The Metabolome Weakens RNA Thermodynamic Stability and Strengthens RNA Chemical Stability

Abstract

We examined the complex network of interactions amongst RNA, the metabolome, and divalent Mg²⁺ under conditions that mimic the E. coli cytoplasm. We determined Mg²⁺ binding constants for the top 15 E. coli metabolites, comprising 80% of the total metabolome by concentration, at physiological pH and monovalent ion concentrations. These data were used to inform the development of an artificial cytoplasm that mimics in vivo E. coli conditions, which we term “Eco80”. We empirically determined that the mixture of E. coli metabolites in Eco80 approximated single-site binding behavior towards Mg²⁺ in the biologically relevant free Mg²⁺ range of ~0.5 to 3 mM Mg²⁺, using a Mg²⁺-sensitive fluorescent dye. Effects of Eco80 conditions on the thermodynamic stability, chemical stability, structure, and catalysis of RNA were examined. We found that Eco80 conditions lead to opposing effects on the thermodynamic and chemical stability of RNA. In particular, the thermodynamic stability of RNA helices was weakened by 0.69±0.12 kcal/mol while the chemical stability was enhanced ~2-fold, which can be understood using the speciation of Mg²⁺ between weak and strong Mg²⁺-metabolite complexes in Eco80. Overall, the use of Eco80 reflects RNA function in vivo and enhances the biological relevance of mechanistic studies of RNA.

Graphical Abstract

Introduction

RNA serves as the conduit of genetic information in the Central Dogma of Molecular Biology and performs numerous functions in cells owing to its capacity to form complex, diverse, and functional structures.¹ The development of genome wide structure-probing techniques in vivo has provided insight into RNA structure and function in cells.^2–5 However, most experimental techniques that provide insight into the mechanism and function of RNA cannot be readily performed in a cell, and are typically limited to simple conditions, usually 100 to 1,000 mM monovalent metal ions and 0.5 to 50 mM free divalent magnesium ions (Mg²⁺) with a dilute buffer.⁶ In vitro studies of RNA in conditions that mimic a cell, so called in vivo-like conditions, provide a link between experiments that probe RNA structure in vivo and experiments that provide mechanistically-tractable, biologically-relevant insight in vitro.⁶

Many studies have investigated the effects of individual components of the cellular environment on nucleic acid structure, including small molecules and non-biological crowders. Studies that used small molecules that are similar to metabolites indicated that these species interact strongly with the unfolded state of nucleic acids and destabilize secondary structure.^7–10 These destabilizing interactions between RNA and small molecules are analogous to the “quinary” interactions observed between proteins and the cytosol.¹¹ Studies that simulate cellular macromolecules revealed stabilized RNA tertiary structures, increased folding cooperativity, and improved RNA function in crowded enviroments.^12–16 Thermodynamic characterization of RNA and DNA helix formation in crowding conditions indicated that crowders generally destabilize helices.^17–19 In summary, using simple models to simulate the cellular environment has provided valuable insight into how the cell affects nucleic acids and motivated investigation of more complex and realistic artificial cytoplasms.²⁰

A number of studies have provided mechanistic insight into proteins in complex environments,²¹ ranging from cell lysates to live cells.^22–24 However, researchers sacrifice the control over the environment that is provided by a simple system. Mechanistic studies of RNA in cells or lysates have two additional problems. The first is the propensity of cells to degrade foreign RNA.^25,26 The second is the lack of control over Mg²⁺ speciation between free and chelated Mg²⁺.

Control over Mg²⁺ speciation is crucial for mechanistic studies of RNA because of the sensitivity of RNA folding to the concentration of Mg²⁺ in the solution, as demonstrated by thousands of studies and summarized.²⁷ Furthermore, recent studies have demonstrated the importance of metabolite-chelated Mg²⁺ complexes to RNA function.^28–30 These studies considered effects of mixtures of one to three metabolites, which is a step forward, but still far from the true complexity of the cellular environment. In addition, in these studies, Mg²⁺ speciation was approximated assuming single-site binding, meaning that one metabolite interacts with one Mg²⁺ ion, and binding constants were extrapolated from published sources, often reported at disparate ionic compositions and pHs.²⁷

Herein, we take a bottom-up approach that builds up complexity to make an artificial cytoplasm that contains 80% of E. coli metabolites by concentration, with biologically relevant concentrations of monovalent ions and free Mg²⁺ ions. We start by compiling metabolite concentrations in E. coli, simplify to the 15 most abundant metabolites, determine metabolite-Mg²⁺ binding constants at biologically relevant pH and ionic strength, and lastly determine the total Mg²⁺ concentration in the final mixture of metabolites. This bottom-up approach allows us to study the effects of the metabolite and metal ion species that comprise a major portion of the interactions that RNA experiences in E. coli cells.

Results

Eco80: An artificial cytoplasm containing 80% of E. coli metabolites by concentration

E. coli cells contain hundreds of different metabolites (~240 mM total),³¹ which is too many to test systematically. However, the 15 most abundant metabolites in E. coli, an experimentally-manageable number, comprise a full 80% (195 mM) of the total metabolites by concentration (Figure 1A). We thus sought to prepare Eco80, an artificial cytoplasm containing the biological concentrations of the 15 most abundant metabolites in E. coli (Table 1).

Figure 1.

Analysis of Mg²⁺ speciation in E. coli metabolite mixtures. (A) E. coli metabolome molar composition. ‘Eco80’ contains the 15 most abundant metabolites, which comprise 80% of the E. coli metabolome. ‘NTPCM’ contains the four strong Mg²⁺-chelating NTPs, and ‘WMCM’ contains 11 other weak Mg²⁺-chelating metabolites. (B-D) Effect of Mg²⁺ concentration on HQS emission without and with mixtures of metabolites that chelate Mg²⁺. Grey lines represent fits to determine the binding constant between Mg²⁺ and HQS. (E-G) Relationship between free Mg²⁺ concentration and the total Mg²⁺ concentration with mixtures of metabolites that chelate Mg²⁺. Hex bins represent a range of total and free Mg²⁺ concentrations simulated from artificial cytoplasm assuming single-site binding (colors correspond to density of simulated values in a hex bin, with yellow being most dense and purple being least dense). Triangle data points (black) are free Mg²⁺ concentrations calculated using HQS emission. Error bars represent the uncertainty in the free Mg²⁺ concentration from propagating errors in the HQS calibration curve fit. Black lines were generated using polynomial regression. The red shaded region is the biological free Mg²⁺ range of 0.5 to 3 mM. The red line is the approximate free Mg²⁺ concentration in E. coli of 2 mM. Downward red arrows represent the total Mg²⁺required to maintain 2 mM free Mg²⁺.

Table 1.

Eco80: The 15 most abundant metabolites, which comprise 80% of the E. coli metabolome by concentration.

Metabolite	Conc. (mM)a	KD (mM)	Chelation strengthe
ATP	9.63 (0.963)	0.28 (0.01)b	Strong (NTPCM)
UTP	8.29 (0.829)	0.248 (0.004)b	Strong (NTPCM)
GTP	4.87 (0.487)	0.201 (0.007)b	Strong (NTPCM)
dTTP	4.62 (0.462)	0.160 (0.003)b	Strong (NTPCM)
L-Glutamic acid	96 (9.6)	520 (50)c	Weak (WMCM)
Glutathione	16.6 (1.66)	NAd	Non (WMCM)
Fructose 1,6-bisphosphate	15.2 (1.52)	5.9 (0.1)b	Weak (WMCM)
UDP-N-acytylglucosamine	9.24 (0.924)	29 (2)b	Weak (WMCM)
Glucose 6-phosphate	7.88 (0.788)	17.3 (0.2)b	Weak (WMCM)
L-Aspartic acid	4.23 (0.423)	465 (12)c	Weak (WMCM)
L-Valine	4.02 (0.402)	NAd	Non (WMCM)
L-Glutamine	3.81 (0.381)	NAd	Non (WMCM)
6-Phospho-gluconic acid	3.77 (0.377)	14.4 (0.2)b	Weak (WMCM)
Pyruvic acid	3.66 (0.366)	15.8 (0.9)c	Weak (WMCM)
Dihydroxyacetone phosphate	3.06 (0.306)	20 (1)b	Weak (WMCM)

^aUncertainty is in parentheses and is propagated from uncertainties in reagent masses and volumes used during sample preparation. Extra significant digits included to avoid systematic rounding errors in the statistical model.

^bDetermined at 37 °C with ITC as measured in Figure S1 and Table S2. Error is the propagated standard error in the fit parameters.

^cDetermined at 37 °C with HQS emission as measured in the Figure S2 and Table S3. Error is the propagated standard error in the fit parameters.

^dNo binding observed as per Figure S2.

^eMetabolites with K_Ds for Mg²⁺ less than the free Mg²⁺ concentration in E. coli of 2 mM (top rows of table) are considered strong Mg²⁺ chelators and denoted “Strong” and K_Ds greater than 2 mM (bottom rows of table) are considered weak Mg²⁺ chelators and denoted “Weak”. Sub-artificial cytoplasms comprising Eco80, nucleotide triphosphate-chelated Mg²⁺ (NTPCM) and weak metabolite-chelated Mg²⁺ (WMCM) are noted.

Eco80 was prepared at a 2x concentration so that it could be diluted into other reagents and contain physiological concentrations of monovalent metal ions at pH 7.0 (see Supplementary Information (SI) Table S1 for details). Briefly, all metabolites in Eco80 were zwitterions or negatively charged near physiological pH 7, the latter requiring electrostatic neutralization with monovalent metal ions. Metabolite salts and free acids were prepared to a final 2x concentration, and the amount of Na⁺ and K⁺ added with each metabolite was recorded. Next, the pH of the 2x stock was adjusted to pH 7.0 using NaOH, and the amount of Na⁺ was again recorded. Lastly, NaCl and KCl were added to final concentrations of 480 mM Na⁺ and 280 mM K⁺, twice the physiological value of 240 mM Na⁺ and 140 mM K⁺.⁶ The 2x-concentrated artificial cytoplasm was then diluted into other reagents to give the final 1x concentration required for experiments.

Next, we considered how metabolites affect the speciation of Mg²⁺ between free and chelated forms. All 15 Eco80 metabolites have functional groups, phosphates and carboxylates, that drive chelating interactions with Mg²⁺ ions (Table 1).²⁷ We sought to quantify Mg²⁺ chelation by the metabolites in Eco80 at the physiological background, since Mg²⁺ binding affinity is dependent on environmental factors such as pH, ionic strength and identity, and temperature.^32–38

We determined apparent disassociation constants (K_D) for Eco80 metabolites in a background of 240 mM NaCl, 140 mM KCl, and pH 7.0 buffer at 37 °C (Table 1). Isothermal titration calorimetry (ITC) was used to measure the K_D for phosphorylated metabolites (Figure S1, Table S2). A fluorescence assay, which measures the free Mg²⁺ concentration in a sample using the divalent metal ion-binding dye 8-hydroxy-5-quinolinesulfonic (HQS) acid,³⁹ was used to estimate the K_D for metabolites that did not produce enough heat on binding Mg²⁺ to measure with ITC (Figure S2, Table S3). For this assay, Mg²⁺ was titrated into HQS solutions in the absence and presence of Mg²⁺ chelators. First, emission of HQS as a function of the total Mg²⁺ in the absence of chelators was independently performed in each panel and fit to a binding model for the binding of Mg²⁺ to HQS (Figure S2 top rows black data and fit). The free Mg²⁺ concentration, which is equal to the total Mg²⁺ concentration in this case, is then associated with the fluorescence emission for each data point using the binding model. This process is repeated in the presence of chelator, using the no-chelator data to obtain the free Mg²⁺ concentration at any total concentration of Mg²⁺ (Figure S2, bottom rows). Note that free and total Mg²⁺ concentrations are the same, y=x, in the absence of chelators, and that the data were right-shifted in the presence of chelators. The affinity of Mg²⁺ binding by a metabolite was thus obtained by fitting the free Mg²⁺ concentration as a function of the total Mg²⁺ concentration.

The binding affinity between Eco80 metabolites and Mg²⁺ ranged from strong to negligible (Table 1). The four most abundant nucleotide triphosphates, ATP, UTP, GTP, and dTTP, were classified as strong Mg²⁺ binders, with K_D values ranging from 0.160±0.003 mM to 0.28±0.01 mM, less than the approximate free Mg²⁺ concentration in E. coli of 2 mM (Table 1). Conversely, 8 metabolites--L-glutamic acid, fructose 1,6-bisphosphate, UDP-N-acetylglucosamine, glucose 6-phosphate, L-aspartic acid, 6-phospho-gluconic acid, pyruvic acid, and dihydroxyacetone phosphate--were classified as weak Mg²⁺ binders with K_D values greater than 2 mM (Table 1). Three other metabolites--glutathione, L-valine, and L-glutamine--had negligible Mg²⁺-binding properties, as measured with HQS (Figure S2). In an effort to understand the effects of Eco80 on RNA mechanistically, we created two sub-artificial cytoplasms: NTP-chelated Mg²⁺ (NTPCM) and weak metabolite-chelated Mg²⁺ (WMCM), comprised of the strong Mg²⁺ chelators (NTPs) and the weak/non Mg²⁺ chelators, respectively (Table 1).

We used two methods to estimate how Eco80 metabolites affect the speciation of Mg²⁺ between free and chelated. Our first method was the HQS assay that we used to estimate binding constants for metabolites, based on calculating the free Mg²⁺ concentration in the presence of metabolites using HQS fluorescence emission (Figure 1B–D, Table S4). The advantage of this method is that it directly determines the concentration of free Mg²⁺; it does not, however, report on speciation of Mg²⁺ to different metabolites. Our second method used a statistical model that accounts for experimental uncertainties in metabolite concentrations and K_D values and calculates Mg²⁺ speciation assuming single-site binding (meaning that one metabolite associated one Mg²⁺ ion). The advantage of this method is that it approximated Mg²⁺ speciation to different metabolites; it did not, however, directly determine free Mg²⁺ concentrations. This statistical model is described in detail in the Supporting Methods. Briefly, concentration errors were propagated from uncertainties in reagent masses and volumes used during sample preparation, and K_D uncertainties were obtained from the fits (Table 1). Both uncertainties were then randomly seeded into Equation 1 1000 times to create a series of virtual artificial cytoplasm with different errors. [Mg]_T is the total Mg²⁺ concentration, [Mg] is the free Mg²⁺ concentration, N is the total number of metabolites in a mixture, “i” is an integer representing each metabolite in a mixture, [L_i]_T is the total concentration of the i^th metabolite in a mixture, and K_D is the dissociation constant of the i^th metabolite.

$${\left[Mg\right]}_T=\left[Mg\right]+\sum _{i=1}^N\frac{{\left[L_i\right]}_T\left[Mg\right]}{{\text{K}}_{Di}+\left[Mg\right]}$$ (1)

Then, Equation 1 was solved numerically 1000 different times to determine a range of free Mg²⁺ concentrations produced at a given total Mg²⁺ concentration in a virtual artificial cytoplasm.

On the basis of agreement between the HQS data and the statistical simulation, methods 1 and 2, respectively, the two methods supported a model in which Mg²⁺ speciated in artificial cytoplasms largely according to single-site binding within the biological free Mg²⁺ concentration range of 0.5 to 3 mM Mg²⁺. However, at higher free Mg²⁺ concentrations, Mg²⁺ did not speciate according to a single-site model (Figure 1 E–G, black data points and hex bins deviate from each other).

In Eco80, the statistical model indicated that the metabolites should buffer the free Mg²⁺ concentration in the biological Mg²⁺ range, where a 20 mM increase in the total Mg²⁺ from 20 to 40 mM leads to only a 2.5 mM increase in free Mg²⁺ from 0.5 to 3 mM (Figure 1E, hex bins). Free Mg²⁺ concentrations measured over this range with HQS emission were consistent with this single-site behavior (Figure 1E). At higher free Mg²⁺ concentrations, Eco80 was expected to lose its free Mg²⁺-buffering capacity as chelators become saturated, and the free Mg²⁺ should have increased sharply with the total Mg²⁺ (model in Figure 1E, hex bins). However, the free Mg²⁺ concentration measured with HQS did not increase as fast as the statistical model predicted above 3 mM free Mg²⁺ (Figure 1E, compare black data points and hex bins). Free Mg²⁺ in Eco80 is expected to increase from 3 mM to ~100 mM as the total Mg²⁺ concentration is increased from 40 mM to 200 mM (Figure 1E, hex bins). However, the free Mg²⁺ concentration measured with HQS only increased from 3 mM to ~10 mM (Figure 1E, data points). One possibility is that multivalent interactions, where several Mg²⁺-saturated metabolites interact with additional Mg²⁺ molecules, dominate the equilibrium. Such non-single-site behavior above 3 mM free Mg²⁺ was also observed in the NTPCM and WMCM artificial cytoplasms (Figure 1 F & G), and was observed previously.²⁸

Lastly, we sought to empirically determine how much total Mg²⁺ is required to attain a free Mg²⁺ concentration of 2 mM in Eco80, NTPCM, and WMCM artificial cytoplasms. The relationship between the free Mg²⁺ calculated from HQS emission and the total Mg²⁺ concentration in each artificial cytoplasm was fit to a polynomial to empirically approximate the data (Figure 1 E–G, black lines), and the total Mg²⁺ concentration required to produce 2 mM Free Mg²⁺ was calculated from the polynomial fit (see methods for details). This resulted in predicted 31.6, 25.0, and 6.4 mM total Mg²⁺ to produce 2 mM free Mg²⁺ in Eco80, NTPCM, and WMCM, respectively (Table 2).

Table 2.

Total Mg²⁺ concentrations used to obtain 2 mM free Mg²⁺ in artificial cytoplasm.

Condition	[Total Mg2+] (mM)	[Chelated Mg2+] (mM)	[Free Mg2+] (mM)
Eco80	31.6	29.6 (±0.2)	2.0 (±0.2)
NTPCM	25.0	23.0 (±0.2)	2.0 (±0.2)
WMCM	6.4	4.4 (±0.2)	2.0 (±0.2)

^aUncertainty is 10%, double the maximum %uncertainty for HQS determination of the free Mg²⁺ concentration in the biological free Mg²⁺ range, propagated from uncertainties in the fit coefficients (See Supplemental Information 2).

Thermodynamic analysis of RNA helices in Eco80 by fluorescence-detected binding isotherms

We sought to understand how Eco80 affects the thermodynamic stability of RNA. Stability of RNA helices is traditionally measured with UV absorbance-detected melting curves, typically monitored at 260 or 280 nm.^40,41 However, absorbance melting curves could not measure helix stability in Eco80 because of the high absorptivity of the nucleotide metabolites. Thus, we pursued a fluorescence-detected binding isotherm assay.

Helix stability was monitored using emission of a 5′-fluorophore-labeled RNA strand (FAM-RNA) in equilibrium with a complementary 3′-quencher-labeled RNA strand (RNA-BHQ1) (Figure 2A). High emission indicates that the FAM-RNA is single-stranded, while low emission indicates that it is bound in duplex with RNA-BHQ1. We used a binding isotherm method, wherein increasing concentrations of RNA-BHQ1 are titrated into a constant concentration of FAM-RNA (Figure S3), resulting in a binding isotherm (Figure 2B). We favored binding isotherms over fluorescence-detected melts because of the very strong dependence of FAM emission on temperature.^42–44 Emission of FAM was monitored at different temperatures, resulting in an isotherm every 0.5 °C from 20 to 80 °C. Figure 2B shows a subset of these 121 isotherms.

Figure 2.

E. coli metabolite-Mg²⁺ mixtures destabilize RNA helices. (A) Layout of a fluorescence-detected binding isotherm assay in a real-time PCR machine. FAM emission is normalized to a passive ROX reference dye. (B) Fluorescence-detected binding isotherms fit to determine equilibrium constants with MeltR. Data points represent raw fluroescence data. Curves represent curve fits. Colors represent different temperatures (purple: 32.3, blue: 41.8, teal: 51.3, green: 54.6, yellow: 58.4, orange: 60.7, red: 63.1 °C). (C) Van’t Hoff relationship between the helix association equilibrium constant and temperature for helix 2:5′-CGCAUCCU-3′/5′-AGGAUGCG-3′ folding in background, Eco80, NTPCM, and WMCM. All conditions contain 2 mM free Mg²⁺, 240 mM Na⁺, and 140 mM K⁺. Points and error bars represent association constants and standard errors propagated from the fit MeltR. Lines represent fits to the van’t Hoff equation that MeltR used to calculate folding energies. (D) The Gibbs free energy at 37 °C (ΔG°₃₇) in Eco80, NTPCM, or WMCM compared to the ΔG°₃₇ in background for five RNA helices. Errors were propagated assuming 1.5% uncertainty in the ΔG°₃₇ (see methods for error analysis).

Raw fluorescence data were fit with MeltR, a program created by the authors, to determine folding energies. MeltR is a package of functions in the R programming language that allowed facile conversion of raw data to folding energies (see SI methods for details). MeltR calculated folding energies using two Van’t Hoff methods: (1) directly fitting a Van’t Hoff plot as a function of temperature (Figure 2C) and (2) globally fitting raw fluorescence emission.

Eco80 thermodynamically destabilizes RNA helices

We used fluorescence-detected binding isotherms to determine helix folding free energies in a background control of 240 mM NaCl and 140 mM KCl, and either Eco80, NTPCM, and WMCM for a set of five representative eight base-pair RNA helices; all solutions contained 2 mM free Mg²⁺ (Table 2). This helix set was designed to contain one or more representatives of each of the 10 Watson-Crick nearest neighbor parameters and vary in AU content from 25% to 75% (Table 3). Both of the aforementioned methods to determine folding free energies in MeltR agreed (Table S5); the results from the Van’t Hoff plots are reported in Table 3, which is ranked according to the AU content of the duplex. Errors in the main text are reported as 1.5% in terms of the ΔG°₃₇ and a detailed error analysis is available in the SI methods.

Table 3.

Stability of RNA helices in E. coli metabolite mixtures.

Sequencea	AU content (%)	Conditionb	ΔG°37 (kcal/mol)c	ΔΔG°37 (kcal/mol)c
1: 5’-CGGAUGGC-3’ 3’-GCCUACCG-5’	25%	Background	−15.60 (0.23)	--
		Eco80	−14.99 (0.22)	+0.61 (0.32)
		NTPCM	−15.28 (0.23)	+0.32 (0.33)
		WMCM	−15.23 (0.23)	+0.37 (0.33)
2: 5’-CGCAUCCU-3’ 3’-GCGUAGGA-5’	38%	Background	−13.84 (0.21)	--
		Eco80	−12.70 (0.19)	+1.17 (0.28)
		NTPCM	−13.42 (0.20)	+0.42 (0.29)
		WMCM	−14.22 (0.21)	−0.39 (0.30)
3: 5’-CGUAUGUA-3’ 3’-GCAUACAU-5’	63%	Background	−10.85 (0.16)	--
		Eco80	−10.41 (0.16)	+0.44 (0.23)
		NTPCM	−10.30 (0.15)	+0.55 (0.22)
		WMCM	−10.85 (0.16)	0.00 (0.23)
4: 5’-CCAUAUCA-3’ 3’-GGUAUAGU-5’	63%	Background	−12.03 (0.18)	--
		Eco80	−11.38 (0.17)	+0.64 (0.25)
		NTPCM	−11.50 (0.17)	+0.52 (0.25)
		WMCM	−11.53 (0.17)	+0.50 (0.25)
5: 5’-CCAUAUUA-3’ 3’-GGUAUAAU-5’	75%	Background	−10.76 (0.16)	--
		Eco80	−10.15 (0.15)	+0.61 (0.22)
		NTPCM	−10.16 (0.15)	+0.60 (0.22)
		WMCM	−9.96 (0.15)	+0.80 (0.22)

^aThe first sequence in each set was 5′-FAM labeled while the second sequence was 3′-BHQ1 labeled.

^bAll solutions contain 2 mM Free Mg²⁺, 240 Na⁺, 140 mM K⁺.

^cExtra significant digits were included to avoid propagating rounding errors

All five representative helices were significantly destabilized in Eco80 relative to the background condition, meaning the ΔΔG°₃₇ between the background condition and Eco80 was larger than its propagated uncertainty (Table 3, Figure 2D). Destabilization ΔΔG°₃₇ values ranged from +0.44±0.23 to +1.17±0.28 kcal/mol, with an average value of +0.69±0.12 kcal/mol (Table 3). We did not observe a clear relationship between AU content and destabilization in Eco80 (Figure S4). Thus, Eco80 destabilized RNA helices but the underlying sequence dependence was not apparent

To better understand how the various components of Eco80 contribute to destabilizing RNA helices, we analyzed the effects of the strong and weak Mg²⁺-chelating metabolites separately. NTPCM, which is comprised of strong Mg²⁺-chelating metabolites, consistently destabilized RNA helices (Figure 2D), with ΔΔG°₃₇ values ranging from +0.32±0.33 to 0.60±0.22 kcal/mol, with an average value of +0.48±0.12 kcal/mol (Table 3). The destabilizing effect of NTPCM appeares to be related to the AU content of the helix, with destabilization increasing linearly from +0.32 kcal/mol at 25% AU content to 0.60 kcal/mol at 75% AU content (R² = 0.9, Figure S4). This linear relationship could be because the NTPCM is comprised solely of NTPs and these can base pair most readily with A and U (see Discussion).

In contrast, WMCM, which is comprised of weak Mg²⁺-chelating metabolites, destabilized, had no effect, or stabilized RNA helices in a fashion that did not depend on AU content (Figure 2D, Table 3). The ΔΔG°₃₇ values ranged from −0.39±0.30 to +0.80±0.22 kcal/mol, with an average value of +0.26±0.12 kcal/mol, hardly above the noise (Table 3). Similar to Eco80, the sequence dependence of stabilization or destabilization was not clear (Figure S4).

Overall, the net effect of Eco80 on RNA helices was destabilization, with AU-content-dependent destabilizing interactions dominating for strong Mg²⁺-chelating metabolites, and a mixture of stabilizing and destabilizing interactions for weak Mg²⁺-chelating metabolites.

Eco80 protects RNA from chemical degradation

Several studies indicated that weak and strong Mg²⁺-chelating metabolites reduce Mg²⁺-mediated RNA degradation.^28,45 To assess whether Eco80 stabilizes the chemical structure of RNA, we used an in-line probing (ILP) assay, which takes advantage of the natural susceptibility of the RNA phosphodiester backbone to cleavage.⁴⁶ For ILP, the 2′-hydroxyl is deprotonated by a Mg²⁺-hydroxide (Mg²⁺-OH⁻), and serves as a nucleophile to attack the adjacent phosphate in a SN2-like mechanism (Figure 3A). Unstructured nucleotides are more susceptible to cleavage because they are more likely to adopt an in-line conformation that favors cleavage.⁴⁶ For this assay, 5′−³²P RNAs were incubated at 37 °C up to 90 h to facilitate in-line cleavage, with time points taken regularly. RNA fragments were then fractionated on a denaturing PAGE gel (Figure S5), providing single-nucleotide resolution for RNA degradation rates measured by the increase in counts with time for a given band. In-line degradation rates for RNA in Eco80, NTPCM, and WMCM with enough total Mg²⁺ to maintain 2 mM free Mg²⁺, were compared to degradation rates in a 2 mM free Mg²⁺ and a 25 mM free Mg²⁺ condition. All conditions contained 240 mM K⁺ and 140 mM Na⁺. The 25 mM free Mg²⁺ condition was chosen because it is a common free Mg²⁺ condition in vitro and is similar to the 25 and 31.6 mM total Mg²⁺ concentrations used for NTPCM and Eco80, respectively (Table 2).

Figure 3.

E. coli metabolite-Mg²⁺ mixtures stabilized the chemical structure of RNA. (A) ILP degradation mechanism facilitated by free Mg²⁺-OH⁻. (B) Secondary structure of the guanine riboswitch aptamer with tertiary contacts.⁴⁸ (C) Degradation rate for the guanine riboswitch aptamer at each residue in different solution conditions. (D-F) Degradation rate in different conditions grouped by structure. Groupings were based on analysis of crystal structures (Table S7). SS: Single-stranded, the base was not participating in hydrogen bonding interactions with other residues. NC: non-canonical, the base was forming non-canonical hydrogen bonding interactions in the tertiary structure. WC: Watson-Crick, the base was in a helix composed mostly of Watson-Crick base pairs.

We first used ILP cleaved versus time data to determine degradation rates for the guanine riboswitch aptamer (Figure 3B) in different artificial cytoplasms. The guanine riboswitch aptamer has been studied extensively, providing structural and mechanistic information.^47–49 We chose to study the guanine riboswitch in its guanine-ligand-unbound, apo, state for experimental simplicity. The expression platform was removed to prevent structural switching, and the guanine ligand was not added to favor the apo state. Moreover, guanine binding to the aptamer induces structural changes only at nucleotides directly mediating the guanine binding site,⁴⁷ indicating that information provided by X-ray crystal structures of the ligand bound aptamer is relevant for a structural analysis of our degradation rates.

Care was taken in our analysis to confirm that the guanine aptamer adopted a similar structure between conditions. The guanine riboswitch aptamer exhibited similar ILP patterns between the 2 mM free Mg²⁺, Eco80, NTPCM, and WMCM conditions, with high degradation in the 3’-region of the P2 stem and high reactivity in the L3 region, indicating that the apo guanine riboswitch aptamer adopts a similar structure in these conditions (Figure 3C, Figure S6). The 25 mM free Mg²⁺ condition exhibited higher degradation rates than the other conditions in the J2/3 junction (Figure 3C). This pattern was similar to ILP data published for another guanine riboswitch at a higher pH and a Mg²⁺ concentration of 15 mM,⁵⁰ supporting that the increase in degradation rates in the 25 mM free Mg²⁺ condition was dependent on the presence of Mg²⁺-OH⁻ complexes (Figure S6).

To further confirm that the guanine aptamer adopts similar structures in all conditions, we collected small angle X-ray scattering (SAXS) data on the apo form of the aptamer. Bell-shaped Kratky plots indicated that the structure of the guanine aptamer is folded between conditions (Figure S7A). To better understand how the solution state compares to the crystal structure, we generated a modeled SAXS curve from the crystal structure of the guanine aptamer (PDB 4FE5)⁴⁸ using WAXSiS to account for scattering by the ordered-solvent layer (Figure S7A, black line).⁵¹ The predicted 19.7±0.2 Å radius of gyration (R_g) of the model was smaller than the R_g calculated using Guinier analysis of the solution state, 24.5±0.2 Å in 2 mM free Mg²⁺, 23.9±0.2 Å in Eco80, 24.5±0.4 Å in NTPCM, 26.9±0.5 Å in WMCM, and 27.1±0.2 Å in 25 mM free Mg²⁺ (Table S6). These trends in radius of gyration were reproduced using paired-distance, p(r), distribution analysis and ab initio-electron-density reconstruction (Table S6),⁵² indicating that the apo guanine riboswitch aptamer adopts an expanded structure in solution compared to its ligand-bound crystal structure. Likewise, distance distributions in solution are right-shifted in comparison to the crystal structure (Figure S7B), and the shape of electron-density reconstructions (Figure S7C–G) are consistent with the crystal structure of the guanine riboswitch aptamer adopting expanded states in solution.

A more detailed interpretation of the SAXS data between solution conditions was confounded by the high noise in our data, ambiguous determination of the maximum distance between atoms (D_max) (Figure S7B, the distribution does not approach a limit at y=0), and inability to deconvolute scattering due to macromolecular shape and the composition of the ordered-solvent layer. However, there was consistent support for structural compaction in Eco80, with the radius of gyration changing from 24.5±0.2 to 23.9±0.4 Å, the D_max changing from 69.4 to 65.4 Å, and the excluded volume changing from 43,000 to 33,000 ų in 2mM free Mg²⁺ and Eco80, respectively (Table S6). This structural compaction in Eco80 is similar to the structural compaction previously observed in crowded conditions.^13,16

We sought to better characterize the structural dependence of RNA degradation in different conditions. We therefore extended the study to two other RNAs with well-defined tertiary structures, the CPEB3 ribozyme^53,54 and yeast tRNA^Phe (Figure S8, S9, & S10).⁵⁵ We inspected the crystal structures of these two RNAs, plus the original guanine aptamer, and manually classified each residue as single-stranded (SS), meaning that the base was not participating in hydrogen bonding interactions with other residues, non-canonical (NC), meaning that the base was forming non-canonical hydrogen bonding interactions with other residues in the tertiary structure, and Watson-Crick (WC), meaning that the base was in a helix comprised mostly of Watson-Crick base pairs (Table S7). Rates of ILP were then analyzed in box plots (Figure 3D–F).

We begin box-plot analysis with the guanine aptamer (Figure 3D). We had data for 3 single-stranded nucleotides without accompanying non-canonical hydrogen bonding interactions. We observed decreased degradation rates at the single stranded (SS) nucleotides in 2 mM free Mg²⁺, Eco80, NTPCM, and WMCM, in comparison to the 25 mM free Mg²⁺ condition. Likewise, we observed an overall decrease in reactivity for nucleotides involved in non-canonical tertiary interactions (NC) in 2 mM free Mg²⁺, Eco80, NTPCM, and WMCM in comparison to the 25 mM free Mg²⁺ condition. In contrast, degradation rates for nucleotides participating in Watson-Crick base pairing interactions were independent of solution conditions. Thus, we observed a trend of protection from degradation in artificial cytoplasm for SS and NC bases specifically, even with similar amounts of total Mg²⁺ in solution as the 25 mM free Mg²⁺ condition.

We repeated our degradation assay with the self-cleaved CPEB3 ribozyme and yeast tRNA^phe, to test if the protection from degradation in artificial cytoplasm was broadly applicable (Figure S8, S9, & S10). For the cleaved-CPEB3 ribozyme, degradation rates at single-stranded residues were reduced in 2 mM free Mg²⁺, Eco80, and NTPCM conditions in comparison to the 25 mM free Mg²⁺ condition (Figure 3E). Interestingly, the degradation rates of single-stranded residues recovered in WMCM, indicating that degradation rates were partially dependent on the strength of Mg²⁺ chelation. Likewise, the degradation rates for residues that were predicted to participate in non-canonical tertiary contacts were reduced in 2 mM free Mg²⁺, Eco80, and NTPCM, but not WMCM, in comparison to the 25 mM free Mg²⁺ condition, further indicating that degradation rates were dependent on the strength of Mg²⁺ chelation. Degradation rates were similar for nucleotides participating in Watson-Crick base-pairs between all conditions.

Yeast tRNA^phe exhibits almost no in-line degradation except for the single-stranded nucleotides in the P3 stem loop (anticodon loop, Figure S10D). Single-stranded nucleotides showed reduced degradation rates in the 2 mM free Mg²⁺, Eco80, and NTPCM conditions in comparison to the 25 mM free Mg²⁺ condition, while degradation rates once again recovered in WMCM (Figure 3F). Degradation rates were constant across conditions for nucleotides that form Watson-Crick base pairs and non-canonical contacts, which is different than the increased degradation observed for nucleotides that form non-canonical contacts in the guanine riboswitch aptamer and the CPEB3 ribozyme. One possible explanation is that the tertiary structure of tRNA^phe is less dynamic than the tertiary structure of the guanine riboswitch aptamer and the CPEB3 ribozyme, thus reducing the degradation rates in regions that participate in non-canonical tertiary interactions to the baseline levels.

Overall, the in-line degradation assay indicated that Eco80 and NTPCM protect RNA from Mg²⁺-OH⁻-mediated degradation in structural regions that are susceptible to in-line cleavage, even though both artificial cytoplasms have relatively high concentrations of total Mg²⁺. WMCM showed an intermediate effect between the high degradation rates in the 25 mM free Mg²⁺ condition and the low degradation rates in 2 mM free Mg²⁺, Eco80, and NTPCM conditions, indicating that degradation rates were dependent on Mg²⁺-chelation strength (see Discussion).

Eco80 supports RNA catalysis

Weak metabolite-chelated Mg²⁺ is known to promote catalysis by ribozymes. For example, CPEB3 ribozyme catalysis is enhanced ~1.6-fold by an estimated 2 mM free Mg²⁺ in solution with 11.3 mM glutamate-chelated Mg²⁺, in comparison to catalysis in 2 mM free Mg²⁺ alone.²⁸ Thus, we hypothesized that Eco80 metabolites would also promote CPEB3 catalysis.

We compared CPEB3 ribozyme cleavage rates in 2 mM free Mg²⁺ and 25 mM free Mg²⁺ to Eco80, NTPCM, and WMCM containing enough total Mg²⁺ to produce 2 mM free Mg²⁺. All conditions contained 240 mM Na⁺ and 140 mM K⁺. Briefly, we purified full length CPEB3 ribozyme (Figure 4A), incubated it in artificial cytoplasm, fractionated time points on a denaturing acrylamide gel, and calculated the fraction cleaved from the relative intensity of cleaved and uncleaved RNA bands (Figure S11). The fraction cleaved as a function of time was fit to a single-exponential equation to estimate the reaction rate constant (Figure 4B).

Figure 4.

Eco80 supports CPEB3-ribozyme catalysis. (A) Secondary structure of the uncleaved CPEB3 ribozyme. (B) Fraction cleaved CPEB3 as a function of time fit to a single exponential. Four technical replicates are displayed. The labels ‘2 mM Free’ and ‘25 mM Free’ refer to the Mg²⁺ concentration. All conditions contain a background of 240 mM Na⁺ and 140 mM K⁺. Enough total Mg²⁺ was added to Eco80, NTPCM, and WMCM to maintain a 2 mM free Mg²⁺ concentration (Table 2). (C) Rate constant (k) for the CPEB3 ribozyme in different conditions. k_rel is the relative rate constant in comparison to the 2 mM free Mg²⁺ condition. (D) Composition of artificial cytoplasms comprised of 80% of yeast and mammalian iMBK metabolites, termed ‘Yeast80’ and ‘Mammal80’, respectively, compared to the composition of Eco80. Each box represents one abundant metabolite. ‘NTPCM’ represents nucleotide metabolites, and ‘WMCM’ represents metabolites which were expected to weakly chelate Mg²⁺, with K_Ds greater than 2 mM.

In comparison to the 25 mM free Mg²⁺ control, CPEB3 ribozyme catalysis was reduced in all conditions (Figure 4C). Surprisingly, CPEB3 catalysis was reduced in Eco80 by ~1/2 in comparison to the 2 mM free Mg²⁺ control, despite the 31.6 mM total Mg²⁺ in Eco80. CPEB3 catalysis was reduced by ~1/3 in NTPCM in comparison to the 2 mM free Mg²⁺ control, an slightly stronger inhibitory effect than Eco80. In contrast, CPEB3 catalysis was enhanced ~1.3 fold in WMCM, similar to the enhancement observed for glutamate-chelated Mg²⁺.²⁸ In summary, Eco80 supports RNA catalysis albeit not in an enhanced fashion in comparison to the 2 mM free Mg²⁺ condition. CPEB3 reaction rates in Eco80 were between the rates in WMCM and NTPCM. WMCM likely had exposed Mg²⁺ to help fold the RNA, while NTPCM did not, and moreover may denature the RNA as per Figure 2 (see Discussion).

WMCM may be more biologically relevant than Eco80 for studying CPEB3 ribozyme activity. We performed an analysis of absolute metabolite concentrations in yeast and mammalian iMBK cells, which have a closer evolutionary relationship than E. coli to human cells, where CPEB3 exists in nature (Figure 4D). Absolute metabolite concentrations were compiled from the literature⁵⁶ and the 11 most abundant metabolites that comprise 80% of the yeast and mammalian metabolome were selected to assess hypothetical Yeast80 and Mammal80 artificial cytoplasms. Estimated metabolite-Mg²⁺ binding constants²⁷ were used to classify each metabolite in Yeast80 and Mammal80 as strong (NTPCM) or weak (WMCM) Mg²⁺ chelators. We found that Yeast80 and Mammal80 would be depleted in strongly chelated-Mg²⁺, with Yeast80 having no strong Mg²⁺ chelators and Mammal80 having only ~4 mM strong Mg²⁺ chelators. (Figure 4D). Thus, the 1.3-fold rate enhancement in WMCM is more relevant to CPEB3 function in human cells than the rate decrease in Eco80.

Discussion

In this study, we used a bottom-up approach to create a complex, yet still manageable artificial cytoplasm, termed Eco80, which encapsulated 80% of the E. coli metabolome (Figure 5A). In order to provide mechanistic insight into the effects of Mg²⁺ speciation on RNA in cells, we also broke down Eco80 into sub-artificial cytoplasms, which contain either metabolites that strongly chelate Mg²⁺ (i.e. NTPs), or metabolites that weakly chelate Mg²⁺.

Figure 5.

Models describing the destabilization of RNA helices and stabilization of RNA chemical structure by Eco80. (A) Semi-quantitative molecular representation of an RNA helix in Eco80. The average number of molecules (colored sphere models) in Eco80 that would occupy a sphere with a 50 Å radius were placed randomly around an 8 base-pair RNA helix using Pymol (blue cartoon, PDB 1SDR). Mg²⁺ ions are represented with teal spheres. Solvent (red wires) and K⁺ (blue spheres) where modeled using WAXSiS.⁵¹ (B-C) Mechanism for destabilization of helices by metabolites and stabilization of helices by Mg²⁺. Net effect of metabolite-chelated Mg²⁺ combines metabolite interactions (red, white, blue) favoring the unfolded state and Mg²⁺ interactions (green) favoring the helical state. (D-E) In-line degradation of the RNA backbone mediated by Mg²⁺ hydroxide species is inhibited by Mg²⁺ chelation.

Importantly, we adopted the Mg²⁺ sensitive dye, HQS,³⁹ to measure Mg²⁺ speciation in artificial cytoplasms. A key challenge to studying RNA under in vivo-like conditions is knowing how components affect the speciation of Mg²⁺ between free and chelated. Published binding constants for cellular components can be unreliable, as they typically apply only to solutions with specific ionic character,³² and more often, binding constants are not known at all.²⁷ Lastly, predicting Mg²⁺ speciation using binding constants requires making assumptions about the stoichiometry of Mg²⁺-component complexes, which may or may not be valid. For example, in this work, the free Mg²⁺ concentration in Eco80, NTPCM, and WMCM measured using HQS approximated the free Mg²⁺ concentration that was calculated using our measured binding constants, when the free Mg²⁺ was in the biological range of 0.5 to 3 mM. However, the calculation was not accurate at higher free Mg²⁺ concentrations where interactions of Mg²⁺ with more than one metabolite became likely (Figure 1 E–G). Thus, the HQS assay provided invaluable information on Mg²⁺ speciation in biologically-relevant solutions, without requiring assumptions or Mg²⁺ binding constants and interaction coefficients among the many metabolites. Although we used this assay to directly measure Mg²⁺ speciation in mixtures of metabolites, it could be applied to Mg²⁺ interactions with other biological molecules.

Eco80 is a significant step towards reconstituting the cytoplasm in vitro but Eco80 is still a simplification. First, the cell contains 228 other metabolites, the 20% of the metabolome that was not included in Eco80. While it was not feasible to test in this study, we expect the remaining 20% of the metabolome to reinforce the effects of Eco80 because the remaining 20% of the metabolome has a similar, but slightly higher, composition of strong Mg²⁺ chelating metabolites to Eco80, 19% and 14% respectively, and because each individual component makes up less than 1.1% of the total metabolome.²⁷ The remaining 20% of the metabolome also shares structural features with the metabolites in Eco80, with less than five of the remaining 228 metabolites expected to carry a net positive charge at physiological pH. The second simplification is that Eco80 does not contain biological macromolecules such as RNA and proteins, or other biological divalent metal ions such as Zn²⁺ or Ca²⁺. While the effects of these cellular components were outside of the scope of our study, the effects could be interrogated using a similar theoretical and experimental treatment.

Our Mg²⁺ speciation calculations and HQS experiments indicated that metabolites play an important role in buffering the free Mg²⁺ concentration in cells. Recent theoretical and experimental studies have demonstrated that the cellular environment buffers the concentration of biological molecules, effectively reducing concentration noise in vivo.^57,58 In our system, single-site-Mg²⁺ interactions in Eco80 buffers the free Mg²⁺ concentration between just 0.5 and 3 mM Mg²⁺, in the presence of a large total Mg²⁺ change between 20 and 40 mM Mg²⁺. This buffering effect was exaggerated at high total Mg²⁺ concentrations in Eco80, where an increase in the total Mg²⁺ concentration to an astounding 200 mM increased the free Mg²⁺ concentration to only ~10 mM.

Our thermodynamic analysis of RNA helices in Eco80 indicated that the E. coli metabolome had a net destabilizing effect on RNA helices of about +0.69±0.12 kcal/mol, with destabilizing effects dominating for NTPCM at about +0.48±0.12 kcal/mol and a mixture of destabilizing and stabilizing effects observed for WMCM averaging at about +0.26±0.12 kcal/mol (Figure 2D).

The destabilizing effect of Eco80 on RNA helices can be compared to a previous study from our laboratory that demonstrated that amino acid-chelated Mg²⁺ stabilized the tertiary fold and increased the folding cooperativity of RNA structures.²⁸ The earlier study only accounted for amino acid-chelated Mg²⁺, a weakly-chelated Mg²⁺ species, and used approximate binding constants. Our work herein is not inconsistent with increased tertiary structure stabilization, as concomitant secondary structure destabilization and tertiary structure stabilization is an important driving force of cooperative folding for biological RNA.¹⁶ Indeed, our SAXS analysis was consistent with increased tertiary compaction of the apo-guanine aptamer in Eco80 (Table S6), even with the destabilization of RNA helices observed in Figure 2.

This apparently small, +0.69 kcal/mol, destabilizing effect in Eco80 on RNA helices could have beneficial effects on the transcriptome in vivo. First, short, marginally-stable helices would depopulate, leading to accessible RNA regions that could interact with proteins and regulatory small RNA. In contrast, stable RNA secondary structures would still form functional structures. Second, global destabilization of RNA helices should also lead to destabilization of kinetic traps in RNA folding pathways in vivo, such as the misfolds observed for the Tetrahymena ribozyme.⁵⁹

Moreover, the destabilizing effect of Eco80 suggests that RNA structures are more dynamic in the cell than in vitro. Helicases and RNA chaperones have been proposed to play a central role in dynamics in cell because interconversion between RNA structures requires breaking base pairs, which is energetically “expensive”. The weak destabilization we observed in this work, about +0.69 kcal/mol, could have a large effect on RNA dynamics in vivo, when extrapolated to the entire transcriptome, outside of the impact of proteinous RNA chaperones. Another consideration is that helicases require the hydrolysis/expenditure of NTPs. A 0.69 kcal/mol decrease in the energy required for every helix that a helicase must unwind could provide major energetic savings for the cell.

Weakening of RNA helix stability in Eco80 can be understood using a model that combines established effects of polar small molecules and Mg²⁺ on RNA helix stability. Polar small molecules are known to interact favorably with the exposed bases in the unfolded state (Figure 5B).^7–10 Likewise, Mg²⁺ is known to interact favorably with the high density of negative charges in helical RNA. Thus, metabolites may destabilize helices by favoring the unfolded state and Mg²⁺ stabilizes helices by favoring the helical state (Figure 5C). The changes in helix formation energy in Mg²⁺-metabolite mixtures demonstrate a balance between metabolites favoring the unfolded state and Mg²⁺ favoring the helical state (Figure 5C, right). For example, NTPCM strongly chelates Mg²⁺, thus sequestering Mg²⁺ from interacting with the folded state, so that the destabilizing interactions between NTPs and RNA dominate, which lead to a consistent destabilization of RNA helices (Figure 2D). In contrast, WMCM only weakly sequesters Mg²⁺, so that Mg²⁺ is available for favorable interactions with helices. This leads to the inconsistent destabilization and even stabilization of RNA helices observed in WMCM (Figure 2D), dependent on the relative strength of stabilizing Mg²⁺-RNA interactions and destabilizing metabolite-RNA interactions (Figure 5C).

NTPCM destabilized AU-rich helices more than GC-rich helices (Figure S4). A similar destabilizing effect on RNA G-quadruplex structures has been reported specifically for cytidine nucleotides.⁶⁰ Interestingly, in the case of G-quadruplex structures, other nucleotides (A and U) had a smaller destabilizing effect, suggesting that G-quadruplexes are destabilized by favorable base-pairing interactions between cytidine nucleotides in solution and Gs in the unfolded state of the RNA. The NTPCM is comprised mostly of ATP, UTP, and dTTP (22.5 mM total versus 4.9 mM GTP). The ATP, UTP, and dTTP are expected to form stronger hydrogen bonds with Us and As, respectively, than Cs and Gs in the unfolded state of RNA, supporting the AU dependence of helix destabilization by NTPCM.

Our analysis of RNA degradation in Eco80 indicated that metabolites protect susceptible regions of RNA from Mg²⁺-OH⁻ mediated degradation (Figure 3). Eco80 and NTPCM had the strongest protective effects, while WMCM had an intermediate protective effect, indicating that protection from degradation is dependent on the strength of the chelating interaction between metabolites and Mg²⁺. In this model, in-line cleavage of the RNA backbone is limited by the formation of Mg²⁺-OH⁻ species, which is favorable for free Mg²⁺, unfavorable for weakly-chelated Mg²⁺, and negligible for strong NTP-chelated Mg²⁺ (Figure 5D). Thus, RNA degradation rates were weakly reduced by depletion of active Mg²⁺-OH⁻ species in the presence of weak Mg²⁺ chelators and strongly reduced by depletion of active Mg²⁺-OH⁻ species in the presence of strong Mg²⁺ chelators (Figure 5E).

Our analysis of CPEB3 catalysis in Eco80 indicated that metabolite-Mg²⁺ mixtures support RNA catalysis. A previous study of hammerhead ribozyme catalysis in the presence of nucleotides found that reaction rates were enhanced by NDP-chelated Mg²⁺, a weakly-chelated Mg²⁺ species, and that NTP-chelated Mg²⁺ had no effect on reaction rates.²⁹ Similarly, our results in metabolite mixtures found that WMCM weakly enhanced CPEB3 ribozyme catalysis while NTPCM weakly inhibited CPEB3 ribozyme catalysis. A previous study of the CPEB3 ribozyme in the presence of weak amino acid-chelated Mg²⁺ indicated that rate enhancement was not driven by direct interactions between amino acid-chelated Mg²⁺ and the catalytic site, but by stabilization of catalytically relevant CPEB3 ribozyme structure.²⁸ In contrast, we observed thermodynamic destabilization of helices and reduction of CPEB3 catalysis in Eco80 and NTPCM, indicating that reduction in catalysis was caused by destabilization of the catalytically relevant structure. Thus, ribozyme rate enhancement in vivo is likely dependent on the presence of weak Mg²⁺ chelators that stabilize the catalytically relevant structure and depletion of strong Mg²⁺ chelators that destabilize the catalytically relevant structure.

Eco80 had opposing effects on the thermodynamic and chemical stabilities of RNA, which reflected the complexity of the cellular environment. The thermodynamic stability of RNA helices was weakened by Eco80, the chemical stability of RNA was enhanced by Eco80, and the catalysis of RNA was supported by Eco80. These seemingly contradictory effects can be understood using the speciation of Mg²⁺ between weak and strong Mg²⁺-metabolite complexes. The effects of Eco80 reflect RNA function in vivo, enhance the biological relevance of mechanistic studies of RNA, and suggest possible ways to enhance the storage of mRNA vaccines.

ABBREVIATIONS

RNA

ribonucleic acid

DNA

deoxyribonucleic acid

Mg²⁺

divalent magnesium ion

HQS

8-Hydroxy-5-quinolinesulfonic acid

Eco80

80% of all E. coli metabolites

NTPCM

nucleotide triphosphate-chelated Mg²⁺

WMCM

weak metabolite-chelated Mg²⁺

ITC

isothermal titration calorimetry

SAXS

small angle X-ray scattering

ILP

in-line probing