Nature - USA (2020-01-23)

Nature | Vol 577 | 23 January 2020 | 567

RREB1 as a RAS-regulated SMAD cofactor

TGF-β binding to the receptor kinases TGFBR1 and TGFBR2 acti-
vates SMAD2–SMAD3–SMAD4 (SMAD2/3/4) trimeric complexes,
which target specific promoters and enhancers by interacting with
context-determining transcription factors^9. SMAD2/3 chromatin
immunoprecipitation and DNA sequencing (ChIP–seq) in PDA cells
treated with TGF-β revealed binding motifs for various RAS tran-
scriptional effectors (FOS and JUN AP-1 components and ELK3 and
the SMAD binding motifs CAGAC and 5GC^24 within SMAD2/3 peaks
independently of KRAS(G12D) (Extended Data Fig. 1h–j). Notably,
RREB1 motifs were specifically enriched within SMAD2/3 peaks
in KRAS(G12D)-dependent TGF-β targets (Extended Data Fig. 1h).
Although EMT is generally pro-tumorigenic in carcinoma cells, TGF-β
triggers apoptosis in KRAS-mutant pancreatic progenitors owing to
simultaneous induction of SNAIL and the pro-epithelial transcription
factor SOX4^12. We used this property of KRAS-mutant pancreatic pro-
genitors to screen a shRNA library targeting 40 transcription factors
expressed in PDA cells using shRNAs targeting the TGF-β receptors as
positive controls (Fig. 1d). Rreb1 and Sox4 were the only transcription
factor transcripts for which two independent shRNAs were enriched
more than twofold (Fig. 1d).
RREB1 contains 15 zinc fingers^21 , but little is known about its function
and regulation^25 –^27. In PDA cells in which SMAD4 expression has been
restored and that also express haemagglutinin (HA)-tagged RREB1
(residues 1–1291 mouse isoform) (Extended Data Fig. 2a), ligation assays
showed close proximity between nuclear RREB1 and SMAD2/3 following
TGF-β treatment (Extended Data Fig. 2b, c). Co-immunoprecipitation
revealed interactions between SMAD3 and HA–RREB1 (Extended Data
Fig. 2d). The genome-binding pattern of HA–RREB1 overlapped with
that of SMAD2/3 in TGF-β-treated cells (Fig. 1e, f, Extended Data Fig. 2e),

including in Snai1 and Has2 but not in Smad7 (Fig. 1g). HA–RREB1 bound

to these loci in the absence of TGF-β signalling (Fig. 1e–g, Extended

Data Fig. 2e). MAPK signalling has previously been implicated in RREB1

regulation^28. Treatment of SMAD4-restored PDA cells with the ERK

inhibitor SCH772984 (ERKi) or the MEK inhibitor AZD6244 (MEKi)

did not alter nuclear localization (Extended Data Fig. 3a) or levels of

RREB1 (Extended Data Fig. 3b, c), but diminished binding of HA–RREB1

to Snai1, Has2 and Il11 cis-regulatory regions (Extended Data Fig. 3d).

HA–RREB1 immunoprecipitated from PDA cell lysates bound double-

stranded DNA probes corresponding to Snai1 enhancer and Has2

promoter regions; ERKi treatment decreased this activity (Extended

Data Fig. 3e). We identified four ERK-dependent phosphorylation

sites in HA–RREB1 immunoprecipitated from SMAD4-restored PDA

cells (Extended Data Fig. 3f, g); all were situated between zinc-finger

domains (Extended Data Fig. 3h). S161 and S970 fit the MAPK phospho-

rylation motif PX(S/T)P, whereas S1138 and S175 may represent indirect

phosphorylation by other kinases. RREB1 with S161 or S970 alanine

substitutions was deficient in restoring Snai1 and Has2 TGF-β responses

to Rreb1-knockout cells and in binding to these loci, compared with

vectors encoding RREB1 with phosphorylation-mimicking aspartate

substitution (Extended Data Fig. 3i, j).

RREB1 and TGF-β-dependent EMT

Rreb1 knockout in SMAD4-restored PDA cells (Extended Data Fig. 4a–c)

reduced TGF-β-dependent binding of SMAD2/3 to regulatory regions

in Snai1 and Has2, and abolished their induction and EMT (Fig. 2a–c,

Extended Data Fig. 4d, e). Rreb1 knockout had limited effects on

the binding of SMAD2/3 to, and induction of, Smad7 (Fig. 2c, Extended

Data Fig. 4f ). Restoration of RREB1 rescued induction of Snai1, Has2

and Il11 by TGF-β in Rreb1-knockout cell lines (Extended Data Fig. 4g).

a b Pancreatic epithelial cell organoids

Pdx1-Cre;

Cdkn2afl/fl;

LSL-YFP

Pancreatic epithelial

cells

tetON

KrasG12D

Kras(G12D) Kras(G12D)

Organoid

culture

CIY organoids

Pdx1-Cre;

LSL-KrasG12D;

Cdkn2afl/fl;

Smad4fl/fl

PDA cancer cells

SMAD4

vector

Monolayer

culture

SMAD4-restored cells

c

E-cadherin

DAPI

ZEB1

Kras(G12D)-On

SB505124 d

e

2

4

RREB1 ChIP-SB

RREB1 ChIP-TGF-β

SMAD2/3 ChIP-SB

SMAD2/3 ChIP-TGF-β

–600 0 600

Distance from RREB1

peak centre (bp)

–600 0 600

Distance from SMAD2/3

peak centre (bp)

Peak density(tags per bp)

fg

TGF-β

15,136

15,136

SB

28,316

24,861

6,497

TGF-β

SB

TGF-β

SB

6

6

6

6

Snai1

12

12

22

22

Has2

DE1DE2 PP1 UE1

TGF-β TGF-β

SMAD2/3 ChIP RREB1 ChIP

SMAD2/3 ChIP RREB1 ChIP

5 kb 10 kb

SMAD2/3ChIP

RREB1ChIP

00

10 15 20

–4

–2

0

2

4

shRreb1

shTgfbr2

shSox4

shTgfbr1

Rreb1 Sox4

25

log

(enrichment TGF- 2

β/SB)

log 2 (mean signal)

0

10

20

30

40

Snai1

OffOn

Relative mRNA

levels

0

2

4

6

Smad7

SB TGFβ

OffOn

25

25

25

25

0

10 kb

Smad7

NS

****

6,315

1

2

****

***

Kras(G12D)-Off

TGF-β

Fig. 1 | RREB1 is a KR AS-dependent SMAD cofactor. a, Source and generation
of CIY pancreatic epithelial organoids and SMAD4-restored PDA cells. b, Snai1
and Smad7 mRNA levels in pancreatic epithelial organoid cultures. Cells
engineered to express doxycycline-inducible KR AS(G12D) treated with the
TGF-β and Nodal receptor inhibitor SB505124 (SB, 2.5 μM) or TGF-β (10 pM) for
1.5 h. Data are mean ± s.d.; n = 4; two-way analysis of variance (ANOVA),
****P < 0.0001. c, E-cadherin, ZEB1 and DAPI immunof luorescence in CIY
pancreatic organoids with or without KR AS(G12D) expression, treated with
SB505124 or TGF-β for 2.5 days. Scale bars, 30 μm. Images are representative of
two independent experiments. d, Screening of pancreatic progenitor
transcription factor shRNA library for mediators of TGF-β-induced lethal EMT.
Dot plot of shRNA enrichment in TGF-β-treated versus SB505124-treated
SMAD4-restored PDA cells. Sox4^12 and Rreb1 transcription factor genes score

positive in the screen. shRNAs targeting Tg f b r1 and Tg f b r 2 are included as

positive controls. e, Position of RREB1 peak summits relative to summits of

overlapping SMAD2/3 peaks (left), and position of SMAD2/3 peak summits

relative to summits of overlapping RREB1 peaks (right), based on ChIP–seq

analysis (Extended Data Fig. 2e). f, Venn diagram showing overlap between

SMAD2/3 and RREB1 ChIP–seq peaks, based on ChIP–seq analysis in Extended

Data Fig. 2e. g, Gene track view of SMAD2/3 and HA–RREB1 ChIP–seq tags at

indicated loci and experimental conditions. Gene bodies are represented

below the track sets. PP, proximal promoter; DE, downstream enhancer; UE,

upstream enhancer. ChIP–seq was performed once and an independent ChIP

was performed in which selective genomic regions were confirmed by

quantitative PCR (qPCR). See also Extended Data Figs. 1–3 and Supplementary

Video 1.