
protein-coding genes, we used H3K4me3 mark as a reference.
Although the SE binding pattern and H3K4me3 profile are
distinct, SE is highly enriched at regions proximal to transcrip-
tion start sites (TSS) (Figures 3D and S3E). This result suggests
a regulatory role of SE in transcriptional events at the TSS for
protein-coding genes. To investigate whether SE recognizes
any consensus DNA sequences, we conducted a motif-
searching analysis and found that SE was indeed preferentially
associated with numerous motifs (Figure S3F). Notably, these
motifs also could be recognized by some other transcription
factors from public databases (e.g., ABI4, AP2, and ERF2)
(Figure S3F). Such coincidence implies that SE might coordi-
nate with the transcriptional factors to fulfill its regulatory role
at the target loci. How SE regulates the non-TE loci at the tran-
scriptional level will be studied elsewhere.
SE Directly Enhances the ATXR5 Activity
Given that SE promotes ATXR5/6-catalyzed H3K27me1 in vivo,
we asked whether SE or ATXR5/6 is required for each other’s
association with the target loci. We generated transgenic plants
expressing PSE-FM-SE in the se-2 and se-2 atxr5 atxr6 back-
ground and PATXR5-FM-ATXR5 in the atxr5 and se-2 atxr5 back-
ground, respectively (Figures S4A and S4B). ChIP-qPCR assays
showed that the SE occupancy at the tested TE loci was not
decreased in the atxr5 atxr6 mutant compared with that of the
complementation line (Figure S4C). Similarly, the enrichment of
ATXR5 was essentially identical in the se-2 mutant relative to
control plants (Figures S4D and S4E). Thus, the occupancy of
SE and ATXR5 to the target loci does not depend on each other.
Since SE acts as a scaffold protein to facilitate the assembly of
ribonucleoprotein complex to promote RNA processing, we hy-
pothesized that SE could facilitate the supply of H3 substrate to
the ATXR5/6 enzymes in vivo. To test this, we first conducted
in vitro pull-down assays and found that SE, like ATXR5, could
directly interact with H3 (Figure S4F). We then performed co-IP
assays with formaldehyde-cross-linked materials and were
able to detect this interaction in vivo (Figures S4G and S4H).
These results show that both ATXR5 and SE could bind to H3,
suggesting that SE might facilitate the access of ATXR5 to the
H3 substrate in vivo.
Next, we asked whether SE could directly enhance the enzy-
matic activity of ATXR5/6 in vitro. We performed histone methyl-
transferase (HMTase) assays on calf thymus-histone mixture
using the full-length recombinant GST-ATXR5, incubated with
His-Sumo, His-Sumo-SE, or His-Sumo-SE C-terminal truncation
that does not interact with ATXR5. Incubation of ATXR5 with
increasing concentrations of SE augmented the H3K27me1
signal (Figures 4A and 4B). Notably, at a molar ratio of two
ATXR5 to one SE, the enzyme activity was doubled, whereas in-
cubation of ATXR5 with the SE truncation did not produce
detectable changes. To gain more insight on the mode of action,
we performed time course experiments of ATXR5 with either
His-Sumo-SE or His-Sumo (Figures 4C and 4D). Consistent
with the previous result, incubation of ATXR5 with SE resulted
in a larger than 2-fold increase in the H3K27me1 signal, while in-
cubation with His-Sumo had no noticeable effects. Remarkably,
while the H3K27me1 signal of ATXR5 control reactions reached
their maxima at 30  60 min, incubation of ATXR5 with SE
continued to increase up to the end of the assay at 90 min.
Together, these results indicate that SE directly enhances the
activity of ATXR5 in vitro.
We also assessed the impact of SE on ATXR6 activity.
Notably, the full-length ATXR6 had different substrate prefer-
ence from ATXR5 (Figure S4I). Consistent with previous studies
(Jacob et al., 2009, 2014), ATXR6 had stronger enzymatic
activity than ATXR5 on recombinant human H3.1 in vitro. But
the full-length ATXR6 was not active with the calf thymus-histone
mixture, which ATXR5 favored (Figure S4I). We then screened 48
combinations of reaction mixtures and identified an optimal
reaction condition, as detected by a strong signal in western
blot assays, for ATXR6 (Figure S4J). ATXR6 was incubated

(B) ChIP-qPCR verification of FM-SE binding loci. Tubulin and AT1G31440 serve as negative controls. Error bars indicate standard deviations of three replicates.
(C) Venn diagrams showing the overlapping of the H3K27me1 hypomethylated TEs/genes and the FM-SE binding loci.
(D) IGV view of selected FM-SE ChIP-seq examples. Locus names and chromosome coordinates are shown on top of each panel. Published H3K27me1 peaks
(GSE54894, orange), normalized counts of H3K27me1 on target loci in Col-0 (blue), and the signals of FM-SE/control ChIP with Col-0 are shown.
(E) ChIP-seq signals were plotted for TE and gene bodies with 2-kb upstream and downstream flanking regions. Normalized FM-SE/control ChIP with the Col-0
signal (blue); normalized signals of H3K27me1/H3 or normalized H3K4me3 (orange) are shown.
See also Figure S3 and Table S4.

Developmental Cell 45, 769–784, June 18, 2018 775

A B Figure 4. SE Directly Enhances ATXR5
Activity In Vitro
GST-ATXR5 - + +
(A) SE enhanced ATXR5 activity in a dose-
His-Sumo-SE - + - 2.5
GST-ATXR5
His-Sumo-SE + His-Sumo-SE dependent manner. GST-ATXR5 was incubated
- - +

signal density
truncation
2 with His-Sumo-SE or a His-Sumo-SE C-terminal

Normalized
Molar ratio to GST-ATXR5
GST-ATXR5 0 0.25 0.5 1 0 0.25 0.5 1 + His-Sumo-SE truncation. Histone mixture from calf thymus was
kDa 1.5 truncation
15 H3K27me1 used as the substrate.
1 (B) Quantitative analysis of ATXR5 activity. The
100 His-Sumo-SE
75 GST-ATXR5 relative intensities of the western blot signals were
50 His-Sumo-SE presented as the mean of three replicates ± stan-
truncation 0 0.25 0.5 1
Molar ratio to GST-ATXR5 dard deviations (SD).
(C) Effect of SE on ATXR5 activity in a time course
experiment. GST-ATXR5 was incubated with
C His-Sumo-SE or His-Sumo and histone mixture
GST-ATXR5 + + + from calf thymus was used as the substrate.
His-Sumo-SE - - + (D) Image quantitation of ATXR5 activity in the
His-Sumo - + - time course. The relative intensities of western
Time (min) 0 5 30 60 90 0 5 30 60 90 0 5 30 60 90
kDa
blot signal were presented as the mean of three
H3K27me1 replicates ±SD ATXR5 activity was analyzed by
15
His-Sumo-SE
western blot assay using an anti-H3K27me1 anti-
100
75 GST-ATXR5 body. The proteins in each reaction were moni-
50
tored by Coomassie blue staining.
See also Figure S4.

Histones
15

D GST-ATXR5
Among the 2,207 upregulated loci in
signal density

2 + His-Sumo-SE the se-2 mutant, only 34 were TEs. The

Normalized

GST-ATXR5 significantly underrepresented ratio of

1 + His-Sumo
0
0 30
Time (min)
with His-Sumo-SE or the control protein, His-Sumo, under the
optimal condition. SE only showed marginal enhancement effect
on ATXR6 activity (Figure S4K). This effect was not statistically
significant, likely because ATXR6 already had strong activity,
which might mask the enhancement effect of SE.
SE Is Required for Transposon Reactivation in the atxr5
atxr6 Mutant
To investigate whether SE, like ATXR5/6, represses transcription
of TEs, we performed RNA sequencing (RNA-seq) analysis of
10-day-old seedlings. Expression levels of ATXR5/6 and SE
genes significantly decreased in their corresponding mutants,
indicating the reliable quality of our RNA-seq data (Figure S5A;
Table S5). Notably, the transcript accumulation of ATXR5/6
was not affected in the se-2 mutant, suggesting that the SE regu-
lation of ATXR5/6 function is not through the transcriptional regu-
lation of ATXR5/6 genes (Figure S5A). RNA-seq analysis showed
that the se mutation caused 4,189 differentially expressed loci;
among these, 2,207 were upregulated, whereas 1,982 downre-
gulated (Figure 5A). The comparable numbers of up or downre-
gulated loci suggests that SE could function as both negative
and positive regulators for transcription, depending on biological
contexts. Notably, cbp80 and hyl1 shared a large portion of co-
regulated loci with se-2, whereas each mutant had its own spe-
cific regulated loci (Figure S5B). Our results were consistent with
the fact that SE is involved in pre-mRNA processing, which re-
quires CBC, and in miRNA biogenesis, which entails both CBC
and HYL1 (Figures S5B and S2).
the reactivated TE loci relative to a large
GST-ATXR5 number of H3K27me1 hypomethylation
60 90 loci was reminiscent of the scenario previ-
ously observed in the atxr5 atxr6 mutant
(Stroud et al., 2012). Only 111 of the
3,197 hypomethylated TEs (3.5%) in atxr5 atxr6 were also tran-
scriptionally reactivated. This result suggests that the loss-of-
function mutations in SE and atxr5 atxr6 had negligible effects
on expression of the H3K27me1 hypomethylated TE loci (Fig-
ure 5B, Table S5). One explanation could be that H3K27me1 is
not the only repressive mark on the TEs and other TGS pathways
probably mask the regulatory roles of SE and ATXR5/6 at those
loci. In line with this, the introduction of mutations involved in other
TGS pathways greatly increases the number of reactivated TEs
regulated by H3K27me1 mark (Figure S5C) (Stroud et al., 2012).
To our surprise, there was barely any overlapping between SE-
and ATXR5/6-repressed loci except for a few well-characterized
TEs (Figures 5B, 5C, S5D, S5E, and S5F). Moreover, the intro-
duction of se mutation in the atxr5 atxr6 background did not
have synergistic effect on expression levels of the H3K27me1
hypomethylated TEs. Rather, approximately 60% of 111 TE
loci, which were reactivated in the atxr5 atxr6 mutant, were sup-
pressed in the se-2 atxr5 atxr6 triple mutant. This suppression
was less obvious in the hyl1 atxr5 atxr6 mutant (Figures 5D and
5E). Consistent with the concerted effect of SE and cbp80 on
genome-wide transcription, the introduction of a cbp80 mutation
in the atxr5 atxr6 background could also suppress atxr5/6-
reactivated loci (Figures 5D and 5E). Moreover, the significantly
suppressed TE loci in se-2 atxr5 atxr6 and cbp80 atxr5 atxr6
largely belonged to LTR/Gypsy retrotransposon family with a
small percentage of DNA/En-Spm DNA transposon family (Fig-
ure 5E). Importantly, the SE protein level was largely reduced
in cbp80; thus, the regulatory effect of cbp80 loss-of-function

776 Developmental Cell 45, 769–784, June 18, 2018

A B Expression of ATXR5/6 controlled C
Genes and TEs Upregulated TEs
3000 2548
hypomethylated TEs (n=3197)
2207 1993
se-2
1687 1
1262 1336 32
2
Number

RPM
215
0 15 0.5
696
1033
109
1080
1121 Up
Up
1982 2199 Down
Down
3000 0
atxr5 atxr6

at e-2

at yl1

at 80
cb 1
-0
l1

r6

0
-2
r6

0
at -2

at hyl1

at 80
-2

l
p8
p8

hy
hy

xr c b 6

6
ol

x
se
x

se

5 p
se

r5 h
5 p

s
xr

xr
at
at

xr

xr

xr

x
cb

xr c b

r5
r5

r5
x
x

r5

r5

at
at

x
x

at

at

at
at

at

at

D Upregulated TEs in atxr5 atxr6 E Families of re-suppressed TEs

(n=111) 120 Unassigned
SINE
2.5

Number
80 LTR/Gypsy
0.0 LTR/Copia
40 DNA/MuDR
-2.5
DNA/Harbinger
0
DNA/En-Spm

reactivated
in atxr5 atxr6

at e-2

at hyl1

at 80
at hyl1

at -2

at 80
l1
-0

6
0
-2

6
xr
p8

5 p
hy

5 se
6

6
ol

se

5 p

s
xr

xr

xr
xr cb
xr

xr

xr
at

xr cb
C

cb

5
xr

xr

xr
xr

xr
at

at

at

at
at

at

at

re-suppressed

F 6 6 6 6 G
xr xr xr xr
2 at
at at 80 at 2000
-0 5 e- 5 -0 5 80 b p 5 Col-0 4.9 atxr5 atxr6
ol txr e-2 s txr ol txr bp c txr
C a s a C a c a
Reactivated
in atxr5 atxr6

AT1TE45390
Count of nuclei

AT2TE28020
AT4TE15005
AT3TE51900 13.1
Reactivated
in se

AT2TE26610
AT1TE46475
5.8 15.6
Actin
No RT 0
2 4 8 16C 2 4 8 16C

2000 se-2 se-2 atxr5 atxr6

4.5
H Col-0
r6 r6 atxr5 atxr6
a tx tx se-2
1 a 5.8
-0 xr5 l1 hyl xr5
Count of nuclei

ol se-2 atxr5 atxr6

C at hy at 40
Reactivated
in atxr5 atxr6

AT1TE45390
***
AT2TE28020 30
RPM

AT4TE15005
20
AT3TE51900
10 *** 5.4
Actin 6.2
*
No RT 0 0
RAD51 BRCA1 AtGR1 2 4 8 16C 2 4 8 16C

Figure 5. SE Is Required for the Reactivation of Transposons and Heterochromatic DNA Re-replication Regulated by ATXR5/6
(A) Numbers of differentially expressed genes and TEs in the indicated mutants compared to Col-0.
(B) Boxplot of RNA-seq reads per million (RPM) values in Col-0 and indicated mutants; 3,197 TEs that were hypomethylated in the atxr5 atxr6 mutant were used.
The line in the middle of the box is plotted at the median. The whiskers are drawn from 10th to 90th percentiles.
(C) Overlap of transcriptionally upregulated TEs in the indicated mutants.
(D) Clustered heatmap of TE expression level in Col-0 and the indicated mutants. 111 TEs, which were significantly upregulated in the atxr5 atxr6 mutant
compared with Col-0 identified by RNA-seq, were clustered according to the Log2RPM value in individual samples. The value was scaled by row and sorted by
complete linkage hierarchical clustering with Euclidean distance for samples and TEs.
(legend continued on next page)
Developmental Cell 45, 769–784, June 18, 2018 777
mutation on gene expression might be partially explained
through destabilization of SE protein (Figure S5G).
To further confirm the RNA-seq results, a few loci were
selected and verified by RT-PCR assays. Indeed, the TEs
reactivated in atxr5 atxr6 were substantially suppressed in the
se-2 atxr5 atxr6 mutant and to a lesser extent in the cbp80
atxr5 atxr6 mutant, whereas no obvious repression effect was
found in the hyl1 atxr5 atxr6 mutant. Notably, the long rRNA
gene variant reactivated in the atxr5 atxr6 mutant (Pontvianne
et al., 2012) was also repressed by the se mutation, implying a
general repression effect of the se mutation on ATXR5/6-regu-
lated transcripts (Figure S5H). On the other hand, the reactivated
TEs identified in SE were not affected by ATXR5/6 (Figure 5F),
suggesting that SE could be involved in other pathways to regu-
late the expression of these TE loci. Together, these results
demonstrate that SE and CBP80 are required for the TE reactiva-
tion in atxr5 atxr6.
SE Is Required for Heterochromatic DNA Re-replication
in the atxr5 atxr6 Mutant
ATXR5/6 promote genome stability by suppressing the DNA re-
replication phenotype characterized as the accumulation of
excess DNA from heterochromatin regions (Jacob et al., 2010,
2014). To test whether SE was also involved in DNA re-replication
regulation, we analyzed DNA content of 4-week-old leaf nuclei
from different backgrounds by flow cytometry. Consistent with
previous reports (Jacob et al., 2010), the 8C and 16C peaks
were much broader, as represented by the higher robust CV
values in the atxr5 atxr6 mutant, suggesting an increase in DNA
content. By contrast, no obvious DNA re-replication defects
were found in the se-2 mutant because the robust CV values of
8C and 16C peaks in the se-2 mutant were similar to those of
Col-0 (Figure 5G). Notably, the DNA re-replication was sup-
pressed in the se-2 atxr5 atxr6 triple mutant as compared with
the atxr5 atxr6 mutant (Figure 5G). We also measured the expres-
sion levels of several marker genes involved in the homologous
recombination (HR) pathway that are upregulated in the atxr5
atxr6 mutant. In agreement with the DNA re-replication defect,
the tested HR genes were upregulated in the atxr5 atxr6 mutant,
while these genes showed no obvious change in the se-2 mutant.
Furthermore, consistent with the suppression of re-replication in
the se-2 atxr5 atxr6 mutant, the expression of the tested HR
genes also displayed a significant decrease in the se-2 atxr5
atxr6 mutant compared with that in the atxr5 atxr6 mutant (Fig-
ure 5H). It has been reported that mutations involved in DNA
methylation, nucleosome assembly, and gene expression/RNA
export also cause significant suppression of DNA re-replication
in the atxr5 atxr6 mutant (Stroud et al., 2012; Jacob et al.,
2014; Hale et al., 2016). However, expression levels of these re-
ported suppressors were not decreased in the se mutant (Fig-
(E) Classification of 111 TEs that were reactivated in the atxr5 atxr6 mutant but significantly re-suppressed in the indicated triple mutants. The y axis represents the
number of re-suppressed TEs in indicated triple mutants.
(F) RT-PCR validation of expression levels of selected TEs in Col-0 and the indicated mutants. Actin, with and without RT, served as an internal control.
(G) Flow cytometry profiles of Col, atxr5 atxr6, se-2, and se-2 atxr5 atxr6 plants. The y axis represents the number of nuclei. The x axis shows ploidy levels of the
nuclei. The numbers above the peaks (robust CV, widths of the peaks) represent the DNA re-replication levels.
(H) Normalized expression levels of DNA repair genes in the indicated mutants from RNA-seq data. The y axis represents the number of RPMs. ***q < 0.001;
*q < 0.05.
See also Figure S5 and Table S5.
778 Developmental Cell 45, 769–784, June 18, 2018
ure S5A), suggesting the suppression of DNA re-replication by
se mutation is not through these reported genes. In summary,
SE is not only required for the TE reactivation, but also essential
for the DNA re-replication in the atxr5 atxr6 mutant. How se mu-
tation suppresses the atxr5 atxr6-associated DNA re-replication
phenotype will be the focus of future studies.
Suppression of TE Reactivation by se Mutation in the
atxr5 atxr6 Mutant Is Not through DNA Methylation or
H3K9me2 Pathways
Previous reports show that non-CG DNA methylation and
H3K9me2 mutants could largely alter the expression of atxr5
atxr6-reactivated TE loci (Stroud et al., 2012). Moreover,
H3K27me1 levels of some TE loci are also decreased in DNA
hypomethylation and nucleosome assembly defect mutants
(Jacob et al., 2014; Wierzbicki et al., 2008). To test whether
se-2 suppression of atxr5/6-reactivated loci involves DNA
methylation, we examined DNA methylation levels of representa-
tive loci in the se mutant with a drm1 drm2 cmt3 mutant as a pos-
itive control. Chop-PCR, locus-specific bisulfate sequencing,
and Southern blot analysis all showed that the se-2 mutation
did not cause significant changes in the DNA methylation status
of the tested loci (Figures S6A–S6C). We also performed whole
genome bisulfite sequencing (WGBS) in se-2 and related triple
mutants. Unlike met1 and cmt3 met1 mutants, the se-2 mutant
did not display obvious changes in genome-wide DNA methyl-
ation in CG, CHG, and CHH contexts (Figure S6D). We further
analyzed the subgroup of H3K27me1 hypomethylated TEs that
were shared by atxr5 atxr6, se-2, and se-2 atxr5 atxr6 mutants
as well as the TEs that were suppressed in the se-2 atxr5 atxr6
mutant. Such analysis did not reveal a significant difference in
DNA methylation levels among the tested mutants (Figures 6A,
6B, and S6E). Thus, the H3K27me1 hypomethylation and sup-
pression of atxr5/6-reactivated loci in the se-2 mutant were not
through DNA methylation defect.
In Arabidopsis, non-CG DNA methylation is generally corre-
lated with H3K9me2 marks. The tight coordination results from
a self-enforcing loop consisting of H3K9me2 histone methyl-
transferase, Su(var)3–9 homolog 4/Kryptonite (SUVH4/KYP),
CMT3, and CMT2 (Du et al., 2015; Stroud et al., 2014). To inves-
tigate whether se mutation affects H3K9me2 levels, we per-
formed genome-wide H3K9me2 ChIP-seq. Consistent with
previous reports, H3K9me2 marks were not significantly altered
in the atxr5 atxr6 mutant (Jacob et al., 2010). In contrast to the
positive control kyp, H3K9me2 levels were not affected in se-2
and se-2 atxr5 atxr6 mutants genome-widely; nor in the loci
that were reactivated by atxr5 atxr6 mutations but suppressed
by se mutation (Figures 6C–6E, S6F, and S6G). Thus, se sup-
pression of TE reactivation in the atxr5 atxr6 mutant unlikely in-
volves DNA methylation or H3K9me2 pathways.
 A B C
1.0 Chr1 Chr2 Chr3 Chr4 Chr5
mCG level (%)
0.0
se-2

mCG
Col-0

atxr5 atxr6

0.9
mCHG level (%)

se-2 atxr5 atxr6

0.0

mCHG
Col-0 for kyp

kyp

0.4
mCHH level (%)

0.0

D
mCHH

Col-0 se-2
for kyp kyp Col-0 atxr5 atxr6 se-2 atxr5 atxr6
4.0
2.0
0.0
-0.4
-0.7
Col-0

Col-0
hyl1

hyl1
hyl1 atxr5 atxr6

hyl1 atxr5 atxr6

atxr5 atxr6

se-2 atxr5 atxr6

atxr5 atxr6

se-2 atxr5 atxr6

se-2

se-2

-1000 TEs body 1000 -1000 TEs body 1000 -1000 TEs body 1000 -1000 TEs body 1000 -1000 TEs body 1000 -1000 TEs body 1000

E
Actin TA3 AT1TE22850 AT1TE57220
GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks

Col-0
atxr5 atxr6
se-2
se-2 atxr5 atxr6
Col-0 for kyp
kyp

AT2TE35840 AT4TE15820 AT4TE28870 AT5TE77610

GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks GSE54894 H3K27me1 peaks

Col-0
atxr5 atxr6
se-2
se-2 atxr5 atxr6
Col-0 for kyp
kyp

Figure 6. Regulation of TEs by se Mutation Is Not through the DNA Methylation or H3K9me2 Pathways
(A and B) Box plots (A) and heatmap diagram (B) of CG, CHG, and CHH methylation levels at the 1,132 H3K27me1 hypomethylated TEs shared by atxr5 atxr6,
se-2, and se-2 atxr5 atxr6 mutants in Col-0 and the indicated mutants. In (B), the rows were sorted by complete linkage hierarchical clustering with Euclidean
distance.
(C) Chart diagram of genome-wide H3K9me2 levels in Col-0 and the indicated mutants. The y axis represents normalized reads of H3K9me2 ChIP-seq signal/
input DNA for kyp and its Col-0 control and the ChIP reads of H3K9me2/H3 for other indicated mutants and their Col-0 control.
(D) Heatmap of H3K9me2 ChIP-seq signals for 1132 H3K27me1 hypomethylated TEs as in (A and B) with 1-kb upstream and downstream flanking regions.
ChIP signals of kyp and its Col-0 control were normalized to input DNA, whereas the ChIP signals were normalized to H3 for other mutants indicated and their
Col-0 control.
(E) Selected TE examples show distribution patterns of H3K9me2 ChIP-seq counts (shown in blue) that were normalized to input for kyp and its control and to H3
for the rest lines. Chromosome coordinates and H3K27me1 peaks identified by previously published data (GSE54894, orange) are shown on top of each panel.
See also Figure S6.

Developmental Cell 45, 769–784, June 18, 2018 779

 se-2 rdr6-11
se-2 rdr6-11
A D F

atxr5 atxr6
atxr5 atxr6

atxr5 atxr6

atxr5 atxr6
1.5 Mock

rdr6-11

rdr6-11

Relative autographic intensity

α-HA IP
Col-0

se-2

se-2
HA-SE

35S:HA-RDR6
35S:HA-SE
TSI
1.0
AT1TE45390
AT2TE28020 kDa *
AT4TE15005 0.5
125
AT3TE51900 100
AT1TE46475 75

AT5TE43260 50 0
AT5TE15820 20min 40min 80min

AT3TE63540 E
AT2TE15880 0 min 20 min 40 min 80min 80min
AT1TE46665 HA-RDR6 + + + + + + + + - -
Actin HA-SE + - + - + - + - + -
No RT

300 bp

B
sRNA level change
1, 4:
1.0

P = 0.6166 se-2 atxr5 atxr6

Log2RPM(mutant/mutant)

se-2
0.5

2, 5:
hyl1 atxr5 atxr6
hyl1
0.0

3, 6:
cbp80 atxr5 atxr6
cbp80 G DNA methylation RNA transcripts
-0.5

Histone methylation sRNA

P = 1.134e-08
Histone 3
-1.0

P = 0.002683

1 2 3 4 5 6
21 nt 21/22/24 nt

SE
C ATXR5
YFP SE-CFP Merged

Polymerase

YFP-RDR6 SE-CFP Merged SE

CBP80

RDR6

Figure 7. SE Represses RDR6-Dependent Posttranscriptional Gene Silencing to Protect TE Transcripts

(A) RT-PCR assays show expression levels of selected TEs in Col-0 and the indicated mutants. Actin, with and without RT, served as an internal control.
(B) Changes of 21nt sRNA level in the indicated groups. The y axis represents the Log2RPM of sRNA reads ratio in the indicated groups.
(C) Microscopy image of co-localization of RDR6 with SE in nuclei in N. benthamiana. Bar, 5 mM.
(legend continued on next page)
780 Developmental Cell 45, 769–784, June 18, 2018
se Suppression of atxr5/6-Reactivated TEs Is
through RDR6
Growing evidence suggests that the RDR6-mediated PTGS
pathway could also regulate silencing of TEs in plants (Creasey
et al., 2014). The RDR6 transcript level was not affected in the
se-2 mutant (Figure S5A). We next introduced a rdr6 mutation
into se-2 and its derived higher-ordered mutants. Interestingly,
the TEs that were suppressed in the se-2 atxr5 atxr6 mutant
were restored to the level of the atxr5 atxr6 mutant in the
rdr6-11 se-2 atxr5 atxr6 quadruple mutant. In contrast, there
was not any enhanced expression of the TEs in rdr6-11 atxr5
atxr6 relative to atxr5 atxr6 or enhanced expression in rdr6-11
se-2 relative to se-2 (Figure 7A). These results suggest that
RDR6 is required for the suppression of TEs in the se-2 atxr5
atxr6 mutant but has a lesser effect on the reactivation of TEs
in atxr5 atxr6 and se-2 mutants (Figure 7A). As RDR6 is required
for the production of secondary siRNAs, we predict that in the
absence of SE function, RDR6 activity on the atxr5/6-reactivated
TEs would result in accumulation of small RNAs (sRNAs) that
mapped to these loci. Next, we examined whether there was
an increased accumulation of these sRNAs in these mutants
by sRNA-seq analyses. Neither the strand distribution nor the
profiles of sRNAs that mapped to the target loci were notably
different among the tested mutants (Figures S7A and S7B). Inter-
estingly, the enhanced accumulation of RDR6-derived 21 nucle-
otide (nt) sRNAs that mapped to these TEs in the se-2 atxr5 atxr6
and cbp80 atxr5 atxr6 mutants but not in the hyl1 atxr5 atxr6
mutant was clearly observable at the genome-wide scale,
when normalized to their respective single mutants. However,
this pattern was not evident for 22nt or 24nt sRNAs (Figure S7C).
These results suggest that the increased 21nt-sRNAs are more
likely the DCL2/4 processed products, but not the 24nt sRNAs
that participate in the RdDM pathway (Figure S7C). Notably,
the portion of 21nt sRNAs that mapped to the suppressed TEs
in se-2 atxr5 atxr6 and cbp80 atxr5 atxr6 as compared to their
respective single mutants, was significantly higher than that of
the total 21/22/24nt sRNAs. However, this increase was not de-
tected in hyl1 atxr5 atxr6 as compared to hyl1 (Figure 7B). These
results suggest that the accumulation of these 21nt sRNAs,
which processed by the RDR6-DCL2/4 pathways, are repressed
by SE and CBP80.
RDR6 is not only a component of the cytosolic PTGS pathway
(Kumakura et al., 2009) but also localized in the nucleus and acts
on unspliced primary transcripts (Hoffer et al., 2011). Moreover,
SE could directly bind to RNA substrates in vivo (Raczynska
et al., 2014). We hypothesized that SE could compete with
RDR6 for the TE transcripts, preventing their processing. To test
this hypothesis, we checked the subcellular localization of YFP-
RDR6 and SE-CFP in N. benthamiana. Although a large portion
of the YFP-RDR6 signal is detected in the cytoplasm, a subset
of YFP-RDR6 was found in the nucleus. When co-expressed,
(D) Western blot analysis of immunoprecipitated HA-tagged RDR6 and SE using anti-HA antibodies.
(E) Semi-in vitro assays of RDR6 activity in a time course with and without SE ssRNA substrate and HA-RDR6 were incubated with and without HA-SE with the
indicated time points. Signals were detected by phosphor imaging.
(F) Statistics of image quantitation of RDR6 activity. The relative autographic intensities were presented as the mean of three replicates ±SD. *p < 0.05.
(G) Working model of dual roles of SE not only promotes ATXR5/6 activity in epigenetic silencing of TEs but also protects TE transcripts from RDR6 activity to
attenuate RNA silencing effect.
See also Figure S7.
SE-CFP and YFP-RDR6 clearly overlapped in the nucleus (Fig-
ures 7C and S7D). To further investigate whether SE and RDR6
compete for the transcripts, we performed semi-in vitro RDR6
enzymatic activity assays as previously described (Curaba and
Chen, 2008). We immunoprecipitated HA-RDR6 and HA-SE pro-
teins (Figure 7D), and incubated the proteins with single-stranded
RNA (ssRNA) substrates. Semi-in vitro assays showed that incu-
bation with SE IP products notably reduced the RDR6 activity, as
compared with incubation with IP from mock materials. More-
over, this inhibitory effect was cumulative in time and was statis-
tically significant (Figures 7E and 7F). These results not only
indicate that the suppression of TEs in the se-2 atxr5 atxr6 mutant
is dependent on RDR6 but also suggests that SE competes with
RDR6. Hence, the TE suppression in se-2 atxr5 atxr6 is due to the
activity of RDR6 on the unprotected TEs transcripts. Taken
together, we propose that SE safeguards RNA transcript surveil-
lance by preventing TE transcripts from entry into the RDR6-
dependent silencing pathway in vivo (Figure 7G).
DISCUSSION
Here we report a novel function of SE, as it directly associates
with the H3K27me1 writers, ATXR5/6, and promotes the accumu-
lation of the ATXR5/6-catalyzed H3K27me1 mark. As such, SE is
a positive regulator of transcriptional silencing of the ATXR5/
6-regulated loci. On the other hand, this very protein can also
antagonize the activity of RDR6 by protecting the TE transcripts
from entering the RDR6-dependent PTGS pathway. In this
scenario, the normal function of SE will allow the presence of
the transcripts by protecting them from the PTGS pathway, while
dysfunction or downregulation of SE will result in the channeling
of the over-accumulated transcripts to RDR6 and further PTGS
pathways for degradation. Thus, the dual roles of SE enable the
fine-tuning of the accumulation of TE transcripts in the cell.
How does SE promote the accumulation of ATXR5/6-cata-
lyzed H3K27me1? One possibility is that SE serves as a hub
for the supply of the H3 substrate to ATXR5/6 for modification.
Two pieces of evidence supported this model: (1) SE could
bind to H3 in vitro and in vivo; and (2) SE could directly promote
the ATXR5 activity. Another possibility is that SE could increase
the binding affinity of ATXR5 to the cofactors like the methyl-
group donor. In addition, SE might serve as a scaffold for the
recruitment of yet-unidentified cellular factors to facilitate
ATXR5/6 action in vivo. For example, it is reported that the
enrichment of ASH2R, a core component of H3K4 methyltrans-
ferase complexes at specific loci is dependent on the CBC
(Li et al., 2016). The mode of action of SE is reminiscent of two
scenarios reported in other organisms: in yeast, Ars2 positively
regulates H3K9 methylation at the pericentromeric loci and
contributes to heterochromatic silencing. Such function of
Ars2 appears to be through two interconnected machineries
Developmental Cell 45, 769–784, June 18, 2018 781
called Red1-containing nuclear RNA silencing (NURS) and Clr4–
Rik1–Cul4 (CLRC) methyltransferase complexes that catalyze
H3K9me2/3 to initiate heterochromatin formation (Egan et al.,
2014). In the other scenario, Ars2 cooperates with carbon catab-
olite repressor 4 (CCR4) Ccr4-NOT complex, in addition to the
Red1-containing core complex, to promote meiotic mRNA
decay and facultative heterochromatin assembly at retrotrans-
posons and at developmentally regulated genes (Sugiyama
et al., 2016). However, there are some fundamental differences
between SE and Pir2/Ars2 functions in the TGS: first, Pir2/Ars2
targets H3K9 methyltransferase in yeast and mammals, whereas
SE acts on H3K27me1 writers in Arabidopsis. Moreover, the SE
promotion of ATXR5/6-catalyzed H3K27me1 might be specific,
as SE appears not to interact with SUVH4/KYP (data not shown),
which deposits H3K9me2/3, or with PRC2 components, which
catalyze H3K27me2/3 in plants. Second, Pir2/Ars2 participation
in epigenetic silencing engages in an RNAi-dependent process.
However, SE promotion of ATXR5/6 activity in plants does not
appear to involve the DNA methylation or H3K9me2 pathways.
The precise mechanism of how SE promotes ATXR5/6 activity
awaits future crystal or Cyro-EM structure analysis.
Although se mutation causes genome-wide reduction of
H3K27me1, this reduction does not cause genome-wide reacti-
vation of TEs as well as DNA re-replication as observed in the
atxr5 atxr6 mutant (Figure 5). We proposed several non-exclu-
sive hypotheses: first, there might be a threshold of H3K27me1
depletion for triggering TE reactivation and DNA re-replication.
The se mutant has less severe H3K27me1 defect relative to
atxr5 atxr6, and might not reach the threshold to induce the
other two phenotypes (Figures 2C and S2G). Second, ATXR5/
6-regulated TEs are probably controlled by multiple silencing
machineries, and downregulation of the H3K27me1 mark itself
is probably insufficient to release their expression, or its effects
are masked by other epigenetic events. Compared with the large
number of hypomethylated TEs, only a small subset of TEs was
reactivated even in the atxr5 atxr6 mutant. A similar scenario is
observed when a change of repressive marks such as
H3K9me2 does not correlate with transcriptional changes at
the targeted loci (Egan et al., 2014). Relative to the atxr5 atxr6
mutant, many more reactivated TEs are detected in the triple mu-
tants of atxr5 atxr6 with DNA methylation-deficient mutants
(Figure S5). Here, SE is a multifunctional protein that might also
be involved in other TGS/PTGS pathways, leading to a low
extent of TE reactivation in the se mutant. In addition to the pos-
itive regulation of H3K27me1, SE might fulfill other functions that
counter the silencing of ATXR5/6-regulated loci. Such function is
actually highlighted in the se-2 atxr5 atxr6 triple mutant where the
TEs and DNA re-replication have been suppressed.
se mutation suppresses reactivation of TEs in atxr5 axtr6 but
not in rdr6 atxr5 axtr6 (Figure 7). Such result indicates that the
suppression of ATXR5/6-regulated TEs in the absence of SE is
clearly caused by RDR6 activity. Several pieces of evidence sup-
port this notion: (1) expression of the TEs repressed in the se-2
atxr5 axtr6 mutant is reactivated in the rdr6-11 se-2 atxr5 axtr6
quadruple mutant; (2) a fraction of RDR6 protein is localized to
the nucleus, suggesting that RDR6 has uncharacterized func-
tions in addition to its canonical activity in cytosol; (3) co-localiza-
tion of RDR6 and SE suggests the functional connection between
the two proteins; and (4) SE and RDR6 compete for the same
782 Developmental Cell 45, 769–784, June 18, 2018
substrates in semi-vitro. One plausible model of how SE sup-
presses RDR6 function would be that SE inhibits RDR6 activity
through its interaction with CBC and the spliceosome. It has
been reported that transcripts from transgenes, relative to
intron-containing mRNAs, are more effective substrates for
RDR6-mediated gene silencing. The intron-based suppression
of RNA silencing depends on SE and CBP80, presumably due
to their competition for the same substrates (Christie et al.,
2011). TE transcripts resemble those from transgenes; thus,
they are suitable substrates of RDR6. CBP80 and SE probably
bind to the transcripts generated from the ATXR5/6-regulated
TE loci, despite their lack of introns. In this scenario, SE/CBP80
might impede the polymerase activity of RDR6 by depleting the
pool of available substrates. This would probably explain why
both mutations of se and cbp80, but not hyl1, could suppress
atxr5/6-reactivated loci, although to different extents. Alterna-
tively, SE might coordinate with RNA surveillance factors to sup-
press RDR6-mediated RNA silencing in Arabidopsis. Numerous
RNA quality control pathways compete with endogenous RDR
activities to deter RNA silencing. This has been exemplified by
members of the 50 -30 XRN exonuclease family (Gy et al., 2007),
CCR4 (Moreno et al., 2013), and SKI-containing exosome com-
plex (Zhang et al., 2015), among others (Liu and Chen, 2016),
which have been shown to directly compete for RDR6 substrates
in Arabidopsis. In fact, our proteomics analysis of the SE complex
also recovered XRN3, which has been associated with the clear-
ance of debris from pri-miRNA processing in the nucleus; thus, it
is plausible that SE cooperates with these RNA quality control
factors to impede RDR6 function on TE transcripts in plants
(Lee et al., 2013). Of course, SE might couple with yet uncharac-
terized RNA processing pathways to inhibit RDR6 function. In line
with this model, a recent genetic screening for suppressors of
ATXR5/6 identified new mutants involved in RNA processing
(Hale et al., 2016). Given that these mutants have a similar pheno-
type as the se mutation regarding the suppression of atxr5/6-
reactivated TEs, they might act in concert with each other to
antagonize RDR6 function and attenuate RNA silencing.
In summary, we found that the multifunctional protein SE has a
non-canonical role, as a coordinator of ATXR5/6-mediated TGS
and RDR6-mediated PTGS silencing of TEs. It should be noted
that apart from repressing TE expression, ATXR5/6 also contrib-
utes to the maintenance of genome stability (Hale et al., 2016;
Jacob et al., 2009, 2010, 2014; Stroud et al., 2012). SE is also a
suppressor of the DNA re-replication phenotype in atxr5 atxr6.
How SE participates in the ATXR5/6-mediated maintenance of
genome stability and whether such process is coupled with SE
regulation of TE expression awaits future efforts. Furthermore,
we noticed that SE is predominantly localized to protein-coding
gene regions besides TE loci, but how is SE recruited to the pro-
tein-coding genes and whether it regulates the transcription of
the targeted loci will also be an exciting research topic.
STAR+METHODS
Detailed methods are provided in the online version of this paper
and include the following:
d KEY RESOURCES TABLE
d CONTACT FOR REAGENT AND RESOURCE SHARING
 d EXPERIMENTAL MODEL AND SUBJECT DETAILS
d METHOD DETAILS
B DNA Constructs and Plant Transformation
B Mass Spectrometry Analysis
B Confocal Fluorescence Microscopy
B Yeast Two-Hybrid Assays
B Co-immunoprecipitation (Co-IP) Assays
B Expression and Purification of Recombinant Proteins
B In Vitro Pull-Down Assays
B Southern and sRNA Blot Analyses
B Histone Methyltransferase Assay
B RNA-Dependent RNA Polymerase (RdRP) Reconstitu-
tion Assay
B Locus Specific Bisulfite Sequencing and Chop-PCR
B Chromatin Immunoprecipitation (ChIP) Assays
B Quantitative PCR and RT-PCR
B Flow Cytometry
B Illumine Sequencing Library Preparation
B Illumina Sequencing and Analysis
d QUANTIFICATION AND STATISTICAL ANALYSIS
d DATA AND SOFTWARE AVAILABILITY
SUPPLEMENTAL INFORMATION
Supplemental Information includes seven figures and six tables and can be
found with this article online at https://doi.org/10.1016/j.devcel.2018.05.023.
ACKNOWLEDGMENTS
We thank Drs. S. Michaels, Y. Jacob, R. Martienssen, and X. Chen for their
generous sharing of ATXR5/6 and RDR6 plasmids and mutant alleles, Y.
Chen and J. Yuan for bioinformatics advice, and L. Zeng and the Microscopy
and Imaging Center facility at Texas A&M University for imaging facilities. We
also thank T. Flusche for careful proofreading of this manuscript. The work was
supported by a grant from NSF (MCB-1716243) to X.Z., D.S., and X.S. were
supported by a China Scholar Council fellowship.
AUTHOR CONTRIBUTIONS
X.Z. conceived the project. Z.M. and X.Z. designed the study. Z.M., C.C.-G.,
Z.W., X.H., X.S., and D.S. performed the experiments. M.E.P. provided exper-
imental materials and intellectual input. Z.M., C.C.-G., and X.Z. analyzed data
and wrote the paper.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: November 19, 2017
Revised: April 4, 2018
Accepted: May 20, 2018
Published: June 18, 2018
REFERENCES
Andreu-Agullo, C., Maurin, T., Thompson, C.B., and Lai, E.C. (2012). Ars2
maintains neural stem-cell identity through direct transcriptional activation of
Sox2. Nature 481, 195–198.
Cao, X., Aufsatz, W., Zilberman, D., Mette, M.F., Huang, M.S., Matzke, M., and
Jacobsen, S.E. (2003). Role of the DRM and CMT3 methyltransferases in
RNA-directed DNA methylation. Curr. Biol. 13, 2212–2217.
Castillo-Gonzalez, C., Liu, X., Huang, C., Zhao, C., Ma, Z., Hu, T., Sun, F.,
Zhou, Y., Zhou, X., Wang, X.J., et al. (2015). Geminivirus-encoded TrAP
suppressor inhibits the histone methyltransferase SUVH4/KYP to counter
host defense. Elife 4, e06671.
Christie, M., Croft, L.J., and Carroll, B.J. (2011). Intron splicing suppresses
RNA silencing in Arabidopsis. Plant J. 68, 159–167.
Creasey, K.M., Zhai, J., Borges, F., Van Ex, F., Regulski, M., Meyers, B.C., and
Martienssen, R.A. (2014). miRNAs trigger widespread epigenetically activated
siRNAs from transposons in Arabidopsis. Nature 508, 411–415.
Curaba, J., and Chen, X. (2008). Biochemical activities of Arabidopsis RNA-
dependent RNA polymerase 6. J. Biol. Chem. 283, 3059–3066.
Deal, R.B., and Henikoff, S. (2011). Histone variants and modifications in plant
gene regulation. Curr. Opin. Plant Biol. 14, 116–122.
Du, J.M., Johnson, L.M., Jacobsen, S.E., and Patel, D.J. (2015). DNA methyl-
ation pathways and their crosstalk with histone methylation. Nat. Rev. Mol. Cell
Biol. 16, 519–532.
Egan, E.D., Braun, C.R., Gygi, S.P., and Moazed, D. (2014). Post-transcrip-
tional regulation of meiotic genes by a nuclear RNA silencing complex. RNA
20, 867–881.
Fang, Y., and Spector, D.L. (2007). Identification of nuclear dicing bodies
containing proteins for microRNA biogenesis in living arabidopsis plants.
Curr. Biol. 17, 818–823.
Gruber, J.J., Zatechka, D.S., Sabin, L.R., Yong, J., Lum, J.J., Kong, M., Zong,
W.X., Zhang, Z., Lau, C.K., Rawlings, J., et al. (2009). Ars2 links the nuclear
cap-binding complex to RNA interference and cell proliferation. Cell 138,
328–339.
Gy, I., Gasciolli, V., Lauressergues, D., Morel, J.B., Gombert, J., Proux, F.,
Proux, C., Vaucheret, H., and Mallory, A.C. (2007). Arabidopsis FIERY1,
XRN2, and XRN3 are endogenous RNA silencing suppressors. Plant Cell 19,
3451–3461.
Hale, C.J., Potok, M.E., Lopez, J., Do, T., Liu, A., Gallego-Bartolome, J.,
Michaels, S.D., and Jacobsen, S.E. (2016). Identification of multiple proteins
coupling transcriptional gene silencing to genome stability in Arabidopsis
thaliana. PLoS Genet. 12, e1006092.
Hallais, M., Pontvianne, F., Andersen, P.R., Clerici, M., Lener, D.,
Benbahouche Nel, H., Gostan, T., Vandermoere, F., Robert, M.C., Cusack,
S., et al. (2013). CBC-ARS2 stimulates 3’-end maturation of multiple RNA
families and favors cap-proximal processing. Nat. Struct. Mol. Biol. 20,
1358–1366.
Hoffer, P., Ivashuta, S., Pontes, O., Vitins, A., Pikaard, C., Mroczka, A.,
Wagner, N., and Voelker, T. (2011). Posttranscriptional gene silencing in nuclei.
Proc. Natl. Acad. Sci. USA 108, 409–414.
Iwasaki, M., and Paszkowski, J. (2014). Epigenetic memory in plants. EMBO J.
33, 1987–1998.
Jacob, Y., Bergamin, E., Donoghue, M.T.A., Mongeon, V., LeBlanc, C., Voigt,
P., Underwood, C.J., Brunzelle, J.S., Michaels, S.D., Reinberg, D., et al. (2014).
Selective methylation of histone H3 variant H3.1 regulates heterochromatin
replication. Science 343, 1249–1253.
Jacob, Y., Feng, S., LeBlanc, C.A., Bernatavichute, Y.V., Stroud, H., Cokus, S.,
Johnson, L.M., Pellegrini, M., Jacobsen, S.E., and Michaels, S.D. (2009).
ATXR5 and ATXR6 are H3K27 monomethyltransferases required for chromatin
structure and gene silencing. Nat. Struct. Mol. Biol. 16, 763–768.
Jacob, Y., Stroud, H., Leblanc, C., Feng, S., Zhuo, L., Caro, E., Hassel, C.,
Gutierrez, C., Michaels, S.D., and Jacobsen, S.E. (2010). Regulation of hetero-
chromatic DNA replication by histone H3 lysine 27 methyltransferases. Nature
466, 987–991.
Jiao, L., and Liu, X. (2015). Structural basis of histone H3K27 trimethylation by
an active polycomb repressive complex 2. Science 350, aac4383.
Krueger, F., and Andrews, S.R. (2011). Bismark: a flexible aligner and methyl-
ation caller for Bisulfite-Seq applications. Bioinformatics 27, 1571–1572.
Kumakura, N., Takeda, A., Fujioka, Y., Motose, H., Takano, R., and Watanabe,
Y. (2009). SGS3 and RDR6 interact and colocalize in cytoplasmic SGS3/
RDR6-bodies. FEBS Lett. 583, 1261–1266.
Langmead, B., Trapnell, C., Pop, M., and Salzberg, S.L. (2009). Ultrafast and
memory-efficient alignment of short DNA sequences to the human genome.
Genome Biol. 10, R25.
Laubinger, S., Sachsenberg, T., Zeller, G., Busch, W., Lohmann, J.U., Ratsch,
G., and Weigel, D. (2008). Dual roles of the nuclear cap-binding complex and
Developmental Cell 45, 769–784, June 18, 2018 783
SERRATE in pre-mRNA splicing and microRNA processing in Arabidopsis
thaliana. Proc. Natl. Acad. Sci. USA 105, 8795–8800.
Laubinger, S., Zeller, G., Henz, S.R., Buechel, S., Sachsenberg, T., Wang,
J.W., Ratsch, G., and Weigel, D. (2010). Global effects of the small RNA
biogenesis machinery on the Arabidopsis thaliana transcriptome. Proc. Natl.
Acad. Sci. USA 107, 17466–17473.
Lee, N.N., Chalamcharla, V.R., Reyes-Turcu, F., Mehta, S., Zofall, M.,
Balachandran, V., Dhakshnamoorthy, J., Taneja, N., Yamanaka, S., Zhou,
M., et al. (2013). Mtr4-like protein coordinates nuclear RNA processing for het-
erochromatin assembly and for telomere maintenance. Cell 155, 1061–1074.
Li, Z., Jiang, D., Fu, X., Luo, X., Liu, R., and He, Y. (2016). Coupling of histone
methylation and RNA processing by the nuclear mRNA cap-binding complex.
Nat. Plants 2, 16015.
Lisch, D. (2013). How important are transposons for plant evolution? Nat. Rev.
Genet. 14, 49–61.
Liu, C., Lu, F., Cui, X., and Cao, X. (2010). Histone methylation in higher plants.
Annu. Rev. Plant Biol. 61, 395–420.
Liu, J., Jung, C., Xu, J., Wang, H., Deng, S., Bernad, L., Arenas-Huertero, C.,
and Chua, N.H. (2012). Genome-wide analysis uncovers regulation of long in-
tergenic noncoding RNAs in Arabidopsis. Plant Cell 24, 4333–4345.
Liu, L., and Chen, X. (2016). RNA quality control as a key to suppressing RNA
silencing of endogenous genes in plants. Mol. Plant 9, 826–836.
Makarevitch, I., Waters, A.J., West, P.T., Stitzer, M., Hirsch, C.N., Ross-Ibarra,
J., and Springer, N.M. (2015). Transposable elements contribute to activation
of maize genes in response to abiotic stress. PLoS Genet. 11, e1004915.
Margueron, R., and Reinberg, D. (2011). The polycomb complex PRC2 and its
mark in life. Nature 469, 343–349.
Matzke, M., and Mosher, R. (2014). RNA-directed DNA methylation: an epige-
netic pathway of increasing complexity. Nat. Rev. Genet. 15, 394–408.
Matzke, M.A., Kanno, T., and Matzke, A.J. (2015). RNA-directed DNA methyl-
ation: the evolution of a complex epigenetic pathway in flowering plants. Annu.
Rev. Plant Biol. 66, 243–267.
Moreno, A.B., Martinez de Alba, A.E., Bardou, F., Crespi, M.D., Vaucheret, H.,
Maizel, A., and Mallory, A.C. (2013). Cytoplasmic and nuclear quality control
and turnover of single-stranded RNA modulate post-transcriptional gene
silencing in plants. Nucleic Acids Res. 41, 4699–4708.
Peragine, A., Yoshikawa, M., Wu, G., Albrecht, H.L., and Poethig, R.S. (2004).
SGS3 and SGS2/SDE1/RDR6 are required for juvenile development and the
production of trans-acting siRNAs in Arabidopsis. Genes Dev. 18, 2368–2379.
Pontvianne, F., Blevins, T., Chandrasekhara, C., Feng, W., Stroud, H.,
Jacobsen, S.E., Michaels, S.D., and Pikaard, C.S. (2012). Histone methyltrans-
ferases regulating rRNA gene dose and dosage control in Arabidopsis. Genes
Dev. 26, 945–957.
Raczynska, K.D., Stepien, A., Kierzkowski, D., Kalak, M., Bajczyk, M.,
McNicol, J., Simpson, C.G., Szweykowska-Kulinska, Z., Brown, J.W.S., and
Jarmolowski, A. (2014). The SERRATE protein is involved in alternative splicing
in Arabidopsis thaliana. Nucleic Acids Res. 42, 1224–1244.
Robinson, J.T., Thorvaldsdóttir, H., Winckler, W., Guttman, M., Lander, E.S.,
Getz, G., and Mesirov, J.P. (2011). Integrative genomics viewer. Nat.
Biotechnol. 29, 24–26.
Robinson, M.D., McCarthy, D.J., and Smyth, G.K. (2010). edgeR: a bio-
conductor package for differential expression analysis of digital gene expres-
sion data. Bioinformatics 26, 139–140.
Sabin, L.R., Zhou, R., Gruber, J.J., Lukinova, N., Bambina, S., Berman, A., Lau,
C.K., Thompson, C.B., and Cherry, S. (2009). Ars2 regulates both miRNA- and
siRNA-dependent silencing and suppresses RNA virus infection in Drosophila.
Cell 138, 340–351.
Shen, L., Shao, N., Liu, X., and Nestler, E. (2014). ngs.plot: quick mining and
visualization of next-generation sequencing data by integrating genomic data-
bases. BMC Genomics 15, 284.
784 Developmental Cell 45, 769–784, June 18, 2018
Sienski, G., Donertas, D., and Brennecke, J. (2012). Transcriptional silencing of
transposons by Piwi and maelstrom and its impact on chromatin state and
gene expression. Cell 151, 964–980.
Stroud, H., Hale, C.J., Feng, S., Caro, E., Jacob, Y., Michaels, S.D., and
Jacobsen, S.E. (2012). DNA methyltransferases are required to induce hetero-
chromatic re-replication in arabidopsis. PLoS Genet. 8, e1002808.
Stroud, H., Greenberg, M.V., Feng, S., Bernatavichute, Y.V., and Jacobsen,
S.E. (2013). Comprehensive analysis of silencing mutants reveals complex
regulation of the Arabidopsis methylome. Cell 152, 352–364.
Stroud, H., Do, T., Du, J., Zhong, X., Feng, S., Johnson, L., Patel, D.J., and
Jacobsen, S.E. (2014). Non-CG methylation patterns shape the epigenetic
landscape in Arabidopsis. Nat. Struct. Mol. Biol. 21, 64–72.
Sugiyama, T., Thillainadesan, G., Chalamcharla, V.R., Meng, Z.,
Balachandran, V., Dhakshnamoorthy, J., Zhou, M., and Grewal, S.I. (2016).
Enhancer of rudimentary cooperates with conserved RNA-processing factors
to promote meiotic mRNA decay and facultative heterochromatin assembly.
Mol. Cell 61, 747–759.
Tan, M.J., Luo, H., Lee, S., Jin, F.L., Yang, J.S., Montellier, E., Buchou, T.,
Cheng, Z.Y., Rousseaux, S., Rajagopal, N., et al. (2011). Identification of 67
histone marks and histone lysine crotonylation as a new type of histone modi-
fication. Cell 146, 1015–1027.
Trapnell, C., Pachter, L., and Salzberg, S.L. (2009). TopHat: discovering splice
junctions with RNA-Seq. Bioinformatics 25, 1105–1111.
Wang, Z., Li, X., Jiang, Y., Shao, Q., Liu, Q., Chen, B., and Huang, D. (2015).
swDMR: a sliding window approach to identify differentially methylated re-
gions based on whole genome bisulfite sequencing. PLoS One 10, e0132866.
Wang, Z., Ma, Z., Castillo-González, C., Sun, D., Li, Y., Yu, B., Zhao, B., Li, P.,
and Zhang, X. (2018). SWI2/SNF2 ATPase CHR2 remodels pri-miRNAs via SE
to impede miRNA production. Nature. . Published online May 16, 2018. https://
doi.org/10.1038/s41586-018-0135-x.
Wierzbicki, A.T., Haag, J.R., and Pikaard, C.S. (2008). Noncoding transcription
by RNA polymerase Pol IVb/Pol V mediates transcriptional silencing of over-
lapping and adjacent genes. Cell 135, 635–648.
Willing, E.-M., Rawat, V., Mandáková, T., Maumus, F., James, G.V.,
Nordström, K.J.V., Becker, C., Warthmann, N., Chica, C., Szarzynska, B.,
et al. (2015). Genome expansion of Arabis alpina linked with retrotransposition
and reduced symmetric DNA methylation. Nat. Plants 1, 14023.
Yang, L., Liu, Z., Lu, F., Dong, A., and Huang, H. (2006). SERRATE is a novel
nuclear regulator in primary microRNA processing in Arabidopsis. Plant J.
47, 841–850.
Zhang, X., Garreton, V., and Chua, N.H. (2005). The AIP2 E3 ligase acts as a
novel negative regulator of ABA signaling by promoting ABI3 degradation.
Genes Dev. 19, 1532–1543.
Zhang, X., Yuan, Y.R., Pei, Y., Lin, S.S., Tuschl, T., Patel, D.J., and Chua, N.H.
(2006). Cucumber mosaic virus-encoded 2b suppressor inhibits Arabidopsis
Argonaute1 cleavage activity to counter plant defense. Genes Dev. 20,
3255–3268.
Zhang, X., Zhu, Y., Liu, X., Hong, X., Xu, Y., Zhu, P., Shen, Y., Wu, H., Ji, Y.,
Wen, X., et al. (2015). Suppression of endogenous gene silencing by bidirec-
tional cytoplasmic RNA decay in Arabidopsis. Science 348, 120–123.
Zhang, Y., Liu, T., Meyer, C.A., Eeckhoute, J., Johnson, D.S., Bernstein, B.E.,
Nusbaum, C., Myers, R.M., Brown, M., Li, W., et al. (2008). Model-based anal-
ysis of ChIP-seq (MACS). Genome Biol. 9, R137.
Zhu, H., Hu, F., Wang, R., Zhou, X., Sze, S.H., Liou, L.W., Barefoot, A.,
Dickman, M., and Zhang, X. (2011). Arabidopsis Argonaute10 specifically
sequesters miR166/165 to regulate shoot apical meristem development. Cell
145, 242–256.
Zhu, H., Zhou, Y., Castillo-Gonzalez, C., Lu, A., Ge, C., Zhao, Y.T., Duan, L., Li,
Z., Axtell, M.J., Wang, X.J., et al. (2013). Bidirectional processing of pri-
miRNAs with branched terminal loops by Arabidopsis Dicer-like1. Nat.
Struct. Mol. Biol. 20, 1106–1115.
STAR+METHODS

KEY RESOURCES TABLE

REAGENT or RESOURCE SOURCE IDENTIFIER

Antibodies
Anti-c-Myc-agarose beads Sigma-Aldrich RRID:AB_10109522
Anti-FLAG M2 magnetic beads Sigma-Aldrich RRID:AB_2637089
anti-His Sigma-Aldrich RRID:AB_260015
anti-H3 Agrisera RRID:AB_10750790
anti-H3K27me1 Millipore RRID:AB_310623
anti-H3K9me2 Abcam RRID:AB_449854
anti-H3 Millipore RRID: AB_2732036
anti-SE Agrisera RRID:AB_10507110
anti-HA Sigma-Aldrich RRID:AB_260092
Chemicals, Peptides, and Recombinant Proteins
Histones mixture from calf thymus Roche 10223565001
Histone H3.1 Human, Recombinant New England Biolabs Cat# M2503S
Histone H3.2 Human, Recombinant New England Biolabs Cat# M2506S
TRIzol Reagent Invitrogen Cat#15596018
Protease inhibitor cocktail Roche Cat#5056489001
RNase inhibitor Invitrogen Cat#N8080119
Dynabeads Protein G Invitrogen Cat#10003D
Dynabeads Protein A Invitrogen Cat#10001D
Critical Commercial Assays
Ribo-Zero Plant and TruSeq Stranded Total RNA kit Illumina Cat# RS-122-2401
The Ovation Ultralow Library System V2 NuGEN Cat# 0344-32
NEXTFlex Bisulfite-Seq Adapters Bioo Scientific Cat# NOVA-511911
EpiTect Bisulfite Kit Qiagen Cat# 59104
Pfu Turbo Cx Polymerase Agilent Cat# 600410
Agencourt AMPure XP beads Beckman Coulter Cat#A63881
pENTR/D-TOPO Cloning Kit Invitrogen Cat#450218
Gateway LR Clonase II Enzyme Mix Invitrogen Cat#11791-100
SuperScript III Reverse Transcriptase Invitrogen Cat#18080085
SsoAdvanced Universal SYBR Green Supermix Bio-Rad Cat#1725271
Deposited Data
Raw and analyzed data Stroud et al., 2013 GEO: GSE39901
Analyzed data Stroud et al., 2012 GEO: GSE38286
Analyzed data Willing et al., 2015 GEO: GSE54894
Raw and analyzed data This paper GEO: GSE111814
Experimental Models: Organisms/Strains
Arabidopsis: atxr5 ABRC SALK_130607
Arabidopsis: atxr6 ABRC SAIL_240_H01
Arabidopsis: se-2 ABRC SAIL_44_G12
Arabidopsis: hyl1 ABRC SALK_064863
Arabidopsis: cbp80 ABRC SALK_024285
Arabidopsis: rdr6-11 Peragine et al., 2004 N/A
Arabidopsis: drm1 drm2 cmt3 Cao et al., 2003 N/A
(Continued on next page)

Developmental Cell 45, 769–784.e1–e6, June 18, 2018 e1

Continued
REAGENT or RESOURCE SOURCE IDENTIFIER
Oligonucleotides
Primers, see Table S6 This paper N/A
Oligo probes, see Table S6 This paper N/A
Recombinant DNA
Plasmid: PSE-Flag-4Myc(FM)-SE This paper N/A
Plasmid: PATXR5-Flag-4Myc(FM)-ATXR5 This paper N/A
Plasmid: P35S-HA-SE This paper N/A
Plasmid: P35S-Flag-4Myc(FM)- ATXR5 This paper N/A
Plasmid: p35S-YFP-SE This paper N/A
Plasmid: p35S-SE-CFP This paper N/A
Plasmid: p35S-DCL1-YFP This paper N/A
Plasmid: p35S-ATXR5-CFP This paper N/A
Plasmid: p35S-ATXR6-CFP This paper N/A
Plasmid: MBP-ATXR5 This paper N/A
Plasmid: MBP-ATXR6 This paper N/A
Plasmid: His-Sumo-ATXR5 This paper N/A
Plasmid: His-Sumo-ATXR6 This paper N/A
Plasmid: His-Sumo-SE This paper N/A
Plasmid: GST-ATXR5 This paper N/A
Plasmid: pGAD-ATXR5 This paper N/A
Plasmid: pGAD-ATXR6 This paper N/A
Plasmid: pGBD-SE This paper N/A
Plasmid: P35S-HA-RDR6 Curaba and Chen, 2008 N/A
Software and Algorithms
Bowtie v1.1.2 Langmead et al., 2009 http://bowtie-bio.sourceforge.net/index.shtml
FASTX-Toolkit v0.0.14 The Hannon Lab http://hannonlab.cshl.edu/fastx_toolkit/
BEDtools v2.26.0 N/A http://bedtools.readthedocs.io/en/latest/
content/bedtools-suite.html/
TopHat2 v2.1.1 Trapnell et al., 2009 http://ccb.jhu.edu/software/tophat/index.shtml
HTSeq v0.6.1 Simon Anders at https://htseq.readthedocs.io/en/
EMBL Heidelberg
edgeR v3.3 Robinson et al., 2010 https://bioconductor.org/packages/release/
bioc/html/edgeR.html
Bismark v0.14.3 Krueger and Andrews, 2011 https://www.bioinformatics.babraham.ac.uk/
projects/bismark/
MACS2 v2.1.1 Zhang et al., 2008 https://github.com/taoliu/MACS
Homer N/A http://homer.salk.edu/homer/motif/

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact,
Xiuren Zhang (xiuren.zhang@tamu.edu).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

The Arabidopsis materials included the wild-type (WT) Col-0, atxr5 (SALK_130607), atxr6 (SAIL_240_H01), se-2 (SAIL_44_G12), hyl1
(SALK_064863), cbp80 (SALK_024285), rdr6-11 (Peragine et al., 2004) and drm1 drm2 cmt3 (Cao et al., 2003). Plants are all in
Columbia background. Plants were grown under long-day (16 h of light followed by 8 h of darkness) in soil or 12 h of light followed
by 12 h of darkness on standard MS medium plates at 22 C.

e2 Developmental Cell 45, 769–784.e1–e6, June 18, 2018

METHOD DETAILS

DNA Constructs and Plant Transformation

A majority of constructs for transgenic plants and yeast two-hybrid (Y2H) assays were made using a Gateway system (Invitrogen).
The destination vectors (containing the destination cassette DC) pHyg-YFP-DC, pBA-DC-CFP, pBA-DC-3HA (Zhang et al., 2005),
pBA-Flag4Myc-DC and pBA002a-Flag4Myc-DC (Zhu et al., 2011) were used for transient expression in N. benthamiana or stable
transformation of Arabidopsis thaliana; pGADT7-DC and pGBKT7-DC were used for yeast transformation. The cDNAs or DNA
fragments were cloned into pENTR/D-TOPO vectors (Invitrogen) and confirmed by sequencing before being transferred to the
destination vectors by recombination using the LR Clonase (Invitrogen). Native promoters of SE and ATXR5 were amplified using
a KOD polymerase with Col-0 gDNA as a template and digested with BamHI/XbaI. The resultant fragments were ligated by T4 ligase
into pBA002a-Flag4Myc-DC treated with the same set of restriction enzymes. The constructs were sequencing validated before
LR reaction. For expression of recombinant proteins in E.coli, the cDNA fragments were cloned into pET28a-Sumo to express
His-Sumo-tagged proteins by ligation; the cDNAs in pENTR/D vectors were transferred into pMAL-DC by LR reaction to produce
MBP-tagged proteins. The primers for the cloning are listed in the Table S6.

Mass Spectrometry Analysis

Two-step immunoprecipitation and proteomics analysis were conducted as described (Wang et al., 2018). Briefly, 10 grams of
10-day-old seedlings of Col-0 and transgenic plants expressing 35S-Flag-4Myc(FM)-SE were ground in liquid nitrogen into
fine powders followed by lysis with 40 ml (4 fold volumes) of IP buffer (20 mM Tris-HCl pH 7.5, 150 mM NaCl, 4 mM MgCl2,
50 nM ZnCl2, 0.1% Triton X-100, 1 mM PMSF, 1% glycerol, 2 pellets of EDTA-free protease inhibitor (Roche), 15 mM MG132).
After removal of insoluble materials by centrifugation twice at 16,000 g for 15 min at 4 C, the extracts were initially immunopre-
cipitated using 500 ml of Anti-FLAG M2 magnetic beads (Sigma-Aldrich, M8823) and incubated in slow rotation for 2 hr at 4 C, the
nonspecific-bound proteins were removed by three consecutive washes with 15 ml of IP buffer for 10 min incubation each at 4 C.
The protein complexes were then eluted by competition with 100 mg/ml 3x FLAG peptide and subsequently immunoprecipitated
with 50 ml Anti-c-Myc-agarose affinity gel (Sigma–Aldrich, #A7470) at 4 C for 1.5 hr. The nonspecific-bound proteins were
removed by five consecutive washes with the IP buffer with 5 min incubation each at 4 C. The protein complexes were eluted
in 200 ml of elution buffer (5 mM EDTA, 200 mM NH4OH) for 20 min. The supernatant was collected, frozen in liquid nitrogen
and dried using the Savant SpeedVac concentrator. Mass spectrometry analysis was performed in the Taplin Mass Spectrometry
Facility at Harvard Medical School.

Confocal Fluorescence Microscopy

Leaves of tobacco plants (N. benthamiana) were infiltrated with agrobacterium ABI strain carrying plasmids of fluorescent tagged
proteins as previously described (Zhang et al., 2005). After two days, samples were collected and the YFP and CFP signals shown
in Figures 1 and S1 were imaged on an Olympus Fluoview FV1000 confocal microscope. CFP and YFP were detected by excitation
with a 458 nm wavelength laser (emission was detected at 475nm, 25nm wavelength range) and with a 515nm laser (emission was
detected at 530nm, 70nm wavelength range), respectively. Images were taken and processed by Olympus Fluoview FV1000 Toolbox
and Adobe Photoshop software. Images in Figures 7 and S7 were taken using a Nikon inverted microscope Eclipse Ti-E (Nikon,
Japan). Fluorescent signals of CFP and YFP were measured by excitation with the CFPHQ Shutter (emission was detected at
485 nm) and with the YFPHQ Shutter (emission was detected at 540 nm), respectively. Images were taken and processed by NIS
Elements-AR 4.30.01 (Nikon) and Adobe Photoshop software.

Yeast Two-Hybrid Assays

Yeast two-hybrid assays (Y2H) were performed according to the manual of Gold Yeast Two-Hybrid System by the manufacturer
(Clontech). Different combinations of pGADT7 and pGBKT7 with either full length or truncations of cDNA constructs were trans-
formed into the yeast strain AH109. The synthetic dropout medium minus Trp, Leu and His (SD-TLH) and supplemented with
20 mM 3-AT was used for examination assay of protein-protein interaction. The yeast transformants were simultaneously grown
on dropout medium minus Trp and Leu (SD-TL) as controls.

Co-immunoprecipitation (Co-IP) Assays

For Co-IP experiments with transient expression system, N. benthamiana leaves were collected two days after agroinfiltration of the
tested constructs, ground in liquid nitrogen and stored at
80 C until use. Total proteins were extracted from 0.4 g of ground powder
in 1.2 ml (3 fold volumes) of IP buffer (40 mM Tris-HCl pH 7.5, 100 mM NaCl, 5 mM MgCl2, 50 nM ZnCl2, 0.5% Triton X-100, 2 mM
PMSF, 1% glycerol, 1 pellet/25 ml Complete EDTA-free protease inhibitor (Roche)); then, the soluble proteins were cleared twice by
ultracentrifugation at 16,000 3 g for 15 min at 4 C. The protein complexes were immunoprecipitated with 15 ml Anti-c-Myc-agarose
affinity gel (Sigma–Aldrich #A7470) at 4 C for 2 hr. The unspecific-bound proteins were removed by four consecutive washes with the
IP buffer with 10 min incubation each at 4 C. The protein complexes were boiled with 23 SDS-loading buffer for western blot
analyses using an anti-Myc antibody for IP proteins and an anti-HA and anti-Actin antibodies for co-immunoprecipitates. For
Co-IP experiments with Arabidopsis plants, 10-day-old wild-type Col-0 and transgenic seedlings expressing PATXR5-FM-ATXR5
were used. The IP buffer and procedures were identical to the ones in the transient system except for anti-FLAG M2 magnetic beads

Developmental Cell 45, 769–784.e1–e6, June 18, 2018 e3

(Sigma-Aldrich, M8823) were used and the co-immunoprecipitated protein was detected with the antibodies specifically recognizing
endogenous proteins.

Expression and Purification of Recombinant Proteins

The recombinant proteins expressed in E.coli were purified by bench purification or through Immobilized Metal Affinity Chromatog-
raphy (IMAC) followed by size exclusion chromatography (SEC) as described (Castillo-Gonzalez et al., 2015). His-tagged proteins
were purified by the Ni-NTA resin (Qiagen) using lysis buffer (40 mM Tris-HCl pH 8.5; 300 mM KCl; 1% Glycerol; 2 mM PMSF),
and eluted with elution buffer contains 300 mM imidazole. MBP tagged proteins were purified by amylose resin beads using lysis
buffer (40 mM Tris-HCl pH 8.5; 300 mM KCl; 1% Glycerol; 2 mM PMSF) and eluted with 10 mM maltose. Proteins were further dia-
lyzed using a buffer (20 mM Tris-HCl pH 8.5; 300 mM KCl; 2 mM PMSF) at 4 C overnight, concentrated, aliquoted and stored at

80 C until usage.

In Vitro Pull-Down Assays

In vitro pull-down assays were done as described (Castillo-Gonzalez et al., 2015). Briefly, for the interaction between SE and ATXR5/
6, one microgram of His-Sumo-tagged prey proteins were pre-absorbed to 50 ml of the amylose resin (NEB) for 1 hr at 4 C in 1 ml of
binding buffer (50 mM Tris-HCl pH 8.5, 100 mM NaCl, 0.5% Triton X-100, 2 mM PMSF). The supernatant proteins were recovered
after centrifugation at 12,0003g for 2 min, transferred to a second tube containing 1 mg of the MBP-tagged bait protein, and
incubated at room temperature for 2 hr. The protein complexes were harvested by adding 50 ml amylose resin beads, followed by
2 hr incubation at room temperature, and cleaned with six vigorous washes with washing buffer (50 mM Tris at pH 8.5,
250 mM NaCl, 0.5% Triton X-100, 1 mM PMSF). The pulled-down proteins were resolved by SDS-PAGE and the preys were detected
by western blot using an anti-His antibody (Sigma-Aldrich, #H1029).
For the interaction between SE and Histone3, one microgram bait proteins of His-Sumo, His-Sumo-ATXR5, or His-Sumo-SE with
0.5 microgram prey proteins of histone mixture from calf thymus (Roche) in 500 mL binding buffer (50 mM Tris-HCl, pH 8.5; 250 mM
NaCl; 0.5% Triton X-100; 1mM PMSF) and incubate overnight at 4 C; After incubation, an anti-His antibody (Sigma-Aldrich, H1029)
and Dynabeads Protein A Magnetic Beads (Thermo Fisher Scientific) were added, and the incubation continued for 1 hour in the same
conditions; Finally, six further vigorous washes were performed with the washing buffer (50 mM Tris-HCl at pH 8.5, 250 mM NaCl,
0.5% Triton X-100, 1 mM PMSF). Pulled-down proteins were resolved by 12% SDS-PAGE and detected by Western blotting using
the anti-H3 (Agrisera, AS10 710) and anti-His antibodies, respectively.

Southern and sRNA Blot Analyses

Genomic DNA was extracted using CTAB (Hexadecyltrimethylammonium bromide) and RNA was extracted using Trizol reagent. The
Southern and sRNA blots were then performed as previously described (Zhang et al., 2006). Genomic DNA was cleaved with the DNA
methylation-sensitive endonucleases MspI at 37 C for overnight and run on 1% agarose gels. The probe of Ta3 DNA fragment was
body labeled by [a-32P] dCTP with Klenow fragment. sRNA complementary oligo probes were labeled by [g-32P] ATP with PNK.

Histone Methyltransferase Assay

In vitro HMTase reactions were modified from the previous report (Castillo-Gonzalez et al., 2015). For H3 substrate preference test,
30 ml of reaction mixture containing 1-2 mM GST-ATXR5 or His-Sumo-ATXR6, 1 mg recombinant human H3.1 (NEB), 1 mg recombinant
human H3.2 (NEB), or 3.3 mg histone from calf thymus (Roche) were incubated with 400 mM SAM in HMTase buffer (50 mM Tris-HCl
pH 9.0, 10 mM MgCl2, 10 mM ß-mercaptoethanol, 250 mM Sucrose, 100mM KCl and 5% Glycerol) for 3 hr at 30 C. For screening of
ATXR6 reaction buffer, 1 mM His-sumo-ATXR6 was incubated with 1 mg recombinant human H3.1 in different HMTase buffers for
three hours at 30 C. ‘‘-’’ represents 0, ‘‘+’’, ‘‘++’’, and ‘‘+++’’ represent different concentrations of tested components in reaction
buffers (MgCl2 : ‘‘+’’, 10 mM. KCl: ‘‘+’’, 20 mM; ‘‘++’’, 50 mM; ‘‘+++’’, 100 mM. Sucrose: ‘‘+’’, 250 mM. Glycerol: ‘‘+’’, 5%; ‘‘++’’,
10%. ). For the SE effect on ATXR6 activity. 1 mM His-Sumo-ATXR6 was incubated with 1-2 mM His-Sumo-SE or His-Sumo alone
with the optimal buffer condition (50 mM Tris-HCl pH 9.0, 10 mM MgCl2, 10 mM ß-mercaptoethanol, 250 mM Sucrose, 100mM
KCl and 5% Glycerol) for 90 min at 30 C. To assess the impact of SE on ATXR5 activity, 30 ml of reaction mixture containing
3.3 mg Histone from calf thymus (Roche), 1 mM GST-ATXR5, and 400 mM SAM in HMTase buffer (50 mM Tris-HCl pH 9, 10 mM
MgCl2, 75nM ZnCl2, 1 mM ß-mercaptoethanol, 250 mM sucrose, 100 mM KCl and 5% Glycerol) was incubated for 0-1.5 hr at
30 C. Either His-Sumo-SE, His-Sumo-SE-C-terminal Truncation or His-Sumo was pre-incubated with GST-ATXR5 and histone
from calf thymus in different molar ratios, for 30 min at 4 C, then the assays were performed as described above. For time
course assays, molar ratio GST-ATXR5: His-Sumo-SE or controls (2:1) was used. The reaction products were separated by 15%
SDS-polyacrylamide gel electrophoresis (PAGE) and detected by Western blotting using anti-H3K27me1 (Millipore, cat #07-448)
antibody.

RNA-Dependent RNA Polymerase (RdRP) Reconstitution Assay

The RdRP reconstitution assay was performed as described (Curaba and Chen, 2008) with modifications. N. benthamiana leaves
were collected two days after agroinfiltration of the 35S-HA-RDR6 or 35S-HA-SE constructs, ground in liquid nitrogen and stored
at
80 C until use. 200 mg of tissue powder was mixed with 1 ml of protein Lysis buffer (50 mM Tris-HCl, pH 7.6, 100 mM NaCl,
5 mM MgCl2, 0.5% (v/v), Nonidet P-40, 1% (v/v) glycerol, 2 mM PMSF, 5 mM DTT, 2 X protease inhibitor (Roche)) and incubated

e4 Developmental Cell 45, 769–784.e1–e6, June 18, 2018

for 10 min at 4 C; Lysates were centrifuged one time for 30 min at 1,000 g at 4 C, then three times for 30 min each at 16,000 g at 4 C;
50 ml of Dynabeads Protein G beads (Thermo Fisher Scientific) with 5 ml anti-HA (Sigma-Aldrich, H9658) antibody were added to the
supernatant and the mixture was incubated on a rotator for 1.5 h at 4 C; The beads were then washed once with 1 ml of protein Lysis
buffer, 4 times with 1 ml of the Washing Buffer (20 mM Tris-HCl, pH 7.6, 100 mM NaCl, 0.1 mM EDTA, 0.05% (v/v) Tween 20, 2 mM
DTT, 2 X protease inhibitor (Roche)). 30 ml of reaction mixtures containing 50 mM HEPES-KOH (pH 7.6), 20 mM NH4OAc, 2% (w/v)
PEG4000, 16 mM MgCl2, 0.1 mM EDTA, 1mM each of ATP, CTP, and GTP, 0.1 mM UTP, and 1.5 unit/ml of RNasin (Invitrogen) were
prepared on ice. The final quantity of the ssRNA template (AT1TE45390) was 15 pmol and supplemented with 0.15 mCi/ml of [a-32P]
UTP (3000 Ci/mmol). Reactions were pre-incubated with 5 ml of HA-SE bound or mock beads for 30min on the ice, and initiated by
adding 5 ml of HA-RDR6 bound beads. The reaction mixtures were incubated on a shaker at room temperature (25 C). The exper-
iments were repeated three times for statistical analysis.

Locus Specific Bisulfite Sequencing and Chop-PCR

Bisulfite sequencing and Chop-PCR were performed as described previously (Castillo-Gonzalez et al., 2015). One microgram of
genomic DNA was extracted from 10-day-old seedlings using CTAB. After treatment with the EpiTect bisulfite kit (Qiagen), purified
DNA was amplified by locus specific primers listed in Table S6. The PCR product was cloned into the pEasy-T vector (Promega) for
sequencing. For Chop-PCR, genomic DNA was cleaved with DNA methylation-sensitive endonucleases, either HpaII, MspI, HaeIII or
endonuclease McrBC which cleaves DNA containing methylated cytosine at 37 C for overnight. The digestion products were sub-
jected to PCR.

Chromatin Immunoprecipitation (ChIP) Assays

The ChIP assays were performed as described (Castillo-Gonzalez et al., 2015). Three grams of 10-day-old seedlings were cross-
linked with 1% formaldehyde by vacuum infiltration at 4 C (10 min for histone modifications and 30 min for PSE-FM-SE, PATXR5-
FM-ATXR5 transgenic plants and anti-SE ChIP), and the reaction was stopped with 100 mM Glycine. For histone modifications
ChIP, Dynabeads Protein A or G Magnetic Beads (Thermo Fisher Scientific) with the anti-H3K9me2 (Abcam, cat #Ab1220), H3
(Millipore, Cat# MABE923), and anti-H3K27me1 (Millipore, cat #07-448) antibodies were used; Anti-FLAG M2 magnetic beads
(Sigma-Aldrich, M8823), Anti-Myc Affinity Gel (Sigma-Aldrich, #A7470), and Dynabeads Protein A beads with anti-SE (Agrisera,
AS09 532A) were used for FM-SE, FM-ATXR5 and SE ChIP, respectively. At least two biological replicates ChIP experiments
were performed.

Quantitative PCR and RT-PCR

Expression levels of the tested genes were examined by RT-PCR. Total RNAs were prepared from 10-day-old seedlings and treated
with DNase before cDNA synthesis using Superscript III reverse transcriptase (Invitrogen) primed by random primers. The Actin gene
was included as an internal control for normalization. The enrichment levels of specific loci after ChIP assays were also tested by
quantitative PCR with CFX384 Touch Real-Time PCR Detection System (BioRad). Primers are listed in Table S6.

Flow Cytometry
Flow cytometry was performed as described (Jacob et al., 2010) with some modifications. Third and 4th leaves of 4-week-old plants
were used for nucleus extraction. The leaves were chopped in Galbraith buffer (45 mM MgCl2, 20 mM MOPS, 30 mM sodium citrate,
0.1% Triton X-100, 5 mM sodium metabisulfite, 100 mg/ml of RNase A) to release the nuclei. The lysate was stained with propidium
iodide with a concentration of 50 mg/ml. Flow cytometry profiles were obtained on a Becton Dickinson FACSCalibur Analyzer at
Flow Cytometry Core Laboratory in Texas A&M University. For each sample at least 20,000 nuclei were analyzed, and the widths
of peaks (Robust CV, robust coefficient of variation values) were calculated using FlowJo V10 software.

Illumine Sequencing Library Preparation

For sRNA-seq libraries, total RNA was extracted from 10-day-old seedling using Trizol (Invitrogen). sRNAs of 19–24 nt were sepa-
rated by 15% denature PAGE, gel-purified, and used for library construction. Three biological replicates were performed for each
sample. sRNA libraries were generated as described (Zhu et al., 2011).
For stranded RNA-seq libraries, total RNA was extracted by RNeasy Plant Kit (Qiagen) with DNaseI treatment, and ribosome deple-
tion libraries were generated by Ribo-Zero Plant and TruSeq Stranded Total RNA kit (Illumina) according to the manufacturer’s
manual. RNA-seq was also performed with three biological replicates for each line.
ChIP-seq libraries were generated using the Ovation Ultralow DR Multiplex System (NuGEN, Cat# 0344-32). The libraries were
generated according to the manufacturer’s manual. Two to 10 microgram of ChIPed DNA was used as the starting DNA and 18 cycles
were used for the library amplification step.
Bisulfite sequencing (BS)-seq libraries were generated using the pre-methylated adapter as described previously (Castillo-
Gonzalez et al., 2015). Briefly, total DNA from 10-day-old seedlings was extracted by CTAB followed by RNase A treatment and
purification via ethanol precipitation. Approximately one microgram of genomic DNA was fragmented by Bioruptor Sonicator
(Diagenode) and ligated to the pre-methylated adapter (NEXTFlex Bisulfite-Seq Adapters, Bioo Scientific). The adaptor-ligated
fragments were purified by QIAQuick column (Qiagen) and bisulfite converted using the EpiTect Kit (Qiagen) following the
manufacturer’s instructions. The converted DNA was later enriched by PCR using the Pfu Turbo Cx Polymerase (Agilent) with

Developmental Cell 45, 769–784.e1–e6, June 18, 2018 e5

18 cycles. The libraries were finally purified with Agencourt AMPure XP beads (Beckman Coulter) according to manufacturer’s
instruction.

Illumina Sequencing and Analysis

All libraries were sequenced on the Illumina HiSeq 2500 platform with single-end 50 bp read length. Base calling and FastQC were
performed with standard Illumina software and then the clean reads were mapped to the Arabidopsis genome TAIR10 (http://www.
arabidopsis.org/).
Adaptors of sRNA reads were trimmed by FASTX-Toolkit (ver. 0.0.14). sRNA with length between 19- to 28-nt were selected and
mapped to the genome using Bowtie (version 1.1.2) (Langmead et al., 2009) with perfect matches. Only reads that did not match
structural RNAs (rRNAs, tRNAs, snRNAs, and snoRNAs) were kept and used for downstream analysis. Reads were counted by
overlapping with miRNAs or transposon features using BEDtools (ver. 2.26.0). For RNA-seq, all reads were mapped using TopHat2
(ver. 2.1.1) (Trapnell et al., 2009) and counted by HTSeq (ver.0.6.1) with default parameters. For sRNA and RNA-seq, normalized
expression levels and differential expressed analysis were calculated by edgeR (ver. 3.3) package (Robinson et al., 2010) with
TMM (trimmed mean of M-values) normalization according to the total reads and FDR (false discovery rate) cut-off of < 0.05. TEs
expression data in Figure S6 was obtained from (Stroud et al., 2012).
BS-seq reads were mapped by Bismark (v0.14.3) (Krueger and Andrews, 2011) with default parameters. After removal of the PCR
duplicates, the methylation information was obtained using Bismark with cutoff at least 4 coverage depth. The DNA methylation level
was calculated in every 200 bp window with a step size of 50 bp. Only windows with 4 or more differentially methylated cytosines
(DMCs) were retained. The DMRs were obtained using swDMR (Wang et al., 2015) with Fisher’s exact test, FDR < 0.05 and 2-fold
change in DNA methylation level.
ChIP-seq reads were also mapped with Bowtie (version 1.1.2), allowing up to 2 mismatches. Only uniquely mapped reads were
retained. ChIP binding peaks were called by MACS2 (v2.1.1) (Zhang et al., 2008). DNA reads from Flag4Myc-SE ChIP with transgenic
plants were normalized with the ones from control ChIP with Col-0 control. ChIP reads for histone modifications were normalized with
the ones of H3 ChIP, and only peaks with FDR < 0.05 were identified as enriched regions. The alignments were also converted to bdg
files using MACS2 and imported into the Integrative Genomics Viewer (IGV) (Robinson et al., 2011) for visualization. The plots by
indicated features were generated by NGS-plot (Shen et al., 2014). Searches for enriched DNA motifs with a 500 bp-length flanking
region were performed using Homer (http://homer.salk.edu/homer/motif/) with default parameters.

QUANTIFICATION AND STATISTICAL ANALYSIS

Quantification of HMTase assays was done by measuring the intensities of the bands from Western blot by the Photoshop software.
Quantification of the RdRP reconstitution assays by measuring the intensities of the bands was done by the software ImageJ.
All quantitative PCR data are presented as the mean of three replicates ± standard deviation (SD).
For ChIP-seq, ChIP binding peaks were called by MACS2 (v2.1.1), peaks with FDR < 0.05 were identified as enriched. MACS2
Bdgdiff package was used for discovery of different histone modifications binding regions with the cutoff of likelihood ratio > 10.
Two independent biological replicates were used and only the regions found in both replicates were retained in the analyses.
For RNA-seq and sRNA-seq, edgeR (ver. 3.3) package was used to calculate FDR (false discovery rate). The cutoff for significance
was 0.05.
For BS-seq, the DMRs were obtained using swDMR package with Fisher’s exact test to calculate FDR. The cutoff for significance
was 0.05.
In Figures 2B and 3C, hypergeometric test were used to calculate p-values for analysis of the significant overlapping. The cutoff for
significance was 0.05.
In Figures 4B and 4D, band intensities were normalized to the signal detected at the reactions without His-Sumo-SE/His-Sumo-SE
C-terminal truncation, where the average signal was arbitrarily set as 1 in (B) and normalized to the signal detected at 0 min where the
amount was arbitrarily set as 0 in (D). Standard deviation (SD) was from three biological repeats.
In Figure 7B, P-values were calculated by two-sided Wilcoxon Rank Sum test for analysis of the levels of significant difference
between the two populations. The cutoff for significance was 0.05.
In Figure 7F, band intensities were normalized to the signal detected without addition of HA-SE at each time point where the
amount was arbitrarily set as 1 and judged to be statistically significant when P < 0.05 by two-tailed paired T-test. Standard deviation
(SD) was from three biological repeats.

DATA AND SOFTWARE AVAILABILITY

Raw data files of Illumina sequencing have been deposited in GEO under the series reference GEO: GSE111814.

e6 Developmental Cell 45, 769–784.e1–e6, June 18, 2018