key: cord-0296324-qijj3bl2
authors: Sekhon, Harsimranjit; Loh, Stewart N.
title: Engineering Protein Activity into Off-the-Shelf DNA Devices
date: 2022-01-03
journal: bioRxiv
DOI: 10.1101/2022.01.03.474821
sha: f31ab8a18ff1ed38b46749a08784fee46cc5d426
doc_id: 296324
cord_uid: qijj3bl2

DNA-based devices are relatively straightforward to design by virtue of their predictable folding, but they lack biological activity. Conversely, protein-based devices offer a myriad of biological functions but are much more difficult to design due to their complex folding. This study bridges the fields of DNA engineering and protein engineering to generate a protein switch that is activated by a specific DNA sequence. A single protein switch, engineered from nanoluciferase using the alternate frame folding mechanism and herein called nLuc-AFF, is paired with different DNA technologies to create a biosensor for a DNA or RNA sequence of choice, sensors for serotonin and ATP, and a computational device that processes two DNA inputs. nLuc-AFF is a genetically-encoded, ratiometric, blue/green-luminescent biosensor whose output can be quantified by cell phone camera. nLuc-AFF is not falsely activated by decoy DNA and it retains full ratiometric readout in 100 % serum. The design approach can be applied to other proteins and enzymes to convert them into DNA-activated switches.

The GCN4 binding domain (orange) was 76 inserted between residues 50 -51 of the nLuc domain of the GeNL protein (green and grey) to create the native 77 frame of the switch (green-orange-grey). To establish the CP frame of the switch (grey-red-blue), the segment of 78 nLuc that was N-terminal to GCN4 (residues 6 -50) was then duplicated (residues 6* -50*; blue) and appended 79 to the C-terminus using a 6-AA linker (red). nLuc-AFF can fold either in the native frame (green luminescence) or presence of a target DNA sequence. We chose to base the AFF scaffold on nanoluciferase (nLuc, a 91 blue luminescent protein derived from deep-sea shrimp Oplophorus gracilirostris) because it possesses 92 high thermodynamic stability 16 (which helps it retain function after ligand binding domain insertion and 93 circular permutation) and is the brightest luciferase currently known 17 . To establish the blue and green 94 states of the sensor, we employed the variant of nLuc (GeNL) in which the GFP variant mNeonGreen 95 (mNG) was fused to the N-terminus of nLuc 18 . The blue emission of nLuc is captured by mNG and 96 essentially completely converted to green fluorescence via bioluminescent resonance energy transfer. 97

For efficient resonant coupling, it was essential to bring the two proteins into proximity by deleting the 98 first four amino acids of nLuc and the last ten amino acids of mNG 18 . 99

Our hypothesis was that a green/blue luminescent switch could be engineered by introducing an 100 intramolecular conformational change that extinguishes luminescence of the nLuc domain fused to 101 mNG and turns on luminescence of a second, circularly-permuted nLuc within the same molecule, via 102 the AFF mechanism. As a result of the AFF conformational change, permuted nLuc becomes 103 separated from mNG by a >100 amino acid segment, the length of which is predicted to be >75 Å (vide 104 infra). Permuted nLuc is thus expected to emit blue luminescence (Fig. 1B) . To couple this 105 conformational change to DNA binding, we turned to GCN4 DNA binding domain as the input domain. 106 GCN4 is a 56 amino-acid peptide consisting of a constitutively dimerized, C-terminal leucine zipper and 107 an N-terminal DNA binding region that is largely unstructured in the absence of its 11-nucleotide 108 consensus DNA sequence (AP-1). GCN4 folds into a rigid, rod-shaped molecule of 75 Å length when 109 bound to AP-1 19 (Fig. 1B) . We previously applied this folding reaction to stretch and unfold the enzyme 110 barnase in the GCN4-barnase fusion protein 20 . We incorporate the AP-1 sequence into off-the-shelf 111 DNA devices to construct two-input logic gates as well as biosensors for ATP and serotonin. 112

Beginning with the GeNL protein, we duplicated the N-terminus of nLuc (residues 1 -50; the 115 duplicated sequence is denoted by asterisks) and appended residues 1* -50* to the C-terminus of the 116 molecule using a 12-AA linker to generate the nLuc-AFF scaffold. The native and CP frames are shown 117 in Fig. 1A . We chose to duplicate this segment because position 50 is at a surface loop which was 118 previously shown to tolerate circular permutation 21 as well as insertions 18 . To couple the native-to-CP 119 fold shift to DNA binding, we inserted GCN4 in the surface loop at position 50 of the native fold. In the 120 ligand-free (OFF) state of the switch, GCN4 is predicted to resemble an extended, disordered surface 121 loop in the absence of AP-1. The native fold should tolerate this insertion well and the OFF state is 122 expected to be green. In the AP-1-bound (ON) state, the length of GCN4 will stretch the 6.3 Å C α -C α 123 distance between the ends of the loop, splitting the native frame and triggering the fold shift to the CP 124 frame (Fig. 1B) . The CP form of nLuc is separated from mNG by 118 amino acids, 56 of which are a 75 125 Å rod, and the nLuc-AFF is thus expected to emit blue light in the CP frame. We constructed and 126 purified the individual native and CP folds and verified that they were functional green and blue 127 luminescent proteins, respectively (Fig. 1C) . 128

The first iteration of nLuc-AFF exhibited mostly blue luminescence, indicating that it was 129 functional but already in the ON state in the absence of AP-1 (Supporting Fig. S1 ). We consequently 130 destabilized the CP frame by shortening the linker that connected the N-and C-termini of the CP frame 131 from 12 AA to 8, 6, 4, and 2 AA. Shorter linkers can destabilize a permuted protein by forcing its termini 132 together, a phenomenon previously demonstrated with barnase 22 . Shortening the linkers progressively 133 lowered the blue:green ratio as expected, but the population of CP frame was still too large even at 2 134 AA linker length (Supporting Fig. S2A ). 135

To further destabilize the CP fold, we deleted residues 1* -4* from the CP frame (VFTL; the 136 same residues that had been removed from the native frame to make the GeNL variant). We then 137 joined the termini of the CP frame with 10, 6, and 4 AA (with the VFTL truncation the effective linker 138 lengths were 6, 2, and 0 AA). The blue:green ratio diminished with decreasing linker length as before 139 (Supporting Fig. S2B ), and this ratio increased after addition of AP-1 for all three variants (Supporting 140 Fig. S3B ). This increase was largely due to a decrease in green luminescence without an increase in 141 blue luminescence for the 0 AA linker construct. The 6 AA linker variant had the best combination of 142 high ratiometric change on AP-1 binding and low background signal, and we used this construct for all 143 further experiments. These results demonstrate that the blue/green populations of the nLuc-AFF sensor 144 can be rationally tuned by adjusting linker length in the CP frame. 145

The apparent affinity for the nLuc-AFF for AP-1 was measured by adding various concentrations 147 of AP-1 to a fixed amount of biosensor and measuring the ratiometric change in blue:green emission 148 using a scanning plate reader (dividing intensity at 460 nm by intensity at 520 nm; denoted as 149 L460/L520) or cell phone camera (dividing the intensity of the blue channel by the intensity of the green 150 channel). The apparent K D of 12.3 nM ± 5.2 nM ( Fig. 2A, Supporting Fig. S4A ) is similar to the reported 151 affinity of the isolated GCN4 peptide for AP-1 (22 nM ± 9 nM) 23 luciferase-based sensors, as many existing designs show diminished luminescence in serum due to 165 absorbance by extraneous components 24 . In 10 % serum, nLuc-AFF bound to AP-1 with similar affinity 166 (K D = 37 nM ± 11 nM; Fig. 2B ) as in buffer, as determined by plate reader and cell phone camera. As 167 serum concentration was raised to 100 %, we observed a decrease in overall luminescence intensity 168 ( Fig. 2D) , as expected from serum absorbance, but the change in blue:green ratio due to biosensor 169 activation remained constant within error (Fig. 2C, Fig. 2D ). Raw cell phone images confirm that the 170 samples are blue in the presence of AP-1 and green in the presence of nonconsensus (NC) DNA (Fig.  171   2C ). This result, and the close agreement between the plate reader and cell phone camera data (Fig.  172 2A, Fig. 2B ), demonstrate that it is possible to quantify the turn-on response using only a cell phone 173 camera, by simply dividing the intensities of the blue and green channels without any image 174 processing. 175

Having developed the nLuc-AFF biosensor that is activated by AP-1, our next goal was to 177 couple presentation of the AP-1 sequence to binding of other DNA sequence inputs. Toehold mediated 178 strand displacement (TMSD) 25,26 is a powerful tool that has been extensively used in applications such 179 as signal amplification 6,7 , RNA computation 8, 9 , and many others. It relies on the ability of an invader 180 strand (ssDNA or ssRNA) to hybridize to a short ssDNA (the toehold) that overhangs from a dsDNA 181 stem-loop structure in which the stem and toehold comprise the sequence complementary to that of the 182 invading strand. The invading strand then fully hybridizes with the dsDNA hairpin, displacing the strand 183 of identical sequence to the invader. 184

We selected a 24-nt sequence from the SARS-CoV-2 (SARS2) genome to serve as the 185 invading strands for proof-of-concept. We designed two DNA hairpins in which the 5' stem and the 5' 186 toehold consisted of the sequence complementary to the SARS2 oligonucleotide. The hairpins differed 187 only by having the single-stranded AP-1 sequence (ssAP-1) (probe 1) or its complement (probe 2) 188 embedded in their loops. Binding of the SARS2 oligonucleotide opens both probes, exposing their 189 loops. The ssAP-1 sequences then hybridize, generating the activating complex that contains the 190 duplex AP-1 input ligand for nLuc-AFF (Fig. 3A ). Since AP-1 is a palindromic sequence with a single 191 mismatch in the center, NUPACK 27 simulations predicted that including ssAP-1 in the loop would 192 extend the stem and make it too stable to be opened by the invading strand (Supporting Fig. S5A ). We 193 therefore added a 3-nt 'bulge' spacer between the stem and the ssAP-1 loop (Supporting Fig. S5B ). 194

The bulge was predicted to interrupt base stacking at the stem-loop junction, increasing efficiency of 195 hairpin opening. 

SARS2 oligonucleotide was incubated with either probe 1 or probe 2 alone, a new product of higher 213 MW appeared, corresponding to the expected dimer of invading strand and opened probe (Fig. 3B) . We 214 observed a new, larger MW band upon adding the SARS2 oligonucleotide to both probes, consistent 215 with formation of the activating complex. 216

Biosensor activation was tested by mixing nLuc-AFF with the two probes in the presence or 217 absence of SARS2 oligonucleotide. We detected a decrease in green luminescence and an increase in 218 blue luminescence only in the presence of the SARS2 strand (Fig. 3C ). L460/L520 revealed a 3.2-fold 219 change in biosensor turn-on along with a color change visible by cell phone camera (blue:green ratio 220 change of 12.2) (Fig. 3D ). This finding, along with the PAGE data, verify that the activating AP-1 duplex 221 is formed from two hairpins only in the presence of the initiating SARS2 sequence. These results 222 demonstrate that our biosensor can be used with TMSD-based DNA devices. 223

To demonstrate the adaptability of our system, we applied it as a logic gate to process two 225 ssDNA inputs. The first input strand (S1) was the same SARS2 oligonucleotide in Fig. 3 and the second 226 input strand (S2) was identical except it bore a different 24-nt SARS2 genomic sequence. Using the 227 approach described above, we then designed computational DNA probes to serve as logic gates, 228 turning on via TMSD if: (i) either S1 or S2 are present (S1 OR S2 condition); (ii) both S1 and S2 are 229 present (S1 AND S2 condition); (iii) S1 is present and S2 is not present (S1 NOT S2 condition). Input 230 and probe oligonucleotides are shown in Fig. 4 . 231

The S1 OR S2 logic gate (Fig. 4A ) was composed of two sets of hairpin pairs, one recognizing 232 S1 and the other recognizing S2, each independently capable of opening and generating a separate 233 activating complex. In agreement, we observed a large color shift when either of the two input strands 234 were added to the biosensor-hairpin mix ( Fig. 4A; Supporting Fig. S6A ). The S1 AND S2 gate (Fig. 4B ) 235 contained a single probe that recognizes S1 and another that recognizes S2, such that both SARS2 236 sequences must be present to generate the activating complex. As expected, the biosensor was not 237 activated by either S1 or S2 alone, and addition of both inputs gave rise to a large color shift (Fig. 4B , 238

Supporting Fig. S6B ). The S1 NOT S2 logic gate (Fig. 4C) consisted of a pair of hairpins that both 239 recognize S1 and generate the activating complex upon binding. S2 was designed to base pair with S1 240 and thus inhibit S1 from initiating TMSD. S2 alone did not activate the biosensor-hairpin mix, and S1 241 alone gave rise to a large spectral shift, both as predicted (Fig. 4C, Supporting Fig. S6C ). When we 242 added S1 and S2, the blue:green ratio diminished to the same value that was observed without any 243 inputs, satisfying the S1 NOT S2 condition. 244 245 strands, S1 and S2, were composed of one of two SARS2 genomic sequences (b/b' or y/y') and a toehold (a/a' or 248

x/x'). For the NOT condition, the S1 toehold was extended by 6-nt (t') and S2 was an exact complement of S1.

The additional 6-nt helps clamp S1 and S2 together, reducing fraying of the toehold in the S1/S2 complex that Our nLuc-AFF system addresses the above limitation by acting as a 'universal' adapter that 272 connects aptamer/analyte binding to bioluminescence. It is universal in the respect that it is designed to 273 work with existing aptamers without any changes to nLuc-AFF or chemical derivatization/surface 274 attachment of the aptamer. The only modification is addition of an oligonucleotide sequence to either 275 the 5' or 3' end of the aptamer. To test that assertion, we sought to create a luminescent biosensor for 276 serotonin using an aptamer that was developed elsewhere 29 . We modified the serotonin aptamer 277 (green and red in Fig. 5A ) by adding to its 5' end a sequence consisting of ssAP-1 (blue) and an 278 additional 'clamp' sequence (pink) that is complementary to first 8 nt of the aptamer (red). In the 279 absence of ligand, the aptamer structure is unstable, and the red sequence base-pairs to the clamp, 280 which, along with the blue sequence, forms the stem of a stable hairpin that prevents ssAP-1 from 281 activating the biosensor (Fig. 5A, left structure) . When the aptamer binds serotonin, it folds and 282 reclaims the red sequence, opening the hairpin and exposing the ssAP-1 sequence in the loop (Fig. 5A,  283 middle structure). This is essentially the DNA analog of the AFF mechanism. A second ssDNA 284 oligonucleotide (naked AP-1) consisting of the sequence complementary to ssAP-1 and a truncated 285 clamp (4-nt) is added to generate duplex AP-1 and activate nLuc-AFF (Fig. 5A, right structure) . 286 NUPACK was used to determine the optimal clamp lengths in the modified aptamer (to form a stable 287 hairpin in the absence of ligand) and in naked AP-1 (to minimize false activation of the aptamer) 288 (Supporting Fig. S7A) . 289 Fig. 7B ). The slower species is likely a dimer, which is expected due to its 308 largely palindromic AP-1 sequence along with the hairpin that protects it. We did not observe any 309 additional products when naked AP-1 was added to the aptamer in the absence of serotonin. With 310 serotonin, a new band was observed that likely corresponds to the activating complex depicted in Fig.  311 5A. 312

Having determined that a 1:1 ratio of aptamer:naked AP-1 yielded optimal results (Supporting 313 Fig. S8 ), we evaluated sensor performance by adding 10 μ M serotonin to samples containing 250 nM 314 aptamer, 250 nM naked AP-1, and 30 nM nLuc-AFF. We observed a 2.7-fold ratiometric change in 315 L450/L520 (Fig. 5B, Fig. 5C ) and no change when the sensor was mixed with serotonin in the absence 316 of aptamer (Supporting Fig. S8D ). The color change was visible by cell phone camera (Fig. 5B inset) , 317

with an average of a 3-fold biosensor activation (Fig. 5C) . 318

We next asked if our strategy was generalizable to other aptamers. We modified a well-319 characterized ATP aptamer 30 to contain the ssAP-1 and clamp sequences using the approach 320 described above, except we appended those nucleotides to the 3'-end of the ATP aptamer. As with the 321 serotonin aptamer, we detected a spectral shift upon addition of 2 mM ATP, with a 3.1-fold increase in 322 L450/L520 and a 3.6-fold change from the cell phone image (Fig. 5D, Fig. 5E ). These results suggest 323 that our design can be readily applied to existing aptamers to generate new luminescent sensors for a 324 variety of small molecules and proteins. Moreover, SRAs can also be used to diagnose vaccine induced thrombocytopenia 34 . 335

Discussion 336 protein-based biosensor that is activated by different DNA inputs. The same nLuc-AFF protein can be 338 paired with different DNA technologies for use in several applications. TMSD converts an arbitrary 339 nucleotide sequence into the AP-1 input for the sensor, for the purpose of detecting a DNA/RNA 340 sequence of choice or creating DNA-based computational devices with luminescent readout. Aptamers 341 extend the potential targets of the nLuc-AFF biosensor to small molecules, metabolites, and proteins. 342

We note that nLuc-AFF retains some green emission even when fully saturated with AP-1. This 343 appears due to the presence of truncated, always-green proteins that are generated by proteolytic 344 cleavage within the 6* -50* sequence, which is unpaired in the native fold and thus expected to be 345 relatively unstructured. These fragments likely co-purify with the full-length sensor by means of their 346 dimerizing GCN4 domains. Efforts to remove these truncated products (which appear as green 347 fluorescent bands by SDS-PAGE) were unsuccessful (Supporting Fig. S11) . 348

There are two main drawbacks to our current design. First is the slow turn-on rate (t 1/2 = 8.25 h ± 349 0.37 h, Supplemental Fig. 10A ). Based on our experience with other AFF-modified proteins, slow 350 switching is due to a large kinetic barrier to unfolding of the native frame. Introducing mutations to 351 destabilize the native fold can accelerate the overall switching rate. Care must be taken to preserve the 352 thermodynamic balance between native and CP folds, and this can be accomplished by introducing the 353 mutation in the region shared by the two folds (residues 51-171; Fig. 1B ) or placing identical mutations 354 in each of the duplicated segments. Destabilizing mutations can often be predicted by structural 355 inspection or computational methods, but these do not always accelerate unfolding rates. A more 356 rational approach is to simulate the AFF conformational change using weighted ensemble methods and 357 identify amino acids that, when mutated, destabilize the native and CP folds but not the transition state 358 ensemble 35 . However, even though the switch is slow to reach completion, robust signal change could 359 be seen within 1 h (Supporting. fig. 10B ). The cell phone images also showed a slight change in color in 360 this time frame (Supporting. fig. 10C, D) , suggesting that it may be possible to use nLuc-AFF for rapid 361 diagnosis in low resource settings. 362

The second limitation of our system arises from the palindromic nature of the AP-1 sequence 363 that activates the sensor. Our methods incorporate the activating sequence into various DNA 364 structures, where it is cryptic until the triggering event (TMSD or ligand binding to aptamer) exposes it 365 for hybridizing with its complement. In the case of TMSD, the DNA structures are hairpins, and 366

introducing the palindromic ssAP-1 sequence into the loop can affect their metastability. Moreover, an 367 unprotected ssAP-1 sequence (such as that in the complementary activating strand in our aptamer 368 experiments) can homodimerize, giving rise to nonspecific activation. What is needed to resolve this problem is a DNA binding protein that recognizes a non-palindromic sequence, has a reasonably long 370 N-to-C distance (≥20 Å), and can be engineered to be unstable in the absence of DNA. Zinc-finger DNA 371 binding domains may be able to serve this purpose 36, 37 . It may also be possible to use SELEX methods 372 to identify a ssDNA aptamer that binds specifically to an altogether different input domain, i.e., a small 373 protein that meets the distance and stability requirements mentioned above. This aptamer sequence 374 would take the place of AP-1 and eliminate the need for the complementary activating strand. 375

The nLuc-AFF switch developed here has several notable attributes. It retains the DNA binding 376 affinity and luminescent properties of the parent GCN4 and nLuc proteins, respectively, and it is not 377 falsely activated by decoy DNA sequences. Its ratiometric output enables quantification by cell phone, 378 by simply recording an RGB image and dividing the raw intensity of the blue channel by that of the 379 green channel. The dual-color output establishes an internal normalization of intensity that makes it 380 possible to quantify cell phone images. Finally, nLuc-AFF stands out from other luciferase-based 381 platforms in that it performs well in serum without needing a secondary luminescent protein to calibrate 382 for signal loss, opening the door to its potential use in clinical samples. 383

Nature has provided enzymes-the Cas family of nucleases-that become activated by specific 385 DNA or RNA sequences. It has not yet been feasible to introduce this mode of regulation into other 386 proteins or enzymes. We have introduced a mechanism by which nanoluciferase was converted to a 387 ratiometric biosensor that changes colors in response to a specific DNA sequence. The underlying AFF 388 approach has been successfully applied to numerous other proteins 10, 15, 38, 39 , suggesting that it can be 389 used to create enzymes whose functions are switched on by specific DNA or RNA inputs. 390

All expression constructs here were cloned into pET25 vector, which contains a C-terminal 6x-393

His tag. GeNL/pcDNA3 (gift from Takeharu Nagai, Addgene plasmid # 85200; 394 http://n2t.net/addgene:85200; RRID:Addgene_85200) was used to PCR out the GeNL gene which was 395 inserted into the expression vector at the NdeI/XhoI sites along with a C-terminal 6xHis tag. The was aged for at least 10 min at room temperature prior to recording luminescence spectra. Biosensor 418 performance was quantified by the ratio of luminescence intensity at 460 nm and 520 nm (L460/L520). 419

Quantification of samples by luminescence spectra and cell phone camera was performed on the same 420 samples in the same plate (Corning Costar 96-well assay plates, white polystyrene, round bottom). 421

Cell phone images were taken using a OnePlus 7 Pro with the following settings: ISO 200, f/1.6 422 and auto WB. For quantification, images were split into the blue and green channels using ImageJ, and 423 intensity in each channel was quantified by manually drawing a circle inside each well and calculating 424 the average intensity. The average intensity in the blue channel was then divided by the average 425 intensity in the green channel. 426

The performance of nLuc-AFF biosensor was characterized by the turn-on as well as the 428 binding affinity of GCN4 to its consensus AP-1 sequence. We purified the nLuc-AFF three independent 429 times (biological repeats) and performed three technical repeats of each biological repeat. To integrating the spectra from 400 -600 nm. 446

Oligonucleotides were designed using NUPACK to ensure metastability as well as high 448 efficiency of TMSD. The two sequences of the nCoV genome used in this study were chosen from the 449 CDC RT-PCR Diagnostic Panel: sequence N1 (cagattcaactggcagtaaccaga) and sequence N2 450 (tcagcgttcttcggaatgtcgcgc). The serotonin and ATP aptamers were also designed using NUPACK. 451

Synthetic oligonucleotides were ordered from Eurofins Genomics without purification and were purified 452 in-house using urea-PAGE (see Supporting Fig. S13 for methodology and oligonucleotide sequences). 453

Hairpins were snap-cooled separately in 20 mM Tris (pH 8.5), 150 mM NaCl at a concentration 455 of 3uM, by heating to 95 °C for 3 min and rapidly cooling to room temperature using a thermal cycler. 456

For validation by PAGE, 500 nM of probes were mixed with 500 nM of triggering oligonucleotide in 20 457 mM Tris (pH 8.5), 150 mM NaCl and incubated at room temperature for 3 h. Samples were then loaded 458 on 12% acrylamide:bisacrylamide (19:1) gel in 1x Tris-borate-EDTA (TBE) buffer and run in TBE for 40 459 min at room temperature. Gels were then stained with ethidium bromide and imaged. For logic gate 460 experiments, hairpins and nLucAFF were mixed to a final concentration of 300 nM and 30 nM, 461 respectively, in 20mM Tris (pH 8.5), 150 mM NaCl, 0.1 mg/ml BSA. Triggering oligonucleotides were 462 room temperature before luminescence was recorded. 464

Serotonin and ATP aptamers were purified and snap-cooled before use as described in 466

Supporting Figure S13 . For PAGE assays, 250 nM of aptamer was mixed with 500 nM of naked AP-1 in 467 20 mM Tris (pH 8.5), 0.15 mM NaCl, whereupon either 10 μ M 5serotonin, 2 mM ATP, or DMSO vehicle 468 (final concentration 0.05%) was added, and the reaction was allowed to proceed for 2 h at room 469 temperature. Samples were then loaded on a 12% polyacrylamide gel and run in TBE buffer. Gels were 470 stained with ethidium bromide and imaged. For luminescence assays, 250 nM aptamer, and 250 nM 471 naked AP-1, and 30 nM nLuc-AFF were mixed, and either 10 μ

Self-Assembly at All Scales

Programmable self-assembly of three-dimensional nanostructures from 10,000 unique components

Programming biomolecular self-assembly pathways

Diverse and robust molecular algorithms using reprogrammable DNA self-assembly

Immuno-SABER enables highly multiplexed and amplified protein imaging in tissues

Next-generation in situ hybridization chain reaction: higher gain, lower cost, greater durability

Third-generation in situ hybridization chain reaction: multiplexed, quantitative, sensitive, versatile, robust

Complex cellular logic computation using ribocomputing devices

Ribocomputing: Cellular Logic Computation Using RNA Devices

A Ca2+-sensing molecular switch based on alternate frame protein folding

Converting a Binding Protein into a Biosensing Conformational Switch Using Protein Fragment Exchange

Small Molecule-Induced Domain Swapping as a Mechanism for Controlling Protein Function and Assembly

De novo design of bioactive protein switches

Optogenetic control of cofilin and α TAT in living cells using Z-lock

Engineering an artificial zymogen by alternate frame protein folding

Engineered Luciferase Reporter from a Deep Sea Shrimp Utilizing a Novel Imidazopyrazinone Substrate

NanoLuc: A Small Luciferase is Brightening up the Field of Bioluminescence

Five colour variants of bright luminescent protein for real-time multicolour bioimaging

X-ray structure of the GCN4 leucine zipper, a two-stranded, parallel coiled coil

Modular Enzyme Design: Regulation by Mutually Exclusive Protein Folding

Multiplex detection of proteinprotein interactions using a next generation luciferase reporter

Structural and Thermodynamic Analysis of a Conformationally Strained Circular Permutant of Barnase

GCN4 Binds with High Affinity to DNA Sequences Containing a Single Consensus Half-Site

A plug-and-play platform of ratiometric bioluminescent sensors for homogeneous immunoassays

A DNA-fuelled molecular machine made of DNA

On the biophysics and kinetics of toehold-mediated DNA strand displacement

NUPACK: Analysis and design of nucleic acid systems

Aptamer-Based Fluorescent Biosensors

Aptamer-field-effect transistors overcome Debye length limitations for smallmolecule sensing

A DNA Aptamer That Binds Adenosine and ATP

Characterization of hospitalized cardiovascular patients with suspected heparin-induced thrombocytopenia

Heparin Induced Thrombocytopenia. in StatPearls

Determination of serotonin release from platelets by enzyme immunoassay in the diagnosis of heparin-induced thrombocytopenia

PF4 Immunoassays in Vaccine-Induced Thrombotic Thrombocytopenia

Large enhancement of response times of a protein conformational switch by computational design

Improved DNA binding specificity from polyzinc finger peptides by using strings of two-finger units

In vitro selection of zinc fingers with altered DNA-binding specificity

Stepwise Conversion of a Binding Protein to a Fluorescent Switch: Application to Thermoanaerobacter tengcongensis Ribose Binding Protein

GFP Variants with Alternative β -Strands and Their Application as Light-driven Protease Sensors: A Tale of Two Tails

We thank Jeung-Hoi Ha for discussions. This work was supported by NIH grant R01GM115762 to 496 S.N.L. 497

The authors declare no conflict of interest. 499