Organization of Intracellular Reactions with Rationally Designed RNA Assemblies

See allHide authors and affiliations

Science  22 Jul 2011:
Vol. 333, Issue 6041, pp. 470-474
DOI: 10.1126/science.1206938


The rules of nucleic acid base-pairing have been used to construct nanoscale architectures and organize biomolecules, but little has been done to apply this technology in vivo. We designed and assembled multidimensional RNA structures and used them as scaffolds for the spatial organization of bacterial metabolism. Engineered RNA modules were assembled into discrete, one-dimensional, and two-dimensional scaffolds with distinct protein-docking sites and used to control the spatial organization of a hydrogen-producing pathway. We increased hydrogen output as a function of scaffold architecture. Rationally designed RNA assemblies can thus be used to construct functional architectures in vivo.

In cells, multienzymatic pathways are often physically and spatially organized onto scaffolds or clusters or into microcompartments (1). Spatial organization helps substrates flow between interacting proteins, limits cross-talk between signaling pathways, and increases yields of sequential metabolic reactions (1, 2). The ability to organize protein complexes and biological pathways spatially presents a strategy to engineer cells (3, 4).

The spatial organization of biomolecules has been the focus of DNA nanotechnology (58). This approach makes use of DNA’s base-pairing to generate one-, two-, and three-dimensional (1D, 2D, 3D) assemblies. DNA structures have largely remained limited to in vitro applications (9). RNA provides a compatible material for in vivo nucleic acid–based construction (10). It can be produced via the transcription machinery and forms stable interactions. RNA has been used to build higher-order assemblies in vitro (11, 12) and can potentially be used in vivo to engineer the intracellular environment.

In this work, we engineered synthetic RNA modules that assemble into functional discrete, 1D, and 2D scaffolds in vivo, and we used them to control the spatial organization of bound proteins (Fig. 1A) (see supporting online materials and methods). Scaffold D0 was constructed from a single RNA module d0, which folded into a duplex with PP7 and MS2 aptamer domains that bind PP7 and MS2 fusion proteins (Fig. 1B) (13).

Fig. 1

Design of RNA modules to organize proteins. (A) Proteins A and B scaffolded onto discrete, 1D, and 2D RNA assemblies. (B) D0 is a RNA strand that folds into a duplex with PP7 and MS2 sites. Ferredoxin/MS2 (FM) and hydrogenase/PP7 (HP) bind D0 to generate D0FH. (C and D) RNA with DDs and PDs initiates the formation of extended assemblies. Capping the palindromic sequences in DDs with PDs prevents its collapse (i) and allows for self-assembly (ii) into functioning tiles (iii). (E) D1 is constructed from a RNA strand d1 bearing PP7 and MS2, and it assembles into tile d1-1 (i). d1-1 assembles into a ribbon D12 (ii) or into a nanotube d1-2 (iii) that grows into D1 (iv). D1 organizes FM and HP into D1FH (v). (F) D2 is constructed from d2' and d2" bearing PP7 and MS2, respectively. d2' assembles into the pro-tile d2-1 (i) and interacts with d2" to generate d2-2 (ii). d2-2 self-assembles into a nanotube D22 (iii) or the 2D D2 (iv). D2 organizes FM and HP into D2FH (v).

We developed an approach for the in vivo isothermal assembly of extended RNA scaffolds by constructing sequence-symmetric RNA building blocks (Fig. 1, C and D) inspired by2D DNA analogs (13, 14). These RNA strands possess dimerization domains (DDs) and polymerization domains (PDs). To prevent the formation of ill-defined networks, it was necessary to disfavor the collapse of the palindromic regions (15) and control assembly order by insuring tile formation before polymerization. We achieved this by designing PDs that fold intramolecularly into kinetically protected hairpin structures (Fig. 1D, step i). The stem of these hairpins is an overlapping shared domain with the DD that discourages collapse (Fig. 1D, red segments), allowing the DD to activate the PD upon self-binding (Fig. 1D, step iii). We further destabilized the collapsed state by incorporating wobble pairs and mispairs (figs. S1 to S4).

The 1D RNA assembly D1 was derived from a single RNA d1 with PP7 and MS2 binding domains (Fig. 1E). d1 assembled into d1-1 (step i), which self-assembled into d1-2 (step iii). The torsion in d1-2 induced folding into an RNA nanotube capable of growing into the 1D scaffold D1 (step iv). The 2D RNA assembly D2 was formed from d2' and d2", each carrying a distinct PP7 and MS2 aptamer (Fig. 1F). The dormant tile d2' spontaneously generated the pro-tile d2-1 (step i), which interacted with d2" to generate tile d2-2 (step ii). d2-2 then self-assembled into the 2D RNA scaffold D2 with PP7 and MS2 binding domains (step iv).

We used atomic force microscopy (AFM) to characterize in vitro transcribed RNA modules d1 and d2'/d2". d1 formed 1D RNA fibers (D1), whereas d2'/d2" assembled into 2D extended RNA fibers (D2) (Fig. 2A). The width of D1 (~5 nm, a few tiles wide) is smaller than that of its DNA analog (14) and might also correspond to 1D ribbons (D12) constructed from a continuous line of single tiles (Fig. 1E, step ii). Given that D2 preferentially grows in a single direction when compared with its DNA analog (13), it might also correspond to RNA nanotubes (D22) that are relatively wider than D1 (Fig. 1D, step iii). To confirm the validity of our assemblies, we used analogs of d1 and d2' with a poly-T stretch in place of the DD incapable of assembling; d1T and d2'T did not generate extended assemblies (Fig. 2B).

Fig. 2

Characterization of RNA assemblies. (A) In vitro transcribed d1 and d2'/d2" assemble into D1 and D2 (AFM; phase images; scale bars, 0.25 μm). (B) In vitro transcribed mutated RNA d1T and d2'T/d2" do not assemble. (C) DNA-based precipitation of in vivo RNA assemblies uses DPC (i) and a release probe (DPR) for recovery (ii). (D) Capture and release of substrate DPS (left gel, beads; right gel, solution). Lane 1, conjugation of DPC to streptavidin-coated magnetic beads; lane 2, capture of DPS; lane 3, release of DPS using DPR. (E) AFM analysis of purified assemblies. (F) ISs bind the DDs of d1 and d2' to prevent their assembly into D1 or D2 (circular structures are drying artifacts). (G) When used during the purification of d1 and d2'/d2", D1 and D2 assemblies are still observed. (H) TEM analysis revealed the formation of 1D assemblies for D1 and 2D aggregates for D2 (scale bars, 100 nm). (I) Quantitative real-time fluorescence polymerase chain reaction analysis of in vivo RNA production levels. Error bars indicate SEM.

We developed a DNA-based precipitation (DP) method to purify our RNA assemblies from cells. Streptavidin-coated magnetic beads with a biotinylated DNA capture probe (DPC) were added to cell lysates. The capture domain of DPC binds the T7 terminator in our RNA molecules (Fig. 2C, step i). The RNA assemblies were released upon addition of DPR that bound the release domain of DPC (Fig. 2C, step ii). We were able to capture and release RNA (Fig. 2D).

In vivo synthesized D1 and D2 revealed extended 1D and 2D assemblies (Fig. 2E and fig. S5). Cross-sectional height analysis showed D1 to have two populations of distinct height (3 and 6 nm), which is characteristic of open versus closed nanotubes. In vivo D2 assembled into 2D structures that are smaller and somewhat different than their in vitro counterparts, suggesting that the assembly process in cells is of lower fidelity. To confirm that the assemblies formed in vivo, we engineered a set of inhibitory strands (ISs) that bound the trigger domains of d1' and d2'. The inhibition by these strands was confirmed in vitro (Fig. 2F). The purification of D1/D2 in the presence of excess ISs did not eliminate the observed 1D and 2D assemblies (Fig. 2G), confirming the formation of D1 and D2 in cells pre-lysis.

Transmission electron microscopy (TEM) analysis of whole bacterial cells expressing D1 or D2 confirmed their assembly in cells. The RNA assemblies were tagged with gold-binding metallothionin-PP7 fusion proteins (PAu) that form clusters (Fig. 2H) (16). Cells coexpressing PAu and D1 formed thin filaments with lengths of 200 to 300 nm, whereas cells coexpressing D2 formed compact spherelike structures ~100 nm in diameter. D0, D1, or D2 does not affect cell growth (fig. S12). Cells carrying the D1 and D2 scaffolds had higher RNA levels relative to cells expressing mutated poly-T RNA analogs (Fig. 2I), consistent with the formation of degradation resistant assemblies. Thus, d1 and d2'/d2" assembled in vivo into D1 and D2.

We used fluorescence complementation to detect protein assembly on our RNA scaffolds (Fig. 3). Green fluorescent protein (GFP) split into two halves (FA and FB) fused to the PP7 or MS2 aptamer binding proteins was used (Fig. 3A). Cells expressing FA and FB alone (Fig. 3B) or D0, D1, or D2 without the split GFPs displayed little fluorescence. However, the coexpression of D0, D1, or D2 with the split GFPs showed increased fluorescence (Fig. 3C). Thus, our RNA scaffolds served as docking sites to promote protein-protein interactions in cells.

Fig. 3

Fluorescence protein complementation in vivo. (A) GFP split into two halves, each of which is fused to PP7 or MS2 (FA and FB). FA and FB bind their respective aptamers (i) and reconstruct functional fluorescent GFP (ii). EGFP, enhanced green fluorescent protein. (B) Fluorescence microscopy imaging of cells expressing FA and FB revealed little to no fluorescence (scale bars, 10 μm). a.u., arbitrary units. (C) Cells coexpressing FA and FB with D0, D1, or D2 reveal an increase in fluorescence, indicating that D0, D1, and D2 scaffold PP7 and MS2 protein chimeras. Gray lines in flow cytometry plots separate OFF and ON cells.

Biological hydrogen production has both fundamental and practical implications. Coexpression of [FeFe]-hydrogenase and ferredoxin catalyzes the reduction of protons to hydrogen through electron transfer (17). We used this system to assess the potential of our RNA scaffolds to constrain flux through spatial organization. We fused the hydrogenase to a single copy of PP7 (HP) and ferredoxin to a dimer of MS2 (FM), and we conducted electrophoretic gel-shift analysis of the binding of FM and HP to D0 (Fig. 4A). Addition of HP to D0 resulted in a single product termed D0H. The addition of FP to D0 resulted in the formation of D0F. The addition of HP and FM to D0 resulted in a single product assigned to the protein-RNA assembly D0FH. HP and FM assembled onto D0 in cells to form D0FH (Fig. 4B).

Fig. 4

Scaffolding hydrogen production. (A) In vitro gel shift of HP (lane 1) binds D0 to form D0H (lane 2). FM (lane 3) binds D0 to form D0F (lane 4). HP and FM bind D0 to form D0FH (lane 5). (B) In vivo gel shift of HP and FM (lane 1) and HP and FM in the presence of D0 (lane 2). (C) Hydrogen biosynthesis as a function of scaffold, normalized to unscaffolded cells expressing HP and FM. (D) Mutating aptamer binding sites (E) do not affect self-assembly, (F) but do prevent protein binding (scale bars, 10 μm) and (G) hydrogen production. Error bars indicate SEM. Dashed lines in (C) and (G) denote separation between scaffolded and unscaffolded proteins.

To determine whether our RNA scaffolds increased hydrogen biosynthesis, we used gas chromatography to analyze cells expressing the hydrogen-producing pathway, along with the different RNA assemblies (fig. S11). The relative levels of FM and HP expression in D0, D1, and D2 cells were comparable (fig. S8). D0, D1, and D2 assembled FM and HP into D0FH, D1FH, and D2FH (Fig. 1). D0 resulted in a 4.0 ± 1.3–fold increase in hydrogen production compared with unscaffolded HP and FM (Fig. 4C). Hydrogen output with the extended assemblies D1 and D2 resulted in a 11 ± 2.8– and 48 ± 1.5–fold increase in hydrogen production (Fig. 4C). When normalized against the amount of RNA in cells (Fig. 2I and fig. S7), D0, D1, and D2 resulted in a 4.0-, 6.2-, and 24-fold increase. The increase with D2 is consistent with its assembly in vivo into “organelle-like” structures effective at concentrating proteins and their products (Fig. 2H). Mutating the PP7 and MS2 binding sites prevented protein scaffolding (Fig. 4, D to G). Thus, RNA can be used to organize enzymatic pathways in vivo to increase output as a function of architecture.

We controlled the spatial organization of proteins in cells using RNA molecules that are sequence-programmed to isothermally assemble into predefined discrete, 1D, and 2D structures in vivo. These assemblies scaffolded proteins and were used to organize a hydrogen-producing biosynthetic pathway. Hydrogen production was optimized as a function of scaffold architecture. Unlike protein-based approaches (3, 4, 17), RNA-based scaffolds allow for the formation of complex multidimensional architectures with nanometer precision. In vivo RNA assemblies can thus be used to engineer biological pathways through spatial constraints (18, 19).

Supporting Online Material

Materials and Methods

Figs. S1 to S13

References (20-35)

References and Notes

  1. Acknowledgments: This work was supported by from the Enerbio-Tuck Foundation and the Institut Français du Pétrole Energies Nouvelles (to C.J.D.); Agence Nationale de la Recherche France, Institut National de la Santé et de la Recherche Médicale–Institut National de Recherche en Informatique et en Automatique projet d’envergure, and Axa Foundation Chair on Longevity (to A.B.L.); Natural Sciences and Engineering Research Council of Canada (to F.A.A.); and the Wyss Institute for Biologically Inspired Engineering. We thank the Harvard Medical School Electron Microscopy Center for assistance in electron microscopy, and we thank the reviewers for their insight.
View Abstract

Navigate This Article