A ribozyme that uses lanthanides as cofactor

Abstract To explore how an early, RNA-based life form could have functioned, in vitro selection experiments have been used to develop catalytic RNAs (ribozymes) with relevant functions. We previously identified ribozymes that use the prebiotically plausible energy source cyclic trimetaphosphate (cTmp) to convert their 5′-hydroxyl group to a 5′-triphosphate. While these ribozymes were developed in the presence of Mg2+, we tested here whether lanthanides could also serve as catalytic cofactors because lanthanides are ideal catalytic cations for this reaction. After an in vitro selection in the presence of Yb3+, several active sequences were isolated, and the most active RNA was analyzed in more detail. This ribozyme required lanthanides for activity, with highest activity at a 10:1 molar ratio of cTmp : Yb3+. Only the four heaviest lanthanides gave detectable signals, indicating a high sensitivity of ribozyme catalysis to the lanthanide ion radius. Potassium and Magnesium did not facilitate catalysis alone but they increased the lanthanide-mediated kOBS by at least 100-fold, with both K+ and Mg2+ modulating the ribozyme's secondary structure. Together, these findings show that RNA is able to use the unique properties of lanthanides as catalytic cofactor. The results are discussed in the context of early life forms.


INTRODUCTION
The RNA world hypothesis describes an early life form in w hich RN A served as the genome and as the only genomeencoded catalyst (1)(2)(3)(4). To study how an RNA based organism could function, r esear chers have developed catalytic RNAs (ribozymes) in the lab, using in vitro selection (5)(6)(7). The most central activity for self-replication is ribozymecatal yzed RN A pol ymerization ( 8 ), w hich r equir es chemically activated nucleotides. Nucleoside 5 -triphosphates (NTPs) could have been used as chemically activated nucleotides because NTPs are used in all of biology as activated nucleotides, with their involvement in di v erse, fundamental metabolic and signaling processes pointing to an ancient origin ( 9 , 10 ) and because known chemical pathways could have led to prebiotic activation of nucleotides in the form of 5 -triphosphates (11)(12)(13).
Any self-replicating molecular system requires a source of energy from the environment to dri v e structure formation, dictated by the second 'law' of thermodynamics ( 14 ).
In other words, for any molecular system, the free energy released by the chemical conversion of that energy source (or the free energy provided by physical forces such as light, or temperature changes) needs to exceed the free energy that is r equir ed to r eplicate the structur ed, information-rich system. Cyclic trimetaphosphate (cTmp) could have served as such an early, external energy source because it is is prebiotically plausible ( 15 , 16 ), and it can serve as polyphosphorylation reagent to generate NTPs ( 11-13 , 17 ), which are then able to drive polymerization or other reactions due to the highly exergonic hydrolysis of pyrophosphate bonds. This energy source can also be used by ribozymes, as shown by an in vitro selection of ribozymes that react their 5hydr oxyl gr oup with cTmp and genera te 5 -triphospha tes ( 18 ). Such 5 -triphosphorylated ribozymes were then able to energetically dri v e otherwise endergonic ligation between 5 -phosophate and 3 -hydroxyl groups. This previous selection was performed with Mg 2+ cofactors because Mg 2+ is known to aid ribozyme catalysis for a wide range of activities ( 7 , 19-23 ), and because Mg 2+ likely existed at concentrations around 10-20 mM in the prebiotic ocean ( 24 , 25 ).
Our interest to explore lanthanides as ribozyme cofactors stems from their ability to activate the phosphorus atoms of cT mp f or nucleophilic a ttack by wa ter ( 26 ). Importantly, the rate of this catalysis is enhanced if the lanthanide is coordinated not only by the negati v ely char ged o xygens of cTmp but also with additional ligands such as the carboxy groups of EDTA (ethylenediaminetetraacetate) or NTA (nitrilotriacetate). We hypothesized that analo gousl y, the lanthanide should also accelerate the nucleophilic attack of RNA 5hydr oxyl gr oups to the cTmp phosphorus a toms, and tha t a beneficial effect similar to that of EDTA and NTA could also be mediated by ribozymes, which often coordinate a Mg 2+ ion at their catalytic site to support acid / base catalysis (27)(28)(29)(30)(31). Lanthanide (III) ions are excellent catalysts for many reactions in water also due to their near-neutral hydrolysis constants of coordinated water molecules, and their high water exchange rate ( 32 ). Together, these ideas motivated us to explore the potential of lanthanides as ribozyme cofactors.
It is possible that lanthanide cofactors could have been used in an RNA world because they can be sufficiently abundant and accessible in specific environments, and because they are used as enzyme cofactors in contemporary organisms. First, lanthanides are not as rare as the name 'rare earth elements' suggests: Their abundance in the Earth's crust is in the range of se v eral mass ppm, similar to cobalt and molybdenum ( 33 ). Due to their unique chemical behavior, they are highly enriched in specific ores and pegmatites ( 34 ). While lanthanide minerals usually have low mobility, their mobility can increase via chelation by organic compounds in the environment ( 35 ). Second, in today's biology, se v eral methylotrophic bacteria use lanthanides as cofactors for methanol dehydrogenases that are closely related to enzymes using Ca 2+ as cofactor (36)(37)(38). When both lanthanides and calcium are available then the lanthanide using enzyme is used pr efer entiall y, likel y because the lanthanide provides greater catalytic potential due to its stronger Lewis acidity and higher ligand turnover ( 39 ). The gene encoding the lanthanide dependent methanol dehydrogenase is widespread in se v eral coastal marine envi-ronments ( 40 ). Ther efor e, the lanthanide's unique catalytic pr operties pr ovide enough evolutionary benefit in some environments to make them useful as catalytic cofactors in today's biology, and perhaps e v en in early stages of life.
To test whether ribozymes could use lanthanides as cofactor, we performed an in vitro selection for selftriphosphorylation ribozymes as described previously ( 18 ), but in the absence of Mg 2+ and in the presence of the lanthanide Yb 3+ . After fiv e rounds of selection, many isolated clones showed detectable activity with Yb 3+ . The ribozyme with the highest activity was characterized in more detail. This ribozyme used Yb 3+ for catalysis and gained > 100-fold in activity from the additional presence of K + and Mg 2+ , which modulated the ribozyme's secondary structure. Ca talysis was media ted only by the heaviest lanthanides with the smallest ion radii. Together, these results showed that ribozymes are able employ the strong Lewis acidity of lanthanides for use as catalytic cofactors.

In vitro selection
The in vitro selection was performed essentially as described ( 18 ), except that the triphosphorylation reaction step was performed with 3 mM Yb 3+ and 10 mM cTmp instead of 100 mM Mg 2+ and 50 mM cTmp. The Yb 3+ concentration of 3 mM was chosen because this concentration did not result in detectable RNA degradation at neutral pH and 22 • C ov er se v eral hours, while some degradation was visible under the same conditions at 10 mM Yb 3+ . The starting DNA library was synthesized based on a custom ultramer single-str anded DNA (Integr ated DN A Technolo gies, IDT) with the sequence 5 -GCTGGA GCTT AA CTGGCG -(N150)-AA CA T CT CGGT CT C GACTG-3 (lower strand) with hand-mix ed pr epar ation of phosphor amidites for the randomized region to reduce nucleotide bias. This sequence was used as template for PCR amplification with PCR primers 5 -AATT TAAT ACGACT CACT AT AGGGCGGT CT CCTGACGAGCTAAGCGAAACTG CGGAAACGCAGTC GAGACCGAGATGTT -3 and 5 -GCTGGA GCTT AA CT -3 to generate the double-stranded DNA library (italicized T7 promoter, underlined hammerhead ribozyme, bold constant regions). The hammerhead ribozyme cleaved itself off co-tr anscriptionally, gener ating a 5 -hydroxyl group at the RNA library. After PAGE purification, 100 nM library RNA was incubated with 50 mM Tris / HCl pH 8.3, 3 mM Yb(Tf) 3 , 10 mM Na 3 cTmp, 150 mM NaCl and 5 mM NaOH (to account for the pH drop due to cTmp chelating metal ions) in a volume of about 100 mL at room temperature. The RNA was then ethanol precipitated, desalted by size exclusion chromato gra phy (P30 spin-columns, Bio-Rad) and ethanol precipitated again. The r ecover ed RNA was liga ted to a biotinyla ted oligonucleotide (5 -biotin-d(GAACT GAAGT GTAT G)rU-3 ) using the R3C ligase ribozyme ( 41 ) with its arms adjusted to anneal to the 5 -constant region of the RNA library and to the biotinylated oligonucleotide (800 nM library RNA, 1000 nM ligase ribozyme, and 1200 nM biotinylated oligonucleotide in 100 mM KCl, 100 mM Tris / HCl pH 8.5 and 2 pM of r andomized libr ary RNA transcribed without a hammerhead ribozyme, which means Nucleic Acids Research, 2023, Vol. 51, No. 14 7165 that it had a 5 -triphosphate). The triphosphorylated RNA was included for the first 4 selection rounds to reduce the number of PCR cycles in each selection round such that no PCR artifacts (i.e. shorter amplicons) were detected. This 5 -triphosphorylated RNA contained the same N150 randomized region as the starting library. Ther efor e, it generated a low amount of 'noise' in the selected sequences but it did not introduce a bias from a specific sequence that could generate a false positi v e. The ligation products wer e captur ed on streptavidin-coated magnetic beads (Promega Z5481) and washed with water, 50 mM NaOH, and again water. After elution with 96% formamide at 65 • C for 5 min the RNA was ethanol pr ecipitated, r e v erse transcribed with the 3 PCR primer (Superscript III Re v erse Transcriptase, Invitrogen), and PCR amplified in two steps. The first PCR used 5 -primer 5 -GAACT GAAGT GTAT GT GAGACCGAGA-3 and 3 -primer 5 -GCTGGAGCTTAACT-3 , and the number of r equir ed PCR cycles was used to monitor the progr ess of the selection. The second PCR used the same PCR primers as the generation of the initial DNA library to complete one round of selection. The selection to optimize the N20 sequence in the truncated ribozyme 15 ( Figure 2 ) was performed identicall y, onl y the conditions during the selection step were adjusted to aid activity of ribozyme clone 15. Specifically, the pH was lowered to pH 7.3, and the concentrations of Yb 3+ and cTmp were adjusted to 6 mM each (round 1), 3 mM each (round 2), 1 mM each (round 3) and 0.3 mM each (round 4).

Self-triphosphorylation activity assays
The assa y f or self-triphosphorylation acti vity of indi vidual ribozymes was performed essentially as described ( 18 ). The DNA sequence containing the T7 promoter, hammerhead ribozyme, and the sequence of a specific, selected ribozyme sequence was transcribed by runoff transcription and the RNA was purified by denaturing PAGE. The standard triphosphorylation conditions before optimizing conditions for ribozyme 51 included 50 mM Tris / HCl pH 7.3, 150 mM NaCl, 3 mM Yb(Tf) 3 ,10 mM Na 3 cTmp and 5 M ribozyme RNA. After incubation at 22 • C for 3 h, an aliquot of 2 l was removed and added to 8 l of a pre-mixed solution such that the 10 l combined solution contained 100 mM Tris / HCl pH 8.0, 100 mM KCl, 5.6 mM Na 2 EDTA, 1 M of R3C ligase ribozyme, 1 M of the ribozyme RNA, 1 M of 5 -[ 32 P] radiolabelled oligonucleotide 5d(GAACT GAAGT GTAT G)rU-3 . The mixture was heated 2 minutes to 65 • C, then cooled at ∼0.1 • C / s to 30 • C to anneal the ligase ribozyme with its substrates. To start the ligation reaction, an equal volume of a solution containing 50 mM MgCl 2 , 4 mM spermidine and 40% (w / v) PEG 8000 was added, incubated for 3 h at 30 • C, and ethanol precipitated. The products were separated on 7 M urea 10% PAGE, exposed to phosphorimaging screens, detected by scanning on a Typhoon phosphorimager (GE), and quantified using the rectangle tool in the Quantity One software (Bio-Rad). Because triphosphorylation ribozyme, ligase ribozyme, and biotinylated oligonucleotides were equimolar the fraction of the ligated, short radiolabeled oligonucleotide was equal to the fraction of ligated ribozyme RNA, and ther efor e a measure of the fraction of self-triphosphorylated ribozyme. In the first screen of 27 selected RNA clones under selection conditions (at pH 8.3), only few clones showed acti vity. Howe v er, the acti vity of the most acti v e ribozyme was much higher at pH 7.3, where about half of the clones showed activity (Supplementary Figur e S2). Ther efor e the conditions for the initial triphosphorylation assays were pH 7.3 (including 6 mM Yb(Tf) 3 , 6 mM Na 3 cTmp, 50 mM HEPES / NaOH pH 7.3, and 150 mM NaCl, for 3 h at 22 • C) before the concentrations of Yb 3+ , cTmp, Na + / K + and Mg 2+ as well as temperature and pH were optimized.

SHAPE probing
The ribozyme secondary structure was studied by SHAPE chemical probing with 1-methyl 7-nitro isatoic acid anhydride (1M7), which reacts with 2 -hydroxyl groups of flexible nucleotides ( 42 ). Final concentrations during the probing were 50 mM MOPS / KOH pH 6.8, 330 M Yb(Tf) 3 (if included) and 3.3 mM cTmp (if included), 200 nM ribozyme 51, and the specified amounts of KCl and MgCl 2 . These compounds were dissolved in 49 l water and immediately mixed with 1 l of a solution containing either 100 mM 1M7 in dry DMSO or dry DMSO without 1M7. After incubation at 25 • C for three minutes the RNA was ethanol pr ecipitated, r e v erse transcribed using the manufacturer's instructions (Superscript III, Invitrogen) using an 8 nucleotide 5 -[ 32 P] radiolabeled DNA primer complementary to the ribozyme's 3 -terminus. This was the shortest primer that resulted in full e xtension. Re v erse transcription products were heated with 750 mM NaOH at 80 • C for 5 minutes to hydrolyze RNA, buffered to pH 5 by the addition of two equivalents of acetic acid over NaOH, and ethanol precipitated. Reaction products were separated by 7 M urea 10% polyacrylamide gelelectrophoresis, exposed to phosphorimager screens, scanned on a Typhoon Phosphorimager (GE), and quantified using the rectangle tool in the software Quantity One (Bio-Rad). For each band, the signal in the lane without 1M7 was subtracted from the signal in the lane with 1M7. Re v erse transcription stops at the nucleotide before the 1M7 modified nucleotide, which allowed assigning the bands to positions in the ribozyme, and thereby generating a SHAPE reaction profile for each experiment. Since each experiment used slightly different amounts of radioactivity, each profile was normalized, and the averages and standard deviations from three experiments wer e r eported. The secondary structur es wer e based on computational predictions by unafold ( 43 ); during these pr edictions, nucleotides wer e constrained as unpair ed if they showed high SHAPE signals.

RESULTS
To test whether ribozymes would be able to use lanthanides as cofactors we first tried to identify lanthanide-using ribozymes from a library that was previously selected from random sequence by 8 rounds of selection in the presence of 100 mM Mg 2+ and 50 mM cTmp ( 18 ). This library shows activity also at 51 mM Mg 2+ and 1 mM cTmp ( 44 ) and contained hundreds of different ribozyme clusters ( 45 ). To identify lanthanide-using ribozymes from this library, the library was subjected to four rounds of selection in the absence of Mg 2+ , but in the presence of 3 mM Yb 3+ and 5 mM cTmp ( Figure 1 ). Four selection rounds were previously sufficient to enrich acti v e ribozymes from a library with more than 10 14 different sequences to dominate the population ( 18 , 45 ), ther efor e we expected that lanthanideusing ribozymes would dominate the library within one or at most two selection rounds from the pre-selected library.
Howe v er, e v en after fiv e selection rounds in the presence of Yb 3+ no ribozyme activity was detected with Yb 3+ (Supplementary Figure S1), suggesting that the previously selected library with hundreds of Mg 2+ -using ribozymes did not contain a single ribozyme that was able to alternati v ely use Yb 3+ as cofactor.
To test more generally whether Yb 3+ could be used by ribozymes as cofactor for triphosphorylation we conducted a new in vitro selection, starting from random sequence. The starting library was the same, 'round 0 library with 150 randomized positions that was used as starting library for the previous selection of ribozymes in the presence of Mg 2+ ions. The selection in the presence of Yb 3+ ions used this library with an effecti v e comple xity of 2.0 × 10 14 different sequences. After four rounds of selection, the library appeared to be dominated by acti v e sequences because the r equir ed number of PCR cycles after re v erse transcription decreased sharply (Supplementary Figure S2A). The selection was continued until selection round 8 with shorter incubation time (initially 3 h reduced to 2 min) to select for the most acti v e ribozymes.
To identify acti v e ribozymes, 29 clones were arbitrarily chosen from selection round 8 and analyzed for selftriphosphorylation activity. For this assay, individual sequences were generated with a 5 -hydr oxyl gr oup, incubated with cTmp and Yb 3+ , then ligated to an equimolar concentration of a 5 -[ 32 P] radiolabeled oligonucleotide using the R3C ligase ribozyme ( 41 ). Because only 5triphosphorylated RNAs could be ligated, the gel-shifted fr action of r adiola beled oligonucleotide informed a bout the extent of self-triphosphorylation (Supplementary Figure  S2B). The fraction of ligated oligonucleotide was equal to the fraction of ligated ribozyme because these two RNAs were employed in equimolar concentra tions. W hen the incubation with cTmp was conducted under selection conditions (pH 8.3), only weak activity was detected. Howe v er, acti vity was significantly higher at pH 7.3, especially when 150 mM NaCl was added (final 6 mM YbCl 3 , 6 mM Na 3 cTmp, 50 mM HEPES / KOH pH 7.3, 150 mM NaCl). Here, at least 19 of the 29 clones showed activity, and one clone showed more than 10-fold higher activity than any of the others. This clone 15 was chosen for further analysis.
To identify the minimal size of ribozyme clone 15 (182 nucleotides in length), its 3 -terminus was truncated in 10nucleotide increments. Howe v er, e v en the truncation of 10 nucleotides abolished activity (Figure 2 ), suggesting that the 3 -terminus was r equir ed for activity. This finding matched the computationally predicted secondary structure of clone 15 ( 43 ), which positioned the ribozyme 3 -terminus close to the reacti v e site at the 5 -terminus. The same structur e pr ediction was used as guide to delete internal regions of the ribozyme. Indeed, r emoving thr ee differ ent segments of the region between position 49 and 163 (T1, T4 T5 in Figure 2 ) retained most of the activity but activity was reduced significantl y w hen all thr ee r egions wer e r emoved simultaneously (T6 in Figure 2 ). To determine whether this central region could be replaced by a shorter fragment, we inserted 20 randomized nucleotides between position 53 and 159 of the ribozyme, and subjected this library to an additional four rounds of selection, with each successi v e selection round decreasing the concentration of Yb 3+ and cTmp. Each round r equir ed only 7-8 PCR cycles to amplify the selected sequences, suggesting that the majority of sequences in the initial N 20 library mediated activity. Twenty-three clones were arbitrarily chosen from the fourth round of this selection and analyzed for activity (Supplementary Figure S3). The most acti v e variant, 'ribozyme 51 with a length of 95 nucleotides, was chosen for further analysis. While ribozyme 15 and ribozyme 51 are related in sequence, ribozyme 51 is much smaller and was used for the further experiments.
To identify the optimum reaction conditions for ribozyme 51, the concentrations of Yb 3+ and cTmp were covaried, and reaction kinetics recorded (Figure 3 ). The highest reaction amplitude (57%) was reached at 3 mM cTmp and 0.3 mM Yb 3+ , while the fastest rate (0.65 min −1 ) was seen at 10 mM cTmp and 1mM Yb 3+ (full data shown in Supplementary Figure S4). Yb 3+ concentrations as low as 33 M led to detectab le acti vity. The ratio of cTmp:Yb 3+ was optimal at 10:1 for most tested combinations, resulting in the highest amplitude and the fastest rate. The drop in activity at higher ratios was most dramatic at Yb 3+ concentrations of 0.1 and 0.3 mM, where a cTmp:Yb 3+ ratio  Figure S4). Rates are only shown for reactions with an amplitude of at least 1% to avoid background noise. Averages from three experiments were used as basis for the curve fits. of 10:1 led to yields of 28% and 57%, while a cTmp:Yb 3+ ratio of 30:1 led to yields of 0.0% and 0.1%. This behavior is consistent with the ribozyme binding Yb 3+ as a complex with two or three cTmp molecules. More cTmp molecules may pre v ent the ribozyme from coor dinating Yb 3+ , and the coordination of only one cTmp molecule (with three negati v e charges) by one Yb 3+ ion would lead to an uncharged, insolub le comple x ( 26 ).
The variation of r eaction temperatur e, r eaction pH, and concentrations of other cations identified an optimum for each parameter ( Figure 5 ). The temperature showed an optimum around 25 • C, similar to the tempera ture a t which the in vitro selection of the ribozyme was performed (22 • C). The pH was optimal in the neutr al r ange between 6.3 and 7.3, which is close to the hydrolysis constant of the inner sphere lanthanide-coordinated water / hydroxyl around 7.7 ( 32 ). Under these conditions (50 mM MOPS / NaOH pH 6.8, 3.3 mM cTmp, 0.33 mM Yb 3+ , 25 • C) the reaction yield was highest with 300-700 mM NaCl or KCl, or 5 mM MgCl 2 , with a rise in reaction yield from about 5% to 68%, and 50%, respecti v ely. The beneficial effect of Yb 3+ , Mg 2+ and K + was confirmed in the reaction kinetics, where dif ferent combina tions of 500 mM KCl, 5 mM MgCl 2 , and 0.33 mM Yb 3+ were employed (Figure 5 E). The ribozyme was inacti v e without Yb 3+ , indica ting tha t the other ca tions were not able to replace Yb 3+ in catalysis. Ribozyme activity was maximal with all three cations (Yb 3+ , K + , Mg 2+ ) and decreased in the order of Yb 3+ / K + / Mg 2+ > Yb 3+ / Mg 2+ ∼ Yb 3+ / K + > Yb 3+ . Curve fitting of the kinetics in the presence of Mg 2+ and / or K + r equir ed two exponentials, which resulted for all three reactions in a fast rate around 1.1 min −1 and a slow rate around 0.1 min −1 . The highest amplitude for the fast rate was in the presence of all three ions. These results suggested that the ribozyme requires Yb 3+ for catalysis and is dependent on K + (or Na + ) and Mg 2+ to fold into its acti v e structure. While the functional roles of these cations may be overlapping (for example, in shielding the negati v e charges) the data show that each ion has a sufficiently distinct role to be r equir ed for optimal activity.
The ribozyme's secondary structure, and its dependence on Yb 3+ , cTmp, K + and Mg 2+ was studied by SHAPE probing in the absence and presence of these ions ( Figure 6 ). The shown structure was consistent with the SHAPE probing data except positions 53-72, which may form a more complicated arrangement (see below). The influence of 500 mM K + and 5 mM Mg 2+ on the secondary structur e (Figur e 6 A) was studied without Yb 3+ or cTmp by SHAPE probing at 50 mM MOPS / NaOH pH 6.8. Ther efor e, the influence of K + and Mg 2+ shows the status bef ore Yb 3+ / cT mp substrate binding. The data showed that positions 25-28 were more accessible in the presence of K + or Mg 2+ , which is consistent with K + and Mg 2+ modulating the structure to present the loop 25-28 for coordination with Yb 3+ / cTmp complex (see below). While K + increases the flexibility of the four positions 10, 36, 38, and 59, Mg 2+ makes nucleotide 57 more fle xib le and nucleotide 71 less fle xib le. These observations suggest specific binding sites that mediate these ion's support in the ribozyme's structure and catalysis (Figure 5 E).
The addition of Yb 3+ and cTmp to the ribozyme in presence of K + and Mg 2+ led to a rigidification of fiv e nucleotide positions 25-29, position 36 near the central loop (  Figures S2 and S3), the fle xib le positions 47-49 and 63-69 may r epr esent linker regions that allow the protected nucleotides 52, 53 and 56 to remain in place while the ribozyme undergoes a conformational change due to binding of the Yb 3+ / cTmp complex. In contrast, cTmp alone did not influence the SHAPE accessibility in the absence of Yb 3+ , which suggested that the ribozyme binds cTmp only as a complex with Yb 3+ , consistent with the observed co-dependence of ribozyme activity on cTmp and Yb 3+ (Figure 3 ).
Since the ribozyme's loop at positions 25-29 was affected by Yb 3+ / cTmp binding (figure 6 C, D), we tested whether this loop is indeed important for the ribozyme's function (Figure 7 ). To do this, we introduced mutations into the loop and measured the ribozyme's activity. The loop sequence AGAUU was converted to Aucgg (mutated positions lowercase), AuAUU, AGAaU or AucaU. All four variants showed a decrease in ribozyme activity, with AGAaU maintaining about half of its activity, AuAUU generating about a 10-fold decrease in activity, and the activity mediated by Aucgg and AucaU below the detection limit. In addition, we tested the importance of three bulged adenosines (A36U , A38U , A44U) that also changed SHAPE accessibility upon Yb 3+ / cTmp binding (Figure 6 C, D). While the mutation of the bulged A44 retained half of the ribozyme's (red), Yb 3+ and Mg 2+ (blue), Yb 3+ and K + (green), Yb 3+ (black), and Mg 2+ and K + (gr ey). Curved lines r epr esent doub le-e xponential equations fitted to the data by least squares fitting, with the rates and amplitudes for each condition gi v en to the right. Note that the reaction time in the graph is displayed in logarithmic order. Error bars in all panels r epr esent standard deviations from triplicate experiments. activity, mutations A36U and A38U showed a > 10-fold reduction in activity. These data confirm that positions 36, 38, and the loop at positions 25-29 are central to the ribozyme's function, and may be directly involved in Yb 3+ / cTmp binding.
To test whether ribozyme 51 could function on a substra te in tr ans , the loop a t positions 10-15 was split in half, and the stem extended to a length of 8 base pairs so that the ribozyme could use it as binding site for a 14-nucleotide long substrate RNA (Supplementary Figure S6). Howe v er, this did not result in detectab le acti vity e v en after 16 hours under optimal conditions. In contrast, in side-by-side experiments the trans-reacting variant of ribozyme TPR1e ( 44 ) gave near-complete substrate conversion after 1 minute reaction time under its own optimal conditions. These results suggested that the lanthanide-using ribozyme 51 r equir ed the stem-loop from nucleotides 7-18 for its activity and cannot be easily converted to the trans-format.
To better understand the importance of the inserted region 53-72 (20 nucleotides inserted into the T6 construct with an internal truncation, Figure 2 ), the activity of ribozyme 51 with these 20 nucleotides, and the T6 variant without these 20 nucleotides were measured under optimized conditions (50 mM MOPS / NaOH pH 6.8, 500 mM KCl, and 5 mM MgCl 2 , and a 1:10 molar ratio of Yb 3+ :cTmp) across a range of Yb 3+ / cTmp concentrations ( Figure 8 ). The 20 nucleotide insert increased ribozyme activity across the tested Yb 3+ / cTmp concentrations, with a similar increase at 33 M and 330 M Yb 3+ . The observa tion tha t the increase was similar a t sub-sa tura ting, and sa tura ting concentra tions of Yb 3+ / cTmp suggested that the 20-nucleotide insert did not increase the ribozyme's affinity for the Yb 3+ / cTmp complex but that it generally stabilized the ribozyme's acti v e conformation. Additionally, these results showed that the ribozyme can show detectable activity with as little as 18 M Yb 3+ and 180 M cTmp. The  sigmoid shape of the curve suggested cooperati v e behavior of added cTmp and Yb 3+ , consistent with the idea that cTmp and Yb 3+ are bound as a complex.

DISCUSSION
In this study, we describe the in vitro selection, and analysis of a ribozyme that uses Yb 3+ as a catalytic cofactor. Three pieces of evidence support the idea that Yb 3+ participa tes in ca talysis: First, e v en without the context of a ribozyme, lanthanides strongly coordinate the three negati v ely charged oxygens of cTmp and activate the phosphorus atoms for nucleophilic attack ( 26 ). Second, Yb 3+ is r equir ed for ribozyme-mediated self-triphosphorylation catalysis, e v en in the presence of Mg 2+ , Na + or K + ( Figure  5 E). Third, ribozyme activity is co-dependent on the concentration of cTmp and Yb 3+ with an optimal 10:1 ratio of [cTmp] : [Yb 3+ ] (Figure 3 ). This is expected if cTmp and Yb 3+ are bound as a complex at the catalytic site.
The lanthanide-using ribozyme 51 showed a pH optimum around pH 6.8 (Figure 5 B), not far from the hydrolysis constant of the inner sphere lanthanide-coordinated water / hydr oxyl ar ound 7.7 ( 32 ). This is consistent with a lanthanide-coordinated water / hydroxide being used in Figure 8. Influence of the inserted 20 nucleotide region at different Yb 3+ / cTmp concentrations. Two constructs were compared: Ribozyme 51 with a selected N20 region (black filled circles), and ribozyme T6, which is lacking this region (empty circles). The constructs are shown in Figure 2 . The ribozyme activity (percent ligated) is shown as a function of the concentration of Yb 3+ , which was added with a 10-fold stoichiometric excess of cTmp. Error bars are standard deviations from triplicate experiments.
Lewis acid / base catalysis of the reaction. In contrast, previously characterized self-triphosphorylation ribozymes that use Mg 2+ as cofactor show an increase of k OBS beyond pH 8, suggesting that for these ribozymes the rate-limiting step is the deprotonation of a group with a much higher p K A , such as the ribozyme's 5 -hydr oxyl gr oup ( 18 ). This different use of the catalytic cations could explain the failure to select lanthanide-using ribozymes from a library that was previously selected in the presence of Mg 2+ ions (Supplementary Figure S1). In contrast, a previously generated RN A-cleaving DN Azyme that was selected in the presence of Mg 2+ was acti v e with lanthanides as cofactors ( 46 ). These da ta suggest tha t the promiscuity in the use of Mg 2+ versus lanthanides differs between the catalyzed reaction and / or the catalyst (triphosphorylation vs. RNA cleavage). The sensitivity of ribozyme 51 to the lanthanide ion radius is consistent with the size-selecti v e coor dination seen for lanthanides in minerals ( 47 ), protein enzymes ( 48 ) and deoxyribozymes ( 49 ).
Contemporary organisms use lanthanides as cofactor for se v eral methanol dehydrogenases (MDHs), which are evolutionarily related to calcium-using MDHs ( 36 , 38 ). While the calcium-using MDHs oxidize methanol to formaldehyde, the lanthanide-using MDHs (with a pr efer ence for Ce 3+ ) are catal yticall y more efficient and oxidize methanol to the less toxic formic acid ( 37 ). In both enzymes, the ca tions are coordina ted a t the ca talytic site using the keto oxygens O5 and O7 as well as the aromatic N6 of pyrroloquinoline quinone (PQQ). Howe v er, these enzymes differ in their additional coordination of the cations by amino acid side chains as well as the backbone geometry near the catalytic site ( 37 ). When both Ca 2+ and Ce 3+ are accessible, methanotrophic organisms seem to pr efer entially use the lanthanide-dependent MDH ( 39 ). To scavenge lanthanides in environments of low lanthanide abundance, the protein lanmodulin binds lanthanides with picomolar affinity ( 50 ). The evolution of lanthanide using MDHs and lanmodulin underscore the high value of lanthanide cations as catalytic cofactors in biological enzymes.
Could lanthanides have been used as catalytic cofactors in prebiotic RN A catal ysis? In contrast to Mg 2+ , which was likely present at about 10 mM concentration in the Hadean ocean ( 25 ), lanthanide 3+ ions would have been present in much lower concentrations, and the lanthanides are often hard to mobilize. However, the lanthanides in modern soils (around 1-100 mg / kg) can be mobilized by lanthanidecoordinating humic acids ( 33 , 35 ), and the lanthanide's poor bioavailability under neutral pH is enhanced by acidic environments such as the volcanic mud pot water where biological lanthanide use was first discovered, and which contained lanthanides at 2-3 micromolar concentration ( 51 ). Since volcanic environments were likely among the few land forms on early Earth ( 52 ) there could have been ecological niches in which early RNA-based life forms used ribozymes with lanthanides as catalytic centers. Howe v er, the current study is focused on the broader picture to explore the chemical space accessible for RNAs to achie v e ef ficient ca talysis. In addition to the eight metal ions in the main groups (Li, Na, K, Mg, Ca, Sr, Ba, Pb) and se v en transition metal ions (Mn, Fe, Co, Ni, Cu, Zn, Cd) that have been known to serve as cofactors for ribozymes (53)(54)(55)(56), this study demonstrates with four additional elements (Lu, Yb, Tm, Er) that e v en the strong Lewis acidic lanthanides can be used as cofactors for ribozymes.

DA T A A V AILABILITY
All data necessary for the reproduction of the described study are described in the manuscript text, figures, and supplemental figures.