Molecular recruitment and release using DNA host condensates

Abstract

Artificial biomolecular condensates are a promising tool to achieve spatial self-organization in active materials and synthetic cells. DNA branched motifs known as nanostars are a particularly versatile tool to build artificial condensates with customizable phase diagrams and viscoelastic properties. Here, we characterize how the inclusion of aptamers in DNA nanostars makes it possible to engineer condensates with the capacity of recruiting biomolecules of interest to the dense phase. Furthermore, we demonstrate how biomolecules can be released from condensates by supplying “kleptamer”, an oligonucleotide that is complementary to the aptamer sequence. We focus specifically on incorporating a DNA aptamer that recruits streptavidin (SA), and investigate how the positioning of the aptamer—whether at the nanostar junction, or at the middle or tip of one of the nanostar arms—affects the formation of condensates and the recruitment efficiency of SA. We find that the aptamer's location within the nanostar does not significantly influence condensation and recruitment, nor the capacity of kleptamers to release the protein. These results provide new insight into the design of synthetic DNA condensates for the uptake and release of target molecules and demonstrate their robustness with respect to nanostar design, potentially broadening the use of such condensates in synthetic biology applications.

---

New concepts

The ability to build synthetic condensates with recruitment capacities comparable to natural biomolecular assemblies represents an exciting frontier in engineering biology. Here, we use DNA branched motifs (“nanostars”) as synthetic host condensates and systematically investigate how the spatial placement of functional domains controls ligand recruitment. We incorporate an aptamer domain binding to streptavidin (SA) at distinct positions within the nanostar motif: at the junction of the nanostar, in the middle of the nanostar arm, or at the tip of the arm. We find that aptamer placement has a modest influence on the speed of phase separation and on the client recruitment efficiency. We extend our framework by considering recruitment of additional clients, the P22 N peptide and biotinylated polystyrene beads. Together, these results establish DNA nanostars as a robust and modular platform for constructing synthetic condensates and contribute structural constraints for engineering programmable molecular recruitment.

Introduction

Biomolecular condensates are compartments without a membrane that form through phase separation.¹ Due to their dynamic nature and ability to recruit molecular guests, condensates are promising tools for biomanufacturing, diagnostics, and the development of responsive materials.² It is possible to build artificial biomolecular condensates out of several components, including proteins, nucleic acids, and lipids, when participating molecules or subunits present sufficiently weak, reconfigurable bonds.³ A successful approach is offered by DNA nanotechnology, which makes it feasible to build synthetic DNA and RNA condensates by programming base-pairing interactions.^4–7 While unstructured DNA strands can lead to spontaneous condensation in some cases,^8,9 the use of structured DNA motifs has proved to be remarkably more controllable and modular.¹⁰ One such example is DNA branched motifs, termed DNA nanostars, that are composed of a few strands assembled into star-shaped motifs and interact through sticky end domains at the tip of their arms.^3,5,6,11,12 Under appropriate temperature and ionic conditions,¹³ nanostars spontaneously phase separate into condensates and form DNA-rich compartments. The viscoelastic properties of the condensates and their kinetics can be tuned through systematic design of the nanostar features, such as the length and sequence of the sticky ends, as well as the number and length of the arms.^6,12,14 This creates new possibilities for controlling the spatial organization of biomolecules in vitro, which could be harnessed for purification, separation, and enhancement of catalysis.¹¹

DNA nanostars can include a variety of domains that allow for recruitment of target or guest molecules without compromising condensation, thus generating “host” condensates. Other oligonucleotides can be easily recruited by simply incorporating complementary domains.¹⁵ For instance, nanostars can be designed to have an overhang that binds to long double-stranded DNA, which changes surface tension and dynamics of the condensates, operating as “DNA surfactants.”¹⁶ With the same strategy, nanostars can recruit biotinylated DNA oligos, which then bind to streptavidin (SA) and subsequently partition it into the condensates.⁴ In order to recruit peptides, proteins, or small molecules, nanostars are often modified to include DNA or RNA aptamers, which are short nucleic acid strands, typically 20- to 60-nucleotide (nt) long, that bind to a target ligand with high specificity and affinity. Aptamers are optimized through in vitro evolution methods like SELEX^17,18 and are utilized in a variety of binding assays, as well as biosensors and diagnostics, such as the electrochemical detection of SARS-CoV-2¹⁹ and sensing toxins and pathogens.^20,21 In recent studies, our group and others have demonstrated that the inclusion of aptamer domains makes it possible to recruit peptides of interest to condensates formed by RNA nanostars.⁷ Moreover, we developed hybrid host nanostars composed of three DNA strands and one RNA strand with an SA aptamer, to recruit SA into the hybrid condensates on-demand through physical and chemical stimuli.²² SA was selected, as it is one of the most commonly used proteins in binding assays, and DNA and RNA aptamers for SA have been extensively characterized.^23,24 While recruitment of guest molecules through aptamers has been demonstrated, it is unclear how the aptamer placement on the nanostars affects their condensation and protein recruitment, and whether the guest molecules can be easily released after sequestration.

Building on the foregoing results, here we use DNA aptamers to form host DNA condensates and explore how positioning of an aptamer on the body of a nanostar affects condensation and recruitment of guests, focusing on SA as a test case. We design three nanostar variants (Fig. 1), each with four arms (valency four) that have the aptamer placed: (1) at the proximal end of two arms (junction design), (2) in the middle of an arm (middle design), and (3) at the distal end of an arm (tip design). We then demonstrate that different aptamer locations yield small, yet observable differences in SA recruitment efficiency, as well as condensate properties such as growth rate, molecular mobility, and phase transition temperatures. We find that middle and junction designs have similar SA recruitment levels, while the tip design is the least effective one, likely due to steric hindrance near the adjacent sticky end. We subsequently use “kleptamers,” oligonucleotides complementary to the aptamer domain and derived from the ancient Greek word kléptes (thief), to competitively displace the aptamer from its target²⁵ and trigger SA release from the condensates, and we evaluate their performance across different nanostar variants. We find that the addition of kleptamers at proper concentrations promotes the complete release of SA from the condensates in all cases. Finally, we demonstrate the generalizability of DNA condensates for the recruitment of diverse targets by using DNA aptamers to recruit and release biotinylated beads and RNA aptamers for the uptake of the P22 bacteriophage N peptide.²⁶ Our results show that DNA nanostars are a robust platform to build host condensates for molecular separation, partitioning, and release, and set the foundation for our results to be used in other classes of host molecules and their targets for spatiotemporal localization of biomolecules and chemical reactions.

Fig. 1 Protein recruitment via nanostars containing DNA aptamers at different locations. (A) Overview of our experimental protocol to make host condensates. Three of the four single-stranded DNA (ssDNA) are annealed from 90 °C to either 37 °C or 25 °C, depending on the experiment. Once the sample reaches the desired temperature, S4 including the aptamer is added isothermally together with SA. (B) Schematic of 4-arm DNA nanostars with various aptamer placements: junction, middle, and tip. (C) Colocalization of SA and condensates is quantified through confocal microscopy images. In this example image of the junction design, 60 minutes after the addition of S4 and SA, the Pearson's correlation coefficient, R = 0.97, indicates high colocalization.

Results and discussion

Designing DNA nanostars with aptamers for protein uptake

We designed 4-arm DNA nanostars that interact with one another and phase separate via palindromic sticky ends at the distal end of each arm, and that contain a DNA aptamer domain which can specifically bind and recruit streptavidin (SA) (Fig. 1A). Each nanostar is made of four DNA strands, each partially complementary to two other strands, that hybridize and form the four arms. Each arm is composed of a 16-nucleotide (nt) long stem and a 4-nt long sticky end. We introduced three nanostar variants, with different aptamer placements: (1) at the proximal end of two arms (junction design), (2) in the middle of an arm (middle design), and (3) at the distal end of an arm, appended to the 3′-end of the strand (tip design) (Fig. 1B). The nanostars were designed with GCGC sticky ends that are well characterized under standard buffer conditions, based on prior works by us and others.^4,11,27 In order to form condensates, DNA nanostars need to have a valency of three or more;²⁷ thus, we employed nanostars with four arms and identical, complementary sticky ends (valency of four). This ensures that even if aptamer placement interferes with the ability of one of the sticky ends to bind, condensates can still form with the other three arms that are not affected by the aptamer placement. Finally, the arm length of 16-nt in nanostars leads to a mesh that is large enough for SA to diffuse into the condensate and be efficiently recruited into it.^14,28,29

To characterize condensate formation and recruitment efficiency, we prepared nanostars with a two-step process. First, only three of the four nanostar strands (S1, S2, and S3) were annealed from 90 °C (at −1 °C minute⁻¹) to 25 or 37 °C in a buffer including 20 mM Tris-HCl and 12.5 mM MgCl₂ to form the nanostar core, which does not assemble into condensates. The second step involves triggering the formation of condensate by isothermally adding the fourth strand (S4), which has the SA aptamer domain (Fig. 1A), in tandem with the addition of an equimolar amount of SA. This approach makes it possible to control the onset of condensation and does not noticeably alter the condensation dynamics nor SA recruitment (Fig. S1).²⁷ In all our experiments, the concentration of each strand–and subsequently nanostars–was either 1.25 µM or 2.5 µM. To measure condensate formation and SA recruitment through microscopy, we doped the strand mix with 17.5% of a FAM-labeled variant of S1, and SA was labeled with Alexa 647 (see the Methods section). We used Pearson's correlation coefficient (R) of microscopy images to measure colocalization and recruitment. This coefficient quantifies the spatial correlation of the pixel intensities in the FAM (nanostar) and Alexa 647 (protein) channels (Fig. 1C). A value of R close to 0 suggests weak colocalization of SA with the condensates, whereas a value near 1 indicates strong colocalization.

Target recruitment and growth speed of DNA host condensates

We first tested whether all our nanostar variants form condensates with the capacity to recruit SA based on condensate population analysis. Each design results in DNA condensates (FAM) at 37 °C that are colocalized with SA (Alexa 647), as shown in Fig. 2A (additional example images at 25 °C are in Fig. S2). We measured high Pearson's correlation coefficient (R) values consistently over the observation time, as we imaged condensates 10 minutes and 60 minutes after the aptamer-containing strand (S4) and SA were added isothermally (Fig. 2A and S4). At 37 °C, the average condensate size increases throughout the experiment, as confirmed by measuring the condensed areas in epifluorescence microscopy images (Fig. S5). Naturally, condensates are expected to be bigger at higher nanostar concentrations, as shown by the box plots in Fig. 2B, which compare condensate areas between 1.25 and 2.5 µM concentrations.

Fig. 2 Host condensate variants and their growth dynamics. (A) We designed three variants of host nanostars by changing the position of the aptamer on the nanostar. Confocal microscopy images were taken at 10 minutes and 60 minutes after the isothermal addition of S4 and SA. We calculated Pearson's correlation coefficient (R) to determine each design's capability in recruiting and localizing SA. (B) The size of the condensates formed by each nanostar design in the presence of SA was measured at concentrations of 2.5 µM (left) and 1.25 µM (right). (C) Representative epifluorescence microscopy images of the host condensates at 60 minutes, formed by the junction, middle, and tip designs and incubated at 37 °C, with equimolar amounts of nanostars and SA, 2.5 µM.

The speed at which condensates grow depends on many parameters, in particular nanostars valency and arm length, which govern condensate formation, coarsening, and fusion.¹⁴ The inclusion of an aptamer domain bound to its target could influence condensate growth, as it can change the nanostar's secondary structure and induce steric hindrance among the arms. We found that the tip design shows the smallest average area across conditions because SA binding near one of the sticky ends reduces its binding capability, effectively undermining the nanostar valency. During our observation window, the condensate size increases only slightly faster in the absence of SA when compared to the condensate size in the presence of SA, possibly due to the conformational changes of the nanostar when bound to it. This trend is observed in microscopy images (compare Fig. 2 and Fig. S3), in the box plots measuring condensate area (Fig. S6), and by the average areas measured in Fig. S5. A power law model fits well to the average area data in Fig. S5 (parameters are reported in Tables S4 and S5). Looking at the condensate eccentricity, there is minimal variation among designs, and a greater difference between incubation temperatures, 25 °C versus 37 °C (Fig. S7). Condensate sizes generally follow an exponential distribution across designs and temperatures (Fig. S8). Condensate growth speed, area, and eccentricity exhibit slightly different trends at 25 °C. Condensates forming at this temperature are larger at 60 minutes, which could be caused by kinetic trapping (Fig. S5 and S8). Generally, the junction design forms the largest condensates, followed by the middle and tip designs.

Partitioning efficiency and diffusivity of host condensate variants upon recruitment of SA

To assess how aptamer positioning influences SA recruitment, we employed confocal microscopy and examined individual condensates by measuring the partition coefficients (PC) of the nanostar and SA across three designs. In each case, the partition coefficient was calculated as the ratio of mean fluorescence intensity within the condensate (I_in) to that in the surrounding dilute phase (I_out). Twenty condensates in total were measured for each condition. For the junction design, we found nanostars partitioned with PC_J = 40.1 ± 5.15 and SA partitioned with PC_J,SA = 8.84 ± 1.02. For the middle design, this is PC_M = 41.3 ± 11.75 and PC_M,SA = 8.61 ± 1.45, and for the tip design, PC_T = 26.3 ± 5.85 and PC_T,SA = 4.96 ± 0.58 (Fig. 3A). The junction and middle designs have a very similar partition coefficient and recruitment efficacy, while the tip design has significantly lower PC values. This aligns with the plausible interference of one of the sticky ends when the aptamer at the tip design induces steric hindrance and binds to SA, inhibiting its ability to phase separate to an equivalent level as the junction and middle designs. With more nanostars incorporated into the dense phase, there are subsequently more aptamers available to recruit SA into the condensate. This results in greater levels of SA recruitment, therefore the design with the highest partition coefficient of nanostar should also recruit the most SA.

Fig. 3 Host condensates recruitment efficiency and dense phase mobility. Nanostars exhibit different efficiency in SA recruitment depending on the location of the aptamer on the arm. (A) The partition coefficient calculations for different condensates (20 condensates from one experimental replica) reveal that the middle design has the greatest recruitment capacity. A confocal microscope was used for imaging. (B) When looking at the mobility of the condensates, recruitment of SA affects each case differently, with the tip and middle designs having increased fluidity and the junction design decreasing in fluidity after SA recruitment. Images were obtained using an epifluorescence microscope. (C) Normalized FRAP data were plotted as the mean of three experimental replicas (n = 3) with error bars corresponding to the standard deviation of the mean.

To quantify the effect of SA recruitment on the dynamic exchange (fluidity) of the DNA condensates, we measured fluorescence recovery after photobleaching (FRAP) on condensates incubated for 60 minutes (Fig. 3B). Recovery curves were fitted to a model to extract characteristic time constants (τ). The differences in the time constants of the condensates with and without SA were measured for the three aptamer designs (Fig. 3C). For the tip design, τ in the absence of SA was τ_T = 315 s compared to τ_T,SA = 206 s in the presence of SA. Similarly, the middle design exhibited τ_M = 415 s which decreased to τ_M,SA = 290 s in the presence of SA. These results suggest that protein recruitment increases mobility for these two designs. In contrast, the junction design displayed τ_J = 369 s and τ_J,SA = 500 s, implying that SA incorporation decreases fluidity in this architecture.

The measured increase in condensate fluidity upon SA binding to the middle and tip designs is likely due to the fact that SA binding may impact the mobility of the arms and the binding dynamics of the sticky ends, weakening inter-nanostar interactions and promoting the formation of more dynamic, liquid-like condensates. Conversely, when SA binds to the junction design, it is likely sequestered toward the center of the nanostar, minimizing its interaction with the sticky ends. As a result, we observe a decrease in fluidity, suggesting that SA is tightly incorporated without perturbing the condensate integrity or inter-nanostar connectivity. We examined our hypothesis about the arms mobility of each nanostar design by performing coarse-grained molecular dynamics (MD) simulations using oxDNA^30,31 and measuring the angles between each pair of arms, θ_i,j, where i and j represent arm indices (Fig. S9 to S11). On average, the junction placement of the aptamer leads to a smaller angle between the arms, which could result in reduced flexibility of the bound arms of two nanostars, and consequently the mobility of nanostars in the dense phase, when compared to the other two designs. We noticed that the distribution of some angles in the middle and tip designs exhibits a leptokurtic shape, suggesting that the corresponding arms are distant enough from the other arms to engage in sticky end hybridization more freely. Additionally, in the middle design, we observed and quantified the bending of the arm with the aptamer (Fig. S12) by defining the angle between the two parts of the arm segmented by the presence of the aptamer (Ψ). We noticed that the aptamer presence severely impacts the arm's collinearity by bending the distal end of the arm, leading, on average, to an acute Ψ and a leptokurtic angular distribution.

Estimating phase transition temperatures

We used temperature-dependent absorbance measurements to probe the structural transitions associated with nanostar folding and condensate formation. The melting temperature of nucleic acid nanostructures (T_m) is a key parameter that depends on thermodynamic properties,³² and is defined as the point at which half of the DNA bases are paired.⁴ We set up annealing experiments wherein 260 nm absorbance was measured as the sample temperature decreased from 70 °C to 25 °C: we found that annealing presents two distinct regions, respectively marking the sequential transition from single-stranded DNA (ssDNA) molecules to assembled nanostars, and from nanostars to condensates (Fig. 4A). For each of these two regions, we identified a melting temperature at the transition midpoint (where half of the relevant DNA domains should be hybridized).

Fig. 4 Temperature-dependent transitions from melted to condensed nanostars. (A) Annealing curves for the three nanostar designs (concentration of 2.5 μM) show two main transitions. The upper-temperature region marks the formation of DNA nanostars from ssDNA. The lower-temperature region corresponds to the formation of condensates from nanostars. The points with the steepest slope in each region indicate when about 50% of the population has formed either nanostars or condensates. (B) Annealing experiments in emulsion droplets show visible condensates at a temperature that correlates with the “shoulder” of each annealing curve (nanostar concentration of 7.5 μM). Additional example images are in Fig. S14. Annealing curves are obtained from the mean of three experimental replicas (n = 3) at each measurement time, with the shaded areas corresponding to the standard deviation.

Among the three designs, the tip design transitioned from ssDNA to nanostars the fastest, followed by the junction and then the middle designs. This trend likely reflects the aptamer's position on the nanostar arm. The tip design with the aptamer located at the distal end of an arm allows the arm sequences to hybridize and the aptamer to self-fold with minimal interference. Similarly, the junction design, positioned at the proximal end of the arm, leaves the core sequence largely unobstructed. In contrast, the middle design disrupts the middle of the arm, reducing hybridization efficiency and delaying nanostar formation.

At lower temperatures, condensate formation follows a different order: the junction design forms condensates first, followed by the middle, and then the tip design. This is probably due to varying degrees of interference with sticky end interactions. The aptamer in the junction and middle designs is far from the sticky ends, thus has little influence on condensation. In contrast, the tip design causes the most interference, compromising valency due to steric hindrance, and results in the lowest condensate formation temperature. The melting curve was also measured, where we raised the sample temperature from 25 °C to 80 °C (Fig. S13). This curve shows a single transition temperature, indicating that melting and annealing follow different kinetic pathways. In this case, the estimated “reverse” melting temperature of the junction design is the highest, followed by the middle and tip designs (Fig. S13). To validate our finding via microscopy and confirm the melting behavior, we encapsulated nanostars in water-in-oil droplets, which makes it possible to observe the sample over time as temperature changes while evaporation is minimal. Individual droplets were tracked through epifluorescence microscopy as the chamber temperature was progressively increased.³³ These experiments reveal that the “shoulder” of each annealing curve corresponds closely to the formation of condensates in the water-in-oil droplets (Fig. 4B and Fig. S14).

Controlling protein release from host condensates

To control SA release from condensates, we designed a system that uses a ssDNA sequence complementary to the aptamer, dubbed “kleptamer,” to switch the aptamer from an active to an inactive state and release the bound SA (Fig. 5A and B). We designed the kleptamer to be complementary to the aptamer binding pocket and largely to its stem region (Tables S2 and S7). Kleptamer was added at 0.5×, 1×, and 2× molar ratios relative to the SA-aptamer complex, one hour after the isothermal addition of the aptamer-containing strand and SA. Using confocal microscopy, the partition coefficient was measured at 60 minutes post-kleptamer addition (Fig. 5C). The samples were incubated at 37 °C throughout the experiment. At 1× and 2×, where theoretically each aptamer is targeted by at least one kleptamer, we observed consistent and efficient SA displacement. In contrast, the 0.5× condition, where only half of the aptamers are targeted, showed greater variability in partition coefficients (Fig. 5C). At this concentration, microscopy images show diverse patterns of SA localization as it diffuses out of condensates. In some droplets, SA release appeared to proceed with a uniform diffusion from the inside outwards; in other cases, SA release proceeded from one side of the droplet to the other. We even found instances in which SA release appeared to follow distinct patterns as two droplets fuse together (Fig. 5B and Fig. S15). A polyT strand with the same length as the kleptamer strand was used as a control, and no SA displacement was observed (Fig. 5C and Table S2). The partition coefficient values when adding the polyT strand are similar to those without kleptamer addition, shown in Fig. 3A. Differences in kleptamer displacement can be seen when comparing the tip design with the junction and middle designs. When 1× polyT and 0.5× kleptamer are added, the tip design exhibits approximately half the partition coefficient values of the junction and middle designs. While at 1× and 2× kleptamer concentrations, these differences become less apparent, the 1× kleptamer concentration for the tip design is slightly less effective. Among the partition coefficients of the three aptamer designs compared at each condition, control, 0.5×, and 2× show statistically significant differences, indicating that at least one design differs from the others (Fig. S16A). Tukey post-hoc analysis reveals that under the control condition, the tip design behavior is significantly different from the junction and middle designs (Fig. S16A and Table S8). At 0.5×, a significant difference is observed between the junction and middle designs, while the tip design is not distinguishable from either. At 2×, a significant difference emerges between the middle and tip designs. Overall, these results demonstrate the potential use of aptamer-enhanced DNA nanostructures not only for biomolecular recruitment and separation, but also for the release of target biomolecules on-demand.

Fig. 5 Release of protein from host condensates using kleptamers. (A) Schematic of the kleptamer–aptamer binding reaction and SA release. (B) 60 minutes after addition, the kleptamer fully displaces SA at 1× and 2× concentrations. At 0.5× kleptamer, there was incomplete release of SA across all designs 60 minutes after the kleptamer strand addition. (C) The degree of SA incorporation into the condensate is measured by the partition coefficient, with a polyT strand instead of klemptamer as a control (number of analyzed condensates, n = 20). The statistical significance of the different conditions across designs can be found in Fig. S16 and Table S8.

Generalization of recruitment and release across designs

An important advantage of SA is that it can be used to recruit virtually any biotin-labeled particles (enzymes, proteins, antibodies),² so the designs presented here could be used directly to recruit a variety of other molecules. To test this, we demonstrate the recruitment of biotinylated beads as a model target, building on our previous work¹ (Fig. 6A). Each design achieves recruitment, as illustrated in example confocal microscopy images (Fig. 6C); as beads appear to tend to aggregate when incorporated into condensates, we quantified recruitment by counting the density of beads in the dispersed phase (background) (Fig. 6B and Fig. S17). As a control, the DNA aptamer strand was replaced with a DNA strand lacking an aptamer and added isothermally along with SA and beads (Fig. S18), which resulted in no recruitment. Following recruitment, the same kleptamers described in the previous section were added to displace SA from the aptamers and consequently from the beads (Fig. 6C). Condensates were imaged 1 and 3 hours after kleptamer addition, with displacement observed at 3 hours (Fig. S17B). Displacement of beads was observed across all three aptamer designs, with the highest number of beads per 100 µm² observed in the Tip design, followed by the middle and junction designs (Fig. 6B). We note that to achieve bead displacement, we adopted a monovalent salt buffer (NaCl) as discussed in the Methods section.

Fig. 6 Recruitment and release of biotinylated beads. (A) Schematic illustrating the recruitment of a biotinylated bead using the SA-bound DNA aptamer. (B) Bead displacement is quantified by measuring the number of beads in the background for the junction, middle, and tip designs before and 3 hours after kleptamer addition. The experiments were repeated three times (n = 3). Background bead density values were grouped by aptamer design, averaged across replicas, and plotted as mean ± standard deviation for pre- and post-kleptamer conditions. (C) Prior to kleptamer addition, beads are bound to the condensates. Three hours after kleptamer addition, SA displacement is observed, resulting in the release of beads into the dispersed phase. Insets at the bottom left of each image show the SA channel for the regions marked by the orange box.

To further extend our results, we modified the nanostar designs to recruit P22 N peptide^7,26 by replacing the SA aptamer with the boxB RNA aptamer and forming hybrid DNA-RNA host nanostars (Fig. S19). We observed that colocalization occurs in each nanostar variant, under different buffer conditions (MgCl₂ and NaCl buffer). When NaCl was used, we observed increased Pearson's correlation coefficient, as well as more spherical, well-formed condensates (Fig. S19). These experiments indicate that different nanostar designs can be adapted to work with RNA aptamers, which might be necessary if a DNA aptamer targeting the protein of interest is not available.

Conclusions

Our findings provide design guidelines for the engineering of synthetic DNA condensates to recruit and localize guest molecules. This study demonstrates that the positioning of a DNA aptamer in DNA nanostars can lead to the formation of “host” condensates capable of localizing streptavidin (SA), which we used as a model example. Regardless of the aptamer placement (i.e., junction, middle, and tip designs), all variants form condensates that successfully recruit SA, illustrating the robustness of our approach. Nonetheless, aptamer placement was found to have a measurable impact on condensate growth rate, fluidity, and recruitment efficacy. The tip design generally produced smaller condensates, consistent with the hypothesis of sticky-end binding interference. FRAP analysis revealed that SA recruitment influenced condensate fluidity based on aptamer location: with the middle and tip designs, the fluidity increased upon binding of SA, while the opposite was observed for the junction design. Analysis of annealing curves further showed that the aptamer location can affect the transition temperature from ssDNA to nanostars, as well as nanostars to condensates. Among the three configurations tested, the middle and junction designs exhibited the highest SA recruitment efficiency, as shown by partition coefficient measurements. In contrast, the tip design showed the lowest recruitment efficiency. This was likely due to the steric hindrance from the aptamer's proximity to the sticky end, which may impair binding and condensate formation.

We successfully displaced SA from host condensates using a kleptamer strand that is complementary to the SA aptamer. The functionality of recruiting and releasing a molecule is critical for signaling, communication, and information processing in living cells.⁹ Integrating this process into DNA-based condensates, as we have demonstrated here, is a crucial step in creating DNA-empowered synthetic cells. We also envision that DNA condensates may become useful tools for the separation of biomolecules, which could then be released on-demand simply by the addition of kleptamers.

Finally, we demonstrated the generalizability of our design through the recruitment and displacement of biotinylated beads using DNA aptamers and the recruitment of P22 N peptide using RNA aptamers. Overall, our results underscore that aptamers are a versatile and useful tool to build DNA-based host condensates that can spatiotemporally organize specific particles. This work offers valuable insights into the design of programmable host condensates for applications in synthetic biology, molecular compartmentalization, and biomolecular assembly. Our study may be further expanded to investigate the effects of nanostar design parameters, such as sticky end sequence, arm length, and arm number, on the physical properties and thermodynamics of nucleic acid-based host condensates.

Methods

Oligonucleotides and protein

All DNA strands were purchased from Integrated DNA Technologies (IDT). Sequences can be found in Tables S1 and S2. NUPACK prediction results are shown in Table S3. The fluorophore-labeled strand was purified using high-performance liquid chromatography (HPLC), and all other strands were purified using standard desalting. The oligonucleotide concentrations were measured using a NanoDrop 2000 Spectrophotometer (Thermo Fisher Scientific). Streptavidin (SA) was purchased from Fisher Scientific (S32357, Molecular Probes™ Streptavidin, Alexa Fluor™ 647 Conjugate). P22 N peptide was obtained from GenScript (P22-N6FAM, 0.9 mg, Table S2).

RNA synthesis

The synthetic gene strands utilized to transcribe RNA were added to 1X transcription buffer (AmpliScribe 10X Reaction Buffer) and were annealed from 90 °C to 25 °C at a rate of −1 °C min⁻¹. The desired RNA strands (with P22 aptamer, named boxB, for the junction, middle, and tip designs) were then transcribed using the AmpliScribe T7-Flash Transcription Kit (Biosearch Technologies) for about 3–4 hours, incubated at 37 °C. Afterward, the reaction was quenched by adding DNAse I (5% v/v) at 37 °C for another 15–30 minutes. RNA was purified using the Monarch® RNA Cleanup Kit (New England Biolabs), and quantitated using the Nanodrop 2000 Spectrophotometer.

Sample preparation and condensate formation

The nanostar core strands (S1, S2, and S3) were mixed at the desired concentrations in 20 mM TrisHCl and 12.5 mM MgCl₂ buffer. The S1 fluorophore-labeled strand was added at a 17.5% molar ratio. The sample then underwent a thermal annealing from 90 °C to either 37 °C or 25 °C (depending on the experiment) with a −1 °C minute⁻¹ rate, using a Bio-Rad T100™ Thermal Cycler. Afterward, the aptamer-containing strand was added isothermally to the sample to reach the final concentration of 2.5 µM or 1.25 µM. An equimolar amount of SA was added at the same time to the sample, followed by a brief vortexing of the sample for 1–2 s. The samples were incubated at 37 °C unless otherwise noted. All the experiments were performed in triplicate (n = 3) unless otherwise noted. As for the P22 N peptide recruitment experiments with three designs, all the strands (the core strands plus the RNA strand with boxB) were mixed in 20 mM trisHCl and either 12.5 mM MgCl₂ or 350 mM NaCl, depending on the experiment. The solution was then annealed from 72 °C to 25 °C at a rate of −1 °C min⁻¹, followed by the addition P22 and incubation at 37 °C. Note that nanostars were labeled with Atto 647.

Fluorescence recovery after photobleaching (FRAP)

After 60 minutes of incubation, condensate samples were loaded into a chamber made with microscope slides (Fisherbrand Premium Superfrost™, 25 × 75 mm, 1.0 mm thick) that has channels made with parafilm stripes sealed by coverslips and epoxy (Gorilla, 5-minute set). All experiments were done in triplicate (n = 3).

Kleptamer experiments

Condensate samples were incubated at 37 °C for 60 minutes, and the kleptamer or polyT strand was then added at a certain concentration relative to the nanostar concentration (i.e., 0.5×, 1×, and 2×) to reach the final concentration of 2.25 µM. All the experiments were performed in duplicate (n = 2).

Emulsion droplet experiments

The oil-surfactant mixture was prepared by mixing 2% (w/v) mixture of surfactant (008 FluoroSurfactant, RAN Biotechnologies) and oil (FC-40 Fluorinert™, Sigma-Aldrich). Using Eppendorf DNA Lobind® tubes, 20 µL of pre-mixed nanostar samples at 7.5 µM (heated to 55 °C for a few minutes to melt condensates) were added onto 80 µL oil-surfactant mixture (also heated to 55 °C). Next, the sample was shaken for about 45–50 s using a bench vortexer in order to generate a large pool of isolated, immiscible droplets of various sizes. The vortexed sample was equilibrated at room temperature for about a minute prior to transferring an aliquot of it (milky fraction of the sample) into an Ibidi chamber (µ-Slide VI 0.4). The channel was sealed using grease and coverslips to prevent evaporation and to limit droplet motion due to pressure differences. The emulsion experiments were performed in duplicate (n = 2).

Bead recruitment and release

Biotinylated carboxylate-modified polystyrene beads solution, FluoSpheres™ Biotin-Labeled Microspheres (#F8767), 0.2 µm, yellow-green fluorescent (505/515), 1% solids, was purchased from Thermo Fisher Scientific. For the bead recruitment experiments, condensates were prepared as described before, except with the use of 350 mM NaCl, which was used in place of 12.5 mM MgCl2. The sample was annealed until 37 °C, and then the DNA-aptamer strand was added isothermally along with SA and the beads. All experiments were done in triplicate (n = 3). Condensates were prepared at a final concentration of 2.5 µM, biotinylated beads were added at 0.5 pM. All condensates were incubated at 37 °C and imaged 1 hour after the isothermal addition of the DNA-aptamer strand, SA, and beads. DNA nanostars were labeled with Cy3, SA was labeled with Alexa 647, and beads were labeled with yellow-green fluorescent (505/515) to measure bead recruitment. Kleptamer experiments for bead displacement were done with a final nanostar concentration of 2.0 µM and 10× kleptamer relative to the final nanostar concentration. These experiments were done in triplicate (n = 3).

We note that recruitment can be obtained with a MgCl₂ buffer; however, under these conditions, displacement of beads with kleptamers was not achieved. This is likely due to weaker electrostatic screening in the presence of a monovalent salt (NaCl) compared to a divalent salt (MgCl₂).

Microscopy imaging

Most of the samples were imaged with a Nikon Eclipse Ti, using a Nikon CFI Plan Apo Lambda 60X Oil objective, and an Eclipse filter cube with excitation wavelength of 448 nm for FAM-labeled nanostars and 646 nm for Alexa 647-labeled SA, when applicable (exposure time of 90 ms). For the majority of experiments, at least 8 images were taken at each time point. Water-in-oil droplets were imaged using a Nikon Plan Fluor 20X/0.5 NA objective on a fixed field of view (exposure time of 1 s). For the FRAP experiments, a Nikon Eclipse Ti-2 integrated with a FRAP module, using a Nikon CFI Plan Apo Lambda 60X Oil objective, was employed. For the protein-release experiments, partition coefficient measurements, and Pearson's correlation coefficient measurements, a Nikon Eclipse Ti integrated with an NL5+ confocal module (Confocal.nl) was used. Images were obtained using a Nikon CFI Plan Apo 60X Oil objective, and the excitation wavelength of 488 nm was used for the FAM-labeled nanostars and 638 nm for Alexa 647-labeled SA (exposure time of 40 ms and 10 ms, respectively). P22 N peptide recruitment experiments were conducted using a Nikon Eclipse Ti-2 integrated with an AX NSPARC confocal module.

Before imaging, coverslips (Fisherbrand™, 60 × 22 mm, 0.13 to 0.17 mm thick) were used to make small imaging chambers, prepared with a parafilm cut piece with a hole punched in the middle, adhered to the coverslips by heating to 50–60 °C. After cooling the coverslip, 2.5 µL of the sample was loaded inside the punched region. The sample was covered with another coverslip (Fisherbrand™, 22 × 22 mm, 0.13–0.17 mm thick) to minimize sample evaporation.

UV-vis spectrophotometry

Melting (heating) and annealing (cooling) curves were obtained using an HP UV-vis spectrophotometer. For the melting experiments, the temperature changed from 21.5 °C to 93.5 °C at a rate of 0.5 °C minute⁻¹, with a hold at each step for 30 s. Annealing curves were acquired by applying the inverse thermal ramp. Absorbance was monitored at 260 nm with a spectral bandwidth (filter width) of 7 nm. Temperature-dependent absorbance changes were calculated within a window of 25 °C to 90 °C to evaluate melting behavior. Due to the stabilization period at initial higher temperatures, we only included up to 70 °C for annealing and 80 °C for melting curves to ensure accurate results. Each sample (volume of 125 µL at 2.5 µM) was loaded into a quartz cuvette and blanked against an equivalent volume of buffer consisting of 12.5 mM MgCl2, 20 mM Tris-HCl. To minimize evaporation over the ∼4 h measurement period, 50 µL of hexadecane was carefully layered over each sample.

Image processing

Condensate size and eccentricity were measured from epifluorescent microscopy images using in-house Python scripts (see code availability section). We consistently selected eight images per each experiment and ran the scripts over their green channel, corresponding to the FAM-labeled nanostars. Colocalization analysis was performed using a Fiji ImageJ plugin, Just Another Colocalization Plugin (JACoP).^34,35 The pixel intensities of both channels (X_i) were obtained, normalized as x_i = (X_i − X_min)/(X_max − X_min), and plotted using Python (see Code availability). Pearson's correlation coefficient (R) corresponding to each microscopy snapshot shown in the figures was calculated using Python's Scipy library.

Partition coefficients were calculated for each condition with condensates from one experiment. The selected region of interest (ROI) of each droplet was drawn on both the nanostar and SA channels, with the assumption that FAM-labeled nanostars are uniformly dispersed inside and outside the condensates. We also selected a background ROI across all samples. The partition coefficient was calculated using the following equation: PC = I_i/I_o, where I_i is the mean fluorescent intensity inside the condensate ROI, and I_o is the mean fluorescent intensity of the background ROI.³⁶

FRAP experiments were analyzed as follows: an initial image was taken before photobleaching, which occurred for 200 ms with a 488 nm laser (intensity of 2%). The recovery was then observed for 180 s. An ROI was placed at the location of bleaching, and a duplicate ROI was made and placed on a non-bleached condensate to offset any bleaching induced by imaging. Recovery was calculated as:

(I_bleach,t/I_bleach,max)/(I_unbleach,t/I_unbleach,max)

where I is the mean pixel intensity, t is the current time point, and max index attributes to the highest intensity over time. The data was then normalized between 0 and 1 using the following equation:

We fit the normalized mean data with this equation:^12,37

where I(t) is the intensity at time t and τ_1/2 is the time constant, and a and b are the minimum and maximum normalized intensity.

MD simulations

To perform the molecular dynamics (MD) simulations, we utilized oxDNA2 compiled on a local machine with the Linux operating system. The nanostar variants were first created using oxView, a browser-based visualization tool.³⁸ The nanostars then underwent a standard, short-term Monte Carlo (MC) relaxation, followed by an MD relaxation to eliminate steric clashes, structural artifacts, and unrelaxed backbone potentials. Relaxations were done at 290 K and monovalent salt concentration of 1 M. Afterward, we performed a long-term MD simulation on the relaxed nanostars at 37 °C and monovalent salt concentration of 1 M to obtain their conformational changes over 1.818 ns. The parameters used in each simulation are listed in Table S6. We used in-house MATLAB scripts to quantify the angular distributions of each nanostar design (see Code availability).

Code availability

The condensate size and eccentricity measurement Python scripts can be found at: https://github.com/klockemel/Condensate-Detection.

The MATLAB scripts used in this study to process and quantify data are available at: https://github.com/mdizani/ProteinRecruitmentV2.git.

The source code for colocalization measurements using JACoP can be found at: https://imagej.net/plugins/jacop.

Conflicts of interest

There are no conflicts to declare.

Data availability

Unprocessed microscopy images are available at: https://ucla.box.com/s/ernvu62b7kwqr29znix1nqewhyvd1n05.

Supplementary information (SI) is available. See DOI: https://doi.org/10.1039/d5nh00586h.

Acknowledgements

We thank Deborah Fygenson, Jaimie M. Stewart, Eli Kengmana, and Thomas Reese for helpful discussions and advice. We also thank Neil Lin for assistance with confocal microscopy instruments. This research was supported by the National Science Foundation (NSF) grant FMRG: Bio award 2134772 and by the Sloan Foundation through award G-2021-16831 to EF. This work was supported by the National Science Foundation Graduate Research Fellowship Program to HRM.

References

1. Y. Shin and C. P. Brangwynne, Science, 2017, 357(6357), eaaf4382.
2. Y. Elani, R. V. Law and O. Ces, Nat. Commun., 2014, 5, 5305.
3. D. L. J. Lafontaine, J. A. Riback, R. Bascetin and C. P. Brangwynne, Nat. Rev. Mol. Cell Biol., 2020, 22, 165–182.
4. Y. Sato, T. Sakamoto and M. Takinoue, Sci. Adv., 2020, 6, eaba3471.
5. S. Biffi, R. Cerbino, F. Bomboi, E. M. Paraboschi, R. Asselta, F. Sciortino and T. Bellini, Proc. Natl. Acad. Sci. U. S. A., 2013, 110, 15633–15637.
6. D. T. Nguyen and O. A. Saleh, Soft Matter, 2017, 13, 5421–5427.
7. J. M. Stewart, S. Li, A. A. Tang, M. A. Klocke, M. V. Gobry, G. Fabrini, L. Di Michele, P. W. K. Rothemund and E. Franco, Nat. Commun., 2024, 15, 1–13.
8. W. Liu, J. Deng, S. Song, S. Sethi and A. Walther, Commun. Chem., 2024, 7, 1–10.
9. A. Samanta, L. Baranda Pellejero, M. Masukawa and A. Walther, Nat. Rev. Chem., 2024, 8, 454–470.
10. A. R. Strom, A. V. Emelyanov, M. Mir, D. V. Fyodorov, X. Darzacq and G. H. Karpen, Nature, 2017, 547, 241–245.
11. S. Agarwal, D. Osmanovic, M. Dizani, M. A. Klocke and E. Franco, Nat. Commun., 2024, 15, 1–13.
12. Y. Sato and M. Takinoue, Nanoscale Adv., 2023, 5, 1919–1925.
13. B.-J. Jeon, D. T. Nguyen, G. R. Abraham, N. Conrad, D. K. Fygenson and O. A. Saleh, Soft Matter, 2018, 14, 7009–7015.
14. S. Agarwal, D. Osmanovic, M. A. Klocke and E. Franco, ACS Nano, 2022, 16(8), 11842–11851.
15. S. Do, C. Lee, T. Lee, D.-N. Kim and Y. Shin, Sci. Adv., 2022, 8, eabj1771.
16. D. Gao, S. Wilken, A. B. N. Nguyen, G. R. Abraham, T. Liedl and O. A. Saleh, Soft Matter, 2024, 20, 1275–1281.
17. S. D. Jayasena, Clin. Chem., 1999, 45, 1628–1650.
18. K. Sefah, D. Shangguan, X. Xiong, M. B. O’Donoghue and W. Tan, Nat. Protoc., 2010, 5, 1169–1185.
19. J. A. Zakashansky, A. H. Imamura, D. F. Salgado, H. C. Romero Mercieca, R. F. L. Aguas, A. M. Lao, J. Pariser, N. Arroyo-Currás and M. Khine, Anal. Methods, 2021, 13, 874–883.
20. E. A. Lamont, L. He, K. Warriner, T. P. Labuza and S. Sreevatsan, Analyst, 2011, 136, 3884–3895.
21. K. L. Hong and L. J. Sooter, BioMed Res. Int., 2015, 2015, 419318.
22. M. Dizani, D. Sorrentino, S. Agarwal, J. M. Stewart and E. Franco, J. Am. Chem. Soc., 2024, 146, 29344–29354.
23. K. Leppek and G. Stoecklin, Nucleic Acids Res., 2014, 42, e13.
24. T. Bing, X. Yang, H. Mei, Z. Cao and D. Shangguan, Bioorg. Med. Chem., 2010, 18(5), 1798–1805.
25. J. Lloyd, C. H. Tran, K. Wadhwani, C. Cuba Samaniego, H. K. K. Subramanian and E. Franco, ACS Synth. Biol., 2018, 7, 30–37.
26. H. Ma, L.-C. Tu, A. Naseri, M. Huisman, S. Zhang, D. Grunwald and T. Pederson, Nat. Biotechnol., 2016, 34, 528–530.
27. S. Agarwal, M. Dizani, D. Osmanovic and E. Franco, Interface Focus, 2023, 13, 20230017.
28. C. M. Dundas, D. Demonte and S. Park, Appl. Microbiol. Biotechnol., 2013, 97, 9343–9353.
29. D. T. Nguyen, B.-J. Jeon, G. R. Abraham and O. A. Saleh, Langmuir, 2019, 35(46), 14849–14854.
30. T. E. Ouldridge, A. A. Louis and J. P. K. Doye, J. Chem. Phys., 2011, 134, 085101.
31. E. Poppleton, M. Matthies, D. Mandal, F. Romano, P. Šulc and L. Rovigatti, J. Open Source Softw., 2023, 8, 4693.
32. J. M. Majikes, M. Zwolak and J. A. Liddle, Biophys. J., 2022, 121, 1986–2001.
33. N. Conrad, G. Chang, D. K. Fygenson and O. A. Saleh, J. Chem. Phys., 2022, 157, 234203.
34. S. Bolte and F. P. Cordelières, J. Microsc., 2006, 224, 213–232.
35. S. V. Costes, D. Daelemans, E. H. Cho, Z. Dobbin, G. Pavlakis and S. Lockett, Biophys. J., 2004, 86, 3993–4003.
36. E. Kengmana, E. Ornelas-Gatdula, K.-L. Chen and R. Schulman, J. Am. Chem. Soc., 2024, 146(48), 32942–32952.
37. S. Sahu, P. Chauhan, E. Lumen, K. Moody, K. Peddireddy, N. Mani, R. Subramanian, R. Robertson-Anderson, A. J. Wolfe and J. L. Ross, PNAS Nexus, 2023, 2, gad231.
38. J. Bohlin, M. Matthies, E. Poppleton, J. Procyk, A. Mallya, H. Yan and P. Šulc, Nat. Protoc., 2022, 17, 1762–1788.

---

Footnote

† These authors contributed equally.

---