Ribozyme-phenotype coupling in peptide-based coacervate protocells

Condensed coacervate phases are now understood to be important features of modern cell biology, as well as valuable protocellular models in origin of life studies and synthetic biology. In each of these fields, the development of model systems with varied and tuneable material properties is of great importance for replicating properties of life. Here, we develop a ligase ribozyme system capable of concatenating short RNA fragments into extremely long chains. Our results show that formation of coacervate microdroplets with the ligase ribozyme and poly(L-lysine) enhances ribozyme rate and yield, which in turn increases the length of the anionic polymer component of the system and imparts specific physical properties to the droplets. Droplets containing active ribozyme sequences resist growth, do not wet or spread on unpassivated surfaces, and exhibit reduced transfer of RNA between droplets when compared to controls containing inactive sequences. These altered behaviours, which stem from RNA sequence and catalytic activity, constitute a specific phenotype and potential fitness advantage, opening the door to selection and evolution experiments based on a genotype – phenotype linkage.


Introduction
Many biological biomolecular condensates are formed from RNA and proteins or peptides 1,2 , and coacervate phases formed by charge-mediated phase separation are part of the mechanism that drives the development of membraneless organelles in modern biology 3,4 .
Condensed phases studied in the context of origin of life can also be formed from RNA and peptides. Catalytic nucleic acids, including ribozymes and DNAzymes, are central to 'Nucleic Acid World' hypotheses, where they act as both the medium of information storage and the catalyst for its replication in early life-like systems. Peptide-RNA condensation forms discrete protocellular compartments 5,6 . RNA-peptide interactions are in many cases beneficial for the function of these early catalysts 7,8 , and may promote the folding and oligomerisation of certain peptides 9 . Perhaps surprisingly, the function of ribozymes and other nucleic acid enzymes in coacervate phases has only been recently established, with catalytic rate being enhanced and enzyme function altered in some cases [10][11][12][13][14] . For example, coacervation with poly(L-lysine) shifts the reaction equilibrium of a minimal hairpin ribozyme from cleavage to ligation 15 .
The ability of coacervate phases to strongly partition a wide range of molecular and macromolecular species 16 , and to support a variety of complex enzymatic processes 17 , makes them appealing proto-and artificial cell models in origin of life studies 18 , synthetic biology 19 and modern biology 20 . Simple coacervates that form from oppositely charged polymers have been rigorously investigated for their ability to selectively partition biomolecules and host a variety of different chemistries and reactions in vitro 20,21 . However, biological condensates formed by protein-protein or protein-nucleic acid interactions are dynamic systems, and their formation, dissolution and physical properties are subject to spatiotemporal regulation 22 .
Similarly, in order to realise a truly convincing model proto-or artificial cell, life-like behaviours such as growth, division and other dynamic, responsive or non-equilibrium processes are essential. Several dynamic and responsive coacervate systems have been established, which are characterized by a phase change in response to environmental stimuli such as light exposure [23][24][25] , changes in temperature 26 or pH 27,28 . Furthermore, non-equilibrium environments formed by gas bubbles inside heated rock pores have been shown to drive the growth, fusion, and division of otherwise inert coacervate microdroplets 29 .
Enzymatic processes that alter the properties of the coacervate components also affect coacervate properties and behaviour. In charge-based condensates, inducing phase change via enzymatic processes has been achieved by, for example, the conversion of ADP into charge-dense ATP 30 , or the alteration of peptide charge state by phosphorylation 31 . This has allowed the reversible generation of coacervate droplets by enzymatic networks. In addition, the polymerisation of UDP in U20-spermine coacervates by polynucleotide phosphorylase has been shown to induce transient non-spherical coacervate morphologies 32 . All these systems depend on the action of proteinaceous enzymes, which presumably emerged later in molecular evolution, perhaps after the first protocells. However, droplets capable of dynamic change via the action of nucleic acid enzymes such as ribozymes have not been previously reported, partially because the catalytic repertoire of these catalysts is limited when compared to their proteinaceous counterparts 33 . Beyond environmental factors such as solution pH and salt concentration 34 , the physical properties and association behaviour of coacervate droplets are determined by factors such as component chain length 35 and charge density 36 . Of these factors, we noted that polymer chain length could potentially be addressed by the action of a nucleolytic or ligase ribozyme. Thus, in a coacervate composed of RNA and a positively charged polymer, ribozyme catalysed RNA cleavage or ligation might alter the physical properties or association behaviour of the system if a sufficiently large change in average RNA length were achieved. In particular, we hypothesised that the elongation of the RNA component could lead to the formation of a denser separated phase with altered physical properties 34,37 .
In this work, we demonstrate that coacervate microdroplets formed from a ligase ribozyme and lysine-based peptides display reciprocal behaviour, in which coacervation enhances and modulates ribozyme activity, and ribozyme activity modulates droplet properties and therefore phenotype. We employ a modified ligase ribozyme that ligates short substrate strands into long concatemers, thus increasing the length of the polyanionic coacervate component. We find that coacervation enhances the rate of substrate ligation 50-fold and inhibits the formation of circular reaction products. In turn, the activity of the ribozyme imparts specific physical properties or phenotypes to the droplets it is contained within, which are not observed in droplets containing inactive ribozymes. These altered behaviours include the inhibition of droplet growth, surface wetting and content exchange, which could provide fitness advantages under certain conditions. Connecting the sequence information of the ribozyme RNA to the physical properties of the resulting microdroplets is an example of a phenotype-genotype linkage, which is a fundamental requirement for the Darwinian evolution of protocellular systems 38 .

Results
To investigate the effect of ribozyme activity on coacervate behaviour, we initially sought to design a ribozyme system capable of increasing RNA chain length via concatenation.
Although several examples of RNA ligase ribozymes have previously been reported [39][40][41] , these typically catalyse the ligation of a single junction, which results in only a modest overall increase in average RNA chain length. To achieve greater product lengths, we harnessed the catalytic core of the R3C ligase ribozyme (ER), whose RNA ligation activity is based on 5'triphosphate activated substrates 39 . This system has recently been shown to be active in the presence of poly(L-lysine) under certain conditions, and so was a promising starting point when considering compatibility with coacervate systems 14 . We previously redesigned the ribozyme-substrate complex to iteratively produce long RNA concatemers from short oligonucleotides 42 . In our final design, the ribozyme (EL) catalyses concatenation of a 31 nt substrate (Figure 1a). Screening of reaction conditions established that strong activity was observed at pH 8.6 in the presence of 10 mM MgCl2 at a range of temperatures (30, 37 and 45 °C) (Figure 1supplement 1). The strongest activity was observed with a 1:1 monomer concentration ratio of substrate:ribozyme, while excess substrate inhibited the formation of longer length products at lower temperatures, most likely because the ribozyme has a higher probability of binding unligated substrates than already growing chains. In order to reduce hydrolytic degradation of RNA, we chose to maximise activity at 30 °C with a 1:1 substrate:ribozyme ratio in all subsequent experiments.
The phase separation behaviour of the ribozyme system in the presence of poly(L-lysine) was initially investigated by titrating increasing amounts of (Lys)19-72 into a fixed concentration of RNA (total monomer concentration = 1 mM). The activity of the ribozyme was inhibited in the presence of excess peptide ((Lys)19-72:RNA > 1) (Figure 1b). The reported concentration ratios are calculated from RNA and peptide monomer unit concentrations, and as such are also equivalent to charge ratios. The addition of (Lys)19-72 led to a gradual increase in turbidity due to phase separation above the critical coacervation concentration of CCC19-72 ≈ 0.14:1 (Lys)19-72:RNA (Figure 1c). These experiments were repeated with a shorter peptide ((Lys)5-24) ( Figure 1supplement 2), for which the onset of coacervation occurred at higher peptide:RNA ratios (CCC5-24 ≈ 0.93:1 (Lys)5-24:RNA), corroborating previous observations with poly(L-lysine) and the HPz ribozyme 15 . In this case, the activity of the ribozyme was not inhibited in the presence of excess peptide. For further experimentation, we selected specific peptide:RNA ratios of 0.75:1 (Lys)19-72:RNA and 3:1 (Lys)5-24:RNA. Whilst these two conditions are not directly comparable due to their differing concentration ratios, the difference in critical coacervation concentration between the two peptide lengths meant that no single ratio supported both droplet formation and ribozyme activity in both systems. Both points occur shortly before the respective maxima in the peptide titration turbidity curve, allow the formation of liquid coacervate droplets and do not suppress ribozyme activity (Figure 1c and Figure 1 -supplement 2b). Fluorescence imaging of samples at these ratios confirmed the formation of phase separated coacervate droplets that strongly partitioned the Cy5-labelled RNA substrate (Figure 1d and Figure 1 -supplement 2c).
The kinetics of chain elongation and final product length at these ratios was greatly enhanced in the coacervate phase compared to solution (Figure 1e-f). The kinetic analyses show that the addition of either (Lys)n peptide resulted in an approximately 50-fold increase in the rate of concatenation and led to the formation of substrate chains with an average length approximately 30 nt greater than those produced in solution (Figure 1e). In solution, the kinetics of chain elongation were best approximated using a first order model (k = 1.5 x 10 -2 ± 1.0 x 10 -3 min -1 ), and a final average product length (n) of 52.2 ± 0.5 nt was observed ( Figure   1e). In the (Lys)19-72 coacervate phase, the ribozyme kinetics were best described by a twophase model (kfast = 7.7 x 10 -1 ± 6.6 x 10 -2 min -1 , kslow = 6.6 x 10 -2 ± 1.3 x 10 -2 min -1 ), with a final product length of n = 77.8 ± 0.4 nt for (Lys) . Similar values were obtained for the short peptide (Figure 1 supplement 2e, source data). We have previously observed a coacervation-induced shift from monobasic to biphasic kinetic behaviour for the hammerhead ribozyme, albeit without an associated increase in rate 10 . The increase in rate observed here is most likely due to the high concentration of ribozyme and substrate that occurs in coacervates directly formed from catalytic RNA and peptides 15 .
In solution, the EL ribozyme produced both linear and circular concatenates, the latter of which are visible on the urea -PAGE gel as additional bands visible in between and above the regularly spaced linear ladder of products, due to the different mobility of linear and circular products through the gel matrix. The presence of circular products was confirmed by treatment of reacted RNA with RNAse R, a 3' to 5' exoribonuclease that only digests linear strands  Having demonstrated that the EL ribozyme is capable of substrate concatenation in both solution and in (Lys)n coacervate droplets, and that its activity is enhanced within the coacervate environment, we aimed to determine if the elongation of the RNA component would lead to differences in droplet properties, thus altering phenotype. To identify changes due solely to chain concatenation, we developed an inactive mutant of the ribozyme by This indicates that the material properties of the droplets change over time.
Whilst surface passivation is used as standard in the imaging of coacervate systems to prevent wetting and adhesion effects, surface interactions can affect droplet formation and morphology, and therefore may also influence the observed protocellular phenotype.
Consequently, we repeated the previous experiment on an unpassivated polystyrene surface (Greiner µclear microplate, medium binding). The coacervate droplets containing the active ribozyme behaved as previously observed, with discrete round morphologies ( Figure 2b) and little change in both average droplet area and population density over the course of the experiment (Figure 2g, h). These droplets exhibited greater average areas than those in the passivated environment, likely due to wetting onto the surface. In contrast, droplets containing the inactive ribozyme wet the surface and rapidly spread, eventually merging to form a film of the condensed coacervate phase on the bottom of the well (Figure 2d). The measured average particle area therefore increased greatly over the course of the experiment, whilst the number of particles decreased (Figure 2g, h).
These differences in behaviour indicate that the generation of longer RNA within the coacervate environment imparts altered physical properties on the droplets, despite maintaining a spherical morphology typical of a liquid system. To further investigate this phenomenon, we added the (Lys)5-24 and (Lys)19-72 peptides to pre-reacted RNA mixtures.
Here, we observed that mixtures containing the active ribozyme and therefore preconcatenated RNA initially formed non-spherical gel-like condensates with both peptides, which relaxed to form spherical droplets over the course of 24 h, whilst mixtures with inactive ribozyme yielded spherical droplets from the outset (Figure 2supplement 5). This suggests that large morphological differences between active and inactive systems are not observed when peptide is added to ribozyme and substrate mixtures without pre-reaction because the average RNA length is initially identical. As the reaction proceeds in the active systems, a transition to a more viscous or gel-like state may occur whilst maintaining the initially formed spherical morphology. All experiments were performed with 1 mM total RNA monomer concentration, a 0.75:1 ratio of (Lys)19-72:RNA monomers, and at 30 °C, pH 8.6 and 10 mM MgCl2. The RNA reaction mixture contained 10 % Cy5-labelled substrate for fluorescence imaging. Particles were measured from at least 9 separate images, except for unpassivated samples for which a single image was captured. Error bars are standard errors. Droplet areas and particle counts were measured using the CellPose segmentation algorithm. Data for condensates formed from the shorter (Lys)5-24 peptide are shown in Figure 2 -supplement 2.
Given the effect of RNA concatenation on the coacervate droplets, we asked whether the activity of the EL ribozyme could affect the interactions between different populations of droplets. Using the previously established conditions and concentration ratios, we mixed populations of droplets containing either a Cy5-or FAM-tagged substrate (10 % total substrate concentration) and monitored mixing and content exchange over the course of 24 h. These effects can be quantified by the calculation of a Pearson correlation coefficient (PCC), which measures the correlation of pixel intensities between the two fluorescence channels 44  A similar yet weaker trend was observed with short (Lys)5-24 peptide droplets, which exhibited greater content exchange overall between both active and inactive droplets (Figure 3supplement 1-3). Here, the presence of the active ribozyme slowed the rate of coalescence and therefore content exchange rather than completely suppressing it, but still allowed active populations to maintain a limited degree of identity over the course of the experiment.

Discussion
Our original goal was to develop a ribozyme capable of RNA chaining that could significantly increase the average RNA length in a coacervate system, and thereby allow us to study the influence of such a reaction on the material properties of the coacervates. Although ribozyme systems that concatenate a short substrate have been previously reported, these do so via a reversible cyclic-phosphate mediated mechanism, and as such the concentration of extended products decreases rapidly as product length increases 45 . The R3C ligase ribozyme was thus an attractive starting point for the development of our system as the 5'-triphosphate activated reaction it catalyzes is quasi-irreversible, and the ribozyme is already optimized for high rate and yield 39 . Indeed, the EL ribozyme was able to generate RNA strands of > 500 nt in length, including circular species, and this activity was enhanced in terms of both rate and yield in (Lys)n coacervates. Although the observation of circular products in concatenation reactions has been previously reported in ribozyme systems 46 , the inhibition of this behaviour by coacervation is notable. Whilst the mechanism of this inhibition is unknown, the absence of circular products suggests that the ends of a substrate strand do not meet, hinting at a decrease in RNA mobility, increased chain stiffness or reduced release of ligated substrate from the ribozyme within the condensed phase compared to the bulk solution. The suppression of circularisation should lead to higher overall RNA lengths as the concentration of RNA ends available for ligation is not decreased over the course of the reaction. It is not possible to quantify this difference as measurement of the average chain lengths requires the use of a 5' fluorescent tag, which itself prevents formation of circular products due to the blocked 5' end.
We have previously reported that the activity of the hairpin ribozyme (HPz) is greatly enhanced in condensed phases comprised of catalytic RNA and (Lys)n peptides 15 . The increase in turbidity observed on titrating either (Lys)19-72 or (Lys)5-24 into a fixed concentration of R3C ribozyme and substrate was similar to that observed for HPz, although in the present study the resultant condensed phase appeared liquid rather than gel-like. This variation in condensed phase morphology may be due to the influence of RNA structure, sequence and hybridisation state on condensation 47,48 , with a higher degree of single stranded or loop structures in the EL ribozyme system. The inhibition of ribozyme activity at excess (Lys)19-72:RNA ratios was comparable to that observed in our previous study 15 , and is attributed to a peptide length dependent melting or misfolding of native nucleic acid tertiary structures at high concentrations of longer peptides 49,50 . Our previous study into the function of the R3C replicase ribozyme in RNA-peptide coacervates reported a similar trend: catalytic activity was inhibited at excess (Lys)n:RNA ratios for a range of peptides (n = 7, 8, 9, 10, 18, and 19-72) 14 .
The present study extends the peptide concentration range in which the R3C ribozyme functions, with robust catalytic activity at ratios as high as 4:1 (Lys)5-24:RNA. A possible explanation is that commercially available (Lys)n deviates from the manufacturer's stated size range, with the (Lys)5-24 used here being predominantly comprised of oligomers between n = 3 to n = 9 in length 15 . These short fragments may interact only weakly with the RNA, effectively reducing the concentration of peptide oligomers that can form condensates. In general, it should be noted that the effect of phase separation on ribozyme activity in a given system is highly specific to the identity of the ribozyme and peptide, their relative concentrations, and environmental conditions such as buffer, magnesium concentration and temperature.
The difference in growth behaviour between active and inactive droplets is a striking demonstration of how the presence of a catalytic species can modulate the physical properties and behaviour of a model protocellular system, in this case by reducing coalescence. The coarsening of liquid coacervate phases is not inevitable: Growth can be also be suppressed by active chemical processes hosted within droplets that produce the polymer components 51 , although in our experiments the action of the ribozyme does not increase the concentration of RNA or peptide in the system. The formation of kinetically trapped states can also allow the persistence of static droplet populations over time 52 . However, increasing polymer lengths in a coacervate system results in stronger cooperative electrostatic interactions between polymer components 34 , and has been shown to affect physical parameters such as droplet water content, critical salt concentration and condensed phase polymer concentration 35 , leading to denser droplets and a more depleted dilute phase. Simulations have shown that increasing RNA length modulates both the material and interfacial properties of RNApeptide condensates, in particular increasing density and surface tension 53 . RNA length also has been shown to modulate viscosity in biomolecular condensates 54 , which in turn determines the droplet fusion dynamics 55 . Thus, increasing overall RNA length by R3C substrate concatenation may inhibit growth via coalescence by increasing condensate density and viscosity, and reduce wetting affects through the alteration of surface tension. Whilst RNA sequence and secondary structure have also been shown to influence the viscoelastic behaviour of RNA peptide condensates 56 , the active and inactive ribozyme systems used in the present study differ by only few nucleotides.
Despite the spherical morphology of the RNApeptide condensates reported here, the inhibition of growth and wetting may also be due to a phase transition to a gel state. The formation of non-liquid RNA-peptide condensates has been previously reported for poly-rA:rUpeptide mixtures, where RNA base pairing leads to the formation of a kinetically arrested gellike solid 48 , as well as in our previous work 15 . These gel condensates melt to form spherical particles upon thermal denaturation and annealing, which disrupts the networked structures formed from complementary RNA strands and peptides. In our R3C system, complementary base-pairing interactions between substrate and ribozyme should also permit the formation of large, non-covalently assembled networked structures due to the repeating nature of the substrate. Ribozyme activity is expected to increase the stability of these structures, as substrate ligation increases the free energy of association between ribozyme and substrate.
We hypothesise that in the inactive system, and before significant concatenation occurs in the active system, RNA-RNA interactions are sufficiently weak to allow the formation of liquid coacervate droplets. In the active system, liquid droplets initially form because the timescale of coalescence is faster than that of RNA ligation. However, as ligation proceeds the stability of networked assemblies of ribozyme and substrate increases, resulting in a transition from a liquid to either a highly viscous liquid state or solid gel state that is both unable to grow and or wet the unpassivated surface on the timescale of the experiment. This hypothesis is supported by our observation that mixtures of pre-concatenated RNA and (Lys)n initially form nonspherical gel-like aggregates, which then relax to form spherical droplets over the course of 24 h.
The reduction of mixing between populations of droplets containing orthogonally labelled RNA likely results from the changes in material properties and growth behaviour caused by RNA concatenation. The coalescence of the coacervate droplets observed in all inactive systems is likely to be the primary mechanism of mixing in this scenario. Indeed, greater droplet growth was observed in the two-population system than for single population experiments, which may be due to the necessity of pipetting each droplet population into the sample environment sequentially, thus increasing mechanical agitation and contacts between droplets. All populations at the initial timepoint contain fusion droplets, which contain separate and unmixed areas of each fluorescent RNA. Notably, fusion droplets in all systems appear evenly mixed after 24 h, suggesting that significant equilibration occurs before any transition to a highly viscous liquid or solid gel. Mixing may also occur by diffusive transfer of RNA between droplets. As the fluorescently labelled substrate is ligated onto the growing concatemer chain, the diffusion coefficient of the labelled material is expected to decrease over the course of the reaction as its effective length increases, which would in turn slow diffusive transfer in the active system.
Taken together, the results reported here describe a range of altered droplet behaviours stemming from ribozyme-driven changes in coacervate physical properties. The ability of a population of protocells to resist passive coarsening, surface wetting and content exchange can be considered a fitness advantage 57 , and the emergence of fitness differences from varying protocellular compositions and phenotypes is a key step on the path towards Darwinian evolution with suitable selective pressures 58 . Furthermore, in this system the phenotypic difference originates from the ribozyme-driven concatenation reaction, which increases the length of the anionic coacervate component. The sequence of the ribozyme in effect comprises the genotype of our protocellular system, with different variants (e.g. active or inactive sequences) producing different protocellular phenotypes. The realisation of a phenotype-genotype linkage is essential for true open-ended evolution, otherwise phenotypes with obvious fitness advantages have no way of being propagated to future generations. We envision that recent methodological developments in RNA sequencing from single coacervate microdroplets will enable future selection experiments on populations of protocells with varying RNA genotypes and varying degrees of fitness with respect to environmental pressures such as heat or salt concentration 59 . It should be noted that as a ribozyme developed by in vitro selection, the R3C ligase is already a highly optimized system and offers little potential for further enhancement via selection. Nevertheless, the reciprocal modulation of ribozyme activity and coacervate properties reported here furthers the argument for the early coevolution of RNA and peptides [7][8][9]18,60 . Whilst the magnitude of behavioural change was greater for active and inactive coacervate populations formed with the longer peptide, clear phenotypic differences were nonetheless also present in the (Lys)5-24 system. This short peptide, predominantly composed of n = 3 -9 residue oligomers, is of a length that could be produced by prebiotically plausible processes such as wet-dry cycling 61 The data were fitted in GraphPad Prism using both a first order ('one-phase association') and second order ('two-phase association') kinetic model. First and second order kinetics were discriminated between using the extra sum-of-squares F test for nested models. The simpler first order model was rejected when P < 0.05.    . The ribozyme reaction buffer contained 10 mM MgCl2 and 50 mM tris pH 8.6, and the reaction was stopped after two hours. Circular bands are marked with an asterisk. After two hours the reaction was stopped and the extracted RNA was digested with RNAse R, leaving only circular products. The formation of circular products is observed in solution but is suppressed in the presence of poly(lysine).      . Each population in the set contained either 10 % FAM-or 10 % Cy5tagged substrate (green and blue, respectively). The two populations in each set were mixed shortly after preparation and then imaged over 24 h in a passivated environment. All experiments were performed at 30 °C, pH 8.6 and 10 mM MgCl2 with a 1 mM total RNA monomer concentration and a 0.75:1 ratio of (Lys)19-72:RNA. The colocalization of the two fluorophores within single droplets is measured by the droplet Pearson coefficient (PCPdroplet) (c), whilst the colocalization of fluorophores in the overall population of droplets is measured by the population Pearson coefficient (PCCpop) (d).
The average particle area and number of particles per unit area over time are shown in e and f respectively. Particles were measured from at least 6 separate images. Error bars are standard errors. Scale bars = 10 μm.   Figure 3 and S10.     (Figure 1d), imaged using 10 % Cy5-tagged substrate strand.