Thermodynamic aspects of the self-assembly of DsrA , a small noncoding RNA from Escherichia coli

DsrA is an Escherichia coli small noncoding RNA that acts by base pairing to some mRNAs in order to control their translation and turnover. It was recently shown that DsrA is able to self-associate in a way similar to DNA and to build nanostructures. Although functional consequence of this RNA self-assembly in vivo is not yet understood, the formation of such an assemblage more than likely influences the noncoding RNA function. We report here for the first time the thermodynamic basis of this natural RNA self-assembly. In particular we show that assembling of the ribonucleic acid is enthalpy driven and that the versatility of the RNA molecule is important for the polymerisation; indeed, an equivalent DNA sequence is unable to make a nanoassembly. The origin of the difference is discussed herein.


INTRODUCTION
Small RNAs (sRNAs) are noncoding RNAs involved in regulation of gene expression in bacteria (Storz et al., 2004;Gottesman & Storz, 2011).We recently showed that, similarly to DNA that is well known to self-associate (Wei et al., 2012), sRNAs are also able to polymerize and that this self-assembly could play a critical role in some important regulatory pathways (Busi et al., 2009;Cayrol et al., 2009).Among them, DsrA, an 87 nucleotides long sRNA, regulates at least two mRNA targets in Escherichia coli.These targets are transcripts of two genes coding for important transcriptional regulators, the stationary-phase and stress response σ s RNA polymerase subunit (the rpoS gene transcript), and H-NS, a histone-like nucleoid protein and transcriptional repressor (the hns gene transcript) (Lease et al., 1998;Majdalani et al., 1998).The action of DsrA stems from sequence-specific RNA-RNA interactions, as it basepairs with both hns and rpoS mRNAs to affect their translation and turnover.
We observed previously that large DsrA polymers exist, resulting from the auto-assembly of two minimal regions of 14 and 8 nucleotides (nt).This 22 nt region is located in the central part of the sRNA (Cayrol et al., 2009) (Fig. 1A).In particular, nuclease footprinting showed that DsrA monomer can adopt three distinct secondary structures (Majdalani et al., 1998;Lease & Belfort 2000;Rolle et al., 2006) characterized by two 5' and 3' stem-loops (SL1 and SL3, see Fig. 1A) but with variability in the central region.This variability probably rep-resents the possibility of the SL2 sequence being able to "breathe", thus allowing an intramolecular stem to convert into intermolecular one with another DsrA sRNA (Cayrol et al., 2009;Lease et al., 2012) (Fig. 1B).Furthermore, due to these 14 and 8-nt regions being adjacent, the interaction with another dimer is favored and results in the formation of a polymer (Fig. 1B).
Because the auto-assembly property could have major effects on DsrA functions in vivo, we investigated here the basis of the minimal sequence auto-association.RNA self-assemblies have been previously observed and characterized with artificial sequences specifically designed to self-assemble, creating the concept of "RNA tectonics" (Jaeger & Chworos, 2006).But this work reports for the first time the thermodynamics of a natural RNA self-assembly.In comparison to protein or deoxyribonucleic self-assembling building blocks, ribonucleic nanostructures combine the advantages of both as it can play a role in the transmission of genetic information or of catalyst.Thus, understanding the basis of the assemblage of natural RNA opens perspectives for the future in vivo developments (Lease, et al., 2012;Afonin et al., 2011).AAGTGCTTCTTGCTTAAGCAAG, DsrA 14 DNA CTTGCTTAAGCAAG, DsrA 8 DNA AAGT-GCTT, DsrA 22 DNA/GC AAGCGCTTCTTGCTTAAG-CAAG and DsrA 8 DNA/GC AAGCGCTT.Concentrations of oligonucleotides were estimated by UV absorption measurements in water at 20°C susing a nearest-neighbor approximation for the absorption coefficients of the unfolded species (Cantor et al., 1970).All nucleic acids concentrations throughout the manuscript will be expressed in strand molarities.

Synthetic oligonucleotides.
Thermodynamics of nucleic acids self-assembly.The self-assembly of minimal sequences was analyzed by UV spectroscopy.To achieve the measurements, all experiments were performed in 10 mM sodium cacodylate buffer pH 7.2 containing 140 mM KCl and 10 mM NaCl.Strand concentrations ranged from 0.5 µM to 6 µM.The dependence of UV absorbance on temperature was measured by a temperature-controlled spectrophotometer along heating scans.The rate of temperature change should not exceed 0.5ºC/min to guarantee complete thermal equilibrium of the cell.At each temperature, absorbance measurements were done at 257 nm.Data were extracted from the profiles recorded and normalized to OD=1.
The energies involved in self-assembly were determined by a van't Hoff plot, i.e. lnK eq = ∆S°/R -∆H°/ RT.Concentrations of single stranded (C ss ) and double stranded (C ds ) species were calculated from the fraction observed on melting curves as C ss = f ss .C total and C ds = f ds .C total = (1-f ss )C total where f ss and f ds are the fractions of single and double strands, respectively.C total is the total strand concentration.
For the formation of the duplex, M + M ↔ M 2 model was applied (M is the monomer and M 2 the doublestranded duplex).The equilibrium association constant K duplex is calculated as For the formation of the hairpin, M ↔ M' model was applied (M the single-stranded monomer and M' the double-stranded hairpin).The equilibrium constant K hair- pin was calculated as In the case of polymerization, the equilibrium can be described by P n + M ↔ P n+1 (P i is the polymer formed by i subunits and M the monomer).K polymer , the equilibrium constant for the association of each subunit to the polymer, can be calculated as (Fontanille & Gnanou, 2010).
Taking into account that P n and P n+1 are values at equilibrium, we can consider that for a large polymer, P n ~ P n+1 and that K polymer = 1/[M] = 1/C ss = 1/(f ss .C total ) (Fontanille & Gnanou, 2010).
For all reactions, ΔGr°, the effective change in standard free energy of reaction, is defined as: where ΔHr° is the change in standard enthalpy of reaction, ΔSr° the change in standard entropy of reaction, R the gas constant and T the absolute temperature (expressed in Kelvin).
∆Sr° and ∆Hr° were extracted directly from the van't Hoff plot where R lnK was plotted against 1/T (R lnK = ∆Sr° -∆Hr°x1/T), with -∆Hr° the slope and ∆Sr° the intercept at origin of the straight-line.Note that thermodynamic parameters have an incertitude of ~10% due to the uncertainty for the lower and upper limit of the melting curves.

Estimating the enthalpic and entropic contributions of RNA self-assembly
As previously reported (Cayrol et al., 2009), DsrA 8 RNA presents an auto-complementary sequence that can selfassemble and form a duplex (Fig. 2, note the presence of 2 G.U wobble base pairs (Varani & McClain, 2000)).Accordingly, the melting temperature Tm (i.e. the temperature at which half of the DNA duplex will dissociate to become single stranded) observed on the thermal denaturation of DsrA 8 RNA was shown to depend on the strand concentration and thus corresponds to a bimolecular transition (Fig. 2A).Here, we extracted from the melting curves a van't Hoff plot and calculated the changes in standard enthalpy (∆Hr°) and in standard entropy (∆Sr°) of reaction for DsrA 8 RNA duplex formation (Fig. 2B).Our analysis indicates that ∆Sr° and ∆Hr° are -0.75±0.1 kJ mol -1 K -1 and -260±25 kJ mol -1 respectively (note that these values are indicative and may vary with experimental conditions; Tm and ∆G are for instance very sensitive to salt concentration).
Conversely, the thermal denaturation of DsrA 14 RNA shows two transitions (Fig. 3A) (Cayrol et al., 2009).As previously reported, the first transition is clearly concentration dependent and corresponds to an intermolecular association, the self-complementary duplex formed by DsrA 14 RNA .By contrast, the second transition is independent of strand concentration and corresponds to the formation of an intramolecular hairpin (Fig. 3A) (Cayrol et al., 2009).Indeed, such a self-complementary sequence commonly adopts two structures: a bimolecular duplex as well as a monomolecular hairpin (Nakano et al., 2007).From the first region of the melting curve, a van't Hoff  (Cayrol et al., 2009).DsrA sRNA 5'-end is denoted by a ball and 3'-end by an arrowhead.Nuclease footprinting shows that this central region can adopt different secondary structures (Majdalani, et al., 1998;Lease & Belfort, 2000;Rolle, et al., 2006), which probably represents the possibility of the SL2 sequence being able to "breathe", allowing formation of intermolecular base-pairs with another DsrA.(B) Minimal sequence allowing DsrA self-assembly.DsrA self-assembly originates from two adjacent regions of 14 (light orange) and 8 nt (dark orange), resulting in a sequence of 22 nt that could polymerize by itself in the absence of the whole RNA sequence (Cayrol et al., 2009).The 22, 14 and 8 nt sequences of DsrA are annotated as DsrA 22 , DsrA 14 and DsrA 8 , respectively, throughout the manuscript.Figure adapted from (Cayrol et al., 2009).Thermodynamics of DsrA self-assembly analysis was made and ∆Sr° and ∆Hr° of DsrA 14 RNA duplex assembly were calculated as -1.2±0.1 kJ mol -1 K -1 and -420±40 kJ mol -1 , respectively (Fig. 3B).From the second part of the curve, ∆Sr° and ∆Hr° for DsrA 14 RNA hairpin formation were calculated as -0.9±0.10 kJ mol -1 K -1 and -335±30 kJ mol -1 , respectively (Fig. 3C).Note the higher stability of the hairpin (higher Tm resulting in lower ∆Gr°) as compared with that of the duplex in our experimental conditions.This has however been previously reported for other RNAs and changes in experimental conditions (especially the concentration of salts) could reverse this tendency (Sun et al., 2007).
Finally, Fig. 4A depicts the melting curve of DsrA 22 RNA .As already described (Cayrol, et al., 2009), in spite of the two self-complementary regions, there is only one concentration dependent transition around 40ºC, demonstrating that 8 bp duplex is well stabilized in the context of DsrA 22 RNA polymer (+ about 20°C).The melting of the 8 and 14 bp regions thus occurs simultaneously.From the first part of the van't Hoff plot, ∆Sr° and ∆Hr° of DsrA 22 RNA self-assembly can be calculated as -1.35±0.15kJ mol -1 K -1 and -300±30 kJ mol -1 , respectively (Fig. 4B, note that these values characterize the association of one subunit to the polymer).
In all, our analysis shows that the energy associated with DsrA self-assembly is enthalpy driven (∆Hr°<0), while the entropic term (∆Sr°) is negative and unfavorable to the self-assembly.This is not surprising as the entropy change can be largely attributed to the cost of assemblage (i.e.mainly the reduction of the conforma-tional freedom of the phosphodiester backbone), thus entropy favors single-stranded conformation.

Importance of RNA features for the self-assembly
Next, we decided to test if the RNA feature of DsrA sequence is important for the formation of the nanostructure.Synthetic DsrA 22 DNA , DsrA 14 DNA and DsrA 8 DNA DNA sequences have been used for this goal (uracils were replaced by thymines).
First, we analyzed the self-assembly of DsrA 8 DNA .The transition observed on Fig. 5A represents the assembly of the duplex; nevertheless, this duplex is unstable (Tm ~ 15ºC) and is probably scarce at room temperature.Indeed, it is likely that the difference in stability between DsrA 8 DNA and DsrA 8 RNA duplex results from the formation of a non-Watson-Crick base pair in the RNA oligonucleotide, the G.U base pair (Fig. 2C), that cannot be formed in DNA between G and T bases.To test this hypothesis, we thus tried to restore the 2 central base pairs by replacing the 4th T of DsrA 8 DNA by a C, thus creating two GC base pairs.This oligonucleotide, called DsrA 8 DNA/GC (AAGCGCTT) forms a stable duplex as seen on Fig. 5B.The concentration dependency of the Tm (~ 35ºC) confirms that the self-assembly corresponds to an intermolecular duplex with this mutation, not to a hairpin (Fig. 5B).We also observe that this duplex is significantly more stable than that formed by DsrA 8 RNA as seen by the higher Tm (~ 25ºC for DsrA 8 RNA vs ~ 35ºC for DsrA 8 DNA/GC ).However, this difference is not surprising as G-C Watson-Crick base pair (A) UV spectroscopic analysis.Absorbance measurements were performed at 257 nm.Strand concentrations ranged from 0.5 µM (light grey) to 6 µM (dark grey).The reaction is intermolecular as shown by its concentration dependency.The resulting duplex is characterized by a Tm around 20°C. (B) van't Hoff plot.The thermodynamic parameters are calculated from (A).Briefly, concentrations of single stranded (C ss ) and double stranded (C ds ) species are extracted from the fraction observed on the thermal denaturation curve (f ss and f ds ).f ss corresponds to the distance between the lower baseline and the curve and f ds is the distance between the upper asymptote and the curve.The equilibrium constant is calculated from C ss and C ds as described in methods and the effective change in standard free energy of reaction is calculated as ∆Gr° = -RT ln K. ∆Sr° and ∆Hr° are extracted from the van't Hoff plot (R lnK = ∆Sr° -∆Hr°x1/T).The curve shown corresponds to a concentration of 1 µM of strand and mean error from 2 independent samples is indicated.(C) Structure of the G.U wobble base-pair, which is the most common non-Watson-Crick base pair present in RNA (Varani & McClain, 2000).Part A and C are adapted from (Cayrol et al., 2009).
Next, we analyzed the melting curve of DsrA 14 DNA .As seen on Fig. 5C, this DNA oligonucleotide forms a duplex, but this duplex is not as strong as that made by DsrA 14 RNA (compare with the first region of Fig. 3A and the thermodynamic parameters calculated from this curve in Fig. 6C with those of Fig. 3B).However, this result is also not surprising as RNA duplexes are known to be more stable than DNA duplexes, due to a slightly better base-stacking in the A-conformation for RNA base pairing (Ebel et al., 1994).
Then, we analyzed the melting curve of DsrA 22DNA .In this case we observed only one concentration dependent transition around 25ºC (Fig. 5D).This result was expected taking into account the low stability of DsrA 8 DNA duplex (see Fig. 5A).Indeed, DsrA 22 DNA is not able to form stable structures similar to those made by DsrA 22 RNA (Cayrol et (A) The first transition for DsrA 14 is clearly concentration dependent and corresponds to the intermolecular association of the self-complementary duplex (Tm around 45°C).By contrast, the second transition is independent of strand concentration and is characterized by a Tm around 60°C.This corresponds to the formation of an intramolecular hairpin.(B) and (C) van't Hoff plots.Thermodynamic parameters are calculated from (A) and correspond to formation of the intermolecular duplex (B) or to formation of the hairpin (C).As two equilibriums are observed in (A), the curve has been separated into two regions (~35-55°C and 55-70°C), which have been analyzed separately.C ss and C ds have been determined in each case independently for the duplex and for the hairpin and then used for calculation of the K D corresponding to each equilibrium.Part A is adapted from (Cayrol et al., 2009).(A) In spite of the two self-complementary regions, there is only one concentration dependent transition in the first part of this curve, corresponding to the self-assembly (Tm around 40-45°C).The dissociation above 40-50°C corresponds to the same intramolecular hairpin as that of DsrA 14 (Tm around 60°C).(B) van't Hoff plot corresponding to self-assembly.Thermodynamic parameters are those corresponding to the association of one subunit to the polymer.Part A is adapted from (Cayrol et al., 2009).Thermodynamics of DsrA self-assembly al., 2009), as seen on a native gel (Fig. 5F) and observed by molecular imaging with a transmission electron microscope (TEM).Thus, we tried to see if stabilization of the 8 bp duplex with G-C base pairs (equivalent position to that of DsrA 8 DNA/GC ) could allow the self-assembly.For this goal we used a DsrA 22 DNA/GC oligonucleotide.As seen in Fig. 5E, the melting of the 8 bp and 14 bp regions occurs simultaneously in this condition (~35ºC).Nevertheless, polymers were also not assembled efficiently in this case (Fig. 5F).This implies that reinforcing the stability of the 8 bp region with a dG-dC base pairing does not allow the building of the nanostructure, even if the 8 bp duplex is formed, and that the characteristics of the RNA molecules (A-form or presence of oligonucleotide not able to self-associate.As observed in this gel, DsrA DNA oligonucleotides are not able to form a stable structure in significant amounts.This result can be confirmed by a TEM analysis.DsrA 22 RNA spontaneously forms a lot of long polymers (TEM image from (Cayrol et al., 2009)); in the case of DsrA 22 DNA sequence, these structures are extremely rare and cannot be observed readily; this result is in agreement with our gel analysis indicating that DsrA 22 DNA is not able to form such a structure in significant amounts, in contrast to Ds-rA 22 RNA .Scale bar 100 nm.non-Watson-Crick base pairs) are also important for the construction of such a self-assembly.Indeed, the versatility of G.U wobble base-pair presumably gives more flexibility to the nucleic acid structure (Demeshkina et al., 2013) and likely favors the self-assembly, which is not possible within an equivalent DNA sequence.

CONCLUSIONS
The importance of noncoding RNAs in biology has become evident through numerous recent discoveries detailing their control of gene expression, especially in response to stress.We have previously shown that DsrA can self-assemble in vitro, resulting in long fibers that can be observed by molecular imaging (Cayrol et al., 2009).This polymeric structure indeed results from a 22 nt region including two minimal zones of dimerization (one of 14 and one of 8 bp) (Cayrol, et al., 2009).We present here for the first time the thermodynamic characterization of the self-assembly of a fragment of this natural sRNA, in particular the 14+8=22 nt region responsible of DsrA assembly.Two important conclusions can be drawn from this analysis: (i) the energy associated with DsrA self-assembly is enthalpy driven and relies on the stacking of base pairs one above the other and in the formation of H-bonds between bases.Conversely the entropic term is unfavorable to the self-assembly; (ii) the RNA feature of the nucleic acid is of primary importance for the self-assembly.This results both from the higher stability of RNA duplexes with an A-type helix relative to DNA and from the formation of non-conventional base pairing as the G.U wobble base-pair (Varani & McClain, 2000).These results thus encourage using ribonucleic acids to build nanostructures for in vivo applications.

Figure 1 .
Figure 1.The model for DsrA sRNA self-assembly.(A)Schematic drawing illustrating how DsrA sRNA self-assembly occurs.The 22 nt central region of DsrA (orange) located between two 5' and 3' stem-loops (SL1 in red and SL3 in pink) allows the self-assembly of the whole 87 nt sRNA(Cayrol et al., 2009).DsrA sRNA 5'-end is denoted by a ball and 3'-end by an arrowhead.Nuclease footprinting shows that this central region can adopt different secondary structures(Majdalani, et al., 1998;Lease & Belfort, 2000;Rolle, et al., 2006), which probably represents the possibility of the SL2 sequence being able to "breathe", allowing formation of intermolecular base-pairs with another DsrA.(B) Minimal sequence allowing DsrA self-assembly.DsrA self-assembly originates from two adjacent regions of 14 (light orange) and 8 nt (dark orange), resulting in a sequence of 22 nt that could polymerize by itself in the absence of the whole RNA sequence(Cayrol et al., 2009).The 22, 14 and 8 nt sequences of DsrA are annotated as DsrA 22 , DsrA 14 and DsrA 8 , respectively, throughout the manuscript.Figure adapted from(Cayrol et al., 2009).

Figure 2 .
Figure 2. Thermal denaturation of DsrA 8 RNA .(A)UV spectroscopic analysis.Absorbance measurements were performed at 257 nm.Strand concentrations ranged from 0.5 µM (light grey) to 6 µM (dark grey).The reaction is intermolecular as shown by its concentration dependency.The resulting duplex is characterized by a Tm around 20°C. (B) van't Hoff plot.The thermodynamic parameters are calculated from (A).Briefly, concentrations of single stranded (C ss ) and double stranded (C ds ) species are extracted from the fraction observed on the thermal denaturation curve (f ss and f ds ).f ss corresponds to the distance between the lower baseline and the curve and f ds is the distance between the upper asymptote and the curve.The equilibrium constant is calculated from C ss and C ds as described in methods and the effective change in standard free energy of reaction is calculated as ∆Gr° = -RT ln K. ∆Sr° and ∆Hr° are extracted from the van't Hoff plot (R lnK = ∆Sr° -∆Hr°x1/T).The curve shown corresponds to a concentration of 1 µM of strand and mean error from 2 independent samples is indicated.(C) Structure of the G.U wobble base-pair, which is the most common non-Watson-Crick base pair present in RNA(Varani & McClain, 2000).Part A and C are adapted from(Cayrol et al., 2009).

Figure 3 .
Figure 3. Thermal denaturation of DsrA 14 RNA .(A)The first transition for DsrA 14 is clearly concentration dependent and corresponds to the intermolecular association of the self-complementary duplex (Tm around 45°C).By contrast, the second transition is independent of strand concentration and is characterized by a Tm around 60°C.This corresponds to the formation of an intramolecular hairpin.(B) and (C) van't Hoff plots.Thermodynamic parameters are calculated from (A) and correspond to formation of the intermolecular duplex (B) or to formation of the hairpin (C).As two equilibriums are observed in (A), the curve has been separated into two regions (~35-55°C and 55-70°C), which have been analyzed separately.C ss and C ds have been determined in each case independently for the duplex and for the hairpin and then used for calculation of the K D corresponding to each equilibrium.Part A is adapted from(Cayrol et al., 2009).

Figure 4 .
Figure 4. Thermal denaturation of DsrA 22 RNA .(A)In spite of the two self-complementary regions, there is only one concentration dependent transition in the first part of this curve, corresponding to the self-assembly (Tm around 40-45°C).The dissociation above 40-50°C corresponds to the same intramolecular hairpin as that of DsrA 14 (Tm around 60°C).(B) van't Hoff plot corresponding to self-assembly.Thermodynamic parameters are those corresponding to the association of one subunit to the polymer.Part A is adapted from(Cayrol et al., 2009).

Figure 5 .
Figure 5. Analysis of DsrA DNA sequences self-assembly.(A) to (D) Thermal denaturation of DsrA DNA sequences: (A) DsrA 8 DNA ; (B) DsrA 8 DNA/GC ; (C) DsrA 14 DNA ; (D) DsrA 22 DNA ; (E) DsrA 22 DNA/GC .(F) Synthetic DsrA 22 self-assembly observed in a gel in native conditions: (1) DsrA 22 RNA ; (2) DsrA 22 DNA ; (3) DsrA 22 DNA/GC ; (4) Negative control of a dA 20oligonucleotide not able to self-associate.As observed in this gel, DsrA DNA oligonucleotides are not able to form a stable structure in significant amounts.This result can be confirmed by a TEM analysis.DsrA 22 RNA spontaneously forms a lot of long polymers (TEM image from(Cayrol et al., 2009)); in the case of DsrA 22 DNA sequence, these structures are extremely rare and cannot be observed readily; this result is in agreement with our gel analysis indicating that DsrA 22 DNA is not able to form such a structure in significant amounts, in contrast to Ds-rA 22 RNA .Scale bar 100 nm.