Ca²⁺ Induces the Formation of Two Distinct Subpopulations of Single Group II Intron Molecules

Abstract

The folding pathway of the Sc.ai5γ derived group II intron ribozyme D135 is highly specific to the correct M²⁺ cofactor. Upon partial replacement of Mg²⁺ with Ca²⁺, the molecules split into two distinct static subpopulations that are not interchangeable. Type 2 molecules thereby form a compact but misfolded state.

The phosphate sugar backbone imposes a large negative charge on RNA and DNA that needs to be neutralized for three dimensional-structure assembly. In addition to proteins and polyamines, the most important cofactors that bind nucleic acids are metal ions. They help to overcome repulsion forces and mediate higher order structure formation, and they frequently participate directly in the chemical reaction of catalytic nucleic acids, i.e. ribozymes.[1] Catalytic RNAs show a distinct specificity for metal ions, both with respect to the kind as well as their concentration.[2–6] Mg²⁺ is the most abundant divalent metal ion in the cell and is often considered the natural cofactor for ribozymes.

Self-splicing group II introns rank amongst the largest ribozymes known and are found in organellar genes of lower eukaryotes, fungi, plants and bacteria.[7] They represent large molecular machines able to perform autocatalysis.[7] In vitro, the D135 ribozyme derived from the S. cerevisiae group II intron Sc.ai5γ displays an optimal activity at unphysiologically high metal ion concentrations (500 mM KCl, 50–100 mM MgCl₂).[8] This ribozyme consists of domains 1, 3, and 5, contains all the elements necessary for activity and represents the best investigated system for group II intron folding.[9,10] D135 folds in a Mg²⁺ dependent fashion.[10] Starting from the unfolded state U in the presence of monovalent metal ions only, two transient on-pathway intermediates I (extended intermediate) and F (folded intermediate) are observed before reaching the native state N by the addition of Mg²⁺. I and F are in fast dynamic equilibrium separated by low energy barriers.[10] Interestingly, increasing [Mg²⁺] activate the structural dynamics, directly linking dynamics to ribozyme activity.[10]

The hammerhead ribozyme retains activity with Ca²⁺ only,[3,11] and the Tetrahymena group I intron was shown to globally fold with Ca²⁺ only, but requires small [Mg²⁺] for catalysis.[12] Group II introns instead are inhibited already by very low [Ca²⁺];[4] Ca²⁺ thereby actively replaces Mg²⁺ ions in the folded state.[4] Calcium inhibition is particularly interesting because the cox1 gene coding for Sc.ai5γ is located in mitochondria and these cellular compartments are not only involved in Ca²⁺ homeostasis, serving as Ca²⁺ storage pools, but both the cox1 gene and altering Ca²⁺ levels are also involved in apoptosis.[13–15] Therefore, the ribozyme splicing activity and thus correct cox1 expression is potentially regulated by local [Ca²⁺] in mitochondria.

Until now it was impossible to distinguish whether the role of Ca²⁺ is reflected in the disturbance of the overall ribozyme structure or the replacement of one or more Mg²⁺ ions directly involved in catalysis. Both of these effects would result in a similar, undistinguishable inhibition in a standard cleavage assay. To address this, we have characterized the folding of the group II intron D135 ribozyme in the presence of Ca²⁺ independently of catalysis using single-molecule Förster Resonance Energy Transfer (smFRET).

Calcium(II) ions affect the overall distribution of conformational states in the D135-L14 ribozyme

The fluorophore labeled D135-L14 ribozyme has been previously characterized[10] and shown to be catalytically competent in cleaving substrate RNA (Supporting Figure S1). We have now carried out smFRET experiments in the presence of 0, 2, 5, 7, or 10 mM Ca²⁺ in a background of Mg²⁺ to give a total [M²⁺] of 100 mM. Cumulative FRET distribution histograms from single-molecule time trajectories are shown in Figure 1a (30–50 molecules each). In the presence of Mg²⁺ only, the two intermediate folding states (I and F, FRET ≈ 0.25 and 0.4) are equally populated, while the native state (N, FRET ≈ 0.6) is clearly a minor population (Figure 1a, top).[10] Upon addition of Ca²⁺ a distinct population transfer takes place: The magnitude of the 0.4 FRET state decreases as the 0.6 FRET state increases dramatically (Figure 1a). Interestingly, the peak value of the highest FRET distribution also shifts from ~0.60 to ~0.54 as [Ca²⁺] increases to 10 mM.

Figure 1.

Distribution histograms of smFRET time traces in the presence of Ca²⁺. a) Upon addition of up to 10 mM Ca²⁺ (in a total divalent metal ion concentration of 100 mM completed with the addition of Mg²⁺), the contribution of the high FRET state at ~0.6 dramatically increases. Concomitantly, the mean maximal FRET value shifts from 0.6 to 0.54 and the peak height of the 0.4 FRET state decreases. b) Histograms of single molecule trajectories separated into subpopulations type 1 (left) and type 2 (right) at the indicated Mg²⁺: Ca²⁺ mixtures. The two subpopulations are distinctly different.

The presence of Ca²⁺ leads to the appearance of two distinct subpopulations and a new misfolded species

To characterize the increase of the 0.6 FRET state in the presence of Ca²⁺, we analyzed the single time trajectories individually (Figure 2). In the presence of Mg²⁺ only, three reoccurring conformational states of D135-L14 are observed at FRET values ~0.25, 0.4 and 0.6, corresponding to I, F and N. These three states are in fast equilibrium and the 0.6 state is only transiently occupied (Figure 2, top). In the presence of Ca²⁺ two distinctively different trace types appear. Type 1 traces are similar to those observed with Mg²⁺ only (Figure 2, middle). In contrast, type 2 traces show a completely new behavior: The molecules predominantly populate the high FRET state, and exhibit direct transitions from 0.25 to 0.6 FRET and fewer transitions to 0.4 FRET (Figure 2, bottom). The 0.4 state apparently describes a conformation along a minor folding pathway, although it is not clear whether the transitions from 0.6 to 0.25 are direct or if they include a very short intermission in the 0.4 FRET, which we miss because of our 33 ms time resolution. The intermediate 0.4 FRET state is reached from both the 0.25 and 0.6 states. Yet FRET 0.4 is not an on-pathway state as the molecules fall back to their original state rather than continuing to the respective third state. Interchanges from single molecules from type 1 to type 2 were also not observed within the ~1 min observation window. This reflects a certain memory effect of the single molecules similar to earlier observations.[16]

Figure 2.

Exemplary single molecule traces at 100 mM Mg²⁺ (top) and of the two subpopulations at 90 mM Mg²⁺: 10 mM Ca²⁺ (middle and bottom). The FRET histograms (right) reveal two distinct types of traces in the presence of Ca²⁺.

Cumulative histograms of the single molecule traces split into type 1 and type 2 traces strongly support the existence of two distinct subpopulations (Figure 1b). Type 1 histograms show large peak distributions at 0.25 and 0.4 FRET and a consistently minor distribution at ~0.6 FRET, which only increases slightly with [Ca²⁺]. In contrast, type 2 traces primarily populate the low and high FRET states, but hardly the intermediate state. Careful inspection of the type 2 histograms also reveals a shift in the maximum of the high FRET distribution from 0.58 to 0.53, whereas the peaks at the two lower FRET states remain unchanged (Figure 1b). This can be explained by concluding that type 2 molecules no longer reach the native state N. Instead, they form a new conformational (misfolded) species M with a FRET state that is clearly distinguishable from the native state N at 0.6 FRET.

We further determined the relative amount of type 2 traces as a function of [Ca²⁺] (Figure 3). The fraction of type 2 molecules increases linearly up to 50% in 10 mM [Ca²⁺], which correlates nicely with the loss of function in the presence of Ca²⁺ ions.[4] This further supports the existence of two subpopulations and explains the observed progression in the overall FRET histograms that include both type 1 and type 2 traces (Figure 1a).

Figure 3.

Fraction of type 2 molecules as a function of [Ca²⁺]. A linear relationship is revealed (slope = 4.5 ± 0.4 mM⁻¹).

Folding rates in the presence of Ca²⁺ support the accumulation of the molecules in the high FRET state

Folding rates for type 1 molecules were determined in a range of Mg²⁺/Ca²⁺ ratios (25–50 molecules each) using dwell time analysis as described.[10] Both forward (k₁₁ and k₁₂) and both backward rates (k₋₁₁ and k₋₁₂) either increase linearly with increasing [Ca²⁺] or stay the same within the error limits (Figure 4a,b). The ratios of forward and backward rates agree with the observation that the high FRET state remains a minor species but its contribution increases in the presence of Ca²⁺ (Figure 1b).

Figure 4.

Folding pathway of D135-L14 in the presence of Mg²⁺ and Ca²⁺. a) In the presence of Mg²⁺ the ribozyme shows a linear folding pathway from the unfolded state U via the on-pathway intermediates I (unfolded intermediate) and F (folded intermediate) to the native state N.[10] b) In the presence of Ca²⁺ two different types of subpopulations appear. Type 1 traces depict a pathway similar to the Mg²⁺ case with slightly altered rates. Molecules of type 2 predominantly fold directly from I to a misfolded species M. Some type 2 molecules follow a minor pathway passing F on the way from I to M. The maxima of the FRET distributions are given in blue. All rates are in s⁻¹ and correspond to 10 mM Ca²⁺/90 mM Mg²⁺. The individual error limits are estimated[17] to be about 50% of the given values. The first step of folding, i.e. the addition of monovalent ions to the RNA, is not shown. Note that we cannot distinguish if I is the dividing point or if it already differs for the two subpopulations (see also text).

For type 2 molecules, we assigned new folding rates describing the transition between 0.25 and 0.4 FRET as well as from the 0.25 or 0.4 states to the highest FRET state at 0.53 (k₂₁/k₋₂₁, k₂₂/k₋₂₂, and k₂₃/k₋₂₃, respectively, Figure 4). Roughly one third of the type 2 molecules show transitions to and from 0.4 FRET, whereas all other traces show a direct I-M transition. Due to the rare occurrence of the transitions involving the F state (especially at low [Ca²⁺]), the rates k₂₁, k₋₂₁, k₂₂ and k₋₂₂ were estimated as the inverse of the respective averaged dwell times. k₂₁ and k₋₂₁ could be calculated at 10 mM Ca²⁺: 90 mM Mg²⁺ only but need to be interpreted with great care due to their large error. k₂₂ and k₋₂₂ are independent of the [Ca²⁺]. Instead, k₂₃ increases with higher [Ca²⁺] and is the fastest rate under all conditions. The opposite transition (k₋₂₃) becomes slower at high [Ca²⁺]. Overall, the rapid accumulation of the molecules in the 0.53 FRET state is well explained by these rates.

Conclusion

This study is the first investigation at the single molecule level on the effect of an M²⁺ other than Mg²⁺ on RNA folding. The group II intron derived D135-L14 ribozyme is the largest protein-free RNA investigated by smFRET and represents a perfect model system for the investigation of folding of large RNAs. Its splicing activity is very sensitive to trace amounts of Ca²⁺.[4] Although an (additional) inhibition of the catalytic step itself by Ca²⁺ cannot be ruled out, our results reveal an unprecedented behavior of RNA folding upon Ca²⁺ binding.

The addition of Ca²⁺ to the D135-L14 ribozymes induces the division of all single molecules into two distinct subpopulations:

1. Type 1 molecules fold very similarly to those in the Mg²⁺-only pathway, the individual folding rates increasing only moderately (if at all) by adding Ca²⁺. We ascribe this type to a structure where Mg²⁺ still occupies the key sites of the threedimensional architecture. Ca²⁺ might bind to some places in the RNA, but presumably only diffusely for charge compensation, thus having hardly any effect on the global structure. The faster ligand exchange rate and larger ionic radius of Ca²⁺ compared to Mg²⁺ might be reflected in the slight increase in folding rates, the higher dynamics of domain assembly, and the slightly less compact high FRET state.

2. In type 2 molecules, 0.4 FRET is highly destabilized and almost inexistent. It is an open question whether the molecules fold directly from the 0.25 state to the most compact and misfolded ~0.53 state, or if the 0.4 state is transiently reached but unobservable due to an occupancy shorter than the experimental time window of 0.33 ms. The high FRET state F is slightly less compact than the native one, and hence, within this second subpopulation, Ca²⁺ obviously occupies one or more key sites in the folded structure.

The differences in folding dynamics of the D135-L14 ribozyme with increasing [Ca²⁺] concur and can thus be explained with the different coordinating properties of Mg²⁺ and Ca²⁺: Ca²⁺ is larger, has faster ligand exchange dynamics, and can enlarge its coordination number to 8, instead of only 6 with Mg²⁺. In addition, Ca²⁺ has an intrinsic lower affinity towards nucleic acids than Mg²⁺. As Mg²⁺ is in excess under all conditions, at least at one site Ca²⁺ must bind more tightly than Mg²⁺.

One can speculate on the distinct role of Ca²⁺ for the individual steps of folding: The 0.25 FRET state I, which is attributed to a folded domain 1,[10] does not seem to be influenced by Ca²⁺ in its global fold. The split into two coexisting but distinct subpopulations upon Ca²⁺ addition becomes noticeable only after the subsequent docking of domains 3 and 5. It was previously shown[10] that high [Mg²⁺] increase the dynamics of group II introns, i.e. the 0.25 state is occupied on a regular basis. Presumably, one (or possibly several) Ca²⁺ ion bind to a specific junction within the I state, which controls the structural dynamics leading to an inevitable branching-off of the pathway towards the misfolded structure M. Maybe binding of the first Ca²⁺ leads to cooperative binding of subsequent ions, which would explain the coexistence of type 1 and type 2 molecules. Such a cooperative binding is corroborated by the linear increase in molecules of type 2 with increasing [Ca²⁺]. Yet we do not observe traces that interchange from type 1 to type 2 behavior during our experimental time window. Although I shows the same FRET value for the two types of traces in our experiments, we cannot rule out that it differs already in the two subpopulations.

Our findings are well in line with previous observations that Ca²⁺ inhibition takes place in prefolded D135 molecules.[4] In earlier biochemical studies, the catalytic rate k_cat is reduced to 50% at 5 mM Ca²⁺ and splicing completely inhibited at 20 mM Ca²⁺. A quantitative comparison between splicing inhibition and the occurrence of two subpopulations with increasing [Ca²⁺] is not possible, because in the previous experiments D5 was added in trans, i.e. D5 is not covalently linked to the rest of the ribozyme. Such a two-piece setup obviously strongly impedes the last step of folding to the active structure.

Proving a functional role for folding heterogeneity has so-far not been achieved. A recent bulk FRET study on the extended Schistosoma hammerhead ribozyme concluded that divalent metal ions other than Mg²⁺ have almost no effect on the global RNA fold, but strongly regulate catalysis.[6] However, the small differences in maximum FRET intensities found in our study would not be detected in bulk experiments aimed to study folding heterogeneity. This illustrates that only single molecule spectroscopy is able to reveal the subtle effect of M²⁺ binding during RNA folding.

Group II intron folding is of general interest because (i) no kinetic traps exist on the native folding pathway, (ii) the active state N is reached only transiently from a collapsed near-native state F, (iii) N is stabilized by substrate binding, and (iv) Mg²⁺ not only induces folding but also increases the inherent dynamics of the folded RNA.[10] Our presented results add two further aspects to this list: (v) It could be shown for the first time that the binding of specific metal ions has an effect on the global architecture of a large RNA. One can thereby distinguish between Mg²⁺ and Ca²⁺ bound RNA molecules on a single molecule level. (vi) By linking the biochemical data[4] with the here observed separation into two subpopulations, a functional role for folding heterogeneity could be demonstrated for the first time.

Experimental Section

RNA preparation and single molecule experiments

The D135-L14 RNA, derived from the S. cerevisiae intron Sc.ai5γ, was obtained by in vitro transcription under standard conditions with homemade T7 polymerase from HindIII digested plasmid pT7D135-L14 and the T-Biotin, Cy3- and Cy5 DNAs purified as described.[18–20] Single molecule experiments were performed as described.[10,21] Samples were incubated in mixtures containing 100 mM Mg²⁺, 98 mM Mg²⁺: 2 mM Ca²⁺, 95 mM Mg²⁺: 5 mM Ca²⁺, 93 mM Mg²⁺: 7 mM Ca²⁺ or 90 mM Mg²⁺: 10 mM Ca²⁺ to test and compare the influence of increasing amounts of Ca²⁺ under conditions of equal ionic strength. For details see Supplementary Information.