nature nanotechnology

Article https://doi.org/10.1038/s41565-024-01726-x

Co-transcriptional production of
programmable RNA condensates
and synthetic organelles

Received: 12 October 2023 Giacomo Fabrini 1,2,3,4, Nada Farag 3, Sabrina Pia Nuccio1, Shiyi Li 5,
Jaimie Marie Stewart6, Anli A. Tang7, Reece McCoy 3, Róisín M. Owens 3
,
Accepted: 20 June 2024
Paul W. K. Rothemund 6, Elisa Franco 5,7, Marco Di Antonio 1 &
Published online: 30 July 2024 Lorenzo Di Michele 1,2,3

Check for updates

Condensation of RNA and proteins is central to cellular functions, and the
ability to program it would be valuable in synthetic biology and synthetic cell
science. Here we introduce a modular platform for engineering synthetic
RNA condensates from tailor-made, branched RNA nanostructures that
fold and assemble c­o-­tr­anscriptionally. Up to three orthogonal condensates
can form simultaneously and selectively accumulate fluorophores through
embedded fluorescent light-up aptamers. The RNA condensates can be
expressed within synthetic cells to produce membrane-less organelles with
a controlled number and relative size, and showing the ability to capture
proteins using selective protein-binding aptamers. The affinity between
otherwise orthogonal nanostructures can be modulated by introducing
dedicated linker constructs, enabling the production of bi-phasic RNA
condensates with a prescribed degree of interphase mixing and diverse
morphologies. The in situ expression of programmable RNA condensates
could underpin the spatial organization of functionalities in both biological
and synthetic cells.

Membrane-less compartmentalization sustained by biomolecular con- RNA repeat sequences9,13,14 and riboswitches15,16 have highlighted the
densates is recognized as a primary regulatory mechanism in cells1–4. By feasibility of this concept. The generality of these strategies, however,
co-localizing nucleic acids, enzymes and metabolites, membrane-less is hampered by the challenges of protein engineering and the limited
organelles (MLOs) such as nucleoli, Cajal bodies and stress granules programmability of natural RNA constructs.
are believed to regulate biogenesis, transcription, post-transcriptional Leveraging nucleic acid nanotechnology17–19, in this paper we
modification and degradation4–7, while pathological condensates have introduce a systematic method for expressing designer biomolecu-
been linked to neurodegeneration8,9. lar condensates from synthetic RNA nanostructures. Our elemen-
The ability to express ‘designer condensates’ with prescribed prop- tary motifs consist of star-shaped junctions, or nanostars, which fold
erties would be valuable to program cellular behaviour10,11 and engineer co-transcriptionally and assemble driven by selective base-pairing
synthetic cells12. Remarkable examples based on peptides10,11 or natural interactions, forming up to three co-existing but fully distinct

1
Department of Chemistry, Imperial College London, London, UK. 2fabriCELL, Imperial College London, London, UK. 3Department of Chemical
Engineering and Biotechnology, University of Cambridge, Cambridge, UK. 4The Francis Crick Institute, London, UK. 5Department of Bioengineering,
University of California at Los Angeles, Los Angeles, CA, USA. 6Department of Computing and Mathematical Sciences, California Institute of Technology,
Pasadena, CA, USA. 7Department of Mechanical and Aerospace Engineering, University of California at Los Angeles, Los Angeles, CA, USA.
e-mail: [email protected]

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1665

Article https://doi.org/10.1038/s41565-024-01726-x

a A B C c
(i) (ii) (iii)
50

45

Tm (°C)
MGA BrA BrA
40
MGA
A A A A A A A A A
5' U A 3' 5' G C 3' G C
5' 3'
C G U A G C
T7 RNAP G C C G U A
C
G
G
C
G
A
C
U
A
C
U
G 0
3'
A U
5' 3'
C G
5' 3'
C G
5'
A B C
A-T A A A B-T A A A C-T A A A

RNA construct
b Epi Epi Epi d
(i) (ii) (iii) 75

µCLD (µm)
50
A
25 B
C
0
3h Confocal 3h Confocal 3h Confocal
3
10

N
2
10

0 24 48
12 h >48 h 12 h >48 h 12 h >48 h
Time (h)
e A B f
I II III IV I II III IV
A
10,000
B

τc (s)
I
Aspect ratio (a.u.)

Aspect ratio (a.u.)

Data I Data
5,000
2.0 2.0
Fit Fit

1.5 II 1.5 II
III III
IV IV
1.0 1.0
0
0 2 4 6 8 0 1 2 3 4 20 40 60 80

Time (h) Time (h) Characteristic length lc (mm)

Fig. 1 | Condensation of co-transcriptionally folding RNA nanostars. Full CLDs, as extracted from image segmentation, are shown in Supplementary
a, Structure of the RNA motifs. An A-type RNA nanostar includes an MGA (i), Fig. 17 (Supplementary Methods 2). N is not computed for system C, which does
B-type includes a BrA (ii) and C-type includes both aptamers (iii). Variants not form discrete aggregates. Data are shown as mean (solid lines) ± s.d. (top) or
feature mutually orthogonal, self-complementary (palindromic) KLs, whose s.e. (bottom) (shaded regions) of three field of views within one sample. e, Top:
sequences are shown in the insets. RNA nanostars are transcribed from linear epifluorescence micrographs (contrast enhanced) depicting coalescence events
dsDNA templates by T7 RNAP. b, Epifluorescence and confocal micrographs for A (left) and B (right) condensates. Scale bars, 15 μm. Bottom: time-dependent
showing condensate formation and coarsening for all three designs in a at aspect ratio of the condensates above, computed as the ratio between major and
different timepoints of an in vitro transcription reaction. Epifluorescence minor axes of the best-fit ellipse. The dashed line shows an exponential fit with
micrographs have been linearly re-scaled to enhance contrast (Supplementary decay constant τc. f, τc against the characteristic size (lc) of A and B condensates
Methods 2). Pristine micrographs are shown in Supplementary Fig. 5. Scale bars, undergoing coalescence. Linear regression yields inverse capillary velocities
50 μm. Timestamps are reported with respect to the start of time-lapse imaging μ/γ = 127.4 s μm−1 and 152.3 s μm−1 for A and B condensates, respectively46
(Methods and Supplementary Table 5). c, Melting temperatures (Tm) of A–C (Supplementary Methods 2). The dashed lines indicate best fits, with 95%
condensates, determined as discussed in Methods, Supplementary Methods 2, confidence intervals shown as shaded regions. Transcription and coalescence
and Supplementary Figs. 8 and 9. d, Top: mean of the CLD, μCLD. Bottom: average events occurred at constant 30 °C temperature (Methods).
number of condensates per microscopy field of view, N, as a function of time.

condensate types. Expressing the condensates in synthetic cells gen- well-characterized DNA nanostars21–23 (Fig. 1a). The RNA motifs inter-
erates MLOs with controlled size, number, morphology and compo- act via self-complementary HIV-type kissing loops (KLs) present at the
sition. Finally, including RNA aptamers enables selective capture of end of each arm24, rather than via the single-stranded (ss) overhangs or
small molecules and proteins, imitating the ability of natural MLOs hydrophobic modifications adopted for DNA designs21–23. Similar KLs
to recruit clients. have been shown to facilitate condensation in bacterial riboswitches15,16.
Because the RNA nanostars are transcribed in situ from DNA tem- We tested three RNA-nanostar designs, labelled A, B and C, featuring
plates, our platform could be directly applied to express synthetic mutually orthogonal KL sequences (Fig. 1a). In designs A and B, one of
MLOs in living cells, besides its immediate use in synthetic cells. In the double-stranded RNA (dsRNA) arms includes a fluorescent light-up
this context, exploring the design space of RNA nanostars will allow aptamer (FLAP): malachite green aptamer (MGA) for A25,26 and Broccoli
for fine-tuning of condensate properties20. aptamer (BrA) for B27 (Fig. 1a(i),(ii)). FLAPs yield a fluorescent signal
upon binding their cognate fluorophores (malachite green (MG) for
RNA nanostar design and co-transcriptional MGA and 3,5-difluoro-4-hydroxybenzylidene imidazolinone (DFHBI)
condensation for BrA), enabling characterization via fluorescence microscopy and
The four-armed nanostars consist of a single RNA strand that folds fluorimetry (Supplementary Fig. 1). Design C includes both MGA and
co-transcriptionally into the intended star-like shape, inspired by BrA in non-adjacent arms (Fig. 1a(iii)). The arms not hosting FLAPs are

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1666

Article https://doi.org/10.1038/s41565-024-01726-x

25 base-pairs long, and separated by an unbound uracil for flexibility28. concentration in the bulk and within the condensates reached a pla-
The motifs were transcribed with T7 RNA polymerase (T7 RNAP) from teau, indicative of a steady state between dilute and condensed RNA
double-stranded DNA (dsDNA) templates, labelled as A-T, B-T and C-T phases (Supplementary Fig. 16 and Supplementary Methods 2).
for designs A, B and C, respectively (Supplementary Fig. 2). Denaturing We gained further insights on condensate growth and coarsening
polyacrylamide gel electrophoresis confirms the expected electropho- dynamics from chord-length distribution (CLD) analysis of epifluores-
retic mobility for most transcripts, with small amounts of truncated and cence micrographs43–45 (Supplementary Fig. 17 and Supplementary
over-elongated products29,30 (Supplementary Fig. 3). Native agarose gel Methods 2). The CLD provides a time-dependent picture of condensate
electrophoresis suggests that transcripts retain the intended folded length scales in a way that is agnostic of their shape, and is thus equally
monomeric conformation, rather than producing misfolded multim- meaningful for the branched C-type structures and the compact A and
ers15,16 (Supplementary Fig. 4). See Methods for sequence design and B condensates.
transcription protocols, and sequences in Supplementary Tables 1–4. Figure 1d (top) shows the time evolution of the mean of the CLD,
The micrographs in Fig. 1b show that all three designs formed μCLD, which is useful as a proxy for the typical condensate size. For all
aggregates when transcribed in vitro (see also Supplementary Figs. 5 designs, μCLD rapidly increased at early stages, probably sustained by
and 6 (top), and Supplementary Video 1 (top)). Variants A and B formed the active transcription leading to monomer addition (Supplemen-
condensates that nucleated and grew, with frequent coalescence tary Figs. 13 and 16). For A and B, frequent coalescence events also
events, indicating a liquid state. The condensates were roughly spheri- contributed to the increase in μCLD, reflected by a steep decrease in the
cal and did not substantially wet the glass substrate (Supplementary number of condensates (N; Fig. 1d (bottom)). Coalescence appears to
Fig. 7), consistent with the evidence of Brownian motion (Supple- occur more readily in A, given the steeper decrease in N and increase
mentary Video 1). Conversely, design C formed a gel-like percolating in μCLD compared with B. Supplementary Video 1 suggests that early
structure that failed to produce discrete condensates but still grew coalescence may be driven by neighbouring condensates touching as
over time. The higher apparent viscosity of design C compared with A they grow through monomer addition, aided by Brownian motion. Con-
and B correlates with the melting temperature of the materials, which sistently, both the size and number of A and B condensates plateaued
is highest in C, followed by B and A (Fig. 1c, Supplementary Figs. 8 and when transcription slowed (Supplementary Fig. 13). In C aggregates,
9, and Supplementary Video 2). All aggregates showed the intended the increase of μCLD continued at later times, driven by the slow coars-
fluorescence output, namely, in the MGA channel for A (red), the BrA ening of the percolating RNA network (Fig. 1b(iii) and Supplementary
channel for B (cyan) and both channels for C (white). Video 1).
Degradation and bubble formation was observed in the con- The coalescence dynamics of A and B condensates can be fur-
densates over time, ascribed to environmental nucleases31 or ther analysed to determine the inverse capillary velocity of the RNA
photo-degradation (Supplementary Fig. 5). Bubbling was most promi- phases, namely, the ratio between their viscosity (μ) and surface ten-
nent for A, consistent with the lower melting temperature and con- sion (γ)46–48 (Supplementary Methods 2). As summarized in Fig. 1e,f, we
sequent expectation that less damage would be required to trigger find μ/γ = 127.4 s μm−1 and μ/γ = 152.4 s μm−1 for A and B condensates,
disassembly. Sequence differences between the constructs may also respectively. These values are significantly higher compared with
influence their susceptibilities to degradation. DNA-nanostar condensates (0.9 s μm−1 to 26.3 s μm−1; refs. 46–48),
The specificity of KL interactions was confirmed by non-sticky, but compatible with the broad range reported for protein-based and
control designs— Ā and B̄—where KLs were replaced with scrambled biological condensates (~10−2 s μm−1 to ~102 s μm−1; refs. 49,50). Fluores-
sequences. These did not yield condensates and produced only dif- cence recovery after photobleaching (FRAP; Supplementary Methods
fused fluorescence (Supplementary Figs. 6 (bottom) and 10, and Sup- 4), performed using dyes covalently linked to the RNA (Methods),
plementary Video 1 (bottom)). revealed lack of recovery over >500 s for both A and B (Supplementary
The evidence that condensation requires KL complementarity sug- Fig. 18), suggesting a higher viscosity of RNA-nanostar condensates
gests that non-specific, cation-dependent, phase-separation mecha- compared with their DNA counterparts46–48. When performing FRAP
nisms32 are not dominant. To further elucidate the role of cations, using the embedded FLAPs, both A and B condensates showed rapid
we characterized the stability of A and B condensates upon buffer fluorescence recovery, probably due to exchange of dyes with the bulk42
replacement (Supplementary Fig. 11). Condensates remained stable (Supplementary Fig. 18).
after 24 hours in Tris-EDTA buffer supplemented with 5 mM or 10 mM
MgCl2, but disassembled in phosphate-buffered saline, consistent with Orthogonal RNA condensates
previous evidence that divalent cations stabilize KL interactions33. KL orthogonality enables the simultaneous transcription of A and B
We assessed the effect of crowding agents, often introduced to designs, which readily formed distinct, co-existing condensates (Fig. 2a,
aid RNA condensation in vitro34,35, on co-transcriptional assembly of Supplementary Figs. 19 and 20, and Supplementary Video 3), confirm-
RNA nanostars (Supplementary Fig. 12). When including 25% v/v poly- ing the negligible influence of base-pairing-independent condensa-
ethylene glycol (PEG) 200, we observed a reduction in condensate size tion pathways32. Consistently, if one of the RNA motifs was rendered
in both A and B systems, consistent with previous observations that non-sticky, condensates of one species co-existed with dispersed RNA
PEG 200 reduces the T7 RNAP transcription rate36. Non-binding variants nanostars of the other (Fig. 2b, Supplementary Figs. 21 and 22, and
Ā and B̄ remained soluble in the presence of PEG, which is thus insuf- Supplementary Video 4).
ficient to trigger non-specific condensation. The relative size of A and B condensates can be controlled by
Bulk fluorimetry was used to monitor the rate of synthesis of the tuning the ratio between the concentrations of the corresponding
RNA constructs (Supplementary Fig. 13). All designs initially showed DNA templates ([A-T] and [B-T]). Relative-size control is demon-
a rapid signal increase, whose rate scaled (nearly) linearly with template strated visually (Fig. 2a, and Supplementary Figs. 19 and 20), through
concentration (Supplementary Figs. 14 and 15). This initial phase was time-dependent μCLD analysis (Fig. 2c and Supplementary Fig. 23), and
followed by a plateau or slower growth, probably due to loss of poly- through the distribution of the final condensate radii, rc, (Fig. 2d and
merase activity and/or nucleotide depletion37–39. A peak, ascribed to Supplementary Methods 2). Condensate numbers anticorrelate with
aggregate sedimentation, was noted for sticky motifs (A, B and C) but their size. For instance, many small A-type condensates were formed
not for non-sticky designs ( Ā and B̄). Differences in plateauing behav- when [B-T] > [A-T] (Supplementary Fig. 24).
iours could derive from variations in the kinetics of aptamer folding Condensate growth occurred in two stages in all sticky A and B
and/or complexation with fluorophores40–42. Epifluorescence micro- systems: after an initial increase, a brief intermediate plateau was
graphs reveal that, after an initial transient, the ratio between nanostar reached, followed by another growth phase before saturation (Fig. 2c).

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1667

Article https://doi.org/10.1038/s41565-024-01726-x

a [A-T]/[B-T] b
0.25 0.50 1 2 4 A and B A and B
KL A
A A A
U A
5' 3'
C G
G C

3h 3h
C G
3' G C 5'
A U
A A A

KL B
A A A
G C
5' 3'
U A
C G
G C
3' A U 5'
C G
Time
A A A

12 h 12 h
A
A A A
U A
5' 3'
A C
G G
C G
3' A C 5'
C A
A A A

B
48 h 48 h
A A A
A A
5' 3'
C C
G G
A G
3' A C 5'
C A
A A A

c 100 d
A and B 56 131
A and B 100 60 66 99
µCLD (µm)

rc (µm)
50 50

0
155 78 53 56 50
0
0 24 48 0 24 48 0 24 48 0 24 48 0 24 48 0.25 0.50 1 2 4

Time (h) Time (h) Time (h) Time (h) Time (h) [A-T]/[B-T]

Fig. 2 | Co-transcribed orthogonal RNA nanostars form immiscible central panel as dashed lines. See Supplementary Fig. 23 for full CLDs. Red and
condensates of controlled size. a, Epifluorescence micrographs of binary cyan curves are relative to A and B condensates, respectively. Data are shown as
systems of A and B RNA nanostars (see Fig. 1) at various timepoints during the mean (solid/dashed lines) ± s.d. (shaded regions) of three fields of view within
transcription transient. Different ratios between the concentrations of the two one sample. d, Distribution of the radii, rc, of A (red) and B (cyan) condensates as a
templates, A-T and B-T, are tested, while keeping [A-T] + [B-T] constant. function of the template concentration ratio [A-T]/[B-T]. Epifluorescence
b, Epifluorescence snapshots analogous to a, but where either A or B is replaced micrographs in a and b have been linearly re-scaled to enhance contrast
by its non-sticky variant, namely, Ā or B̄. Sketches on the right-hand side show (Supplementary Methods 2). Pristine micrographs are shown in Supplementary
examples of scrambled non-binding KL sequences. c, Time evolution of μCLD Figs. 19 and 21. All scale bars, 50 μm. Timestamps are reported with respect to the
(Supplementary Methods 2), for samples in a and b, the latter shown in the start of time-lapse imaging (Methods and Supplementary Table 5).

This behaviour was not observed in single-component systems (Fig. 1d), spherical condensate in each synthetic cell (Fig. 3b and Supplementary
nor in binary systems with one sticky and one non-sticky species Fig. 27), including design C that only generated extended networks in
(Fig. 2c, middle). Supplementary Videos 3 and 4 reveal that the inter- bulk. The different morphology is rationalized by noting that C aggre-
mediate plateau was reached when same-type condensates became gates need to relax over much smaller length scales within synthetic
temporarily unable to coalesce due to being ‘caged’ by neighbouring cells. Yet, shape relaxation was slower for C compared with A and B
condensates of the opposite type. Coalescence events that still man- (Fig. 3c, and Supplementary Figs. 28 (top) and 29, and Supplementary
aged to occur, however, reduced lateral crowding given that merged Video 6 (top)). Polydispersity in condensate size reflects the variability
condensates occupy less space in the horizontal plane, unjamming the in the size of synthetic cells, with the final volume of the condensates
system and accelerating coalescence. In fact, condensates reached scaling linearly with that of the enclosing W/O droplet (Fig. 3d).
larger dimensions in A and B systems (Fig. 2c) compared with Ā and B, Control experiments with non-sticky designs revealed uniform
and A and B̄ mixtures (Fig. 2b and Supplementary Fig. 23), indicating fluorescence within the synthetic cells, confirming assembly specificity
that steric encumbrance from non-binding condensates of the opposite (Supplementary Figs. 28 (bottom) and 30, and Supplementary Video
phase ultimately facilitates coalescence. Two-step decreasing trends, 6 (bottom)). Fluorimetry can be used to monitor RNA synthesis rates,
consistent with those seen in μCLD, were observed in condensate num- as shown in Supplementary Fig. 31, where the delayed growth in the
bers (Supplementary Fig. 24). MGA signal (A component) is due to initial fluorophore partitioning
Addition of 25% v/v PEG 200 induced non-specific affinity between in the oil phase54,55, rather than to a slower growth of the condensates
A and B condensates51,52 (Supplementary Fig. 12), leading to extended (compare with Fig. 3c). Consistent with bulk data, initial transcription
networks of small alternating A and B domains reminiscent of colloidal rates were found to scale (nearly) linearly with template concentration
gelation53. (Supplementary Figs. 32 and 33).
Condensate co-assembly is also possible with three RNA spe- Supplementary Fig. 34 shows the time-dependent concentration
cies A, B and C (Supplementary Figs. 25 and 26, and Supplementary of B nanostars during transcription transient (Supplementary Meth-
Video 5. After allowing sufficient time for relaxation, all species formed ods 5). The nanostar concentration exceeded 10 μM (or ~1 g l−1) within
spherical condensates, including C, which was unable to do so in 1.5 h of the start of transcription (~15 min in Fig. 3b(ii); Supplementary
single-component samples (Fig. 1b(iii)). The difference in morphology Table 6), consistent with previous reports on DNA nanostars show-
is probably due to A and B stars hindering the formation of a percolat- ing phase separation at concentrations as low as 0.25 μM (ref. 40) or
ing C network in favour of smaller aggregates that relax more readily. ≤ 0.1 g l−1 (ref. 56).
We obtained synthetic cells with two distinct A and B MLOs, as
RNA MLOs in synthetic cells shown in Fig. 3e, Supplementary Figs. 35–39, and Supplementary Vid-
Addressable RNA condensates could be extremely valuable to engi- eos 7 and 8 with microscopy, and Supplementary Fig. 40 with fluorim-
neer compartmentalization in synthetic or living cells, where they etry. In most cases, each synthetic cell contained one condensate of
could operate as MLOs capable of recruiting compounds and under- each type (Fig. 3g) and, like in bulk, we could control the relative size
pinning spatial separation of functionalities. To demonstrate this, of the organelles by changing the template ratio (Fig. 3f). When includ-
we transcribed our condensates within synthetic cells constructed ing component C, we obtained three distinct phases (Fig. 3h, Supple-
from water-in-oil (W/O) droplets (Fig. 3a). All designs formed a single mentary Figs. 41–43 and Supplementary Video 9), exemplifying the

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1668

Article https://doi.org/10.1038/s41565-024-01726-x

a b
Epi Epi Epi
T7 RNAP Encapsulate
dsDNA (i) (ii) 15 min (iii) 15 min
rNTPs

Water
Oil
15 min Confocal Confocal Confocal
c
A
µCLD (µm)

30
B
15 C
0
0 24 48
1h >48 h 1h >48 h 4h >48 h
Time (h)

e [A-T]/[B-T] d
0.25 0.50 1 2 4 2.0 A n = 66/70

VC (x105µm3)
B
1.0

0
0 0.5 1.0

VSynCell (x106µm3)

f g n = 42 42 46 40 32
100 %A
>48 h >48 h >48 h >48 h >48 h 3.0
33 40 28
41 31 75

Fraction of synthetic cells (%)

50

Radial ratio rA/rB

2.0 25

0
>48 h >48 h >48 h >48 h >48 h 100
%B
1.0 75
h
50

25
0
1 1 1 2 4 0
1 2 3
4 2
>48 h 2 weeks 2 weeks [A-T]/[B-T] Number of condensates

Fig. 3 | Membrane-less RNA organelles expressed in synthetic cells. regions (Supplementary Methods 3). MLOs occupy 18.2 ± 0.5% and 15.3 ± 0.3% of
a, Diagram showing MLOs formed in synthetic cells consisting of W/O emulsion the volume of the synthetic cells for A and B systems, respectively. e, Zoomed-
droplets encapsulating transcription machinery, ribonucleotide triphosphates in (top) and larger field-of-view (bottom) confocal micrographs depicting
(rNTPs) and DNA templates. b, Epifluorescence and confocal micrographs synthetic cells co-expressing A- and B-type condensates, with different template
showing MLO formation over time in synthetic cells expressing A-type (i), B-type concentration ratios [A-T]/[B-T] (compare Fig. 2a). f, Distribution of the ratio
(ii) and C-type (iii) RNA nanostars (see Fig. 1). Epifluorescence micrographs between the radii of A and B MLOs (rA/rB) as a function of [A-T]/[B-T] for samples in
have been linearly re-scaled to enhance contrast (Supplementary Methods 2). e (Supplementary Methods 3). g, Percentage of synthetic cells containing a given
Pristine images are shown in Supplementary Fig. 27, alongside images relative number of A-type (top) or B-type (bottom) MLOs. The percentages of synthetic
to additional timepoints. Timestamps are reported with respect to the start of cells containing exactly one A and one B MLOs are 78.57%, 97.62%, 86.96%, 77.50%
time-lapse imaging (Methods and Supplementary Table 6). c, Time-dependent and 87.50% for [A-T]/[B-T] = 0.25, 0.50, 1, 2 and 4, respectively. Colour codes in f
mean of μCLD, computed as discussed in Supplementary Methods 2. Data are and g match those in e. Numbers in f and g indicate sampled synthetic cells.
shown as mean (solid line) ± s.d. (shaded region) from three fields of view h, Confocal micrographs showing synthetic cells expressing three orthogonal
within one sample. d, Scatter plot of condensate volume (VC) versus synthetic MLO-forming RNA nanostars (A, B and C in Fig. 1) at different timepoints. Scale
cell volume (VSynCell) for samples in b(i) and b(ii). Dashed lines indicate best fits bars in e, bottom, and h, left and centre, are 150 μm. All other scale bars, 50 μm.
to linear regression models, with 95% confidence intervals shown as shaded

possibility for scaling up the number of addressable organelles. Syn- demonstrated in ref. 57 with DNA constructs, nanostar L is ‘chimeric’, as
thetic cells often failed to produce exactly three distinct MLOs, prob- it features two A-type and two B-type KLs (Fig. 4a). As shown in Fig. 4b,
ably due to steric effects and the intrinsic slow relaxation of phase C low fractions of the linker template L-T ([A-T]:[L-T]:[B-T] = 10:1:10, or
and consistent with bulk experiments (Supplementary Figs. 25 and 26). linker template fraction (LTF) = 1/21) produced grape-like clusters,
When replacing component C with C,̃ which features identical KLs but blocking the relaxation of A and B domains into two large conden-
lacks any FLAPs, synthetic cells with exactly three MLOs were more sates (Supplementary Figs. 44–46). Arrested coarsening is arguably
common (Supplementary Fig. 42), suggesting that aptamers affect the due to interphase adhesion limiting the ability of the condensates
coarsening kinetics of the material. to slide past each other, as noted for linker-free A–B systems with
We further expanded the possible organelle architectures in syn- crowding agents (Supplementary Fig. 12). At higher LTFs, bigger A-
thetic cells by introducing linker RNA nanostars, dubbed L, modulating and B-rich domains formed, with Janus-like morphologies emerg-
the mixing between the A and B components. Similarly to the strategy ing at [A-T]:[L-T]:[B-T] = 5:1:5 (LTF = 1/11) (Supplementary Fig. 46, and

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1669

Article https://doi.org/10.1038/s41565-024-01726-x

a c
Chimera linker
Fully demixed MLOs (L) Tunable mixing and composition 28 30 32 32
31 24 38
1.0

JA
0.5
Orthogonal KLs Linker template fraction

0
b [A-T]:[L-T]:[B-T] (LTF)
10:1:10 (1/21) 7:1:7 (1/15) 5:1:5 (1/11) 3:1:3 (1/7) 2:1:2 (1/5) 1:1:1 (1/3) 1:2:1 (1/2)

1.0

JB
0.5

1 1 1 1 1 1 1
21 15 11 7 5 3 2
LTF

Fig. 4 | Controlling morphology and composition of MLOs with linker RNA changes in condensate colour occur away from the confocal imaging plane,
nanostars. a, Chimeric RNA linker nanostars (L), with two A and two B KLs, probably due to differences in the extinction coefficients of MG and DFHBI. Scale
enable control over mixing in systems of A and B nanostars by varying the relative bars, 50 μm (top) and 150 μm (bottom). c, Distributions of mixing indices JA and
concentration of the linker template L-T. b, Zoomed-in (top) and larger field-of- JB of the MLOs (calculated as discussed in Supplementary Methods 3) and shown
view (bottom) confocal micrographs, acquired after more than 48 h from the as a function of LTF for samples in b. Low JA and JB are indicative of purer A-rich
start of transcription, depicting synthetic cells producing A, B and L nanostars and B-rich phases, while JA, JB ≈ 1 indicate complete mixing of the two RNA species.
with different ratios between DNA templates ([A-T]:[L-T]:[B-T]). The LTF, shown Numbers indicate examined synthetic cells.
in brackets, is computed as [L-T]/([A-T] + [L-T] + [B-T]). For LTF = 1/3 and 1/2, slight

Supplementary Videos 10 and 11). Here we note the occasional forma- Selective protein capture in RNA condensates
tion of a cavity in the interphase contact region, hinting at an uneven While Figs. 1–4 demonstrate that RNA condensates can selectively
linker distribution. For [A-T]:[L-T]:[B-T] = 3:1:3 (LTF = 1/7) we observe sequester small molecules—the fluorophores associated with MGA
hollow, capsule-like organelles for most of the larger synthetic cells and BrA—imitating natural MLOs requires capturing larger and func-
(Supplementary Videos 11 and 12, and Supplementary Fig. 46). Com- tional macromolecules, particularly proteins. To this end, we modified
position [A-T]:[L-T]:[B-T] = 2:1:2 (LTF = 1/5) produced Russian-doll mor- designs A and B to include a 5′ overhang, to which a protein-binding RNA
phologies with an A-rich outer shell, while single-phase condensates aptamer can connect via base pairing (Fig. 5a(i),(ii)). Nanostructures
were observed for LTFs ≥ 1/3. AYFP and BSTV were thus designed to connect to a yellow fluorescent pro-
The trends observed in condensate morphology are broadly con- tein (YFP)-binding aptamer (YFPapt)59 and a streptavidin (STV)-binding
sistent with a decrease in the interfacial tension between A- and B-rich aptamer (STVapt)60, respectively (Supplementary Fig. 47).
phases (γAB) with increasing LTF, as observed for equilibrium assembly For both designs, co-expressing the modified nanostars and their
of DNA nanostars57. Adhering de-mixed droplets (LTF = 1/21 to 1/11), partner aptamers from distinct templates in synthetic cells (Supple-
are indeed expected at equilibrium when γAB ≈ γA ≈ γB, where γA (γB) mentary Fig. 48) led to the target proteins (enhanced yellow fluores-
is the interfacial tension between the A-rich (B-rich) phase and the cent protein (EYFP) and Alexa Fluor 405 streptavidin (Alexa405-STV)
surrounding buffer2. The Russian-doll morphology (LTF = 1/5) should conjugate) readily partitioning within the formed MLOs (Fig. 5b(i),(ii),
emerge for γA < γAB < γB, with the evidence that γA < γB being consistent Supplementary Figs. 49 and 50, and Supplementary Video 13). Omit-
with trends seen in melting temperatures (Fig. 1c), while full mixing ting the aptamer led to the target proteins remaining in the lumen of
(LTF ≥ 1/3) should occur when γAB ≈ 0 (ref. 57). However, some mor- the synthetic cell. Protein partitioning was quantified through the
phological features, including the cavities seen for LTF = 1/11 and 1/7 parameter ξ, calculated as the ratio between the fluorescence inten-
and the outer layer of small B-rich domains seen for LTF = 1/5, are not sity of the protein recorded within or outside the MLOs (Supplemen-
expected at equilibrium, hinting that these may constitute metastable tary Methods 3). When protein-binding aptamers were present, the
states emerging from isothermal co-transcriptional assembly. median ξ was ~3.5 and ~2 for YFP and STV, respectively, while it fell
Confocal projections reveal a curved morphology for the clusters below 0.5 when no aptamers were present (Fig. 5c(i),(ii)). The sig-
formed at LTF = 1/21, ascribed to sedimentation within the W/O droplet nificant anti-partitioning noted in the absence of aptamers is prob-
(Supplementary Fig. 46). In all other conditions, the compact conden- ably a consequence of excluded volume interactions between RNA
sates appeared unaffected by substrate curvature. and proteins. Indeed, the condensate mesh size, estimated as twice
The abundance of linkers also influences the degree of mixing the RNA-nanostar arm length (~15.7 nm; Supplementary Methods 5),
between the two phases, which we quantified with indices JA and JB. is comparable with the hydrodynamic diameters of the STV (6.4 nm)
Index JA (JB) was computed as the ratio between the fluorescence inten- and EYFP (5 nm)61. Both STV and EYFP have mildly acidic to neutral
sity from nanostars A (B) in the B-rich (A-rich) phase over the signal in isoelectric points62,63, hence Coulomb repulsion towards RNA may also
the A-rich (B-rich) phase (Supplementary Methods 3). We observed enhance anti-partitioning.
limited mixing (JA, JB ≪ 1) for low LTF, followed by a moderate increase As an alternative strategy for STV capture, we replaced STVapt with
and by an abrupt jump to JA, JB ≈ 1 upon reaching the threshold for a biotinylated DNA oligonucleotide complementary to the overhang
complete mixing (Fig. 4c). A similarly sharp mixing transition was in BSTV (Fig. 5a(iii), Supplementary Figs. 49 and 50 and Supplementary
noted for DNA nanostars, remarkably occurring at similar fractions Video 13). With this approach, TexasRed streptavidin (TexasRed-STV)
of tetravalent linkers58. conjugate was distributed non-uniformly within the condensates,

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1670

Article https://doi.org/10.1038/s41565-024-01726-x

a (i) AYFP YFPapt EYFP (ii) BSTV STVapt Alexa405- (iii) BSTV BiotinDNA TexasRed-
STV STV

–YFPapt-T +YFPapt-T –STVapt-T +STVapt-T –BiotinDNA +BiotinDNA

b (i) (ii) (iii)

c n = 22/18 n = 11/28 n = 32/18

(i) (ii) (iii)
–YFPapt-T –STVapt-T –BiotinDNA

+YFPapt-T +STVapt-T +BiotinDNA

0 1 2 3 4 5 0 1 2 3 4 5 0 1 2 3 4 5
ξ ξ ξ

Fig. 5 | Selective protein capture in designer RNA MLOs. a, RNA-nanostar proteins are recruited in the MLOs. Scale bars, 50 μm (top) and 150 μm (bottom).
designs are modified to include a single-stranded 5′ overhang that can connect c, Protein partitioning parameter, ξ, calculated from confocal micrographs in b as
to a protein-binding moiety. Design AYFP is identical to nanostar A (Fig. 1) but can the ratio between the fluorescence signal of the target proteins recorded within
connect to YFP-binding aptamer (YFPapt) (i). Created with BioRender.com. Design and outside the MLOs for individual synthetic cells (Supplementary Methods
BSTV is identical to B but can connect to either an STV-binding aptamer (STVapt) 3). Data are shown for the three systems in a and b, with and without protein-
(ii) or a biotinylated DNA oligonucleotide (BiotinDNA) (iii). The templates of binding moieties. In the box plots, the central line marks the median, the box
protein-binding aptamers (YFPapt-T and STVapt-T) are transcribed in synthetic cells marks the Q1 (first quartile) – Q3 (third quartile) interquartile range (IQR) and the
alongside the associated RNA-nanostar templates (AYFP-T and BSTV-T, respectively), whiskers enclose data points within Q1 − 1.5 × IQR and Q3 + 1.5 × IQR. Data points
while pre-synthesized BiotinDNA is encapsulated alongside template BSTV-T. are relative to individual synthetic cells from a single field of view of one imaged
b, Diagrams (top) and confocal micrographs (bottom) of synthetic cells sample ((i), (ii) and (iii), left) or two fields of view from two technical replicates
expressing RNA MLOs in the absence (left) or in the presence (right) of protein- ((iii), right). All data points are shown except for four outliers with ξ > 5 in (i) and
binding moieties for the systems presented in a. When protein-binding aptamers (iii), omitted for ease of visualization.
are expressed or the biotinylated DNA oligonucleotide is included, target

forming irregularly shaped clusters with solid-like appearance aptamers might be embedded to recruit molecular guests, including
(Fig. 5b(iii) and Fig. 5c(iii)), reminiscent of multi-phase cellular conden- enzymes and metabolites, while ribozymes70 could confer catalytic
sates64–66. The non-uniform protein distribution probably results from properties to the MLOs. Because the condensates are transcribed from
the finite amount of the biotinylated DNA anchor available, which is all DNA templates, their formation could be controlled through standard
sequestered at early transcription stages and thus accumulates at the transcription regulation pathways in both synthetic cells and living
centre of the condensates. The solid-like look of the protein-rich mate- cells. Owing to their open-ended programmability, we expect that the
rial may be a consequence of the tetravalent STV strongly cross-linking RNA-nanostar condensates will constitute a valuable solution for the
multiple RNA nanostars, making the material more viscous. toolkit of synthetic biology.

Conclusion Online content

Our platform enables the expression of synthetic condensates and Any methods, additional references, Nature Portfolio reporting sum-
MLOs with prescribed size, number, morphology and composition, and maries, source data, extended data, supplementary information,
able to capture small guest molecules and proteins. The elementary acknowledgements, peer review information; details of author contri-
building blocks, RNA nanostars, were designed utilizing the rule-based butions and competing interests; and statements of data and code avail-
approaches of nucleic acid nanotechnology, which provide extensive ability are available at https://doi.org/10.1038/s41565-024-01726-x.
opportunities for updates aimed at programming arbitrary charac-
teristics of the designer MLOs. Among the many features that can be References
straightforwardly controlled are nanostar valency, flexibility and arm 1. Hyman, A. A., Weber, C. A. & Jülicher, F. Liquid-liquid phase
length, all known to predictably influence self-assembly in analogous separation in biology. Annu. Rev. Cell Dev. Biol. 30, 39–58 (2014).
DNA systems21,23,28,40,67. Control over condensate size could be achieved 2. Shin, Y. & Brangwynne, C. P. Liquid phase condensation in cell
by co-transcribing surface-passivating RNA constructs68,69. Further physiology and disease. Science 357, eaaf4382 (2017).

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1671

Article https://doi.org/10.1038/s41565-024-01726-x

3. Lyon, A. S., Peeples, W. B. & Rosen, M. K. A framework for 27. Filonov, G. S., Moon, J. D., Svensen, N. & Jaffrey, S. R. Broccoli:
understanding the functions of biomolecular condensates across rapid selection of an RNA mimic of green fluorescent protein
scales. Nat. Rev. Mol. Cell Biol. 22, 215–235 (2021). by fluorescence-based selection and directed evolution. J. Am.
4. Feric, M. et al. Coexisting liquid phases underlie nucleolar Chem. Soc. 136, 16299–16308 (2014).
subcompartments. Cell 165, 1686–1697 (2016). 28. Brady, R. A., Brooks, N. J., Foderà, V., Cicuta, P. & Di Michele, L.
5. Hirose, T., Ninomiya, K., Nakagawa, S. & Yamazaki, T. A guide Amphiphilic-DNA platform for the design of crystalline
to membraneless organelles and their various roles in gene frameworks with programmable structure and functionality.
regulation. Nat. Rev. Mol. Cell Biol. 24, 288–304 (2023). J. Am. Chem. Soc. 140, 15384–15392 (2018).
6. Gomes, E. & Shorter, J. The molecular language of membraneless 29. Hartvig, L. & Christiansen, J. Intrinsic termination of T7 RNA
organelles. J. Biol. Chem. 294, 7115–7127 (2019). polymerase mediated by either RNA or DNA. EMBO J. 15,
7. Molliex, A. et al. Phase separation by low complexity domains 4767–4774 (1996).
promotes stress granule assembly and drives pathological 30. Triana-Alonso, F. J., Dabrowski, M., Wadzack, J. & Nierhaus, K. H.
fibrillization. Cell 163, 123–133 (2015). Self-coded 3′-extension of run-off transcripts produces aberrant
8. Patel, A. et al. A liquid-to-solid phase transition of the ALS protein products during in vitro transcription with T7 RNA polymerase.
FUS accelerated by disease mutation. Cell 162, 1066–1077 J. Biol. Chem. 270, 6298–6307 (1995).
(2015). 31. Saleh, O. A., Jeon, B.-J. & Liedl, T. Enzymatic degradation of liquid
9. Jain, A. & Vale, R. D. RNA phase transitions in repeat expansion droplets of DNA is modulated near the phase boundary. Proc. Natl
disorders. Nature 546, 243–247 (2017). Acad. Sci. USA 117, 16160–16166 (2020).
10. Hastings, R. L. & Boeynaems, S. Designer condensates: a toolkit 32. Wadsworth, G. M. et al. RNAs undergo phase transitions with
for the biomolecular architect. J. Mol. Biol. 433, 166837 lower critical solution temperatures. Nat. Chem. 15, 1693–1704
(2021). (2023).
11. Dai, Y., You, L. & Chilkoti, A. Engineering synthetic biomolecular 33. Horiya, S. et al. RNA LEGO: magnesium-dependent formation of
condensates. Nat. Rev. Bioeng. 1, 466–480 (2023). specific RNA assemblies through kissing interactions. Chem. Biol.
12. Buddingh’, B. C. & van Hest, J. C. M. Artificial cells: synthetic 10, 645–654 (2003).
compartments with life-like functionality and adaptivity. Acc. 34. Van Treeck, B. & Parker, R. Emerging roles for intermolecular RNA–
Chem. Res. 50, 769–777 (2017). RNA interactions in RNP assemblies. Cell 174, 791–802 (2018).
13. Guo, H. et al. Spatial engineering of E. coli with addressable 35. Boeynaems, S. et al. Spontaneous driving forces give rise to
phase-separated RNAs. Cell 185, 3823–3837.e23 (2022). protein–RNA condensates with coexisting phases and complex
14. Xue, Z. et al. Targeted RNA condensation in living cells via material properties. Proc. Natl Acad. Sci. USA 116, 7889–7898 (2019).
genetically encodable triplet repeat tags. Nucleic Acids Res. 51, 36. Takahashi, S., Okura, H. & Sugimoto, N. Bisubstrate function of
8337–8347 (2023). RNA polymerases triggered by molecular crowding conditions.
15. Poudyal, R. R., Sieg, J. P., Portz, B., Keating, C. D. & Bevilacqua, P. C. Biochemistry 58, 1081–1093 (2019).
RNA sequence and structure control assembly and function of 37. Chamberlin, M. & Ring, J. Characterization of T7-specific
RNA condensates. RNA 27, 1589–1601 (2021). ribonucleic acid polymerase: I. General properties of the
16. Bevilacqua, P. C., Williams, A. M., Chou, H.-L. & Assmann, S. M. enzymatic reaction and the template specificity of the enzyme.
RNA multimerization as an organizing force for liquid–liquid J. Biol. Chem. 248, 2235–2244 (1973).
phase separation. RNA 28, 16–26 (2022). 38. Arnold, S. et al. Kinetic modeling and simulation of in vitro
17. Seeman, N. C. & Sleiman, H. F. DNA nanotechnology. Nat. Rev. transcription by phage T7 RNA polymerase. Biotechnol. Bioeng.
Mater. 3, 17068 (2017). 72, 548–561 (2001).
18. Geary, C., Rothemund, P. W. K. & Andersen, E. S. A single-stranded 39. Akama, S., Yamamura, M. & Kigawa, T. A multiphysics model
architecture for cotranscriptional folding of RNA nanostructures. of in vitro transcription coupling enzymatic reaction and
Science 345, 799–804 (2014). precipitation formation. Biophys. J. 102, 221–230 (2012).
19. McRae, E. K. S. et al. Structure, folding and flexibility of 40. Agarwal, S., Osmanovic, D., Klocke, M. A. & Franco, E. The growth
co-transcriptional RNA origami. Nat. Nanotechnol. 18, 808–817 rate of DNA condensate droplets increases with the size of
(2023). participating subunits. ACS Nano 16, 11842–11851 (2022).
20. Stewart, J. M. et al. Modular RNA motifs for orthogonal phase 41. Han, K. Y., Leslie, B. J., Fei, J., Zhang, J. & Ha, T. Understanding
separated compartments. Preprint at bioRxiv https://doi. the photophysics of the spinach-DFHBI RNA aptamer-fluorogen
org/10.1101/2023.10.06.561123 (2023). complex to improve live-cell RNA imaging. J. Am. Chem. Soc. 135,
21. Biffi, S. et al. Phase behavior and critical activated dynamics of 19033–19038 (2013).
limited-valence DNA nanostars. Proc. Natl Acad. Sci. USA 110, 42. Li, X., Kim, H., Litke, J. L., Wu, J. & Jaffrey, S. R. Fluorophore-
15633–15637 (2013). promoted RNA folding and photostability enables imaging of
22. Sato, Y., Sakamoto, T. & Takinoue, M. Sequence-based single broccoli-tagged mRNAs in live mammalian cells. Angew.
engineering of dynamic functions of micrometer-sized DNA Chem. Int. Ed. 59, 4511–4518 (2020).
droplets. Sci. Adv. 6, eaba3471 (2020). 43. Levitz, P. Off-lattice reconstruction of porous media: critical
23. Brady, R. A., Brooks, N. J., Cicuta, P. & Di Michele, L. Crystallization evaluation, geometrical confinement and molecular transport.
of amphiphilic DNA C-stars. Nano Lett. 17, 3276–3281 (2017). Adv. Colloid Interface Sci. 76-77, 71–106 (1998).
24. Liu, D. et al. Branched kissing loops for the construction of diverse 44. Testard, V., Berthier, L. & Kob, W. Influence of the glass transition
RNA homooligomeric nanostructures. Nat. Chem. 12, 249–259 on the liquid-gas spinodal decomposition. Phys. Rev. Lett. 106,
(2020). 125702 (2011).
25. Kolpashchikov, D. M. Binary malachite green aptamer for 45. Di Michele, L. et al. Multistep kinetic self-assembly of DNA-coated
fluorescent detection of nucleic acids. J. Am. Chem. Soc. 127, colloids. Nat. Commun. 4, 2007 (2013).
12442–12443 (2005). 46. Do, S., Lee, C., Lee, T., Kim, D.-N. & Shin, Y. Engineering DNA-based
26. Yerramilli, V. S. & Kim, K. H. Labeling RNAs in live cells using synthetic condensates with programmable material properties,
malachite green aptamer scaffolds as fluorescent probes. ACS compositions, and functionalities. Sci. Adv. 8, eabj1771
Synth. Biol. 7, 758–766 (2018). (2022).

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1672

Article https://doi.org/10.1038/s41565-024-01726-x

47. Sato, Y. & Takinoue, M. Sequence-dependent fusion dynamics and 62. Diamandis, E. P. & Christopoulos, T. K. The biotin-(strept)avidin
physical properties of DNA droplets. Nanoscale Adv. 5, 1919–1925 system: principles and applications in biotechnology. Clin. Chem.
(2023). 37, 625–636 (1991).
48. Jeon, B.-j et al. Salt-dependent properties of a coacervate-like, 63. Whitaker, M. Genetically Encoded Probes for Measurement of
self-assembled DNA liquid. Soft Matter 14, 7009–7015 (2018). Intracellular Calcium Vol. 99 (ed. Whitaker, M.) Ch. 6 (Academic
49. Brangwynne, C. P., Mitchison, T. J. & Hyman, A. A. Active liquid-like Press, 2010).
behavior of nucleoli determines their size and shape in Xenopus 64. Yamazaki, T., Nakagawa, S. & Hirose, T. Architectural RNAs for
laevis oocytes. Proc. Natl Acad. Sci. USA 108, 4334–4339 (2011). membraneless nuclear body formation. Cold Spring Harb. Symp.
50. Wang, H., Kelley, F. M., Milovanovic, D., Schuster, B. S. & Shi, Z. Quant. Biol. 84, 227–237 (2019).
Surface tension and viscosity of protein condensates quantified 65. Schuster, B. S. et al. Biomolecular condensates: sequence
by micropipette aspiration. Biophys. Rep. 1, 100011 (2021). determinants of phase separation, microstructural organization,
51. Atha, D. H. & Ingham, K. C. Mechanism of precipitation of proteins enzymatic activity, and material properties. J. Phys. Chem. B 125,
by polyethylene glycols. Analysis in terms of excluded volume. 3441–3451 (2021).
J. Biol. Chem. 256, 12108–12117 (1981). 66. Yewdall, N. A., André, A. A. M., Lu, T. & Spruijt, E. Coacervates as
52. Asakura, S. & Oosawa, F. On interaction between two bodies models of membraneless organelles. Curr. Opin. Colloid Interface
immersed in a solution of macromolecules. J. Chem. Phys. 22, Sci. 52, 101416 (2021).
1255–1256 (1954). 67. Brady, R. A., Kaufhold, W. T., Brooks, N. J., Foderà, V. & Di Michele, L.
53. Lu, P. J. et al. Gelation of particles with short-range attraction. Flexibility defines structure in crystals of amphiphilic DNA
Nature 453, 499–503 (2008). nanostars. J. Phys. Condens. Matter 31, 074003 (2019).
54. Hansch, C., Leo, A. & Hoekman, D. Exploring QSAR: Hydrophobic, 68. Walczak, M. et al. Responsive core–shell DNA particles trigger
Electronic, and Steric Constants Vol. 2 (American Chemical lipid-membrane disruption and bacteria entrapment. Nat.
Society, 1995). Commun. 12, 4743 (2021).
55. Franco, E. et al. Timing molecular motion and production with 69. Walczak, M. et al. A synthetic signalling network imitating the
a synthetic transcriptional clock. Proc. Natl Acad. Sci. USA 108, action of immune cells in response to bacterial metabolism. Adv.
E784–93 (2011). Mater. 35, e2301562 (2023).
56. Conrad, N., Chang, G., Fygenson, D. K. & Saleh, O. A. Emulsion 70. Tanner, N. K. Ribozymes: the characteristics and properties of
imaging of a DNA nanostar condensate phase diagram reveals catalytic RNAs. FEMS Microbiol. Rev. 23, 257–275 (1999).
valence and electrostatic effects. J. Chem. Phys. 157, 234203
(2022). Publisher’s note Springer Nature remains neutral with regard to
57. Jeon, B.-J., Nguyen, D. T. & Saleh, O. A. Sequence-controlled jurisdictional claims in published maps and institutional affiliations.
adhesion and microemulsification in a two-phase system of DNA
liquid droplets. J. Phys. Chem. B 124, 8888–8895 (2020). Open Access This article is licensed under a Creative Commons
58. Gong, J., Tsumura, N., Sato, Y. & Takinoue, M. Computational DNA Attribution 4.0 International License, which permits use, sharing,
droplets recognizing mirna sequence inputs based on liquid– adaptation, distribution and reproduction in any medium or format,
liquid phase separation. Adv. Funct. Mater. 32, 2202322 as long as you give appropriate credit to the original author(s) and the
(2022). source, provide a link to the Creative Commons licence, and indicate
59. Shui, B. et al. RNA aptamers that functionally interact with green if changes were made. The images or other third party material in this
fluorescent protein and its derivatives. Nucleic Acids Res. 40, e39 article are included in the article’s Creative Commons licence, unless
(2012). indicated otherwise in a credit line to the material. If material is not
60. Leppek, K. & Stoecklin, G. An optimized streptavidin-binding included in the article’s Creative Commons licence and your intended
RNA aptamer for purification of ribonucleoprotein complexes use is not permitted by statutory regulation or exceeds the permitted
identifies novel ARE-binding proteins. Nucleic Acids Res. 42, e13 use, you will need to obtain permission directly from the copyright
(2014). holder. To view a copy of this licence, visit http://creativecommons.
61. Fleming, P. J. & Fleming, K. G. HullRad: fast calculations of org/licenses/by/4.0/.
folded and disordered protein and nucleic acid hydrodynamic
properties. Biophys. J. 114, 856–869 (2018). © The Author(s) 2024

Nature Nanotechnology | Volume 19 | November 2024 | 1665–1673 1673

Article https://doi.org/10.1038/s41565-024-01726-x

Methods Sigma-Aldrich (CalBioChem), already resuspended at 1 mg ml−1 in

Sequence design 50 mM bicarbonate-borate buffer, 0.9% NaCl, 5 mg ml−1 BSA, pH 8.1,
Four-armed RNA nanostars with 25 base pair (bp)-long arms separated and stored at 4°C. Alexa405-STV conjugate was purchased from Inv-
by one unpaired uracil residue were designed starting from the itrogen, resuspended at 1 mg ml−1 in PBS, pH 7.2 (Gibco) with addition
sequences of previously reported RNA junctions71 and fluorogenic MG25,26 of 5 mM sodium azide (0.1 M solution, Sigma-Aldrich), and stored at
and Broccoli27 RNA aptamers. Three main nanostar variants, namely A, −20 °C. The dsDNA ladder for electrophoresis (GeneRuler Ultra Low
B and C, were designed to bind identical motifs, while being Range DNA Ladder) was purchased from Thermo Scientific. The ssRNA
non-interacting towards each other. Intra-population interactions were ladder (RiboRuler Low Range ssRNA Ladder) and 2× RNA Gel Loading
guaranteed by palindromic KLs. KL A (5′-AUCGCGAAA-3′) was adapted Dye were purchased from ThermoFisher.
from the KL domain in the bottom right arm of the T4 tetrahedron in ref.
71, by making it palindromic. Asymmetrical flanking bases (5′-A…AA-3′) PCR amplification and purification of DNA templates
were introduced due to their presence, albeit in reverse order, in the KL Amplification of gBlock DNA templates was carried out using Q5
domains found in the Lai variant of the human immunodeficiency virus-1 High-Fidelity DNA Polymerase (New England Biolabs). PCR mix-
dimerization initiation sequence (HIV-1 DIS)72,73. KLs B and C were tures were prepared on ice, with 10 ng DNA template per mixture,
designed to match the interaction strength of KL A. This was achieved and annealed in a Bio-Rad C1000 Touch Thermal Cycler according
by computationally generating, using Python3, all possible palindromic, to the following protocol: pre-heating at 98 °C; initial denaturation
6 nt sequences with the same GC content as KL A. The resulting set was at 98 °C for 30 s; 30 amplification cycles (98 °C for 10 s, 64–65 °C for
filtered to exclude all sequences with more than two overlapping nucleo- 20 s depending on primer melting temperature, 72 °C for 7 s); final
tides with KL A. Among these, KL C was selected as 5′-AGGUACCAA-3′. extension at 72 °C for 2 min; hold at 4 °C. Samples were stored at 4 °C
To determine KL B, the constraint was relaxed to allow a maximum of 3 and gel-purified within 1 week. Purification was carried out using a 2%
overlapping nucleotides with both KL A and KL C. Among these, KL B w/v agarose (Sigma-Aldrich) gel, prepared in Tris-borate-EDTA (TBE) 1×
was chosen as 5′-AGUCGACAA-3′—a sequence similar to the Mal variant buffer (from TBE 10×, Thermo Scientific) with GelRed nucleic acid gel
of the HIV-1 DIS KL (5′-GUGCAC-3′)72. Design C̃ is similar to C and shows stain (3×, Biotium) and run at 120 V for 90 min (Supplementary Fig. 2).
the same KL domains, but lacks FLAPs. The minimum free energy con- Gels were imaged using a Syngene G:BOX Chemi XRQ gel documenta-
figuration of all designs was evaluated using NUPACK (default Serra and tion system. One PCR reaction (50 μl + 10 μl TriTrack DNA Loading Dye
Turner, 1995 parameters and 1 M Na+)74,75. Kinefold76 was used to test 6×, Thermo Scientific) was loaded in each well, and bands were cut using
co-transcriptional folding. All designs were tested via batch jobs, and a scalpel under the ultraviolet illumination. Gel bands were loaded in
further considered only if the helix tracing graph showed correct folding pairs in 2 ml Eppendorf tubes, treated by adding 4 μl of Monarch Gel
order and reasonable stability of each helix. All nanostar sequences are Dissolving Buffer (New England Biolabs) per mg of gel, and incubated at
provided in Supplementary Table 1. Sequences for the coding/ 50 °C until complete dissolution. The obtained mixtures were purified
non-template DNA strands were obtained by adding a prefix comprising using the Monarch PCR & DNA Cleanup Kit (New England Biolabs). Elu-
the T7 promoter (5′-TTCTAATACGACTCACTATA-3′, 17 nt consensus T7 tion was performed with 12–14 μl per gel-band pair. The concentration
promoter underlined) to the equivalent DNA sequence of the RNA nano- of purified DNA templates was determined by measuring absorbance at
structures71 (Supplementary Table 2). DNA primers for PCR amplification 260 nm (average of 3 repeated measurements) using a NanoDrop One.
of the templates were designed aiming for 40−60% GC content and
length between 18 nt and 26 nt (Supplementary Table 3). DNA primers RNA transcription in bulk
were verified using NUPACK74 and the NEB melting temperature calcula- Transcription was carried out using the CellScript T7-FlashScribe Tran-
tor (https://tmcalculator.neb.com/#!/main) for use with Q5 High-Fidelity scription Kit. The final reaction mixture contained T7 RNAP in the proprie-
DNA Polymerase under standard 500 nM primer concentration. tary transcription buffer, complemented by 9 mM of each ribonucleotide
triphosphate, 0.05 units per μl of RNase inhibitor and 10 mM dithiothrei-
Materials tol. Unless otherwise specified, DNA templates were added to an overall
DNA primers were purchased from and purified by Integrated DNA concentration of 40 nM and both MG and DFHBI dyes were added to
Technologies via standard desalting. Unless otherwise specified, all transcription mixtures (even samples lacking the corresponding
dsDNA templates were purchased from Integrated DNA Technologies as aptamer) in proportions equal to 1 μl of 1 mM dye:20 μl mixture, yield-
gBlocks. Only the shorter STVapt-T was purchased as a double-stranded ing a final concentration of approximately 45.45 μM for each dye. Unless
Ultramer. All DNA strands were received lyophilized and reconsti- otherwise specified, samples were loaded in rectangular glass capillaries
tuted at 100 μM. DNA primers were reconstituted in nuclease-free (either 0.20 mm × 4.00 mm × 50 mm or 0.40 mm × 4.00 mm × 50 mm,
water (UltraPure DNase/RNase-free distilled water, Invitrogen), VitroCom) sealed and glued on a glass coverslips (24 mm × 60 mm, Men-
while gBlock DNA templates were reconstituted in syringe-filtered zel Gläser) via a 2-component epoxy (Araldite Rapid). To avoid the glue
TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0), obtained by diluting coming in contact with the sample, the sides of the capillary were capped
Tris-EDTA 100× (Sigma-Aldrich) in nuclease-free water. DNA primer with mineral oil. Glue was allowed to set for 30 min, during which samples
concentration was determined by measuring absorbance at 260 nm were kept in a dark environment at room temperature.
(average of 5 repeated measurements) using a Thermo Scientific
Nanodrop One Microvolume ultraviolet–visible spectrophotometer RNA transcription in synthetic cells and protein capture
using extinction coefficients provided by the supplier. Primers were Synthetic cells were generated by encapsulating the in vitro transcrip-
then diluted to 10 μM in nuclease-free water as per PCR kit instruc- tion mixture described above within W/O droplets77. Briefly, 22–23 μl of
tions. MG chloride and DFHBI were purchased from Sigma-Aldrich. transcription mixture were added on top of 90 μl of 2% w/w Pico-Surf
The as-received powders were dissolved in nuclease-free water and (Sphere Fluidics), a biocompatible surfactant, in Novec 7500, a fluori-
DMSO to produce 1 mM and 10 mM stock solutions, respectively. nated oil, within an Eppendorf tube. The resulting mixture was vor-
The 10 mM DFHBI solution was then diluted to 1 mM in nuclease-free texed at 2,500 rpm for 30 s and then left to equilibrate for 1–2 min
water. The resulting 1 mM MG and DFHBI solutions were stored at before extracting the top layer containing the synthetic cells. For
4 °C and −20 °C, respectively. Recombinant EYFP was purchased from protein capture experiments, unless otherwise specified, the volume
RayBiotech, resuspended at 0.67 mg ml−1 in nuclease-free water and of transcription mixtures was increased to 23 μl to accommodate
stored at −80 °C. EYFP was used within 2 days from reconstitution to EYFP, TexasRed-STV or Alexa405-STV, each at a final concentration
prevent aggregation. TexasRed-STV conjugate was purchased from of 1.25 μM. For control experiments in the absence of target proteins

Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01726-x

the transcription mixture volume was kept to 22 μl. In assays relying a Hamamatsu Orca-Flash 4.0 v3 camera, and a Lumencor SPECTRA X
on protein-binding aptamers (YFPapt, STVapt), the total DNA template light-emitting diode engine. The following SPECTRA X light-emitting
concentration was kept equal to 40 nM, and the composition ratio was diodes were used to excite the corresponding fluorophores or FLAPs:
chosen to be [nanostar DNA template]/[aptamer DNA template] = 3. 395 nm for Alexa405-STV, 470 nm for DFHBI/BrA, 550 nm for EYFP,
For TexasRed-STV capture assays via BiotinDNA, [BSTV-T] was similarly 575 nm for TexasRed-STV, and 640 nm for MG/MGA. Samples, enclosed
kept at 30 nM, while [BiotinDNA] was chosen to be 10 μM, yielding an in glass capillaries and glued to a microscope coverslip (see above),
approximately 8× excess of biotin compared with streptavidin. MG was were taped to a Peltier-controlled copper temperature stage (Temikra).
omitted in assays including TexasRed-STV due to fluorescence emission Imaging was automated via the ND acquisition module of Nikon’s NIS
overlap. Unless otherwise specified, samples were loaded in capillaries software, with PFS enabled to ensure constant z-height during the
as for bulk samples, but omitting mineral oil capping. time-lapse. Three non-overlapping fields of view per capillary (sam-
ple) were imaged. Epifluorescence z-stacks (120–160 μm z-height,
Effect of buffer exchange on condensate stability distributed from −20/−60 μm to +100 μm around the PFS plane, with a
For buffer-exchange experiments (Supplementary Fig. 11), bulk tran- 3.5-4 μm step depending on the run) were captured at each timepoint.
scription samples (20 μl per sample) were prepared as described above For condensate formation time-lapses, both in bulk and in syn-
and loaded in 384-well plates (black, Greiner Bio-One) to enable buffer thetic cells, the temperature was set to 30 °C. Samples were imaged
exchange. The DNA template concentration was reduced to 2 nM to every 15 min for 10 h, and every 30 min for further 38 h. For binary and
account for the reduced bottom-surface-to-volume ratio of the micro- ternary systems in bulk, automated acquisition terminated at 42 h and
plate wells compared with the capillary chambers, avoiding the forma- data at 48 h were collected manually.
tion of extremely large condensates upon sedimentation. Microplates For melting-temperature determination experiments (Supple-
were sealed with an adhesive aluminium film. Samples were incubated mentary Fig. 8), temperature was set to 25 °C and increased by 1 °C
using a custom-made microplate heated stage with temperature set at every 15 min up to 75 °C. Samples were imaged after 10 min of hold at
30 °C for the plate chamber and at 35 °C for the lid, and then imaged each temperature.
after 24 h. Samples underwent buffer exchange with either PBS 1× Due to the time required for sample preparation and set-up, imag-
pH 7.4 (diluted from 10× PBS, Invitrogen), TE 1× (diluted from 100× TE, ing started 1–2 h after mixing the DNA templates with the rest of the
ThermoFisher) supplied with 5 mM MgCl2 (Sigma-Aldrich) pH 8.0 or transcription mixtures (Supplementary Tables 5 and 6). Micrographs
TE 1× supplied with 10 mM MgCl2 pH 8.0 via four consecutive washes, and videos have been labelled according to imaging time, with time
separated by 30 min intervals, during which the samples were kept at 0 referring to the start of the imaging run, rather than to the start of
30 °C. For the first wash, 70 μl of the desired buffer were added to the transcription. When comparing samples from different runs in the
sample well. For remaining washes, 70 μl supernatant were removed same figures, timepoints have been aligned to reflect any delays in the
before adding an equal volume of fresh buffer. Samples were imaged run starting times (as in Fig. 3b, and Supplementary Figs. 44, 45, 49 and
after the final wash and after an additional 24 h incubation at 30 °C. 50). Conversely, videos have all been labelled independently, that is,
relative to the start of their specific run.
Effect of crowding agents
To test the effect of crowding agents (Supplementary Fig. 12), bulk Confocal imaging
samples were prepared as described above. To reach the indicated final As well as FRAP experiments, laser scanning confocal imaging was per-
volume, samples were supplemented with either RNase-free water (for formed on a Leica Stellaris 8 (DMi8 CS Premium) inverted microscope.
control samples, ‘no PEG’) or PEG 200 (Sigma-Aldrich) at a final concen- The microscope was equipped with a solid-state 405 nm laser as well
tration of 25% v/v. Samples were incubated in a Bio-Rad C1000 Touch as a white-light laser (440–790 nm). The following objective lenses
Thermal Cycler at 30 °C (heated lid at 35 °C) for 24 h before imaging. were used: HC PL APO CS2 10× DRY (NA 0.40) and HC PL APO CS2 20×
(NA 0.75) DRY. For Alexa405-STV, the 405 nm laser was used and emis-
FRAP sion was recorded around 421 nm. The white-light laser was used for
For FRAP experiments conducted on FLAPs (Supplementary Fig. 18(i)), all other dyes with the following excitations/emission wavelengths:
bulk samples were prepared as described above. For FRAP conducted DFHBI/BrA, 447/501 nm; EYFP, 514/527 nm; TexasRed-STV, 595/615 nm;
using the covalently linked fluorescein-12-UTPs (Supplementary MG/MGA, 628/650–660 nm. A line-sequential illumination mode was
Fig. 18(ii)), bulk samples were prepared as described above with adopted; for example, bright-field, 405 nm and 628 nm in sequence
the difference that uridine triphosphate (UTP) concentration was 1, 447 nm in sequence 2, 514 nm in sequence 3. The pinhole was set
reduced from from 9 mM to 8.95 mM and 50 μM of fluorescein-12-UTP to 1 airy unit. Line-averaging was enabled and set to 2–3. The scan
(Sigma-Aldrich) were included for labelling, corresponding to ~0.55% of mode along the x direction was selected to be bidirectional after phase
the total UTP content. MG and DFBHI were not included. Samples were calibration. The scanning speed was set to 400 Hz for high-resolution
incubated in a Bio-Rad C1000 Touch Thermal Cycler at 30 °C (heated stills (4,096 px × 4,096 px or 8,192 px × 8,192 px) captured with the 10×
lid at 35 °C) for 24 h before imaging. or 20× lens, and 1,000 Hz for zoomed-in z-stacks acquired with the
20× lens. For the latter, the zoom factor was tuned to select a single
Fluorimetry condensate or droplet, and top/bottom planes were manually tuned,
For fluorimetry experiments in bulk and in synthetic cells, including with a Nyquist optimized z-step. Unless otherwise stated, all reported
excitation/emission scans reported (Supplementary Fig. 1) and kinet- confocal micrographs are pristine. Two-dimensional orthogonal
ics assays (Supplementary Figs. 13–15, 31–33, 40 and 43), samples cross-sections (XY, XZ, YZ) and volume three-dimensional renderings
were prepared as discussed above and loaded in transparent UV-Star (Supplementary Fig. 46) were generated from z-stacks using ‘Sections’
384-well plates (Greiner Bio-One). in the three-dimensional module of Leica Application Suite (LAS) X.
Clipping of three-dimensional renderings in Supplementary Video 11
Epifluorescence imaging were produced via the ‘Movie Editor’ in LAS X, converted to AVI using
Time-lapse epifluorescence imaging of RNA transcription, in bulk and ffmpeg, collated in FIJI and finally re-exported in MP4 using Permute3.
in synthetic cells, was performed on a Nikon Eclipse Ti2-E inverted
microscope with Perfect Focus System (PFS), equipped with Plan Gel electrophoresis on RNA transcripts
Apochromat λ 10× (numerical aperture (NA) 0.45, working distance Samples were prepared following the bulk transcription protocol
(WD) 4,000 μm) and λ 20× (NA 0.75, WD 1,000 μm) objective lenses, described above, but reducing the sample volume to 10 μl. No MG

Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01726-x

or DFHBI dyes were added. The samples were incubated in a Bio-Rad Code availability
C1000 Touch Thermal Cycler at 30 °C (heated lid at 35 °C) for 16–18 h, The data analysis code used in this publication is available at https://
and then treated with 0.5 μl of the DNase I solution provided with the github.com/ld389/Fabrini-2024-Nat-Nanotechnol.
transcription kit (1 unit per μl) for 30 min at 37 °C (heated lid at 40 °C).
For non-denaturing agarose gel electrophoresis (AGE) (Sup- References
plementary Fig. 4), after incubation, samples were diluted 3× with 71. Li, M. et al. In vivo production of RNA nanostructures via
RNase-free water and mixed with 3 μl of 60% glycerol (used instead programmed folding of single-stranded RNAs. Nat. Commun. 9,
of loading dyes). Six microlitres of each diluted sample was loaded 2196 (2018).
into the gel wells. Ten microlitres RiboRuler Low Range ssRNA Lad- 72. Ennifar, E., Walter, P., Ehresmann, B., Ehresmann, C. & Dumas, P.
der (ThermoFisher) was loaded in the leftmost lane. For the control Crystal structures of coaxially stacked kissing complexes of the
DNA nanostars (Supplementary Table 4), 10 μl of 0.7 μM DNA sample HIV-1 RNA dimerization initiation site. Nat. Struct. Biol. 8,
(annealed in the Bio-Rad C1000 Touch Thermal Cycler) were loaded 1064–1068 (2001).
in the rightmost lane. Agarose gels were prepared at 2% w/v agarose 73. Mundigala, H., Michaux, J. B., Feig, A. L., Ennifar, E. & Rueda, D.
(Sigma-Aldrich) in TBE 1× buffer (from TBE 10×, Thermo Scientific), HIV-1 DIS stem loop forms an obligatory bent kissing intermediate in
with the addition of GelRed nucleic acid gel stain (3×, Biotium). Gels the dimerization pathway. Nucleic Acids Res. 42, 7281–7289 (2014).
were run at 120 V for 45 min. 74. Zadeh, J. N. et al. NUPACK: analysis and design of nucleic acid
For denaturing polyacrylamide gel electrophoresis (PAGE) (Supple- systems. J. Comput. Chem. 32, 170–173 (2011).
mentary Fig. 3), RNA transcripts were prepared and treated with DNase I 75. Serra, M. J. & Turner, D. H. Predicting thermodynamic properties
as for native AGE, then diluted 50× with RNase-free water. Sample (4 μl) of RNA. Methods Enzymol. 259, 242–261 (1995).
and RiboRuler Low Range ssRNA Ladder (2.5 μl) aliquots were each 76. Xayaphoummine, A., Bucher, T. & Isambert, H. Kinefold web server
mixed with equal volumes of 2× RNA Gel Loading Dye (ThermoFisher), for RNA/DNA folding path and structure prediction including
before incubation in a Bio-Rad C1000 Touch Thermal Cycler at 70 °C pseudoknots and knots. Nucleic Acids Res. 33, W605–10 (2005).
(heated lid at 80 °C) for 10 min. All samples were promptly transferred 77. Agarwal, S., Klocke, M. A., Pungchai, P. E. & Franco, E. Dynamic
on ice before loading onto the gel. Eight microlitres of each sample and self-assembly of compartmentalized DNA nanotubes. Nat.
5 μl of RiboRuler Low Range ssRNA Ladder were loaded into the wells of Commun. 12, 3557 (2021).
a 7 M urea, 8% polyacrylamide gel. The gels were prepared by combining
3.7 ml 30% w/v acrylamide/bis-acrylamide partitioned solution (29:1, Acknowledgements
Sigma-Aldrich), 4.0 ml 10× TBE buffer (ThermoFisher), and 6.3 g urea L.D.M. and N.F. acknowledge support from the European Research
(Sigma-Aldrich), to achieve a mixture with 7 M final urea concentration. Council (ERC) under the Horizon 2020 research and innovation
To obtain a 1.5 mm 8% polyacrylamide gel, this volume was adjusted to programme (ERC-STG No 851667 - NANOCELL) and a Royal Society
14 ml using UltraPure RNase-free water in a 50 ml Falcon tube. Following University Research Fellowship (UF160152, URF\R\221009). G.F.
the addition of 75 μl of 10% w/v ammonium persulfate (Sigma-Aldrich) acknowledges funding from the Department of Chemistry at Imperial
and 15 μl N,N,N′,N′-tetramethylethylenediamine (Sigma-Aldrich), the College London. M.D.A. acknowledges support from a Biotechnology
tube was inverted several times. The final mixture was allowed to polym- and Biological Sciences Research Council (BBSRC) David Phillips
erize for 20 min in an assembled gel electrophoresis cassette (Bio-Rad). Fellowship (BB/R011605/1) and a Lister Institute Research Prize.
Gels were run at 150 V for 60 min, followed by post-staining with 1× SYBR S.P.N. acknowledges support from the Engineering and Physical
Gold Nucleic Acid Gel Stain (diluted from 10,000× concentrate in DMSO, Sciences Research Council (EPSRC) (EP/S023518/1). J.M.S. is a Merck
ThermoFisher) in 1× TBE buffer for 15 min. Awardee of the Life Sciences Research Foundation. E.F. acknowledges
Gels were imaged using a Syngene G:BOX Chemi XRQ gel docu- support from the Alfred Sloan Foundation through award G-2021-
mentation system. 16831, and from the US NSF through CAREER award 1938194, FMRG:
Bio award 2134772, and BBSRC-NSF/BIO award 2020039. E.F. also
Statistics and reproducibility acknowledges funding from the UCLA Eli and Edythe Broad Center of
No statistical method was used to determine sample size. The experi- Regenerative Medicine and Stem Cell Research Rose Hills Foundation
ments were not randomized. The investigators were not blinded to Innovator Grant. P.W.K.R. acknowledges support from the Alfred
allocation during experiments and outcome assessment. No data were Sloan Foundation through award G-2021-16831 and from the US
excluded except (in limited cases) when removing artefacts of incorrect NSF through FMRG: Bio award 2134772. R.M. acknowledges funding
segmentation of microscopy images, as specified in Supplementary from the EPSRC Centre for Doctoral Training in Nanoscience and
Methods 2 and 3. Several control experiments were executed (Sup- Nanotechnology (NanoCDT, EP/S022953/1). G.F., L.D.M. and M.D.A.
plementary Information) and found to be consistent. Information on acknowledge the Facility for Imaging by Light Microscopy (FILM)
repeats is provided in the relevant figure captions. No reproducibility at Imperial College London and thank S. Rothery for his invaluable
issues emerged. assistance and the publicly released FIJI macros he developed.

Reporting summary Author contributions

Further information on research design is available in the Nature G.F. and L.D.M. designed the research. G.F. performed the
Portfolio Reporting Summary linked to this article. experiments, aided by N.F. and R.M. S.P.N. developed and performed
the molecular biology protocols. G.F. analysed the data, aided by
Data availability N.F. and L.D.M. G.F. developed the code for analysis, aided by L.D.M.
The raw data underpinning this publication are available, free of charge, G.F. and L.D.M. wrote the paper, aided by E.F. and with input from all
at https://doi.org/10.17863/CAM.108563. For large microscopy datasets other authors. All authors contributed to critical discussion and data
(time-lapses, z-stacks), a representative selection of all the data is pro- interpretation. L.D.M. supervised the research.
vided after binning and time downsampling due to space limitations
on the repository. The full dataset is available from the correspond- Competing interests
ing author. Oligonucleotide sequences generated for this work are E.F., S.L., A.A.T., G.F. and L.D.M., through the Regents of University
provided in the Supplementary Tables 1–4. Source data are provided of California, have filed a patent application in the US Patent and
with this paper. Trademark Office, which includes disclosure of inventions described

Nature Nanotechnology
Article https://doi.org/10.1038/s41565-024-01726-x

in this paper, provisional application serial number 63/588,142, filed Correspondence and requests for materials should be addressed to
on 5 October 2023, and entitled ‘Single stranded RNA motifs for Lorenzo Di Michele.
in vitro co-transcriptional production of orthogonal phase separated
condensates’. The other authors declare no competing interests. Peer review information Nature Nanotechnology thanks Zoher
Gueroui and the other, anonymous, reviewer(s) for their contribution
Additional information to the peer review of this work.
Supplementary information The online version
contains supplementary material available at Reprints and permissions information is available at
https://doi.org/10.1038/s41565-024-01726-x. www.nature.com/reprints.

Nature Nanotechnology
 nature portfolio | reporting summary
Corresponding author(s): Lorenzo Di Michele
Last updated by author(s): Jun 16, 2024

Reporting Summary
Nature Portfolio wishes to improve the reproducibility of the work that we publish. This form provides structure for consistency and transparency
in reporting. For further information on Nature Portfolio policies, see our Editorial Policies and the Editorial Policy Checklist

Statistics
For all statistical analyses, confirm that the following items are present in the figure legend, table legend, main text, or Methods section.
n/a Confirmed
The exact sample size (n) for each experimental group/condition, given as a discrete number and unit of measurement
A statement on whether measurements were taken from distinct samples or whether the same sample was measured repeatedly
The statistical test(s) used AND whether they are one- or two-sided
Only common tests should be described solely by name; describe more complex techniques in the Methods section.

A description of all covariates tested

A description of any assumptions or corrections, such as tests of normality and adjustment for multiple comparisons
A full description of the statistical parameters including central tendency (e.g. means) or other basic estimates (e.g. regression coefficient)
AND variation (e.g. standard deviation) or associated estimates of uncertainty (e.g. confidence intervals)

For null hypothesis testing, the test statistic (e.g. F, t, r) with confidence intervals, effect sizes, degrees of freedom and P value noted
Give P values as exact values whenever suitable.

For Bayesian analysis, information on the choice of priors and Markov chain Monte Carlo settings
For hierarchical and complex designs, identification of the appropriate level for tests and full reporting of outcomes
Estimates of effect sizes (e.g. Cohen's d, Pearson's r),
), indicating how they were calculated
Our web collection on statistics for biologists contains articles on many of the points above.

Software and code

Policy information about availability of computer code
Data collection Leica AS X 5.2.1, Nikon NIS Elements V5.2, BMG Clariostar Software MARS V3.41

Data analysis Python3, Matlab2023b

For manuscripts utilizing custom algorithms or software that are central to the research but not yet described in published literature, software must be made available to editors and
reviewers. We strongly encourage code deposition in a community repository (e.g. GitHub). See the Nature Portfolio guidelines for submitting code & software for further information.

Data
Policy information about availability of data
All manuscripts must include a data availability statement This statement should provide the following information, where applicable:
- Accession codes, unique identifiers, or web links for publicly available datasets
- A description of any restrictions on data availability
- For clinical datasets or third party data, please ensure that the statement adheres to our policy
April 2023

Raw data underpinning this publication are available, free of charge, at https://doi.org/10.17863/CAM.108563. For large microscopy datasets (time-lapses, z-stacks),
a representative selection of all data is provided after binning and time downsampling due to space limitations on the repository. The full dataset is available from
the corresponding author. Oligonucleotide sequences generated for this work are provided in the Supplementary Information, Tables S1-S4.

1
 nature portfolio | reporting summary
Research involving human participants, their data, or biological material
Policy information about studies with human participants or human data. See also policy information about sex, gender (identity/presentation),
and sexual orientation and race, ethnicity and racism.
Reporting on sex and gender N/A

Reporting on race, ethnicity, or N/A

other socially relevant groupings

Population characteristics N/A

Recruitment N/A

Ethics oversight N/A

Note that full information on the approval of the study protocol must also be provided in the manuscript.

Field-specific reporting
Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design

All studies must disclose on these points even when the disclosure is negative.
Sample size No statistical method was used to determine sample size.

Data exclusions No data were excluded except (in limited cases) when removing artefacts of incorrect segmentation of microscopy images, as specified in the
SI Methods (sections 2 and 3).

Replication Several control experiments were executed (SI) and found consistent. Information on repeats is provided in the relevant figure captions. No
reproducibility issues emerged.

Randomization The experiments were not randomised

Blinding The investigators were not blinded to allocation during experiments and outcome assessment

Reporting for specific materials, systems and methods

We require information from authors about some types of materials, experimental systems and methods used in many studies. Here, indicate whether each material,
system or method listed is relevant to your study. If you are not sure if a list item applies to your research, read the appropriate section before selecting a response.

Materials & experimental systems Methods

n/a Involved in the study n/a Involved in the study
Antibodies ChIP-seq
Eukaryotic cell lines Flow cytometry
Palaeontology and archaeology MRI-based neuroimaging
Animals and other organisms
Clinical data
Dual use research of concern
Plants
April 2023

2
 nature portfolio | reporting summary
Plants
Seed stocks N/A

Novel plant genotypes N/A

Authentication N/A

April 2023