Abstract
A challenge in developing proteolysis targeting chimeras (PROTACs) is
the establishment of a universal platform applicable in multiple
scenarios for precise degradation of proteins of interest (POIs).
Inspired by the addressability, programmability, and rigidity of DNA
frameworks, we develop covalent DNA framework-based PROTACs (DbTACs),
which can be synthesized in high-throughput via facile bioorthogonal
chemistry and self-assembly. DNA tetrahedra are employed as templates
and the spatial position of each atom is defined. Thus, by precisely
locating ligands of POI and E3 ligase on the templates, ligand spacings
can be controllably manipulated from 8 Å to 57 Å. We show that DbTACs
with the optimal linker length between ligands achieve higher
degradation rates and enhanced binding affinity. Bispecific DbTACs
(bis-DbTACs) with trivalent ligand assembly enable multi-target
depletion while maintaining highly selective degradation of protein
subtypes. When employing various types of warheads (small molecules,
antibodies, and DNA motifs), DbTACs exhibit robust efficacy in
degrading diverse targets, including protein kinases and transcription
factors located in different cellular compartments. Overall, utilizing
modular DNA frameworks to conjugate substrates offers a universal
platform that not only provides insight into general degrader design
principles but also presents a promising strategy for guiding drug
discovery.
Subject terms: DNA, Proteolysis, Drug discovery and development
__________________________________________________________________
The lack of a universal platform for PROTAC development remains a major
bottleneck. Here, the authors report modular DNA framework-based
PROTACs (DbTACs) that enable precise control of the linker length and
selective degradation of diverse targets in different cellular
compartments using various warheads.
Introduction
Protein degradation is an emerging strategy to treat diseases,
especially proteolysis targeting chimeras (PROTACs)^[38]1. PROTACs are
heterobifunctional molecules that comprise a ligand targeting the
protein of interest (POI), an element that recruits E3 ligases, and a
linker connecting the above two moieties^[39]2–[40]4. Among them, the
linker plays a critical role in bridging these two moieties. While many
traditional linkers, including PEG^[41]5, linear aliphatic
chains^[42]6,[43]7, and more rigid (piperazine-type) linkers^[44]5,
have been studied, designing effective linkers remains challenging.
Flexible linkers can cause configuration changes in PROTACs, leading to
drug resistance^[45]8. Moreover, structure–activity relationship
studies (SARs) on linker lengths are primarily empirical and require
time- and labor-intensive research^[46]5. In some instances, randomly
selected linker lengths have been employed for PROTACs^[47]6,[48]9.
Therefore, identifying accurate and controllable protein degradation
profiles necessitates critical consideration of desirable linker types,
lengths, and even attachment points^[49]10.
To date, improving the linker types has enabled the development of
PROTACs from bivalent to trivalent, but the linker and its connection
with the other three fractions pose significant synthetic
burdens^[50]11,[51]12. Moreover, even trivalent PROTACs have
difficulties in efficiently conjugating multiple types of ligands, such
as aptamers, antibodies, and peptides, or in simultaneously degrading
different kinds of “undruggable” targets involving structural proteins,
kinases, or transcription factors^[52]13–[53]16. Thus, it is imperative
to optimize degradation modes by introducing multiple ligands and
targets simultaneously.
Recently, DNA has been engineered to form DNA frameworks^[54]17, such
as DNA tetrahedra^[55]18, octahedra^[56]19, and icosahedra^[57]20, with
well-controlled surface chemistry. The rigidity^[58]21,[59]22,
addressability^[60]23, and artificial programmability^[61]24,[62]25 of
these DNA frameworks provide them with linker-like properties. The
length of DNA frameworks can be precisely controlled with the spacing
of two adjacent deoxynucleotides (~3.3 Å). Moreover, the
caveolin-mediated endocytosis mechanism of DNA frameworks has been
clarified^[63]26, which is conducive to improving the poor cell entry
efficiency of traditional PROTACs. Although bispecific aptamer
chimeras^[64]27, aptamer-PROTAC conjugates^[65]28, RNA-PROTACs^[66]29,
and O’PROTAC^[67]30 have been developed, the use of DNA frameworks as
linkers has not been previously reported. We propose that the
hybridization of DNA frameworks with PROTACs will provide multiple
benefits: (1) High-throughput synthesis of many variants will be
possible through simple dynamic combinatorial chemistry and
self-assembly, saving time and labor compared to the complicated
synthesis of small-molecule PROTACs. (2) Precise positioning of ligands
will be feasible due to the editability and ease of site-specific
modification^[68]31. (3) The linker length can be easily controlled
through DNA framework-engineered PROTACs. (4) A modular toolkit for
rapidly creating highly selective and specific PROTACs will enable
multi-target hydrolysis of different protein subtypes. (5) This
approach will be ideally compatible with various ligands in the
libraries, allowing for the degradation of different “undruggable”
targets.
Inspired by the unique characteristics of DNA frameworks, we have
developed an innovative strategy for the development of DNA
framework-based PROTACs (DbTACs) by combining computational prediction,
DNA self-assembly, and PROTACs technology (Fig. [69]1). To study linker
length-activity relationships, we employ DbTACs formed with DNA
tetrahedra, the ligands of cyclin-dependent kinase (CDK) family protein
and cereblon (CRBN) E3 ligase as representative templates.
Specifically, the position of the CDK9 ligand is fixed, while the CRBN
ligand is shuttled on DNA tetrahedra to produce DbTACs with linker
lengths ranging from 8 to 57 Å. DbTACs with different linker lengths
are visualized, and an optimal DNA linker length of 26 Å is found to be
the most effective in vitro. The difference in binding affinity
explains the mechanism of linker length–activity relationships.
Furthermore, we successfully demonstrate the feasibility of our idea by
creating bispecific DbTACs (bis-DbTACs) sharing one CRBN ligand, which
selectively degrades CDK6 (a protein involves in cell proliferation and
differentiation) and CDK9 (a protein that participates in
transcriptional regulation, DNA repair, and metabolism) in cancer cell
lines. To expand the scope of our platform, antibody-, and DNA
motif-based DbTACs are also developed that facilitate the degradation
of CDK9 kinase and ETS-related gene (ERG) transcription factor. The
information presented here has great potential to be applied to other
DNA frameworks, known ligands, as well as targets in the libraries.
Fig. 1. Design of the DbTACs platform for selectively targeted protein
degradation.
[70]Fig. 1
[71]Open in a new tab
The schematic diagram of the DbTACs platform shows a click
chemistry-mediated programmable linker. The universality of the
platform is highlighted by the following features: (1) precise
degradation of proteins of interest (POIs); (2) simultaneous selective
multi-target proteolysis; (3) compatibility with various warheads, such
as small molecule, antibody, or DNA motif, for degrading different
kinds of “undruggable” target proteins.
Results
DbTACs design, synthesis, and characterization
In this study, we developed a strategy for the design and construction
of DNA framework-based PROTACs (DbTACs) by combining computational
prediction and click chemistry-mediated programmable linker technology.
We utilized the active ligands (L1 and L2) of CDK9 and CRBN that were
previously investigated^[72]32, which were subsequently modified with
an azide group to serve as our models (Supplementary Fig. [73]1a). The
distance between the two ligands was finely tuned by regulating the
number of deoxynucleotides at a length-step of ~3.3 Å. DbTACs were
designed and constructed with computer-aided technology to completely
expose covalent ligands outside the DNA tetrahedra (Supplementary
Fig. [74]1b). All DNA base sequences used for synthesizing DbTACs were
presented in Supplementary Table [75]1. Herein, dibenzocyclooctyne
(DBCO)-modified DNA strand S1 coupled with azide-modified L1 through
click chemistry (Fig. [76]2a(i)), which was selected as a fixed design.
Meanwhile, the azide-modified L2 sites on DBCO-modified S4 sequences
were reasonably changed (Fig. [77]2a(iii)). After that, the above two
DNA strands were self-assembled with DNA strands S2 and S3 to obtain
various DbTACs with different linker lengths, including 8, 11, 16, 21,
26, and 57 Å (Fig. [78]2a(ii)). We specified a basic structural formula
to define each compound using the term DbTACs-linker length, such as
DbTACs-26 Å. The theoretically calculated linker lengths of DbTACs in
the all-atom model match the prediction (Supplementary Fig. [79]1b).
Successful DNA–ligand covalence was confirmed by molecular weight
conjugation of mass spectrometry (Fig. [80]2b, c). The UV‒visible
spectra further showed characteristic absorption peaks of DNA, L1, and
L2 (Supplementary Fig. [81]2a–g). According to Watson-Crick base
pairing, four DNA sequences were then self-assembled into DbTACs by
thermal annealing. The covalent attachment of ligands and step-by-step
self-assembly of DbTACs was further confirmed via distinct bands with
progressively delayed migration in agarose (Supplementary Fig. [82]3a)
and polyacrylamide gel electrophoresis (PAGE) assays (Fig. [83]2d and
Supplementary Fig. [84]3b). Moreover, the emission peak of
representative DbTACs-26 Å overlapped L1 at 438 nm in fluorescence
spectroscopy (Supplementary Fig. [85]2h). The green fluorescence of L2
was substantially quenched by DbTACs-26 Å (Supplementary Fig. [86]2i).
The morphology of DbTACs-26 Å was clearly observed by atomic force
microscopy (Fig. [87]2e). In conclusion, mounting evidence validates
that DNA frameworks can be applied to bridge ligands. We have thus
developed a strategy for simple and rapid preparing candidate DbTACs,
which holds the potential for efficient protein degradation.
Fig. 2. Preparation and characterization of DbTACs.
[88]Fig. 2
[89]Open in a new tab
a Chemical structures and design rationale of (i) S1-L1, (ii) DbTACs,
and (iii) S4-L2 utilizing click chemistry and self-assembly. The mass
spectra of b S1-L1 and c S4-L2. The theoretical molecular weight of
S1-L1 (20,587.52 Da) = S1 chain (20,061.33 Da) + L1 (526.19 Da), and
S4-L2 (20,250.38 Da) = S4 chain (19,795.19 Da) + L2 (455.19 Da). d PAGE
analysis and e atomic force microscope image of representative
DbTACs-26 Å. f Schematic illustration of the ligand covalent sites of
programmable DbTACs equivalents to a DNA tetrahedral scaffold template
with two polyA domains. The 5 nm and 10 nm Au NPs correspond to L2 and
L1, respectively. g Cartoons and TEM images of predesigned
DbTACs-linker length equivalents obtained by adsorbing Au NPs to a DNA
tetrahedral scaffold with two polyA domains. The scale bars, 75 Å and
150 Å, respectively.
Considering the difficulty of visually characterizing site-specific
ligands and linker lengths within DbTACs, single-strand DNA encoders
containing polyadenine (polyA) of varying lengths (Supplementary
Table [90]3) were employed to prepare overhangs at ligand-modified
locations. These overhangs are capable of adsorbing Au nanoparticles
(NPs) via electrostatic interaction^[91]33–[92]36. Thus, Au NPs 5 nm
(Supplementary Fig. [93]4a) and 10 nm (Supplementary Fig. [94]4b) were
introduced to represent ligands L2 and L1, respectively. Furtherly,
programmable atom equivalents (Fig. [95]2f) were fabricated through
self-assembly and easily observed by transmission electron microscopy
(TEM). TEM images clearly revealed the distribution number and spacing
of Au NPs in DbTACs equivalents (Fig. [96]2g). They were identical to
the predesigned bivalent ligand modification and the angstrom-scaled
linker length (average distance calculated to be 8 ± 2 Å (i), 11 ± 2 Å
(ii), 16 ± 4 Å (iii), 21 ± 6 Å (iv), 26 ± 6 Å (v), and 57 ± 2 Å (vi)).
Notably, these results were highly consistent with the linker lengths
measured in all-atom models of DbTACs built using the aforementioned
computer software. The precise control of linker lengths demonstrates
the reliability and effectiveness of this method for ligand
localization in DbTACs.
To ensure the viability of our in vitro evaluation, it was necessary to
test the stability of DbTACs under physiological conditions. We found
that intact DbTACs-26 Å remained visible in the PBS medium even after
24 h (Supplementary Fig. [97]5a, b). The half-life of DbTACs-26 Å in
10% FBS-contained medium was determined to be 29.2 h, and they remained
stable for up to 6 h, suggesting an optimal incubation time for
subsequent experiments (Supplementary Fig. [98]5c, d).
DbTACs are highly potent CDK9 degraders
After establishing stable DbTACs, their degradation kinetics were
investigated in live cells. We have reason to believe that
water-soluble DbTACs can be efficiently internalized into cells through
caveolin-mediated DNA tetrahedral endocytosis^[99]26. To investigate
the fate of DbTACs after internalization, we conducted a subcellular
localization assay using Cy3-labeled DbTACs-26 Å (DbTACs-26 Å-Cy3) in
HepG2 cells. The cells were treated with DbTACs-26 Å-Cy3 for varying
durations and subsequently stained with DAPI to track the intracellular
localization of the compound. The results indicated that DbTACs-26 Å
was effectively internalized by the cells and subsequently translocated
to the nucleus (Supplementary Fig. [100]6). Therefore, we hypothesized
that DbTACs with different linker lengths could be exploited to
precisely control intracellular protein degradation. A human acute
myeloid leukemia cancer cell line, MV4–11, was incubated with various
DbTACs, and the degradation rates were analyzed using western blot (WB)
(Fig. [101]3a). The results showed that DbTACs were excellent in
inducing CDK9 degradation. Impressively, the degradation rate of
DbTACs-26 Å at 200 nM achieved 75.2%, performing as well as the
positive CDK9 targeting PROTAC B11^[102]32. In contrast, the
degradation rates of DbTACs with linker lengths of 8, 11, 16, 21, and
57 Å were significantly lower than that of 26 Å, ranging from 18.2% to
65.2%. Importantly, a DNA tetrahedral control (tDNA) demonstrated that
the DNA architecture alone was not responsible for the observed protein
degradation. These trends were consistent with the degradation capacity
of most degraders at 80 nM (Supplementary Fig. [103]7). Further
immunofluorescence staining images (Fig. [104]3b) showed that
DbTACs-26 Å induced weaker green fluorescence than the control and
other tested DbTACs-linker lengths groups, indicating decreased
expression of the CDK9 protein in this case. Therefore, the catalytic
degradation activity of DbTACs could be programmed using DNA
tetrahedral linkers. DbTACs sandwiched and consistently recruited the
CRL4^CRBN ubiquitin ligase complex to interact with CDK9 protein in a
dose-dependent manner, with a DC[50] value of 95.15 nM (Fig. [105]3c).
Representative DbTACs-26 Å downregulated CDK9 protein levels from 6 to
12 h (Fig. [106]3d). Although no significant protein degradation before
6 h, we cannot exclude the possibility that DbTACs-26 Å was undergoing
cavitation-mediated endocytosis and ubiquitination^[107]26. The slight
recovery after 24 h may result from counteracting factors such as
protein resynthesis or the presence of feedback mechanisms^[108]37.
Fig. 3. Enhanced protein degradation mediated by DbTACs in cancer cells.
[109]Fig. 3
[110]Open in a new tab
a WB analysis and quantification of CDK9 protein levels in MV4–11 cells
treated with 200 nM DbTACs with different linker lengths (DbTACs-8,
−11, −16, −21, −26, and −57 Å) and a positive compound B11 for 6 h.
GAPDH was used as a loading control. An unpaired two-tailed t-test was
used to evaluate statistical significance. ****P < 0.0001 (Control vs.
DbTACs-26 Å), n.s. represents no significance. The error bars indicate
the mean ± SD values; n = 3. b Immunofluorescence staining images of
human hepatoma cells (HepG2) treated with 200 nM DbTACs with different
linker lengths or control for 6 h. The nuclei were stained with DAPI in
blue, and CDK9 protein was stained in green. The red dotted square in
the merged layer indicates an individual cell at a higher
magnification. Scale bars, 20 and 3 μm, respectively. c
Concentration-dependent degradation and d time degradation
characteristics of CDK9 by representative DbTACs-26 Å analyzed by WB.
GAPDH was used as a loading control. An unpaired two-tailed t-test was
used to evaluate statistical significance. **P = 0.0083, ***P = 0.0009,
****P < 0.0001 (Control vs. 6 h+ and Control vs. 12h+). The error bars
indicate the mean ± SD values; n = 3.
To further investigate the efficacy of DbTACs in degrading CDK9 in live
cells, we transfected human embryonic kidney cells (HEK293T) with
plasmids encoding CDK9 fused with a localization signal (eGFP) and
monitored protein abundance in real time. Successful introduction of
exogenous CDK9 was confirmed by Supplementary Fig. [111]8a (i), and
their gradual degradation was observed in HEK293T cells treated with
DbTACs for 6 h (Fig. [112]4a) and 12 h (Supplementary Fig. [113]8a
(ii)) using inverted fluorescence microscopy. Flow cytometry analysis
(Supplementary Fig. [114]8b) and WB analysis (Supplementary
Fig. [115]8c) confirmed the depletion of CDK9-eGFP but not its control
in HEK293T cells after 12 h. These results demonstrate the
effectiveness of DbTACs in degrading CDK9 in live cells and suggest
their potential utility in therapeutic applications.
Fig. 4. Real-time visualization of protein degradation and mechanism of
DbTACs.
[116]Fig. 4
[117]Open in a new tab
a Live-cell imaging was performed to visualize the real-time
localization of CDK9 in HEK293T cells and to track the decrease in CDK9
after treatment with DbTACs-26 Å for 6 h. The scale bars, 40 μm. b
SEC-HPLC analysis of retention time of DbTACs-26 Å after incubation
with human recombinant CDK9 or CRBN protein or both. c Molecular
docking sites of the ternary complex in an all-atom model. SPR
sensorgrams were employed to monitor the interaction between e
DbTACs-26 Å (binary complexes) or d DbTACs-26 Å, f DbTACs-8 Å, and g
DbTACs-57 Å preincubated with human recombinant CRBN protein (ternary
complexes) and immobilized CDK9 protein. h WB analysis of the selective
degradation potency of DbTACs-26 Å in MV4–11 cells. The cells were
co-incubated with different treatments and collected after 6 h. i A
ligand competition test for the degradation of the target protein by
CDK9 degrader DbTACs-26 Å. Inhibitors of CDK9, CRBN, and proteasome
(BAY., P.M., and MG132, respectively) were used, and all signals of
each band were normalized successively to GAPDH.
The enhanced binding affinity of the ternary complex
We further explored the critical molecular mechanism underlying the
ability of DbTACs to promote CDK9 degradation. Compared with free
DbTACs (7.158 min), size exclusion chromatography-high performance
liquid chromatography (SEC-HPLC) results (Fig. [118]4b) revealed the
shortest retention time (5.350 min) when DbTACs bound to both human
recombinant CDK9 and CRBN proteins. This suggested that ternary
complexes were efficiently formed. Molecular docking studies were then
performed to predict the binding mode of DbTACs-26 Å with the protein
binding sites (Fig. [119]4c). The modeling studies revealed that the
ternary complex had a larger absolute docking score (−80.59)
(Fig. [120]4c) compared to the binary complex (−54.54) (Supplementary
Fig. [121]9). Furthermore, among investigated DbTACs with various
linker lengths, DbTACs-26 Å displayed the most stable ternary
conformation, with a docking score of −80.59 (Supplementary
Fig. [122]10). In contrast, the other linker lengths exhibited
relatively lower docking scores: DbTACs-8 Å: −73.39, DbTACs-11 Å:
−64.90, DbTACs-16 Å: −64.01, DbTACs-21 Å: −63.54, and DbTACs-57 Å:
−75.63. This observation suggested that DbTACs-26 Å tended to form more
stable ternary complexes. To further assess the binding kinetics of the
ternary complex, surface plasmon resonance (SPR) assays were used. We
immobilized CDK9 protein on a chip surface and measured the binding
parameters of DbTACs alone (binary binding) or DbTACs preincubated with
the CRBN protein partner (ternary binding). Multivalent effects may be
responsible for the higher binding affinity of the CDK9-DbTACs-CRBN
ternary complex (K[D]^ternary = 42.2 nM) (Fig. [123]4d, Supplementary
Fig. [124]11a) compared to the CDK9-DbTACs binary complex
(K[D]^binary = 248.8 nM) (Fig. [125]4e, Supplementary Fig. [126]11b)
when excluding feeble interactions between CDK9 and CRBN (Supplementary
Fig. [127]11c, d). The “positive cooperativity”
(α = K[D]^ternary/K[D]^binary, α > 1) thus further stabilizes the
ternary complexes, allowing for the desired biological effects.
Importantly, DbTACs-26 Å (K[D]^ternary = 42.2 nM) exhibited a robust
binding affinity with CDK9 and CRBN proteins, in contrast to DbTACs-8 Å
(K[D]^ternary = 185.4 nM) (Fig. [128]4f, Supplementary Fig. [129]11e)
and DbTACs-57 Å (K[D]^ternary = 120.7 nM) (Fig. [130]4g, Supplementary
Fig. [131]11f). This indicates that the enhanced binding affinity of
DbTACs-26 Å is due to the appropriate spacing provided by the 26 Å
linker length, allowing two ligands to match their corresponding sites
and forming a stable ternary complex that facilitate the degradation of
CDK9.
The selectivity and degradation mechanism of DbTACs
To better understand the mechanism behind the selective degradation
induced by topologically engineered DbTACs, we investigated their
activity towards specific proteins. DbTACs-26 Å demonstrated selective
degradation activity toward CDK9 while exhibiting no potency towards
CDK1/2 and CDK6 (Fig. [132]4h). Conversely, the positive control drug
B11 exhibited non-selective specificity towards CDK6 and CDK9. Thus,
like a “dumbbell”, DbTACs retained high selectivity in inducing
specific protein–ligand and protein–protein contact. This could be
attributed to the high stability and rigidity of DNA tetrahedra, which
differs from traditional linear linkers. To further scrutinize the
pathway of CDK9 degradation induced by the selective modulator, several
rescue experiments were carried out. As expected, the excessive
monovalent ligands BAY-1143572 (BAY.) or pomalidomide (P.M.)
effectively blocked the active regions of the kinase and ligase,
thereby impeding the function of DbTACs-26 Å (Fig. [133]4i,
Supplementary Fig. [134]8d). In this case, the degradation of CDK9 was
also hindered in cells pre-treated with MG132, a proteasome inhibitor.
The results confirmed that the mechanism of DbTACs involves effectively
binding to CDK9 and Cullin RING ligase, as well as proteasome-mediated
degradation. To investigate the mRNA levels of CDK9 and downstream
MCL-1 (an antiapoptotic gene), RT-qPCR assays (Supplementary
Fig. [135]12) were further performed. Importantly, there were no
significant differences in mRNA levels between treated groups and
control groups, reinforcing the mechanisms verified above. Having
proven reductions in disease-associated CDK9 protein via the
proteasomal pathway, DbTACs-26 Å led to substantially more potent cell
apoptosis (Supplementary Fig. [136]13) and cytotoxicity (Supplementary
Fig. [137]14) than the control group. To evaluate the level of
autophagy, we employed a fluorescence-based assay using an Autophagy
Staining Assay Kit with MDC (Supplementary Fig. [138]15). After MV4–11
cells were treated with DbTACs-26 Å at different concentrations, the
mean fluorescence intensity of the MDC probe was significantly
increased compared with the control group, indicating the accumulation
of acid compartments. In particular, the fluorescence intensity of the
200 nM DbTACs-26 Å was similar to that of the autophagy-induced
positive group, indicating that autophagosomes were formed during
autophagy. Together, these findings demonstrate that DbTACs promote
autophagy in our cellular model. Collectively, the biological data
support DbTACs for valid protein degradation and downstream
intervention.
To further investigate the selective degradation mechanism of DbTACs,
4D-FastDIA quantitative proteomic analysis was performed on the MV4–11
cells from different treatment groups. Based on these qualified data
(Supplementary Fig. [139]16), a total of 43,504 peptides, 6294
proteins, and 6275 quantifiable proteins were identified in MV4–11
cells. Principal component analysis (PCA) revealed distinct protein
expression patterns between the control group and the DbTACs-26 Å
group, and they were relatively separated from each other
(Supplementary Fig. [140]17). In the proteins exhibiting significant
changes, the DbTACs-26 Å group displayed 11 upregulated proteins and
131 downregulated proteins (Fig. [141]5a). Notably, among the
downregulated proteins, CDK9 protein exhibited the most significant
downregulation in abundance, while CDK1/2 and CDK6 proteins were not
downregulated. These results are consistent with the aforementioned
cellular experiments. To investigate the potential functional
enrichment of the differentially expressed proteins, subcellular
distribution analysis (Fig. [142]5b) and cellular component analysis
(Fig. [143]5c) were performed on these proteins. The results indicated
prominent localization of these differentially expressed proteins in
the nucleus, supporting the targeting of nuclear proteins by
DbTACs-26 Å. Further functional analysis, including gene ontology (GO)
enrichment analysis, highlighted the involvement of differentially
expressed proteins in critical biological processes related to DNA
damage and cell cycle processes (Fig. [144]5d). Importantly, the
molecular function analysis emphasized the impact on cyclin-dependent
protein kinase activity, aligning with targeted degradation of CDK9 by
DbTACs-26 Å (Fig. [145]5e). Furthermore, the KEGG pathway analysis
revealed the p53 signaling pathway as a key mechanism associated with
the significantly changed proteins in the DbTACs-26 Å group
(Fig. [146]5f). Protein–protein interaction analysis further
demonstrated interactions among different proteins, including CDK9, and
their association with TP53, supporting the selective degradation
mechanism of DbTACs-26 Å (Fig. [147]5g).
Fig. 5. Proteomics analysis of differential protein in MV4–11 cells treated
by DbTACs-26 Å.
[148]Fig. 5
[149]Open in a new tab
a Volcano plot showing fold changes of protein abundance from global
proteomics analysis of MV4–11 cells treated with DbTACs-26 Å for 6 h at
200 nM. Statistical test (t-test analysis). b Subcellular localization
prediction of identified proteins using WoLFPSORT. The subcellular
localization of identified proteins was predicated using the WoLFPSORT
database with amino acid sequences of identified proteins. c Gene
ontology (GO) analysis of a cellular component of samples between
DbTACs-26 Å-treated group and control group. Statistical test (Fisher’s
exact test). Molecular function analysis of compound
DbTACs-26 Å-treated group and control group. Cluster analysis of d
biological process, e molecular function, and f KEGG pathway of samples
between DbTACs-26 Å-treated group and control group. n = 3. Statistical
test (Fisher’s exact test). g Protein–protein interaction network
analysis of CDK9 with other proteins.
Multitargeted proteolysis of DbTACs
Informed by the preparation method of DbTACs, a modular trifunctional
agent, named bispecific DbTACs (bis-DbTACs), was generated to extend
this platform. Herein, a covalent ligand 3 (L3) was designed to target
CDK6 protein and shared one E3 ligase ligand (L2) with CDK9 ligand (L1)
on the DNA tetrahedral template (Fig. [150]6a). Similarly, an all-atom
model of bis-DbTACs was initially constructed to observe its
configuration (Supplementary Fig. [151]18). The ligands loading and the
establishment of bis-DbTACs were confirmed by the presence of a single
band and slow migration in gel electrophoresis (Supplementary
Fig. [152]19a, b) and the characteristic peaks in UV‒visible spectra
(Supplementary Fig. [153]19c). We also patterned 5, 10, and 15 nm Au
NPs (Supplementary Fig. [154]4c) with valence bond analogs using
multiple sticky polyA domains, to generate tandem structures of three
Au NPs with different diameters, corresponding to L2, L1, and L3 based
on molecular sizes (Fig. [155]6b). TEM images showed the successful
generation of the desired structures (Fig. [156]6c). Surprisingly,
bis-DbTACs exhibited concentration-dependent consumption in CDK6
levels, while retaining a 55% degradation rate of CDK9 (Fig. [157]6d).
Notably, the individual tDNA component was not responsible for the
degradation of CDK6 or CDK9. Furthermore, neither of bis-DbTACs nor
tDNA induced degradation of CDK1/2. This phenomenon is primarily due to
target selectivity and differences in ligand–protein affinities.
Finally, double-immunofluorescence studies of HepG2 cells showed that
bis-DbTACs markedly attenuated CDK6 (red) and CDK9 (green) fluorescence
compared to the control group (Fig. [158]6e, Supplementary
Fig. [159]19d). To assess the impact of bis-DbTACs on the cell cycle,
we performed flow cytometry analysis employing propidium iodide (PI)
staining (Supplementary Fig. [160]20). After treating MOLM13 cells with
bis-DbTACs, a significant reduction in the proportion of cells in the S
phase, which is critical for DNA replication, was observed when
compared to the control group. This finding implied that bis-DbTACs
possess the ability to impede DNA synthesis, thus modulating the cell
cycle progression. In summary, these data suggest that bis-DbTACs are
effective degraders of CDK6 and CDK9, providing a basis for their
potential applications in cancer treatment. Moreover, this strategy
holds promise for the development of degraders targeting dual or
multiple targets.
Fig. 6. Design, preparation, characterization, and efficacy of bis-DbTACs.
[161]Fig. 6
[162]Open in a new tab
a Schematic illustration of bis-DbTACs design, which is based on
DbTACs. b Schematic illustration of three ligand covalent sites
of bis-DbTACs equivalent to a DNA tetrahedral scaffold with three polyA
domains. Au NPs (5, 10, and 15 nm) correspond to CRBN, CDK9, and CDK6
ligands, respectively. c Cartoon and representative TEM images of
bis-DbTACs equivalents. Scale bars are 75 Å and 200 Å, respectively. d
WB analysis of the selectively targeted degradation ability of
bis-DbTACs at different concentrations and semiquantitative analysis of
the grayscale. The error bars indicate the mean ± SD values; n = 3. e
Immunofluorescence double-staining images of HepG2 cells treated
with/without bis-DbTACs were recorded by confocal laser scanning
microscopy. The cell nucleus was stained with DAPI. CDK6 and CDK9
proteins were labeled with anti-CDK6 and anti-CDK9 antibodies,
respectively. Scale bars, 10 μm.
Generalizability of DbTACs
Many proteins lack known small-molecule ligands, so we focused on
whether antibodies (Abs) could be used as target ligands since they are
more broadly available for “undruggable” proteins. Firstly, CDK9 Abs
and L2 were selected and conjugated onto DNA tetrahedra to prepare
Abs-based DbTACs (Abs-DbTACs), further providing the universality of
the platform. We coupled commercially available Abs that recognize CDK9
full length (1–372 aa) to 5′-thiol-modified DNA tetrahedra through the
heterobifunctional crosslinker N-succinimidyl 3-maleimidoproppionate
(BMPS) (Fig. [163]7a). After purifying Abs-crosslinker covalent via
NAP-5 desalting column, the absorbance of the product was redshifted
from 465 nm to 595 nm by the Bradford assay (Supplementary
Fig. [164]21), indicating the potential for subsequent coupling with
sulfhydrylated DNA tetrahedra. Agarose gel results further confirmed
that DNA tetrahedra were conjugated to the Abs (Fig. [165]7b). The
characteristic UV‒visible absorption peak of Abs-DbTACs was 267 nm,
located between HS-tDNA (260 nm) and Abs (280 nm) (Fig. [166]7c).
Abs-DbTACs exhibited remarkable stability in PBS for up to 12 h
(supplementary Fig. [167]22a) and in cell medium for up to 6 h
(supplementary Fig. [168]22b), and showed resistance to protease
(supplementary Fig. [169]22c) and DNase I (Supplementary Fig. [170]22d)
degradation within 12 h. Further, SEC-HPLC (Fig. [171]7d) analyzed that
the molecular weight largest Abs-DbTACs exhibited the shortest
retention time (6.490 min). The Abs (7.790 min) and DNA tetrahedra
(10.734 min) used alone, however, were easily intercepted by the gel
chromatographic column due to their relatively small molecular weights.
Additionally, the concentration of Abs-DbTACs in the mixture was
quantified to 9.85 µM according to the peak area ratio (23: 1) of
Abs-DbTACs to HS-tDNA. It is worth noting that Abs-DbTACs were prepared
without site specificity, and different lysine residues may be modified
in this study. Therefore, Abs-DbTACs herein were not a single peak. We
hypothesize that this may be beneficial for recognition, as the mixture
of binders can be delivered in different directions to recruit more
POIs and E3 ligase. To further study antigen-antibody binding
properties, WB data (Fig. [172]7e) confirmed a significant 65.5%
decrease of CDK9 expression in MOLM13 leukemia cells following a 12 h
treatment with Abs-DbTACs. Similarly, the individual tDNA exhibited
negligible influence on the degradation process. Overall, these results
demonstrate that antibodies are preferable for DbTACs to degrade
proteins without known small-molecule ligands.
Fig. 7. Design, preparation, characterization, and efficacy of Abs-DbTACs
formed using antibody as POI ligand.
[173]Fig. 7
[174]Open in a new tab
a Strategy for designing Abs-DbTACs using CDK9 antibody as the POI
ligand. b Self-assembly process of Abs-DbTACs was analyzed by agarose
gel electrophoresis. The preparation of Abs-DbTACs was verified by c
UV‒visible spectra and d SEC-HPLC. e WB analysis of the targeted
CDK9 degradation ability of Abs-DbTACs at different concentrations in
MOLM13 cells. A free CDK9 antibody was chosen as the negative control
(NC). A semi-quantitative analysis of their grayscale values was
performed. The error bars indicate the mean ± SD values; n = 3. An
unpaired two-tailed t-test was used to evaluate statistical
significance. **P = 0.0016, ***P = 0.0002, ****P < 0.0001 (Control vs.
191 nM Abs-DbTACs and Control vs. 497 nM Abs-DbTACs), n.s. represents
no significance.
To further illustrate the “one-for-all” properties of DbTACs platform,
different types of ligands and targets were investigated
simultaneously. Specifically, a DNA motif that specifically recognizes
the ETS-related gene (ERG), a highly cancer-related transcription
factor, was used as a ligand. We designed and synthesized
oligonucleotide-based DbTACs (Oligo-DbTACs) by introducing DNA
tetrahedra as a linker conjugating L2 and ERG ligand, with binding
moiety (purple) and three extra nucleotides (black) for protection of
oligo degradation, according to literature modification^[175]30
(Fig. [176]8a). A non-specific sequence (ACGCGATCGAGATGTACTT) was
chosen as a negative control (NC) based on in silico prediction by the
PROMO software, ensuring that no known proteins would bind to this
sequence (Supplementary Fig. [177]23). Mass spectrometry showed that
the DNA strand successfully loaded the reverse ERG binding consensus
sequence (Supplementary Fig. [178]24). The successful construction of
Oligo-DbTACs was confirmed through the observation of significant
delays in electrophoretic mobility shift assay (Fig. [179]8b, c). These
Oligo-DbTACs were delivered into PC-3 prostate cancer cells, which
overexpressed ERG. The kinetics experiment confirmed that Oligo-DbTACs,
but not the NC or tDNA, effectively degraded ERG protein in a
dose-dependent manner (Fig. [180]8d). In summary, these results suggest
that DbTACs derived from oligonucleotide ligands have the potential to
effectively degrade various “undruggable” targets, such as
transcription factors, DNA and RNA binding proteins.
Fig. 8. Design, preparation, characterization, and efficacy of Oligo-DbTACs
formed by DNA motif as POI ligand.
[181]Fig. 8
[182]Open in a new tab
a Schematic illustration of the design strategy for Oligo-DbTACs formed
using DNA tetrahedra as a linker conjugating L2 and ERG ligand, with
binding moiety (purple) and three extra nucleotides (black) for
protection of oligo degradation. b Agarose gel electrophoresis and c
PAGE images of the self-assembly process of Oligo-DbTACs formed using
DNA motif as POI ligand. d WB analysis of the targeted degradation of
ERG protein by Oligo-DbTACs with different concentrations in PC3 cells.
A non-specific sequence was chosen as the negative control (NC) target
ligand. And the semiquantitative analysis of their grayscales. The
error bars indicate the mean ± SD values; n = 3. A paired two-tailed
t-test was used to evaluate statistical significance. *P = 0.0153
(Control vs. 80 nM), P = 0.0111 (Control vs. 200 nM), **P = 0.0051
(Control vs. 500 nM), P = 0.0025 (Control vs. 1000 nM), and P = 0.0030
(tDNA vs. 1000 nM).
In addition, we explored the degradation of proteins located in
different cellular compartments, including cytoplasmic hematopoietic
progenitor kinase 1 (HPK1), which holds great promise for cancer
immunotherapy. Thus, we developed HPK1-DbTACs that utilized a DNA
tetrahedral scaffold co-conjugated with the azide-modified HPK1 ligand
(L4) and CRBN ligand (L2). Through WB analysis (Supplementary
Fig. [183]25a, b), a concentration-dependent degradation of HPK1 by
these HPK1-DbTACs was observed, in contrast to the HPK1 kinase
inhibitor sunitinib, which showed no degradation. Importantly, negative
controls (Control, tDNA, tDNA-L2, and tDNA-L4) did not induce
degradation either. Furthermore, the DC[50] fit of HPK1 degradation was
determined to be 262.0 nM (Supplementary Fig. [184]25c). These findings
demonstrated the potential of DbTACs to target and degrade cytoplasmic
proteins, expanding the application of the DbTACs platform.
Discussion
The research presented here takes inspiration from a commonly used
Chinese proverb, “all for one, and one for all”, attempting to develop
a modular platform for PROTACs. For this purpose, we outline a highly
modular linker design strategy that utilizes framework nucleic acid as
scaffolds, to which both target ligands and E3 binders can be
conjugated. As an advantageous platform, DbTACs afforded resolution at
the angstrom level for interrogating spatial distance-activity
relationships. Indeed, further mechanistic experiments showed that
representative DbTACs-26 Å exhibit high binding affinity, with K[D]
values that are 2.86- to 4.39-fold lower than those of DbTACs-8 Å and
DbTACs-57 Å. Given this advantage, we are capable of precisely
designing DNA linker lengths and manipulating degradation. While DbTACs
carry the risk of off-target effects, their modularity and potential
for selective protein targeting make them a promising strategy for
protein degradation. Continued refinement and optimization of DbTAC
design principles will enable researchers to harness their full
potential, advancing research of DNA-binding proteins and facilitating
the development of targeted therapies.
Additionally, the transition from conventional bivalent to trivalent
PROTACs^[185]12,[186]38 may not be a beneficial approach to immediately
improve multi-target degradation, given poor water solubility and
complex chemical synthesis. On the one hand, the DbTACs platform
addresses poor water solubility. They were primarily based on DNA
sequences, which are water-soluble molecules. They can be easily
dissolved in aqueous solutions or PBS without the need for an organic
solvent. As a result, DbTACs do not require solvent replacement during
the experimental process. This feature simplifies the experimental
procedures and reduces potential side effects associated with residual
organic solvents. On the other hand, water-soluble DbTACs can be easily
synthesized via high-throughput click chemistry, as well as other
methods such as Staudinger ligation chemistry^[187]39 and activated
esters^[188]40. Moreover, the selectivity of DbTACs enables precise
degradation of multiple targets. We have demonstrated herein that DNA
tetrahedral framework-engineering chimeras can be used to deal with
proteins and initiate numerous future applications. For example, DbTACs
formed with antibodies (Fig. [189]7) or DNA motifs (Fig. [190]8), or
other types of elements as ligands have shown a powerful role in
protein degradation. There is scope for prospective variations in
DbTACs technology, exemplified by using different E3 ligase (CRBN, VHL,
MDM2, etc.) ligands (Fig. [191]9a) and ideal DNA framework (DNA
octahedron, icosahedron, etc.) (Fig. [192]9b). If various types of
target ligands are known (small molecules, aptamers, RNA, polypeptides,
nanoantibodies, etc.) (Fig. [193]9a, c, d), this would allow all
warheads to engage under structural constraints. However, to fully
understand the potential, versatility, and generalizability of DbTACs
technology, a more diverse range of target proteins and animal studies
should be investigated in the future.
Fig. 9. The potential of the DbTACs platform based on computational tools
prediction.
[194]Fig. 9
[195]Open in a new tab
All-atom models of DbTACs for coupling with a other small molecule
ligands and E3 ligands created using PolygenDNA and MOE software. b A
DNA octahedron-based DbTACs model for exploring the diversity of DNA
frameworks. All-atom models of DbTACs for coupling with c aptamers and
d RNA chains as pharmacophores built using PolygenDNA and MOE software.
e All-atom model of DbTACs for simultaneous selective degradation of
multiple targets.
In summary, we have proposed DNA framework-engineered DbTACs with
programmable linkers as an approach for the development of empirical
nature in the linker domain. Compared to conventional PROTACs, DbTACs
present a modular approach to generating degraders. This approach
effectively reduces the number of steps involved in preparing the
linker itself, as well as the subsequent coupling of the linker with
the ligands. This represents a significant step forward in the
development of more efficient drug production methods. We anticipate
that DbTACs will serve as a promising platform for researchers to
screen ligand candidates in the future. Furthermore, we expect the
usage of our platform as a chemical biology tool to continue
interdisciplinary collaboration and the innovation of other bi-, tri-,
and even multi-specific drug models (Fig. [196]9e). We look forward to
further validation of this exciting pharmacological modality.
Methods
Chemical synthesis
Details of the synthesis of the compounds and their intermediates are
provided in the Supplementary Note. MestReNova (v12.0.3) was used for
all chemical NMR analyses.
Equipment
Mass spectra of the samples were acquired using an LTQ XL liquid mass
ion trap mass spectrometer (Thermo, USA). The atomic force microscope
image was taken by Burker AXS Dimension Icon. The particle sizes of the
Au NPs were determined using a Litesizer 500 Particle Size Analyzer
(Anton Paar, Austria). The UV‒visible spectra were recorded with a
NANODROP ONE (Thermo Fisher Scientific, USA). Fluorescence spectra were
obtained on a fluorescence spectrometer (Sartorius, Germany).
Fabrication of DbTACs and HPK1-DbTACs
The DNA strands were purchased from Sangon Biotech Co., Ltd. (Shanghai,
China). The DNA sequences used are shown in Supplementary Table [197]1.
To form DbTACs and HPK1-DbTACs, the single-stranded DNA S1 (50 μM) was
mixed with CDK9 ligand (500 μM) or HPK1 ligand (500 μM) at a 1:1 molar
ratio. Additionally, DNA S4 (50 μM) was mixed with CRBN ligand (500 μM)
at a 1:1 molar ratio. The two solutions were stirred at 28 °C for 2 h
at a speed of 800 rpm. Subsequently, the mixtures were combined in
equimolar ratios with pre-designed single-stranded DNA sequences in TEM
buffer (10 mM Tris, 1 mM EDTA, 20 mM MgCl[2], pH = 8.0). The resulting
solution was heated at 95 °C for 5 min and then annealed at 4 °C for at
least 0.5 h using a T series Multi-Block Thermal Cycler (LongGene,
China) to stabilize the structures. Finally, the prepared DbTACs or
HPK1-DbTACs were stored at 4 °C until further use.
Fabrication of bis-DbTACs
For the formation of bis-DbTACs, the single-stranded DNA S1 (50 μM), S2
(50 μM), and S4 (50 μM) were mixed with CDK9 ligand (500 μM), CDK6
ligand (500 μM), and CRBN ligand (500 μM) at a 1:1 molar ratio,
respectively. The three solutions were stirred at 28 °C for 2 h at a
speed of 800 rpm. The subsequent steps are described as mentioned
above. The DNA sequences used are shown in Supplementary Table [198]2.
Fabrication of Abs-DbTACs
The DNA sequences used are shown in Supplementary Table [199]4. The
preparation of antibody-DbTACs conjugates followed a previously
reported protocol^[200]41 with modifications.
Firstly, 1 eq. CDK9 polyclonal antibody was buffer-exchanged into 1×
PBS buffer (pH 7.2) and treated with 80 eq. of the hetero-bifunctional
N-succinimidyl 3-maleimidoproppionate (BMPS) linker (6.67 µL in DMSO)
overnight at room temperature. The BMPS-modified antibody was then
purified using a NAP-5 desalting column and concentrated by vacuum.
Next, the 5′-thiol-modified DNA S1 strand was treated with tris
(2-carboxyethyl) phosphine hydrochloride (TCEP)-containing TEM buffer
(pH 8.0) for 2.5 h at room temperature. Separately, the DNA S4 strand
containing the DBCO group and azide-modified pomalidomide was catalyzed
to form S4–L2 by copper-free click chemistry. The above solutions and
the other two strands forming DNA tetrahedra were mixed at a 1:1:1:1
eq. ratio to synthesize HS-tDNA through thermal annealing. The HS-tDNA
pellet was then ethanol-precipitated to remove excess TCEP and
dissolved with the solution containing the concentrated
antibody-crosslinker directly. The mixture was incubated overnight at
room temperature, and the constituents of the antibody-modified DbTACs
mixture were analyzed by agarose gel electrophoresis (100 V for
30 min). The prepared Abs-DbTACs were stored at 4 °C for further use.
Fabrication of Oligo-DbTACs
The DNA sequences used are shown in Supplementary Table [201]4. To
synthesize Oligo-DbTACs, the DNA S4 strand containing a DBCO group and
azide-modified pomalidomide was firstly catalyzed to form S4–L2 through
copper-free click chemistry. Next, the S4–L2 strand was mixed with DNA
S1_ERG(R), ERG(F), S2, and S3 chains at an equivalent ratio and
thermally annealed to form Oligo-DbTACs. Similarly, a negative control
was prepared using the same method but with the DNA S1_NC(R) and NC(F)
strands. The resulting Oligo-DbTACs were stored at 4 °C until further
use.
Mass spectrometry analysis
The S1–L1 and S4–L2 samples were obtained with three replicates and
analyzed by Sangon Biotech Co., Ltd. using ESI ion-hydrazine mass
spectrometry (Thermo LTQ). A 30 μL sample injection at a final
concentration of 1–5 μM was performed. Data analysis was conducted
using Origin software (Version 2018, OriginLab Software Inc., USA).
Gel separation of DbTACs, bis-DbTACs, Abs-DbTACs, and Oligo-DbTACs
Agarose gel (2 and 3%) was used for gel separation. In brief, 5 µL each
sample was mixed with 1 μL of DNA loading buffer (6X) and loaded onto
the gel. Electrophoresis separation was performed using a Bio-Rad
electrophoresis system (USA) at 120 V for 20 min. The gels were
visualized using a gel electrophoresis imaging system (Bio-Rad, USA).
For the verification of ligand covalent and self-assembly mechanism,
PAGE (10%) was employed. Five microliters of each sample were mixed
with 1 μL of DNA loading buffer (6×) and loaded onto the gel.
Electrophoresis was conducted in 1× TBE buffer at 80 V for 80 min. The
resultant gels were stained with SYBR green (10,000×, Solarbio Science
& Technology Co., Ltd. (Beijing, China)) for 0.5 h in the dark and
imaged using a Bio-Rad imaging system.
Preparation and enrichment of Au NPs
Synthesis of Au NPs with different sizes (5, 10, and 15 nm) was carried
out via the sodium citrate reduction method. In brief, 1% trisodium
citrate (w/v, 4 mL, 4 mL, 4 mL), 1% tannic acid (w/v, 0.7 mL, 0.1 mL,
0.01 mL), and 0.1 M K[2]CO[3] (0.2 mL, 0.025 mL, 0.0025 mL) were mixed,
and then the mixed solution was diluted to 20 mL with deionized water
and named solution 1. Then, 1% HAuCl[4] (w/v, 1 mL, 1 mL, 1 mL) was
added to deionized water to obtain solution 2, which was mixed with
solution 1 at a ratio of 1:4. The resulting mixture was immediately
heated and stirred at 60 °C for 10 min until the color of the solution
turned bright red. The obtained Au NPs were stored at 4 °C and
characterized by TEM and particle size analysis.
Enriching Au NPs was performed according to guidelines. The prepared Au
NPs were concentrated up to 100-fold with bis(p-sulfonatophenyl)
phenylphosphine dihydrate dipotassium salt (BSPP) following established
protocols^[202]34,[203]35,[204]42. The resulting BSPP-protected Au NPs
were quantified using a UV‒visible spectrophotometer.
Imaging of DbTACs and bis-DbTACs equivalents by TEM
The DNA sequences used are shown in Supplementary Table [205]3.
Firstly, BSPP-protected Au NPs (5, 10, and 15 nm) were mixed with
single-stranded DNA to replace the ligand covalent sites with polyA
domains for 1 min at room temperature. Citrate-HCl buffer was added to
the mixture to a final concentration of 10 mM and incubated for 30 min.
The resulting DNA-encoded BSPP-Au NPs were then mixed with other
single-stranded DNA sequences of predesigned DbTACs or bis-DbTACs
equivalents, heated at 95 °C for 5 min in the T series Multi-Block
Thermal Cycler and annealed at 4 °C for at least 0.5 h to obtain
stabilized DbTACs or bis-DbTAC equivalents.
To observe the generated DbTACs or bis-DbTACs equivalents, they were
dropped onto a dry copper grid (AIied High Tech Products, Inc.,
Beijing, China) for 2 min. The samples were then observed and
photographed using TEM (JEM-2000 EX II (JEOL Company, USA)).
Stability of DbTACs and Abs-DbTACs
Specifically, DbTACs-26 Å (11.36 μM) was incubated with PBS and 10%
FBS-contained IMDM medium at 37 °C for different time points, including
0, 1, 2, 4, 6, 12, and 24 h. At each time point, samples were taken and
subjected to detection using 2% agarose gel electrophoresis to analyze
the stability of the DbTACs.
Abs-DbTACs (7.41 μM) were incubated with PBS, 10% FBS-contained 1640
medium, protease (150 U/mL, Shanghai yuanye Bio-Technology Co., Ltd.,
[206]S10051), or DNase I (150 U/mL, Shanghai yuanye Bio-Technology Co.,
Ltd., S10073) at 1:1 (v/v) at 37 °C for different time points,
including 0, 1, 2, 4, 6, 12, and 24 h. The subsequent procedures were
performed as previously described.
UV‒visible spectroscopy of Abs-DbTACs
The purified Abs-crosslinker was mixed with Bradford reagent (Sangon
Biotech Co., Ltd.) at a 1:1 ratio, and the absorbance was measured by a
NANODROP ONE instrument. In the presence of the antibody, a shift in
the maximum absorption peak from 465 to 595 nm was observed.
Cell culture
The acute myeloid leukemia cancer cell lines MV4–11 (catalog number
CRL-9591), human hepatoma cells HepG2 (catalog number HB-8065) and
human prostate cancer cells PC3 (catalog number CRL-1435), and human
embryonic kidney (HEK) 293 T (catalog number CRL-3216) were purchased
from American Type Culture Collection (ATCC, USA). Human T lymphocyte
cell line Jurkat, Clone E6-1 (catalog number, TCHU123) was from Cell
Bank/Stem Cell Bank, Chinese Academy of Sciences. The acute myeloid
leukemia cancer cell line MOLM13 (catalog number iCell-h423) was
purchased from iCell Bioscience Inc., Shanghai. MV4–11 cells were
cultured in Iscove’s modified Dulbecco’s medium (IMDM, Invitrogen,
Shanghai, China) supplemented with 10% FBS (Gibco) at 37 °C in a
humidified atmosphere containing 5% CO[2]. HepG2, HEK293T, and PC3
cells were incubated in Dulbecco’s modified Eagle’s medium (DMEM,
Invitrogen, Shanghai, China) supplemented with 10% FBS at 37 °C in a
humidified atmosphere containing 5% CO[2]. MOLM13 and Jurkat cells were
cultured in Roswell Park Memorial Institute (RPMI-1640, Invitrogen,
Shanghai, China) supplemented with 10% FBS at 37 °C in a humidified
atmosphere containing 5% CO[2].
Subcellular localization of DbTACs
HepG2 cells were seeded onto confocal dishes and incubated overnight
for adhesion. The cells were then treated with Cy3-labeled DbTACs-26 Å
(DbTACs-26 Å-Cy3) at a final concentration of 200 nM for various
durations (0, 10 min, 3 h, 6 h, 12 h, and 24 h). After exposure to
DbTACs-26 Å-Cy3, the cells were rinsed three times with PBS. To fix the
cells, 4% paraformaldehyde was applied for 20 min, followed by PBS
wash. Subsequently, 100 μL of DAPI dye (Jiangsu KeyGEN Bio TECH Corp.,
Ltd., KGA215-50) was added to the cells and allowed to stain at room
temperature for 10 min, protected from light. The staining solution was
then discarded, and the cells were rinsed twice with PBS. Finally, PBS
was added, and inverted fluorescence microscopy (Leica DMi8, Germany)
was employed to observe and capture images.
Western blotting
MV4–11 cells were seeded at a density of 2 × 10^6 cells/well in 6-well
plates and allowed to adhere and grow overnight. Different
concentrations (final concentrations of 80 and 200 nM) of DbTACs
solutions (DbTACs-8 Å, −11 Å, −16 Å, −21 Å, −26 Å, and −57 Å) were
added to the cells, along with positive control (B11), and negative
controls (Control, tDNA). The cells were then co-incubated for 6 h at
37 °C in an incubator. After the incubation period, the cells were
collected and subjected to WB analysis.
Jurkat cells were seeded at a density of 2 × 10^6 cells/well in 6-well
plates and allowed to adhere and grow overnight. Subsequently, various
concentrations (final concentrations of 50, 100, 200, 400, 800, and
1600 nM) of HPKE-DbTACs solutions were added to the cells, accompanied
by negative controls at a final concentration of 1600 nM (Control,
Sunitinib, tDNA, tDNA-L2, and tDNA-L4). The cells were co-incubated at
37 °C in an incubator for 12 h. The following steps are described
above.
MV4-11, MOML13, PC3, or Jurkat cells were cultured in 6-well plates and
treated with specific concentrations of compounds for specific times at
37 °C with 5% CO[2]. The cells were collected by centrifugation, washed
with ice-cold PBS three times, and lysed with RIPA lysis buffer
(RIPA:PMSF = 99:1) (Hangzhou Fude Biological Technology) for 50 min on
ice. The lysates were centrifuged at 14,847g for 25 min at 4 °C. The
protein concentration of the supernatants was measured using an
enhanced BCA protein assay kit (Beyotime Biotechnology (Shanghai,
China)). The protein extract was mixed with dual-color protein loading
buffer, denatured in a 100 °C bath for 10 min, and loaded onto an
SDS‒PAGE gel. The proteins were then transferred to a nitrocellulose
membrane by electrophoresis. Primary antibodies used in this study were
rabbit GAPDH polyclonal antibody (Proteintech Group, Rosemont, IL, USA,
10494-1-AP, 1:10000), rabbit CDK9 polyclonal antibody (Proteintech
Group, Rosemont, IL, USA, 11705-1-AP, 1:1000), mouse CDK1/2 (AN21.2)
monoclonal antibody (Santa Cruz Biotechnology, sc-53219, 1:250), mouse
CDK6 antibody (Proteintech Group, Rosemont, IL, USA, 66278-1-Ig,
1:1000), rabbit ERG polyclonal antibody (Proteintech Group, Rosemont,
IL, USA, 14356-1-AP, 1:1000), rabbit HPK1 polyclonal antibody
(Proteintech Group, Rosemont, IL, USA, 23950-1-AP, 1:1000). The
secondary antibodies used were HRP-conjugated recombinant rabbit
anti-mouse IgG kappa light chain (Proteintech Group, Rosemont, IL, USA,
SA00001-19, 1:5000) and HRP-conjugated Affinity Pure goat anti-rabbit
IgG (H + L) (Proteintech Group, Rosemont, IL, USA, SA00001-2, 1:10000).
Western blot images were captured using UVITEC Imaging Systems (Uvitec
Ltd., UK). The protein levels were quantified by the gray values of
bands in the resulting images using ImageJ.
Immunofluorescence staining
HepG2 cells were seeded into 10 mm confocal dishes and incubated
overnight. Various DbTACs-linker lengths or bis-DbTACs (200 nM) were
applied for 6 h at 37 °C with 5% CO[2]. Subsequently, cells were
quickly washed thrice with ice-cold PBS and then fixed with 4%
paraformaldehyde for 20 min. After fixation, cells were washed with PBS
and permeabilized with 0.1% Triton X-100 at room temperature for
10 min. The samples were then incubated in a blocking solution for 1 h
at room temperature, followed by incubation with primary antibody
overnight at 4 °C. The primary antibody used was mouse CDK6 antibody
(Proteintech Group, Rosemont, IL, USA, 66278-1-Ig, 1:100), rabbit CDK9
polyclonal antibody (Proteintech Group, Rosemont, IL, USA, 11705-1-AP,
1:100). After washing thrice with PBST (PBS with 0.1% Tween 20), the
samples were incubated with the secondary antibody in the dark for
1.5 h at room temperature. The secondary antibodies used were
coraLite594-conjugated donkey anti-mouse IgG(H + L) (Proteintech Group,
Rosemont, IL, USA, SA00013-7, 1:100) and coraLite488-conjugated donkey
anti-rabbit IgG(H + L) (Proteintech Group, Rosemont, IL, USA,
SA00013-6, 1:100). The nuclei were counterstained with DAPI (Jiangsu
KeyGEN Bio TECH Corp., Ltd, KGA215-50, 1 µg/mL, 100 µL) at room
temperature for 10 min, followed by washing. Images were acquired using
an inverted fluorescence microscope or a confocal laser microscope
(FV1000, Olympus, Japan).
Plasmid construction
The expression plasmid CDK9_pcDNA3.1(+)-N-eGFP was constructed by
inserting a synthesized gene encoding CDK9 into the pcDNA3.1(+) vector.
The vector contains an eGFP signal sequence, and the insertion was
performed using the multiple cloning sites (MCS) BamHI/EcoRI.
Live cell real-time imaging of CDK9 content
HEK293T cells were cultured into 10 mm confocal dishes until they
reached 70% confluency and were then incubated overnight. The CDK9-eGFP
plasmid was constructed and transfected into HEK293T cells using
Lipofectamine^TM 2000 transfection reagent (Thermo Fisher Scientific,
USA) for coculture for 6 h. After 6 h, the medium was changed to
FBS-containing DMEM. The expression of CDK9–eGFP fusion protein with
green fluorescence was recorded by an inverted fluorescence microscope
after 65 h. To investigate the dynamics of CDK9 protein content in
living cells, changes in fluorescence intensity were further observed
by administering DbTACs for 6 h and 12 h.
Flow cytometry of CDK9 on live cells
HEK293T cells were seeded into 6-well plates and incubated overnight to
reach approximately 70% confluency. After successfully expressing
CDK9-eGFP fusion protein in HEK293T cells, the PBS, DbTACs-26 Å, and
B11 (final concentrations of 200 nM) were then added to cells and
incubated for an additional 12 h. After treatment, cells were detached
from the wells using trypsin-EDTA, washed and resuspended with PBS at a
concentration of 1 × 10^6 cells/mL. CDK9-eGFP levels were measured
using flow cytometry (BD Accuri® C6, USA) equipped with a 488 nm laser
for excitation and a 530/30 nm bandpass filter for detection of GFP
fluorescence. Data were analyzed using FlowJo software (BD Biosciences,
USA).
Cell apoptosis assay
The MV4–11 cells (4.0 × 10^5 cells/well) were seeded into each well of
a 6-well transparent plate and incubated overnight. After 24 h of
seeding, the test compounds were added at final concentrations of 0,
80, 100, 150, 200, and 250 nM, and the cells were incubated for another
6 h. Next, the cells were collected by centrifugation at 100g for 5 min
and washed twice with PBS. The staining solution containing annexin
V-FITC and propidium iodide (PI) was prepared according to the
manufacturer’s instructions (Beyotime Biotechnology, Shanghai, China).
Then, the cells were resuspended in 500 μL of binding buffer containing
the staining solution and incubated at room temperature for 20 min away
from light. Finally, the samples were analyzed by flow cytometry in an
ice bath, and the data were analyzed using FlowJo software.
Cell autophagy assay
MV4–11 cells were seeded into a 96-well blackboard and allowed to
adhere overnight. Subsequently, the cells were treated with various
compounds for a duration of 12 h. As a positive control, an autophagy
inducer (Beyotime Biotechnology, Earle’s Balanced Salt Solution) was
applied for 4 h. After treatment, the cells were stained using the
Autophagy Staining Assay Kit with MDC (Beyotime Biotechnology, C3018S)
following the provided instructions. The fluorescence microplate
(SpectraMax® iD3) was employed to evaluate the levels of autophagy in
MV4–11 cells post-treatment. The optical density (OD) values were
measured at an excitation wavelength of 335 nm and an emission
wavelength of 512 nm.
Cell cycle assay
Briefly, MOLM13 cells were seeded in 6-well plates and treated with
bis-DbTACs (final concentrations of 80, 200, 250, 500, and 1000 nM) or
PBS (Control) for 12 h. After the treatment period, cells were
harvested and fixed in ethanol. Fixed cells were then stained using the
Cell Cycle Detection Kit (Jiangsu KeyGEN Bio TECH Corp., Ltd., KGA,
KGA512) following the manufacturer’s protocol. The flow cytometry data
were processed using ImageJ to determine the distribution of cells in
different phases of the cell cycle, including the G1, S, and G2 phases.
qRT‒PCR analysis
The MV4–11 cells were seeded into 6-well plates and allowed to incubate
overnight. DbTACs-26 Å (final concentrations of 0, 80, and 200 nM) was
incubated with cells for 6 h, then cells were collected and RNA was
extracted using the TRIzol method. The RNA purity and concentration
were measured using a NANODROP ONE spectrophotometer. The cDNA was
synthesized from the extracted RNA using the PrimeScript™ RT reagent
kit (Takara Bio, Japan). The primer sequences used for qRT-PCR analysis
were as follows: CDK9_F: 5′-CTCTGCGGCTCCATCAC-3′, CDK9_R:
5′-GCCTGTCCTTCACCTTCC-3′; MCL-1_F: 5′-GATGTGAAATCGTTGTCTCGAG-3′,
MCL-1_R: 5′-GAAATGAGAGTCACAATCCTGC-3′; Actin_F:
5′-CCTCACTGTCCACCTTCC-3′, Actin_R: 5′-GGGTGTAAAACGCAGCTC-3′. qRT-PCR
was performed on a Real-Time PCR Instrument (Thermo Fisher Scientific,
USA) using the following protocol: initial incubation at 37 °C for
15 min, followed by 5 s at 85 °C to inactivate previous amplicons with
uracil-DNA glycosylase and a 2 min incubation at 95 °C to activate the
Taq polymerase. The amplification cycle, consisting of 15 s at 95 °C,
30 s at 60 °C, and 30 s at 70 °C, was repeated 40 times. The relative
expression levels of CDK9 and MCL-1 were determined using the ΔΔCt
method (Ct gene of interest−Ct internal control). Actin was used as the
internal control. The results were presented as the fold change
relative to the control group and plotted. All experiments were
repeated six times.
Quantitative proteomics analysis
The 4D-FastDIA-based quantitative proteomic analysis of human MV4–11
cells was carried out by Jingjie PTM Biolabs Inc. (Hangzhou, China).
Samples were sonicated three times on ice using a high-intensity
ultrasonic processor (Scientz) in lysis buffer (8 M urea, 1% protease
inhibitor cocktail). The remaining debris was removed by centrifugation
at 12,000g at 4 °C for 10 min. Finally, the supernatant was collected,
and the protein concentration was determined with a BCA kit according
to the manufacturer’s instructions.
For digestion, the protein solution was reduced with 5 mM
dithiothreitol for 30 min at 56 °C and alkylated with 11 mM
iodoacetamide for 15 min at room temperature in darkness. The protein
sample was then diluted by adding 100 mM TEAB to urea concentration
less than 2 M. Subsequently, trypsin was added at 1:50
trypsin-to-protein mass ratio for the first digestion overnight and
1:100 trypsin-to-protein mass ratio for a second 4 h-digestion.
Finally, the peptides were desalted by the C18 SPE column.
The tryptic peptides were dissolved in solvent A (0.1% formic acid, 2%
acetonitrile/in water) and directly loaded onto a homemade
reversed-phase analytical column (25 cm length, 75/100 μm i.d.).
Peptides were separated with a gradient from 6 to 24% solvent B (0.1%
formic acid in acetonitrile) over 70 min, 24–35% in 14 min, and
climbing to 80% in 3 min then holding at 80% for the last 3 min, all at
a constant flow rate of 450 nL/min on a nanoElute UHPLC system (Bruker
Daltonics).
The peptides were subjected to a capillary source followed by the
timsTOF Pro (Bruker Daltonics) mass spectrometry. The electrospray
voltage applied was 1.60 kV. Precursors and fragments were analyzed at
the TOF detector, with an MS/MS scan range from 100 to 1700 m/z. The
timsTOF Pro was operated in parallel accumulation serial fragmentation
(PASEF) mode. Precursors with charge states 0–5 were selected for
fragmentation, and 10 PASEF-MS/MS scans were acquired per cycle. The
dynamic exclusion was set to 30 s.
The resulting MS/MS data were processed using MaxQuant search engine
(v.1.6.15.0). Tandem mass spectra were searched against the human
SwissProt database (20,422 entries) concatenated with the reverse decoy
database. Trypsin/P was specified as a cleavage enzyme allowing up to
two missing cleavages. The mass tolerance for precursor ions was set as
20 ppm in the first search, and 5 ppm in the main search, and the mass
tolerance for fragment ions was set as 0.02 Da. Carbamidomethyl on Cys
was specified as a fixed modification, and acetylation on protein
N-terminal and oxidation on Met were specified as variable
modifications. FDR was adjusted to <1%.
Then, PCA was used to evaluate the repeatability of samples from each
group. Differential proteins were obtained after the qualification of
samples, whose differences in relative quantification in two groups
were compared by t-test, and the corresponding p-value was calculated.
In addition, with a criterion of p-value ≤ 0.05, the protein
ratio > 1.5 was regarded as upregulation, while the protein
ratio < 1/1.5 as downregulation.
Based on the identified proteins, the subcellular localization analysis
was performed using the WoLF-PSORT database; GO annotation is to
annotate and analyze the identified proteins with eggnog-mapper
software (v2.1.6). The software is based on the EggNOG database
(v5.0.2, [207]http://eggnog5.embl.de/#/app/home). Extracting the GO ID
from the results of each protein note and then classifying the protein
according to Cellular Component, Molecular Function, and Biological
Process; Kyoto Encyclopedia of Genes and Genomes (KEGG) database (v5.0,
[208]http://www.kegg.jp/kegg/mapper.html) was used for KEGG pathway
enrichment analysis. Fisher’s exact test was used to analyze the
significance of KEGG pathway enrichment of differentially expressed
proteins (using the identified protein as the background), and P
value < 0.05 were considered significant. Furthermore, all
differentially expressed protein database accession or sequence were
searched against the STRING database (v11.5,
[209]https://cn.string-db.org/) for protein–protein interactions. Only
interactions between the proteins belonging to the searched data set
were selected, thereby excluding external candidates. STRING defines a
metric called “confidence score” to define protein-protein interaction
(PPI) confidence; we fetched all interactions that had a confidence
score ≥ 0.7 (high confidence). PPI network form STRING was visualized
using the R package “networkD3” tool.
SEL-HPLC analysis of the formation of ternary complexes
DbTACs-26 Å were preincubated with either human recombinant CDK9
(Solarbio Science & Technology Co., Ltd. (Beijing, China)), CRBN
(Cloud-Clone Corp (Wuhan, China)) or both proteins together at room
temperature for 0.5 h in an equal molar ratio. The formation of ternary
complexes was analyzed by size exclusion high-performance liquid
chromatography (SEC-HPLC) using an FLM Scientific Instrument Co., Ltd.
(Guangzhou, China) column, with the retention time of the samples being
measured at 260 nm. The mobile phase used was Tris-NaCl buffer with a
pH of 7.4, and a flow rate of 0.8 mL/min was maintained throughout the
analysis.
SEL-HPLC analysis of Abs-DbTACs
Abs-DbTACs, HS-tDNA, and free CDK9 polyclonal antibody (final
concentrations of 200 nM) were analyzed by an SEC-HPLC column (FLM
Scientific Instrument Co., Ltd. (Guangzhou, China)). PBS was used as
mobile phase, pH = 7.4, and flow rate: 0.8 mL/min. The retention time
of the samples was monitored and analyzed to determine the composition
of the complexes.
All-atom models of DbTACs
All-atom models of various DbTACs, bis-DbTACs, Abs-DbTACs,
Oligo-DbTACs, etc., were constructed using the PolygenDNA
program^[210]43,[211]44 and MOE software. To generate an all-atom model
of a DNA tetrahedron using the PolygenDNA program, the DNA sequence of
interest was specified as input. A double-stranded DNA helix was then
generated, and the tetrahedral vertices were placed at the desired
positions in 3D space. A series of energy minimization and geometry
optimization steps were applied to refine the initial atom positions
and adjust the geometry of the helix to achieve optimal bond lengths,
bond angles, and dihedral angles. The resulting pdb file contained
all-atom coordinates for all atoms in the DNA tetrahedron, including
hydrogen atoms and other small molecules. Molecular visualization
software, such as MOE, was used to visualize and analyze the model and
make any necessary adjustments. To generate the final pdb file of
series DbTACs by covalently linking the DNA tetrahedra and ligands, MOE
was used. Firstly, the 3D coordinates of the ligands were generated and
optimized using the MOE Builder module. Next, the DNA tetrahedra pdb
file was loaded into MOE, and the 3D coordinates of the DNA tetrahedra
were optimized using the MOE Protein Preparation Wizard. The ligands
were then docked into the optimized DNA tetrahedra structure using the
MOE Dock module, and the covalent bonds between the ligands and DNA
tetrahedra were formed using the MOE Editor module. Finally, the
resulting structure was energy minimized using the MOE Energy
Minimization module to obtain the all-atom model of various DbTACs. MOE
employs various algorithms to minimize the energy of the structure and
optimize the geometry of the molecule. The optimized pdb file of
various DbTACs was used for subsequent analysis and simulations.
Molecular docking analysis
The molecular docking studies of DbTACs-linker lengths with CDK9
protein and DbTACs-linker lengths with CDK9-CRBN proteins were
performed using the MOE software package developed by Chemical
Computing Group. The crystal structures of CDK9 protein (pdb ID: 3BLH)
and CRBN protein (pdb ID: 4CI3) were downloaded from the Protein Data
Bank (PDB) and prepared using the MOE Protein Preparation Wizard.
For the DbTACs-linker lengths with CDK9 docking, the DbTACs pdb file
generated from the previous step was loaded into MOE. Prior to docking,
the protein underwent preprocessing using QuickPrep with default
options, which included rectifying structural inaccuracies,
incorporating hydrogen atoms, optimizing three-dimensional H-bonding
networks, eliminating water molecules beyond 4.5 Å from the protein,
and minimizing within a limited range of 8 Å of the altered base pairs.
Next, the CDK9 ligand on the DbTACs was identified as the Ligand Site
using the MOE Site Finder module, and a docking box was defined around
the site. The native ligand pockets of the CDK9 protein were selected
as the Receptor Site in the docking studies. The MOE Dock module was
then used to dock DbTACs into the binding site of CDK9 using a
Refinement with Rigid Body protocol, which generated several docking
poses based on the predicted interaction energies between the ligand
and receptor. The docking poses were ranked according to their binding
affinities, and their Docking Score S^[212]45 was recorded. The pose
with the highest Docking Score was selected as the final result. The
Docking score S was calculated based on the following formula 1:
[MATH: ΔGBindingCalc=α23
(EInterCoul)
+ΔGBindR⏟Δ<
mrow>GBindElec+
EIntervdW+ΔGBindnpsol⏟Δ<
mrow>GBindNon−polar+c :MATH]
1
The Docking scoring S was primarily composed of
[MATH: ΔGBindingCalc. :MATH]
, which included
[MATH: ΔGBindElec. :MATH]
and
[MATH: ΔGBindNon−polar :MATH]
.
[MATH: EInterCoul. :MATH]
and
[MATH: EIntervdW :MATH]
represented the columbic and van der Waals contribution to binding,
respectively.
[MATH: ΔGBindR :MATH]
was the change in reaction field energy upon binding. The
[MATH: ΔGBindnpsol :MATH]
term represented the change in non-polar solvation (van der Waals and
cavitation cost) upon binding. Furthermore, the scaling factor for
electrostatic interactions was empirically determined to be 2/3, which
improved accuracy compared to the theoretically ideal value of
1/2^[213]46.
SPR binding assay
For binary binding experiments, stock solutions of DbTACs-26 Å or free
CRBN protein were serially diluted in PBS-P (containing 0.5% Surfactant
P20) running buffer (twofold serial dilution). The diluted solutions
were injected over a CM5 chip coated with immobilized CDK9.
For the ternary binding study, DbTACs-8 Å/DbTACs-26 Å/DbTACs-57 Å were
mixed with a solution of CRBN protein to prepare a final solution of
400 nM DbTACs and 800 nM CRBN protein in PBS-P running buffer. The
complexes were preincubated in PBS-P running buffer for 0.5 h, followed
by serial dilutions (six-point twofold serial dilutions).
SPR binding responses for binary and ternary complexes were performed
in multicycle kinetic mode at 298.15 K with a contact time of 60 s, a
flow rate of 30 µL min^−1, and a dissociation time of 60 s. The raw
sensorgram data was processed using Biacore T200 Evaluation Software.
The reference surface and blank injections were subtracted from the raw
data before data analysis. The steady-state affinity (SSA) model was
used to calculate the association rate (k[on]), dissociation rate
(k[off]), and dissociation constant (K[D]) for the binding affinity
between binary and ternary complexes.
Statistics and reproducibility
The results are presented as the mean ± standard deviation (SD).
Differences between groups were analyzed for statistical significance
using t-test analysis in GraphPad Prism (version 8, GraphPad Software
Inc., USA). Statistical significance was accepted with P < 0.05, *:
P < 0.05, **: P < 0.01, ***: P < 0.001, ****: P < 0.0001, and n.s.: no
significant difference. Each experiment was independently repeated
three times, and similar results were obtained. The data analysis was
conducted by using Origin software (Version 2018, OriginLab Software
Inc., USA).
Reporting summary
Further information on research design is available in the [214]Nature
Portfolio Reporting Summary linked to this article.
Supplementary information
[215]Supplementary Information^ (9.4MB, pdf)
[216]Peer Review File^ (7.1MB, pdf)
[217]Reporting Summary^ (2.8MB, pdf)
Acknowledgements