87 128
global RNA processing and translation (Ding et al., 2014; Rous-
kin et al., 2014; Spitale et al., 2015). Most lncRNAs, however, are
not sufficiently abundant for detection in vivo, and in vivo sec-
ondary structure studies can be obfuscated in the cell by the
binding of proteins to RNA. Overall, detailed analysis of the
native structure of individual lncRNAs is still largely lacking and
is necessary to gain deeper insights into their precise roles.
Our prior work identified the mouse lncRNA
Braveheart
(
Bvht
),
which appears to act in
trans
to regulate cardiovascular lineage
commitment (Klattenhoff et al., 2013). Given that lncRNAs are
generally lowly conserved by sequence and that many of these
transcripts are species specific (Johnsson et al., 2014; Ponting
et al., 2009), RNA secondary structure is key for understanding
their broader roles. To investigate the molecular mechanism of
Bvht
action, here we determined the secondary structure of
in vitro transcribed full-length
Bvht
( 590 nt) using SHAPE and
DMS probing and find that the transcript is organized into a
highlymodular structure including a 5
0
asymmetric G-rich internal
loop (AGIL). Using CRISPR/Cas9-mediated homology-directed
repair (HDR), we deleted this loop (denoted
bvht
dAGIL
) in mouse
ESCs (mESCs) and show that the AGIL motif is necessary for car-
diomyocyte (CM) differentiation. Similar to short hairpin RNA
(shRNA)-mediated
bvht
depletion, key cardiac transcription
factors (TFs) fail to activate during the transition from nascent
mesoderm to the cardiac progenitor (CP) state. Using a protein
microarray platform, we demonstrate that the AGIL motif inter-
acts with a small subset of factors including the heart-expressed
zinc-finger TF cellular nucleic acid binding protein (CNBP/ZNF9),
known to bind G-rich single-stranded nucleic acids (Calcaterra
et al., 2010; Chen et al., 2007). Finally, we find that CNBP re-
presses CM differentiation and that loss of CNBP partially res-
cues the
bvht
dAGIL
phenotype, suggesting that these factors
function together to specify the cardiovascular lineage. Our re-
sults show how a small RNA motif in
Bvht
can direct cell fate
and demonstrate that structural studies combined with genetic
perturbation can provide critical insights into lncRNA function.
RESULTS
Braveheart
Is Organized into a Highly Modular Structure
RNA can form complex structures that have catalytic activity or
that act as scaffolds for the binding of metal ions, small mole-
cules, nucleic acids, and proteins (Mondrago´ n, 2013; Noller,
1984; Serganov and Patel, 2007). To obtain the secondary struc-
ture of
Bvht
, we used the shotgun secondary structure determi-
nation strategy (3S) (Novikova et al., 2013), with the goal of
obtaining more detailed mechanistic insight into
Bvht
function.
First, we performed SHAPE probing (Deigan et al., 2009) (Fig-
ure 1A, top) and DMS probing (Tijerina et al., 2007) (Figure 1A,
bottom) on in vitro transcribed full-length
Bvht
(Figure S1; Table
S1, available online). We next repeated the SHAPE and DMS
probing on shorter fragments (Table S1) to identify sub-domains
of
Bvht
. When a region’s reactivity in shorter fragments shows
similarity to the profile in the full-length RNA, it suggests that
this region adopts a modular fold in the context of full-length
RNA structure. As shown in Figure 1B, we generated overlapping
fragments and performed SHAPE probing as above. Detailed
comparisons between each fragment and the full-length tran-
script revealed several regions of similar reactivity (Figure 1B).
For example, the 55 nt stretch at the 3
0
end of
Bvht
exhibited
high reactivity using both SHAPE and DMS probing, indicating
a low probability of being structured, and was left out of the anal-
ysis. We obtained the fold for
Bvht
by coordinating the modular
sub-folds.
The overall secondary structure shown in Figure 1C is most
consistent with both our SHAPE and DMS analysis of full-length
Bvht
and of the shorter fragments.
Bvht
consists of 12 helices,
8 terminal loops, 5 sizeable (>5 nt) internal loops, and a five-
way junction (5WJ).
Bvht
appears to be organized into three do-
mains, roughly corresponding to its three exons: the 5
0
domain
(H1–H2), central domain (H3–H8), and 3
0
domain (H9–H12) (Fig-
ure 1C). The 5
0
domain contains an AGIL between H1 and H2,
consisting of a large single-stranded region (14 nt) on the
5
0
side and very short single-stranded region (3 nt) on the
3
0
side. The central domain consists of a 5WJ (H4, H5, H6, H7,
and H8) connected to the 5
0
domain by H3. The 3
0
domain con-
tains four helices (H9, H10, H11, and H12).
Braveheart AGIL Motif Is Necessary for Proper ESC
Differentiation
To date, lncRNA function has largely been determined by tran-
script knockdown or by genetic deletion of large regions that
may encompass regulatory elements confounding phenotypic
interpretation. We focused on dissecting the function of the
AGIL region because it appeared to be less commonly repre-
sented in known RNA secondary structure databases and
because G-rich regions often play regulatory roles in the genome
(Aguilera and Garcı´a-Muse, 2012; Rhodes and Lipps, 2015). For
example, after searching the Gutell database of secondary
structures of ribosomal and RNase P RNAs (Cannone et al.,
2002), we found that only 13 of >400,000 asymmetric 5
0
internal
loops had similar size and asymmetry. The crystal structure of
one such loop was recently solved, forming an intricate tightly
packed configuration of purines (Ren et al., 2016). Thus, using
Figure 1.
Bvht
Secondary Structure Determination by Chemical Probing
(A) Normalized SHAPE (top) and DMS (bottom) probing reactivity profiles of full-length
Bvht
. Horizontal lines indicate normalized dimensionless reactivity. Both
traces were normalized by the reactivities for highly reactive nucleotides. Nucleotides that have a normalized reactivity >0.5 are considered as highly flexible and
likely represent single-stranded regions. Positions of
Bvht
exons are labeled below the reactivity profile.
(B) Shotgun secondary structure (3S) analysis of
Bvht
. Normalized SHAPE probing reactivity of indicated
Bvht
fragments is compared to full-length transcript. Full
length, 1–590; 5
0
fragment, 1–325; middle fragment, 155–475; 3
0
fragment, 300–590; Half_H9, 282–349; and Half_H10-H11, 380–457. The sub-regions with highly
similar reactivity patterns to full-length transcript are highlighted in purple under the reactivity profile.
(C) Secondary structure of
Bvht
was derived with 3S via SHAPE and DMS chemical probing experiments. The normalized SHAPE or DMS reactivity is represented
by indicated colors. Circle, SHAPE; diamond, DMS. The AGIL motif is highlighted by red dashed lines. H1 to H12 indicates the helices.
See also Figure S1.
Molecular Cell
64
, 37–50, October 6, 2016
39