ARTICLE
Received 22 May 2015
| Accepted 22 Feb 2016 | Published 30 Mar 2016
Sleeping Beauty transposase structure allows
rational design of hyperactive variants for genetic
engineering
Franka Voigt
1
, Lisa Wiedemann
2
, Cecilia Zuliani
1
, Irma Querques
1
, Attila Sebe
2
, Lajos Ma
´
te
´
s
3,w
,
Zsuzsanna Izsva
´
k
3
, Zolta
´
n Ivics
2
& Orsolya Barabas
1
Sleeping Beauty (SB) is a prominent Tc1/mariner superfamily DNA transposon that provides a
popular genome engineering tool in a broad range of organisms. It is mobilized by a
transposase enzyme that catalyses DNA cleavage and integration at short specific sequences
at the transposon ends. To facilitate SB’s applications, here we determine the crystal structure
of the transposase catalytic domain and use it to model the SB transposase/transposon
end/target DNA complex. Together with biochemical and cell-based transposition assays,
our structure reveals mechanistic insights into SB transposition and rationalizes previous
hyperactive transposase mutations. Moreover, our data enables us to design two additional
hyperactive transposase variants. Our work provides a useful resource and proof-of-concept
for structure-based engineering of tailored SB transposases.
DOI: 10.1038/ncomms11126
OPEN
1
European Molecular Biology Laboratory, Structural and Computational Biology Unit, Meyerhofstrasse 1, Heidelberg 69117, Germany.
2
Paul Ehrlich Institute,
Division of Medical Biotechnology, Paul Ehrlich Strasse 51-59, Langen 63225, Germany.
3
Max Delbru
¨
ck Center for Molecular Medicine, Robert Ro¨ssle Strasse
10, Berlin 13092, Germany. w Present address: Biological Research Centre, Institute of Genetics, Temesva
´
ri krt. 62, Szeged 6726, Hungary. Correspondence
and requests for materials should be addressed to O.B. (email: barabas@embl.de).
NATURE COMMUNICATIONS | 7:11126 | DOI: 10.1038/ncomms11 126 | www.nature.com/naturecommunications 1
T
ransposons are natural DNA-mobilizing vehicles that have
an intrinsic ability for genomic integration. In contrast
to other genome engineering tools, such as zinc-finger
nucleases, TALENs or the CRISPR/Cas9 system, transposons
directly insert their genetic cargo into genomes, and can thereby
enable stable gene transfer with high efficiency potentially
alleviating the need for clonal selection in medical applications.
SB is a member of the widespread Tc1/mariner superfamily of
DNA transposons that has been developed as a genome
engineering tool in a broad range of organisms
1,2
. It offers up
to 95% gene transfer efficiency in diverse vertebrate cell types and
is widely used in forward genetic screens
3,4
as well as in ex vivo
human gene therapy trials
5,6
. The transposon (tnp) DNA includes
terminal inverted repeats (TIRs) at its ends and encodes a
transposase protein, the workhorse of transposition, that catalyses
all DNA cleavage and joining reactions required for transposition.
Structural and biochemical studies of model Tc1/mariner
transposons (Tc3, Mos1 and HsMar1)
7–9
revealed a so-called
cut-and-paste mechanism that involves: (i) transposase binding to
the TIRs; (ii) synapsis of the tnp ends generating an intertwined
nucleoprotein complex, the paired-end complex (PEC) (also
called transpososome); (iii) coordinated stepwise cleavage of the
two DNA strands on both tnp ends; (iv) target DNA (tDNA)
binding via recognition of a TA dinucleotide; and (v) integration
into the new genomic location. Integration occurs with a
two-base-pair (bp) stagger at each side of the target TA and
results in single-strand gaps on both transposon flanks that are
repaired subsequently by host enzymes leading to characteristic
target site duplication.
Tc1/mariner transposases contain an N-terminal bipartite
PAIRED-like
10
DNA-binding domain (DBD) consisting of two
helix-turn-helix motifs, and a C-terminal catalytic domain
with an RNaseH-like fold and a catalytic triad of three acidic
residues (DDE)
1
that execute DNA hydrolysis (in excision) and
transesterification (in integration) in a two metal ion-dependent
manner
11,12
. Well-studied Tc1/mariner transposases (for example,
Mos1 and HsMar1) function as homodimers that likely
undergo multiple conformational rearrangements throughout the
transposition reaction
7,8,13–15
. They initially assemble as auto-
inhibited dimers, which rearrange upon DNA binding to allow
hydrolysis of the first DNA strand (non-transferred strand, NTS)
2–3 nucleotides (nt) inside the tnp on both ends
8,14
. Subsequently,
the transpososome changes conformation to bring the second
DNA strand (transferred strand, TS) into the transposase active
site
7,8,15
for cleavage at the exact tnp boundary. Finally, tDNA
capture might induce further conformational changes, but since
the post-TS-cleavage transpososome already holds the terminal
3
0
OH groups of the TS in the transposase active sites, no major
rearrangements seem to be required to proceed to integration.
From all the distinct nucleoprotein assemblies, high-resolution
structural information is so far only available for the Drosophila
mauritiana Mos1 pre-
16
and post-TS-cleavage
7
PECs, both of
which contain the TS ends in the transposase active sites. These
highly similar structures depict the tnp ends in a parallel
arrangement with the 3
0
OH ends of the TS-s situated 24.8 Å
from each other. This arrangement appears consistent with a
concerted attack of the two tnp ends on the tDNA with a 2-bp
stagger
7
, although the 3
0
OH groups are somewhat far, suggesting
that the tDNA may need to be bent for efficient integration. In
accordance, bent tDNA has been shown to be a preferred target for
integration of Tc1/mariners
17–19
, including SB
20,21
and might even
be a general feature of DDE enzymes
22,23
.
The 39 kDa SB transposase (340 amino acids (aa)) features a
similar domain composition to other Tc1/mariner transposases.
However, it has limited (o20%) sequence identity to its relatives
and includes sub-domains with unknown functions, such as the
glycine-rich strip
1
(aa 183–198) that is specific to the Tc1
subfamily at the base of the so-called clamp loop (aa 159–190)
7
.
SB’s transposon end architecture is also unusual: while the
mariner transposons HsMar1 and Mos1 contain short 30–40 bp
almost perfect inverted repeats, SB’s TIRs are B230 bp long and
contain two 30–35 bp highly conserved transposase binding sites
placed in a direct repeat (IR/DR) orientation separated
by DNA segments of variable sequence
1
. To date, little
biochemical data is available for the SB transposon and
structural information is limited to the N-terminal HTH-motif
of the transposase DBD
24
. To facilitate SB’s applications in
genetic engineering, several hyperactive transposase variants
have been developed
2,25–27
, providing up to 100-fold increased
efficiencies. However, the activity of the transposase is cell type
dependent
28
and several applications, including gene transfer into
medically relevant primary human cells, would greatly benefit
from further enhanced transposase variants. However, rational
engineering of novel SB variants is currently hampered by the
lack of specific mechanistic and structural data, particularly for
the transposase catalytic domain. To overcome this limitation,
here we solve the first crystal structure of the SB catalytic domain
and, based on this, generate a model of the SB transpososome
containing transposon end as well as target DNA. The
acquired structural insights enable us to rationalize previous
hyperactive mutations on the transposase as well as design novel
defective and hyperactive SB variants. Our work provides
novel insights into SB’s mechanism and demonstrates how
structure-based design can help generate further designer
transposases.
Results
The structure of the SB100X catalytic domain. To provide
structural insights into SB transposition, we crystallized the
catalytic domain (aa 114–340; including most of the flexible inter-
domain linker that spans aa 110–127) of the current most active
SB transposase variant
2
, SB100X, and solved its structure at 1.4 Å
resolution (Fig. 1a, Supplementary Fig. 1 and Table 1). The core
of the structure reveals a canonical RNaseH-fold, consisting of a
central five-stranded b-sheet (b1–b5) surrounded by five
a-helices (a1–a5). The catalytic residues (D153, D244 and
E279, red in Fig. 1a) are assembled in close proximity
establishing an active site conformation similar to the one
observed in the crystal structure of the homologous Mos1
transposase PEC (grey in Figs 1a,b)
7
.
In addition to the core RNaseH domain, most of the inter-
domain linker (aa 117–127) and also the flexible clamp loop (aa
159–190), which is inserted between b1 and b2 of the RNaseH-
fold
7
and includes part of the glycine-rich strip (aa 183–190)
1
, are
visible in the structure. While the RNaseH core superposes closely
with that of Mos1 in the PEC (r.m.s.d. 1.97 Å for Ca atoms,
Fig. 1b) the linker and the clamp loop assume different
conformations (Supplementary Fig. 2). In the Mos1 PEC, the
clamp loop is extended and interacts with the similarly extended
linker and both tnp ends across the dimer interface, playing an
important role in PEC assembly
7
. In turn, in our SB100X catalytic
domain crystals, the clamp loop is curved (Fig. 1a and
Supplementary Fig. 2b), mostly pivoting on three consecutive
G-s (aa 188–190, marked with red arrow in Fig. 1a) in the
glycine-rich strip, and contributes to an extended protein–protein
interface (2,350 Å
2
surface area; DG ¼27.2 kcal mol
1
)
29
between symmetry related molecules in neighbouring
asymmetric units (Fig. 1b,c and Supplementary Fig. 3). This
interface brings two catalytic domains into close proximity with
their active sites facing each other in an arrangement that
resembles the one in the Mos1 PEC
7
(Fig. 1b), but is more
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11126
2 NATURE COMMUNICATIONS | 7:11126 | DOI: 10.1038/ncomms11126 | www.nature.com/naturecommunications
compact and features several additional tight sequence-specific
interactions (Fig. 1d,e and Supplementary Fig. 3g,h). The clamp
loops of both protomers form reciprocal interactions with the
RNaseH core of the partner molecule, covering the active sites
(Fig. 1c,d and Supplementary Fig. 2b). In addition, the tip of the
clamp loop contains two short antiparallel b-strands (aa 169–171
and 174–176), which form a b-hairpin and interact with the main
chain of the inter-domain linker (aa 119–122) of the partner
molecule (Fig. 1e and Supplementary Fig. 3g). This contact
resembles the b-stranded clamp loop—linker interaction
observed in the Mos1 PEC, but involves a different part of the
linker, which is structurally equivalent to the regulatory
WVPHEL motif of mariner transposases
15,16,30
.
To investigate the functional relevance of the observed
interface, we mutated amino acids N280 and K339 on the
RNaseH core (Fig. 1d) and assessed the effects in transposition
assays in HeLa cells (Fig. 1f and Supplementary Fig. 4). These
amino acids were selected for mutagenesis, because they are
critical to the newly observed interface, but localize distant from
the intermolecular interface in the PEC and have not been
attributed a functional role before. We find that simultaneous
mutation of both residues strongly reduces transposition
(P ¼ 1.7 10
4
; t-test), indicating that their combined interac-
tions are required for efficient transposition (Fig. 1f). However, as
the observed arrangement of the catalytic domains is too compact
to allow DNA to access the active sites (Fig. 1b), it does not
appear compatible with PEC assembly. Instead, it might mediate
the formation of a pre-catalytic dimer that must still undergo
conformational rearrangement upon PEC formation (illustrated
by the arrow in Fig. 1b).
Molecule I Molecule II
ba
β1
α1
β2
β3
β4
β5
α2/3
α4
α5
N
C
Clamp loop
D153
D244
E279
c
Molecule IMolecule II
N280
II
K339
II
P183
I
N180
I
ed
Molecule I Molecule II
Clamp loop
II
Clamp loop
I
Linker
I
Linker
II
Linker
II
Clamp loop
I
L122
II
V168
I
W169
I
K172
I
G173
I
R170
I
E174
I
C176
I
A175
I
K171
I
P121
II
A117
II
R118
II
K119
II
K120
II
Clamp loop
I
f
100
80
60
40
20
0
D3
SB100X
N280A
K339A
N280A/
K339A
Transposition (%)
Figure 1 | The structure of the SB100X transposase catalytic domain. (a) The SB100X catalytic domain (blue) assumes an RNaseH-fold with all catalytic
residues (red) assembled in the active site. Conserved a-helices (a) and b-strands (b) are indicated. Insert: superposition of active-site residues in SB100X
and Mos1 (grey, PDB 3HOS)
6
.(b) The SB100X dimer (dark and light blue) observed in our crystal structure (molecules I and II) superposed onto the Mos1
PEC structure (grey). Arrow illustrates the rearrangement (B50° swing to left with B20° rotation backwards) that can bring Molecule I into the PEC
conformation. (c) Top view of the SB100X catalytic domain dimer highlights the role of the clamp loops in the interaction. (d) Residues N280 and K339 of
the RNaseH core form sequence-specific interactions (hydrogen bonds indicated with dashed lines) with the clamp loops of the partner molecules,
anchoring them to cover the active sites. Insert: close-up of the dimer contacts mediated by N280 and K339, respectively. (e) b-stranded interactions
between the backbones of the clamp loop and the inter-domain linker. (f) Transposition assay for N280A and K339A mutants in HeLa cells. D3 indicates
the negative control using a catalytically inactive transposase mutant, E279D. Error bars represent the means
±
s.e.m. of three independent experiments.
Statistical analysis was performed by a one-sample t-test and P values are as follows: N280A, 0.061; K339A, 0.34; and N280A/K339A, 1.7 10
4
.
Table 1 | X-ray crystallographic data collection and
refinement statistics.
Data collection
Space group I 2
1
2
1
2
1
Cell dimensions
a, b, c (Å) 84.24, 113.98, 144.94
a, b, g (°) 90, 90, 90
Resolution (Å) 50–1.4 (1.50–1.40)*
R
sym
0.081 (1.707)
I/sI 17.9 (2.0)
Completeness (%) 100.0 (100.0)
Redundancy 11.2 (11.2)
No. unique reflections 136,613 (25,318)
CC(1/2) 1 (0.694)
Refinement
Resolution (Å) 50–1.4
No. reflections, all/free set 136,607/6,830
R
work
/R
free
0.171/0.195
No. atoms
Protein 3,782
Ligand/ion 219
Water 721
B-factors 25.40
Protein 22.40
Ligand/ion 39.60
Water 36.80
R.m.s deviations
Bond lengths (Å) 0.020
Bond angles (°) 1.91
*Highest resolution shell is shown in parenthesis.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11126 ARTICLE
NATURE COMMUNICATIONS | 7:11126 | DOI: 10.1038/ncomms11126 | www.nature.com/naturecommunications 3
The SB transposase/tnp end/tDNA complex model. To evaluate
and exploit the predictive value of our SB100X structure for
mechanistic studies and protein engineering, we next aimed to
place it into the functionally relevant context by modelling the SB
tDNA capture complex (TCC), which contains the full-length
transposase in complex with tnp end and target DNA (Fig. 2a).
Since an intact TCC structure is so far not available for any
Tc1/mariner transposon, we used the Mos1 PEC structure
7
to
model the SB100X PEC (combining a homology model of the
DNA-binding domains with the catalytic domain coordinates
determined here), and docked a bent tDNA substrate derived
from the prototype foamy virus (PFV) intasome structure
23
into
the positively charged cleft formed at the base of the catalytic
domains (Fig. 2b).
We used a bent tDNA for our modelling, as previous
biochemical and structural data have indicated that Tc1/mariners
require significant target bending for efficient integration
16–18,21
,
a feature that seems to be broadly conserved among DDE
enzymes
22,23,31
. In accordance, the tDNA-binding groove in the
SB100X TCC model has an inward arc with the active sites
situated at the bottom (Fig. 2b) and cannot fit a straight B-form
DNA without major clashes. In turn, the kinked PFV tDNA fits
well in the cleft with its sugar-phosphate backbones delving into
the active sites. In the resulting TCC model the two tnp ends are
arranged in parallel, with the 3
0
OH groups of the TS approaching
the phosphate groups flanking the central TA on opposite strands
of the tDNA (P-O distances: 4.8 and 5.3 Å, Fig. 2c).
The SB TCC model rationalizes previous hyperactive mutations.
Using the TCC model, we first aimed to rationalize previous
mutations within the SB100X catalytic domain that contributed
to generation of the hyperactive SB100X transposase
2
. These
comprise RKEN214–217DAVQ (ref. 27), M243H, and T314N
(Fig. 3). We find that N314 is exposed on the TCC surface, where
the mutation can improve transposase solubility, which has been
shown to be a limiting factor for transposition
2
. H243 is situated
next to the second catalytic residue D244 (Fig. 3b) and establishes
a parallel-displaced p-stack with H249, structuring the 243–251
loop and helping to position D244 in the active site. Finally,
residues 214–217 are located on the beginning of helix a1,
immediately following the loop connecting b3toa1 (aa
209–214), which forms part of the target-binding groove
(Fig. 3c). D214 and Q217 side chains form hydrogen bonds
with main chain atoms in b3 and a1, intimately connecting these
elements and stabilizing the connecting loop conformation.
Simultaneously, A215 and V216 form hydrophobic interactions
with residues in the neighbouring a3 helix, positioning the loop
on the target-binding surface. Thus, the 214–217 mutations have
likely helped shape and ideally position the b3–a1 linker to
interact with the tDNA at the rim of the binding groove.
In addition, the structure offers a rationale for other previously
reported hyperactive mutations that are not present in
SB100X. For example, the ALHKID205–210KLVRIE mutations
27
probably have a similar effect on the b3–a1 loop as the
RKEN214–217DAVQ mutation; the M243Q mutation in SB11
(ref. 25) likely stabilizes the active site the same way as M243H;
while surface mutations (for example, V253H, V255R
25
, T295N
2
and D260K
26
) improve transposase solubility.
Structure-based design of tDNA interaction mutants. Next, we
used our TCC model to map transposase residues that may be
involved in tDNA-binding (green in Fig. 4a–c), and tested the
effect of their mutations in transposition assays (Fig. 4d and
Supplementary Fig. 4). Overall, we selected two sets of amino
acids for mutation: (i) amino acids located near the integration
site that may be involved in TA recognition and/or tDNA
bending (K186 and H187; Fig. 4b, dark green); and (ii) residues
that are likely involved in unspecific tDNA binding on the flanks
of the integration site (I212, N245, K252 and Q271; Fig. 4c, light
green); then, we designed mutants to either abolish the predicted
function of these residues or to generate a hyperactive phenotype.
In the first set, K186 and H187 localize to the end of the clamp
loop, which delves into the major groove of the tDNA substrate in
our TCC model and might contribute to broadening the groove
b
c
tDNA
TIR
TIR
a
tDNA
3′OH
3′OH
Figure 2 | The SB TCC model. (a) Cartoon representation of the model: SB100X dimer (blue), tnp ends (TIRs, grey) and bent tDNA substrate
(tDNA, dark grey). (b) Surface representation of the tDNA-binding groove coloured by electrostatic potential calculated using APBS
48
at 150 mM salt.
The kinked tDNA (grey cartoon) fits well in the positively charged groove. Pink balls represent the 3
0
OH groups of the tnp TS. (c) The tDNA
(stick representation with atomic colouring) scissile phosphates approach the active sites of the TCC.
ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11126
4 NATURE COMMUNICATIONS | 7:11126 | DOI: 10.1038/ncomms11126 | www.nature.com/naturecommunications
and/or kinking the tDNA substrate via unstacking bases at the
integration site (Fig. 4b). In Mos1, R186 was proposed to be
essential for target recognition and its alanine mutation abolished
integration
7
. Interestingly, the equivalent K186S mutation in
SB100X does not affect transposition (Fig. 4d), indicating that the
K186 side chain is not directly responsible for integration
site recognition in SB, while the K186E mutation abolishes
transposition (P ¼ 10
4
; t-test) confirming that proximity of this
protein segment to DNA is essential. In addition, we find that
mutations of H187 also affect transposition: introduction of an
aspartate (H187D) abolishes transposition (Po10
5
; t-test),
while aromatic residues (H187F/Y)—that can function well in
tDNA bending—support activity and even exhibit hyperactive
phenotypes (P ¼ 0.069 and 0.19 based on t-test for F and Y,
respectively) relative to SB100X. This suggests that in SB, H187
(rather than K186) might fulfil a function in target recognition by
side chain specific interactions with tDNA bases at the integration
site in a way similar to R186 in Mos1.
Our second set of mutants (I212, N245, K252 and Q271)
localizes to the surface of the catalytic domain within the tDNA-
binding groove (Fig. 4c,d). For Q271, the TCC model suggests
that its side chain directly contacts the tDNA backbone and helps
target binding. In agreement, both Q271S and Q271E mutants
exhibit reduced transposition (P ¼ 0.21 and o10
4
, respectively;
t-test). N245 forms a hydrogen bond with S270 and helps
position the loop that contains Q271. Consistently, the N245S
mutant is also practically inactive (P ¼ 0.0022; t-test). In turn,
K252 is situated on the verge of the tDNA-binding groove fairly
H249
H243
E279
D244
D153
Q217
D214
A215
V216
α1
I212
b
c
tDNA
TIR TIR
H243
H243
N314
N314
214
DAVQ
217
214
DAVQ
217
a
tDNA
α3
β3
Figure 3 | SB100X structure explains previous hyperactive mutations. (a) The TCC model with catalytic domain residues mutated for the generation of
SB100X shown in yellow. The localization of the interactions shown in (b,c) is indicated by red and orange rectangles, respectively. (b) H243 stabilizes the
catalytic core (red) via p-stacking with H249 (indicated with dashed lines). (c) Residues
214
DA VQ
217
stabilize a flexible loop (including I212) and tether it
to the tDNA-binding surface. Dashed lines indicate hydrogen bonds.
cd e
H187
H187
K186
K186
tDNA
Q271
S270
N245
I212
K252
tDNA
TIR
TIR
tDNA
I212
K252
N245
Q271
I212
K186
H187
K186
H187
a
b
K252
Q271
TIR
160
140
120
100
80
60
40
20
0
140
120
100
80
60
40
20
0
Transposition (%)
Excision (%)
SB100X
D3
K186R
K186S
K186E
H187Y
H187F
H187D
Q271S
Q271E
N245S
K252S
K252E
I212S
I212T
TS
NTS
In vivo
Figure 4 | Structure-based mutagenesis maps tDNA binding and generates novel hyperactive transposases. (a) TCC model with predicted tDNA-
binding residues selected for mutagenesis. (b,c) Zoom-ups on residues predicted to be involved in tDNA recognition or bending (dark green, b) and
residues lining the target-binding groove (light green, c). (d) Transposition assay for the mutants in HeLa cells. P values (based on one-sample t-test):
K186R, 0.93; K186S, 0.22; K186E, 0.0019; H187Y, 0.069; H187D, 6.9 10
6
; Q271S, 0.21; Q271E, 7.5 10
5
; N245S, 0.0022; K252S, 0.68; K252E, 0.042;
I212S, 0.0047; and I212T, 0.27. D3 indicates the negative control with the inactive E279D transposase. (e) Excision assays with the hyperactive I212S
transposase in vivo and in vitro (monitoring cleavage on both DNA strands: NTS, TS). Results are normalized to SB100X. All error bars represent the s.e.m.
for three independent experiments.
NATURE COMMUNICATIONS | DOI: 10.1038/ncomms11126 ARTICLE
NATURE COMMUNICATIONS | 7:11126 | DOI: 10.1038/ncomms11126 | www.nature.com/naturecommunications 5