Programmable Computational RNA Droplets Assembled via Kissing-Loop Interaction

DNA droplets, artificial liquid-like condensates of well-engineered DNA sequences, allow the critical aspects of phase-separated biological condensates to be harnessed programmably, such as molecular sensing and phase-state regulation. In contrast, their RNA-based counterparts remain less explored despite more diverse molecular structures and functions ranging from DNA-like to protein-like features. Here, we design and demonstrate computational RNA droplets capable of two-input AND logic operations. We use a multibranched RNA nanostructure as a building block comprising multiple single-stranded RNAs. Its branches engaged in RNA-specific kissing-loop (KL) interaction enables the self-assembly into a network-like microstructure. Upon two inputs of target miRNAs, the nanostructure is programmed to break up into lower-valency structures that are interconnected in a chain-like manner. We optimize KL sequences adapted from viral sequences by numerically and experimentally studying the base-wise adjustability of the interaction strength. Only upon receiving cognate microRNAs, RNA droplets selectively show a drastic phase-state change from liquid to dispersed states due to dismantling of the network-like microstructure. This demonstration strongly suggests that the multistranded motif design offers a flexible means to bottom-up programming of condensate phase behavior. Unlike submicroscopic RNA-based logic operators, the macroscopic phase change provides a naked-eye-distinguishable readout of molecular sensing. Our computational RNA droplets can be applied to in situ programmable assembly of computational biomolecular devices and artificial cells from transcriptionally derived RNA within biological/artificial cells.


Sequences
Below are tables of the DNA/RNA sequences used in this study.Sticky ends (SEs) and kissing loops (KLs) are marked by bold fonts.In each motif, hybridized domains are colored in the same color, while the spacers ('TT' and 'UU') are in gray.The DNA sequences were purchased from Eurofins Genomics (Tokyo, Japan).
The RNA sequences and miRNAs were from Sangon Biotech (Shanghai, China) and Hokkaido System Science Co. Ltd. (Sapporo, Japan).Upon arrival, they were dissolved into nuclease-free water (ultrapure DNase/RNasefree distilled water, Thermo Fisher Scientific, MA, US) at a concentration of 100 µM and stored in a -80°C freezer until use.

X-Motif with SEs
Table S1 X-motif DNA with SEs (dSE-X) 1

Name
Sequence (5'-3') This motif was used to numerically predict the melting temperature of the hybridized SEs ( ) and   experimentally obtain the dissolving temperature of the resulting condensates ( in Figure 2.   )

X-Motif with KLs
In addition to KL_Ori-X (Table S3), we designed two X-motif RNAs with weaker KL interaction by adapting the self-complementary subsequence in the KL.We replaced two GC base-pairs (bps) with two AUs for KL_Mut1-X and with two GUs (wobble bps) for KL_Mut2-X.In the KL interaction, two KLs form a helix via the Watson-Crick base-pairing in the self-complementary segments (underlined in the Tables), and then establish the tertiary structure via the coaxial stacking between the two loops.
The omitted segments (…) are equal to the non-modified counterparts of KL_Ori-X.
This motif was used to obtain (KL interaction) and (condensates) in Figure 2.
The omitted segments (…) are equal to the non-modified counterparts of KL_Ori-X.
This motif was used to obtain (KL interaction) and (condensates) in Figure 2.

Branched RNA Motif Capable of Sensing Specific MiRNAs
In the sensing RNA motif (Table S6), STH1,2 include a short single-stranded overhang as a toehold (underlined by dashed lines), which prefers to bind with the complementary region of the target miRNAs (m1 and m2, respectively.See Table S7) and initiate the strand displacement reaction.Only in the presence of m1 and m2, the 4-valence sensing RNA motif splits up into two 2-valence motifs.Otherwise, the valence is conserved, although the structural flexibility increases in the input of only m1 or m2.The displaced strand becomes a single-stranded extension protruding from the neighboring branch.The KL regions were adapted from those of KL_Mut2 (Table S5).
The set of miRNAs (Table S7) is the tumor marker for detecting early-stage breast cancer. 2 the fluorescence recovery after photobleaching (FRAP) experiments described below, we labeled the sensing RNA motif by replacing 0.5 µM of KL_Mut2-Xs2 (10% of the final concentration) with 0.5 µM FAMlabeled ssDNA (Table S8) instead of adding the staining reagents.
Table S6 Sensing RNA motif with KLs (KL_Mut2-Xs) This motif was used to construct the computational RNA droplets investigated in Figures 3, 4, and 5.These RNA sequences were used as inputs for performing the logic operations in Figure 5.This fluorophore-labeled sequence was utilized in the FRAP experiments (Figure 4d).

Design and Specificity Verification of Sequences
Among the DNA/RNA sequences mentioned above, dSE-X (Table S1) is one of the DNA motifs featured and characterized by our previous paper. 1rSE-X (Table S2) is the RNA analog of this DNA motif.The KLended RNA motifs (Table S3-Table S5) were adapted from rSE-X by replacing the SEs with the KLs including the varied palindromic subsequences.To provide KL_Mut2-Xs (Table S6) with the recognition capability, we added the two extensions including the toehold regions to KL_Mut2-X (Table S5), which were designed to initiate strand displacements with the two target miRNAs, respectively.
To verify the specificity in the structure and input-induced rearrangement of KL_Mut2-Xs, we used NUPACK, a well-known software suite for the design and analysis of nucleic-acid constructs. 3Herein, since the other RNA motifs share the stem regions with dSE-X well-studied previously with NUPACK, 1 the following description is focused on the generated analysis data for KL_Mut2-Xs.

Simulated Structure of Six SsRNAs as Components of KL_Mut2-Xs
To simulate the resulting structure from the multiple RNA sequences, we employed a NUPACK 4.0 Python package (Python version: 3.10.12).The model was specified as 'material = RNA', 'celsius = 37', and 'sodium Supporting Information 6 = 0.35'; the complex consisted of the six ssRNAs (Table S6) at a concentration of 5.0 µM in a tube with the max complex size of 6.The tube analysis predicted the designed structure at approximately 5.0 µM (Figure S1a).To visualize the resulting structure of KL_Mut2-Xs, the six ssRNAs were grouped and concatenated into two strands for ease of calculation.The two sequences were entered in the web-based NUPACK utilities, generating a 6-branched folded structure similar to the designed 6-branch KL_Mut2-Xs (Figure S1b).These simulated results satisfied us that the six ssRNAs would constitute the designed 6-branched KL_Mut2-Xs motif with high fidelity.

Input-Induced Rearrangement of KL_Mut2-Xs
Next, we conducted numerical predictions of the possible rearrangement of KL_Mut2-Xs in the presence of the target miRNAs (m1 and m2) to examine how the resulting strand displacement reactions could affect the motif structure.KL_Mut2-Xs and related miRNAs were entered into the NUPACK 4.0 (Python package) at 5.0 µM and 5.1 µM, respectively, with the same model as the construction verification.It was suggested that upon the addition of both m1 and m2, KL_Mut2-Xs favored a distinct separation into two substructures, each comprising two consecutive KL domains binding with the related miRNAs (Figure S2a).This means that the motif reduces the valency from to upon the addition of both m1 and m2.In the presence of m1  = 4  = 2 only (Figure S2b) or m2 only (Figure S2c), the most likely structure apparently experienced only one strand displacement with no structural separation, and thus conserved the initial valency of .These numerical  = 4 results suggested that KL_Mut2-Xs would show the strand displacements sequence-specifically and hence the AND-gate operations.

Construction of Condensates
Throughout our experiments, we constructed condensates from the sequences listed above (Table S1-Table S6).An equimolar mixture of single-stranded sequences (ssDNAs or ssRNAs) was dissolved into an aqueous buffer within a PCR tube, with the final concentration of 5.0 µM for each ssDNA/RNA, 350 mM for NaCl (> 99.5% purity, FUJIFILM Wako Pure Chemical Corp.), and 20 mM for Tris-HCl pH8.0 (UltraPure, Thermo Fisher Scientific).For confocal microscopy observation, the following fluorescent dyes were also added: 1000× Quant-iT TM OliGreen (ssDNA Reagent, Thermo Fisher Scientific) for studying the SE/KL interaction strength (Figure 2) and 10000× SYBR TM Gold (Nucleic Acid Gel Stain, Thermo Fisher Scientific) for demonstrating the AND logic operation (Figure 5).The mixed solutions were incubated at a fixed temperature of 25°C in a plate thermal cycler (Mastercycler® nexus flat, Eppendorf, Hamburg, Germany) for >2 h to allow for the selfassembly of nanostructures into the condensates of DNA or RNA.
A sample solution was applied in a silicon sheet cavity as an observation chamber, where a punch-holed silicon rubber sheet of 1.0 mm in thickness (As One, Osaka, Japan) was affixed onto a 3.0 mm × 4.0 mm glass plate with a thickness of 0.13-0.17mm (No.1, Matsunami Glass Ind., Ltd., Kishiwada, Japan).The glass plate was treated with a BSA (bovine serum albumin, FUJIFILM Wako Pure Chemical Corp.) solution of 5 w/v% BSA and 20 mM Tris-HCl in the experiments using the sensing RNA motif (Figure 4, 5).Finally, the sample solution was covered with mineral oil (Nacalai Tesque, Inc., Kyoto, Japan) to minimize unwanted evaporation.

Sample Heating Procedures
In the thermostability (Figure 2b) and melting (Figure 4a) experiments, a sample confined in an observation chamber was placed on the thermoplate at room temperature (RT) and underwent the following heating procedure.For the thermostability experiments, the applied temperature was increased with a ramp of 5°C in the lower temperature range and a much smaller step in the upper temperature range close to the condensates .After reaching each temperature studied, the sample was imaged twice at elapsed times of 1 h and 2 h with   the confocal microscopy.If the phase of condensates remained, the sample was further heated up to the next

AND Logic Operations
RNA condensates were assembled from KL_Mut2-Xs (Table S6) in a PCR tube at a fixed temperature of 25°C in the plate thermal cycler.Afterward, aqueous buffers containing miRNAs (Table S7) as an input were added, with the final concentration of [KL_Mut2-Xs] = 5.0 µM and [miRNA] = 5.0 µM.Following a two-hour incubation at 25°C, the sensing RNA condensates were visualized with FV1000 on the Peltier heating stage fixed at 30°C. Figure 5d gives plots of the occupancy of bright pixels in a confocal microscopy image to quantify the condensation level of computing RNA condensates at the initial, intermediate, and equilibrium states.Pixels were judged to be 'bright' when their intensity values were higher than a specific threshold value of 8 in the 16bit grayscale.The occupancy was calculated by dividing the number of the bright pixels with the entire area of the obtained image.
In the main manuscript, we stated that inputs of both m1 and m2 created a drastic phase-state change in the RNA condensates from liquid to dispersed states, whereas the input of either m1 or m2 led to no significant phase change.To support this finding, we ascertained that the resulting phase states would not return to the initial state after an extended time period.In Figure S9, we observed the samples used in the input experiments after an overnight incubation at RT.In the inputs of m1 and m2, the dispersed state was conserved well; in the inputs of either m1 or m2, the dense phases were also preserved well, with a slight recovery in the fluorescence.This is firm evidence that the motif restructuring and resulting macroscopic conformations at 30°C suggested in Figure 5 was not a transient but a thermally irreversible process.

Figure S1
Figure S1 Simulated structure of 5.0 µM six ssRNAs constitute KL_Mut2-Xs (Table S6).(a) Screenshot of possible structures generated by NUPACK 4.0 (Python package).The resulting structures are listed in the order of equilibrium concentration.The top structure corresponds to the designed KL_Mut2-Xs.37°C, 350 mM NaCl.(b) Folded structure of the six ssRNAs concatenated into two strands predicted by the web-based NUPACK utilities.37°C, 1 M NaCl.

Figure S6
Figure S6 Native Page analysis of KL_Mut2_X.The time periods mean an incubation time length.

Figure S7
Figure S7 Native Page analysis of KL_Mut2_Xs.The time periods mean an incubation time length.

Figure S8
Figure S8 FRAP procedures.(a) Representative confocal microscopy images of photobleached RNA condensates (in a liquid-like state).Scale bars: 50 µm.(b) Description of the data analysis of captured fluorescence recovery.