Interactions of Protonated Guanidine and Guanidine Derivatives with Multiply Deprotonated RNA Probed by Electrospray Ionization and Collisionally Activated Dissociation

Abstract Interactions of ribonucleic acid (RNA) with guanidine and guanidine derivatives are important features in RNA–protein and RNA–drug binding. Here we have investigated noncovalently bound complexes of an 8‐nucleotide RNA and six different ligands, all of which have a guanidinium moiety, by using electrospray ionization (ESI) and collisionally activated dissociation (CAD) mass spectrometry (MS). The order of complex stability correlated almost linearly with the number of ligand atoms that can potentially be involved in hydrogen‐bond or salt‐bridge interactions with the RNA, but not with the proton affinity of the ligands. However, ligand dissociation of the complex ions in CAD was generally accompanied by proton transfer from ligand to RNA, which indicated conversion of salt‐bridge into hydrogen‐bond interactions. The relative stabilities and dissociation pathways of [RNA+m L−n H]n− complexes with different stoichiometries (m=1–5) and net charge (n= 2–5) revealed both specific and unspecific ligand binding to the RNA.


Introduction
Interactions of ribonucleic acids( RNA) with proteins and small molecules play an importantrole in many fundamental biological processes. [1] RNA-protein [2] and RNA-drug [3] complex interfaces are often stabilized by stacking, cation-p,h ydrogenbond, and salt-bridge interactions [2b, 4] between guanidinium functionalities and RNA. For example, arginine residuesa re frequently found in the RNA-binding regions of proteins, [2b, 5] such as those in the humanr ibosomalp rotein L7 and the human immunodeficiency virus type 1( HIV-1) rev protein, [6] and many pharmaceutically active compounds contain guanidinium moieties, [7] such as antihypertensive drugs (e.g.,a miloride, clonidine, guanethidine), antidiabetics (e.g.,m etformin, buformin, galegin), and antibiotics(e.g.,streptomycin, sulfaguanidine). [8] RNA-ligand complexes can be studied by using computational [9] or experimental approaches such as nuclear magnetic resonance (NMR)s pectroscopy, [3a, 10] X-ray crystallography, [11] biochemical methods, [12] and crosslinking strategies. [12a, 13] Althoughh ighly promising native mass spectrometry (MS) studies of RNA-protein [14] and RNA-drug [15] complexes have contin-ued to appear in the literature over the past 20 years, they are still scarce compared with those for protein-protein interactions. [16] This is quite surprising in view of the high stability of guanidinium-phosphate interactions in gaseous ions [17] and the inherent advantages of native MS, for example, that it does not requires table isotope labeling or crystallization,i s not limited by crosslinkingr eactivity,a nd uses only relatively small quantities of sample materialc ompared with NMR spectroscopya nd X-ray crystallography. Moreover,anumber of laboratories have reported that in the gas phase, the strength of noncovalent bonds between RNA [14g, 15e, 17b, 18] or deoxyribonucleic acid (DNA) [17d] andb asic ligands can even exceed those of covalentb onds. As ac ase in point, we have recently shown that the noncovalent bonds between trans-activationr esponsive (TAR) RNA and ap eptidew ith the arginine-rich binding regiono ft he trans-activator of transcription (tat) protein from HIV-1 are sufficiently strong to survive phosphodiester backbone cleavage of the RNA by collisionally activatedd issociation (CAD), which thus allowedi ts use to probe tat binding sites of TARR NA by top-down MS. [14g] The unusuals tabilityo fn oncovalent bonds in the gas phase has been attributed to strong electrostatic interactions, [14g] such as salt bridges,i onic and neutralh ydrogen bonds, and charge-dipole interactions, [19] of which salt bridges were thought to providet he highest contribution to stability. [20] In support of this hypothesis, calculations suggest that the interaction energy between guanidine andt rifluoroacetic acid, that is, the stabilization achievedw hen the two neutralm olecules are brought from infinite distance to equilibrium distance, is % 70 kJ mol À1 ,w hereas that of protonated guanidine and trifluoroacetate, that is, the stabilization achievedw hen the two Interactions of ribonucleic acid (RNA) with guanidine and guanidine derivatives are important features in RNA-proteina nd RNA-drugb inding. Here we have investigated noncovalently bound complexes of an 8-nucleotide RNA and six different ligands,a ll of whichh ave ag uanidinium moiety,b yu sing electrospray ionization (ESI)a nd collisionally activated dissociation (CAD) mass spectrometry (MS). Theo rder of complex stability correlateda lmost linearly with the number of ligand atoms that can potentially be involved in hydrogen-bond or salt-bridge interactions with the RNA, but not with the proton affinity of the ligands. However,l igand dissociation of the complex ions in CAD was generally accompanied by protont ransfer from ligand to RNA, which indicated conversion of saltbridge into hydrogen-bond interactions. The relative stabilities and dissociation pathways of [RNA + m LÀn H] nÀ complexes with different stoichiometries (m = 1-5) and net charge (n = 2-5) revealed both specific and unspecificl igand binding to the RNA. oppositely charged ions are broughtf rom infinite distance to equilibrium distance, is % 500 kJ mol À1 . [21] These energies differ by almost an order of magnitude, which highlights the fact that the balance between covalenta nd noncovalent bond dissociationc ritically depends not only on the number but also the type of interactions.
Here, we used electrospray ionization (ESI) and low-energy CAD to systematically study the binding of basic ligandst oa n 8-nucleotide (8-nt) RNA.A ll ligands investigated (Table 1), that is, guanidine (Gnd), 1-methylguanidine (meGnd), 1,1,3,3-tetramethylguanidine (tmeGnd), 3-guanidinopropanoic acid (Gpa), l-2-amino-3-guanidinopropanoic acid (aGpa), and l-arginine (Arg), contain (substituted)g uanidinium moieties with pK a values between1 2.6 and 13.8, [22] and vary both in protona ffinity (PA) and the types and number of intra-and intermolecular interactions that they can form. As ab asis for the further development of top-downn ative MS for the detection of RNAprotein complexes and the characterization of their binding interfaces,wea ddress the relative strengths of individual interactions, the competition between noncovalent and covalent bond cleavage, binding specificity, the energetics of intermolecular proton transfer,a nd the effect of the complex net charge.

ESI MS of RNA-Ligand Complexes
Noncovalently bound complexes of the 8-nt RNA (GGCUAGCC, 5'-OH and 3'-OH termini)a nd guanidine or guanidine derivatives (Table 1) were produced by electrospraying solutions of 1 mm RNA and 5-100 mm ligand (L) in 1:1H 2 O/CH 3 OH at pH % 7.5, adjusted by the addition of piperidinea nd imidazole ( % 1mm each). The additive mixture of piperidine and imida-zole was used because it very efficiently suppresses the formation of sodium and potassium adducts without promoting formationo fh ighly charged ions. [23] 8-nt RNA contains all canonical nucleobases and, accordingt ot heoretical predictions (http://rna.tbi.univie.ac.at/cgi-bin/RNAWebSuite/RNAfold.cgi), [24] shouldn ot form any stable secondary structures in solution. However,t he RNA sequence is self-complementary and ah igh methanolc ontent along with al ow RNA concentrationw as used to prevent dimer formation; [25] dimer ions were not observed in any of the ESI spectra recorded in this study.T he near-neutral pH of % 7.5, at which the guanidine moiety of all ligandss hould be protonated (Table 1) and the RNA phosphodiesterm oieties deprotonated, [26] was chosen to promote the formation of intermolecular salt bridges between the ligand guanidiniuma nd RNA phosphodiester moieties in solution. Undert hese conditions, RNA-ligand complex ions, [RNA + m LÀn H] nÀ ,f rom ESI wereo bserved for all ligandss tudied, as illustrated for guanidine in Figure 1A. The net charge n on the RNA-ligand complexes ranged from 2t o5( Figure1 and Figure S1 in the Supporting Information), and the proportion of [RNA + m LÀn H] nÀ ions generally increased as n decreased, that is, 0% for n = 6a nd > 70 %f or n = 2 ( Table S1), which is consistentw ith each protonated ligand compensating one of the negative charges of the RNA in the associationr eaction in solution[ Reaction (I)]: In agreement with previous studies of guanidinium derivatives binding to DNA, [27] the maximum number of boundl igands (Table 1) did not exceed seven,t he number of phosphodiesterm oieties in the 8-nt RNA, at the highest ligand concentration used (100 mm;F igure S1), except for Gpa. At this concentration,R NA complex ions with up to 14 Gpa ligands were detected, along with abundantG pa cluster ions. The latter were not observed for Gnd, meGnd, and tmeGnd,a nd were found in much lower abundance for aGpa and Arg at ac oncentration of 100 mm than for Gpa at 10 mm ( Figure S1). Ap ossible rationale for the different behavior of Gpa regarding cluster formationa nd bindingt oR NA is its net charge, which should be zero assuming that the guanidinea nd carboxylic acid groups are protonated and deprotonated, respectively,a t the solutionp Ho f% 7.5 used. By contrast,G nd, meGnd, tmeGnd, aGpa,a nd Arg should each carry an et charge of + 1 at this pH assuming that both the guanidine and amino groups are protonated and the carboxylic acid moieties are deprotonated. In this case, Coulombic repulsion limits cluster formation and bindingt oR NA for all ligands studied,e xcept for the overall neutral Gpa. Although the RNA complex and ligand cluster ions found in the ESI spectrad on ot necessarily reflect the speciesp resenti ns olution, these data suggest that the [RNA + m LÀn H] nÀ ions predominantly originate from association reactions in solution.  [21] The ligand PA values are 986, 1032, and 1051 kJ mol À1 for Gnd, tmeGnd, and Arg, respectively [28] (experimentalP Av alues for meGnd, Gpa, and aGpa have not been reported, but calculations [29] suggest that they also lie in this range), and that of dimethyl phosphate, (CH 3 O) 2 PO 2 À ,a samodel for the deprotonated phosphodiester moiety,i s1 387 kJ mol À1 . [30] However, the PA values of the deprotonated phosphodiester moieties in Scheme1.Schematicdiagram of the energies associated with ligand dissociation from [RNA + (mÀ1) LÀ(n + 1) H] (n + 1)À ·[L + H] + ions for m = 1, similar to that for ion pairs in Ref. [21];the minimum energyrequired for complexd issociation is indicated in blue. RNA can be different from that of dimethyl phosphate as a result of internal hydrogen bonding and charged elocalization. [31] For example, aP Avalue of 1279 kJ mol À1 was derived in bracketing-type experiments for adenosine monophosphate, [32] the gas-phase structure of which features ionic hydrogen bondingb etween the phosphate and the 3'-hydroxyl group, which in turn is hydrogen bondedt ot he 2'-hydroxyl group; [33] that of phosphate is 1383 kJ mol À1 . [30] From the above PA values, PT from protonated ligand to dimethyl phosphate, that is, [L + H] + + (CH 3 O) 2 PO 2 À !L + (CH 3 O) 2 PO 2 H, or to adenosine monophosphate in transiently formed, unstable complexes is highly exothermic (DPA, Scheme 1) by 336 to 401 or 228 to 293 kJ mol À1 ,r espectively. By contrast, PT from protonated ligand to deprotonated RNA within stable RNA-ligand complexes is an endothermic reaction (DH PT,complex > 0kJmol À1 )t hat requires an energyi nput to proceedb ecause the binding energies of [RNA + (mÀ1) LÀ (n + 1) H] (n + 1)À ·[L + H] + ions are generally far higher than those of [RNA + (mÀ1) LÀn H] nÀ ·Li ons, as ar esult of the far higher electrostatic interaction energies of the former.
Therefore, the energy providedb ys low ion heatingi nC AD causes PT within the [RNA + (mÀ1) LÀ(n + 1) H] (n + 1)À ·[L + H] + ions to produce [RNA + (mÀ1) LÀn H] nÀ ·Li ons that can further dissociate into [RNA + (mÀ1) LÀn H] nÀ and L[ Reaction(II), Scheme 1] unless the interconversion barrier between saltbridge binding motifs (protonated ligand and deprotonated phosphodiesterm oiety) and hydrogen-bond motifs (both ligand and phosphodiester moiety uncharged) is sufficiently high to preventP Tont he timescale of the experiment. Calculated interconversion barriersa re far smaller than PT reaction exothermicities; [21] up to about 18 kJ mol À1 for protonated dimers of betaine and ammonia [34] and1 5t o3 0kJmol À1 for overall neutral dimers of guanidine and formic acid. [35] Al-thoughP Tb arriers in the larger structures studied here likely differ from the above values, [36] the lack of products from Reaction (IV) suggestst hat the barriers for interconversion between salt-bridge and neutral-binding motifs in the RNA-ligandc omplexes are too small to prevent PT from protonatedl igand to deprotonated RNA.
The branching ratio of products from loss of [LÀH] À by Reaction (III) versus loss of neutralligand Lb yReaction (II) was affected by the complex ion net charge n,the number of ligands m bound to the RNA,t he ligand identity,a nd the energy available for dissociation.F or n = 2t o3a nd all ligands studied,t he only products from CAD of [RNA + m LÀn H] nÀ ions were from successive losses of neutral ligand [Reaction (II)].M oreover,n o deprotonated ligand,[ L ÀH] À ,w as detected for tmeGnd, meGnd, or Gnd irrespective of the net charge n and the CAD energy used, which suggests that the PAso f[ tmeGndÀH] À , [meGndÀH] À ,a nd [GndÀH] À far exceed those of the [RNA + (mÀ1) LÀn H] nÀ ions;ac orrespondingly high pK a value of 28.5 was reported for Gnd. [37] However,u pt o1 %[ RNA + (mÀ1) LÀ(nÀ1) H] (nÀ1)À ions were detected for tmeGnd, meGnd, and Gnd at n = 4, which can be attributed to PT from evaporated solvent to [RNA + m LÀn H] nÀ ions duringt he 1s ion accumulationt ime in the collisionc ell.L ikewise, CAD of [RNAÀ4H] 4À and [RNA + LÀ4H] 4À ions of tmeGnd, meGnd, and Gnd showed < 1% [RNAÀ3H] 3À ions irrespectiveo fthee nergy used. PT to [RNAÀ5H] 5À ions during the 1s accumulation periodw as even highera tu pt o3 0%,w hereas no PT was observedf or n = 2a nd 3. These data indicate an increasing proton affinity of the [RNAÀn H] nÀ ions with increasing n,s imilar to the increasing PA of peptidea nd protein [M + n H] n + ions with decreasing n. [38] Consistent with the PAso fl igand anionsc omprising carboxylates, for example, % 1385 kJ mol À1 for [ArgÀH] À , [30] that are comparable to that of the deprotonated phosphodiester moiety,C AD of [RNA + LÀn H] nÀ ions with Gpa, aGpa, andA rg for n = 4t o5didp roduce [LÀH] À ions, butb ecause our FT-ICR instrumentr elies on charged etection, the [LÀH] À ions were detectedw ith as ensitivity up to four times lower than the corresponding multiply charged RNA (complex) ions with n = 3 and 4. Moreover,t ime-of-flight differencesi nt he transfer of ions with low and high m/z values ( % 58 to % 173 for [LÀH] À versus % 630 to % 1261 for the corresponding RNA or RNA complex ions) from the collision to the ICR cells complicate quantitative detection of complementary ionic products from Reaction(III). We thus used only the signals of RNA (complex) ions for further data analysis.
For Gpa, aGpa, and Arg at n = 4, the fraction of [RNA + (mÀ1) LÀ(nÀ1) H] (nÀ1)À and [RNA + (mÀ2) LÀ(nÀ2) H] (nÀ2)À ions from Reaction (III) (of all products from Reactions (II) and (III)) generally increased with increasing energy used for CAD ( Figure 2). For Gpa above % 10 eV,h owever,t he fractiono f products from Reaction (III) substantially decreased again in favor of those from Reaction (II). At energies above 20 eV, c, y, a,a nd w fragments from RNA backbone cleavage and loss of chargeda nd neutral RNA nucleobases wereo bserved ( Figure 2), but these cannot accountf or the decrease in products from Reaction (III) because they were also observed in highly similar yields for aGpa and Arg. Moreover,t he RNA ions from Reaction (II) have ah igher net chargea nd thus are more prone to covalent-bond cleavage than those from Reaction (III) ( Figure S2), which should increase and not decrease the fraction of productsf rom Reaction (III). Instead, we proposet hat the observed partitioning between products from Reactions (II) and (III) results from different energy requirementsf or the different PT reactions associated with ligand dissociation,a si llustrated in Scheme 2f or Gpaa nd Scheme 3f or aGpa;r eactions for Arg should be similar to those of aGpa.
As discussed above, Gpa, aGpa,a nd Arg have zwitterionic structures at the solution pH of % 7.5 used here, and probably bind to deprotonatedR NA by the formation of salt bridges. Based on the energies in Scheme 1, we proposethat the PT reactions associated with ligand dissociation [Reactions (II) and (III)] occur in the gas phase. In Scheme 2A, the salt bridge between the guanidinium moiety of Gpa and ad eprotonated RNA phosphodiester moiety is converted into af ar weaker ionic hydrogen bond [20b] by an intramolecular PT from the guanidinium to the carboxylate moieties of Gpa, which dissociates into Reaction (II) products at relativelyl ow energy (E 1 ). At elevated energy (E 2 ), an intermolecular PT between the guanidinium moiety of Gpa and the deprotonated RNA phosphodiester moietyb ecomesc ompetitive and more products from Reac-  Scheme2.Proposed PT reactionsa ssociated with ligand dissociationf rom [RNA + GpaÀ4H] 4À ions with energyr equirements of A) E 1 ,B)E 2 ,and C) E 3 , for which E 1 < E 2 < E 3 .Ate levated energy E 3 ,changesi nthe higher-order RNA structure allow for PT from an RNA phosphodiester moiety (shown in violet) that was not initially bound to Gpa.O ther RNA-ligand interactions that potentiallystabilize the complex structures beforeand after PT are omitted for clarity.
Scheme3.Proposed PT reactionsa ssociated with ligand dissociation from [RNA + aGpaÀ4H] 4À ions with energyr equirements of A) E 1 ',B )E 2 ',and C) E 3 ',for which E 1 ' % E 1 < E 2 ' % E 2 < E 3 < E 3 '.Atelevatede nergy E 3 ',changes in the higher order RNA structure allowf or PT from an RNA phosphodiester moiety (violet) that was not initially bound to aGpa.O therR NA-ligand interactions that potentially stabilize the complexs tructures before and after PT are omitted for clarity.  2B). The proposed order of energies, E 2 > E 1 ,i sc onsistentw ith the higherP Ao fa cetate ( % 1454 kJ mol À1 ) [30] comparedw ith that of dimethylphosphate (1387 kJ mol À1 ), [39] according to which PT from the guanidinium to the carboxylate moiety is energetically favored by 67 kJ mol À1 over PT to ad eprotonated phosphodiester moiety. Although the PAso fs mall model compoundsg enerally differ from those of the corresponding sites in [RNA + m LÀn H] nÀ ions, primarily as ar esult of hydrogen bonding and the presence of multiple charges, [36] they can still reflect the competition for protons between different sites.
Importantly,t he PT reactions in Scheme 2A and Bd on ot require any changes in the RNA-ligand complex structure, whereas protonation of the carboxylate group in Scheme 2C assumest hat an uncharged RNA phosphodiester moiety (or,alternatively,anucleobase with relativelyh igh gas-phasea cidity, such as guanine or adenosine) [40] comesi nto sufficiently close proximity to the carboxylate duringe xtension of the RNA structure [41] at even highere nergy (E 3 ), which makes another proton available for intermolecular PT and Gpa dissociation by Reaction(II). The latter PT reactionw as negligible for m = 2t o 3, which we attribute to higher energy requirements for structural transitions in the [RNA + 2GpaÀ4H] 4À and[ RNA + 3GpaÀ4H] 4À ions that are stabilized by additional electrostatic interactions. Likewise, CAD of [RNA + GpaÀ5H] 5À ions (Figure S3) showed only very few products from Reaction (II) at higher energy (Scheme 2C), which can be rationalized by the smaller number of protons in the [RNAÀ5H] 5À ions compared with that of the [RNAÀ4H] 4À ions, and an inherently more extended structureo ft he more highly charged nucleic acid anions. [42] The proposed interactions and PT reactions associatedw ith aGpa dissociation from [RNA + aGpaÀ4H] 4À ions are illustrated in Scheme 3; those of Arg should be similar. In addition to the guanidinium moiety,b oth aGpa and Arg have an amino group that is protonated at pH 7.5 [43] and can form an additional salt bridge with another,n ot necessarily adjacent, deprotonated phosphodiesterm oiety.H owever,t he PA of methylamine (899 kJ mol À1 ) [30] as am odel for the amino group is substantially smaller than that of methylguanidine (1002 kJ mol À1 ) [44] as a model for the guanidinium moiety,a nd we propose that facile PT occurs at approximately the same energy as that required for intramolecular PT from the guanidinium to the phosphodiesterm oiety (E 1 ' % E 1 ). At elevated energy (E 2 ' % E 2 ), two protons are transferredt ot he RNA and [aGpaÀH] À dissociates. At energy E 3 ',w hich is significantly higher than E 3 because extension of the RNA structure requires more energy when the additional amino group also forms ah ydrogen bond with the RNA, ap roton is transferred from ar emotes ite and neutral aGpa dissociates.
In summary,d issociation of [RNA + m LÀn H] nÀ ions of tmeGnd, meGnd, and Gnd at n = 2t o5a nd all energies used gave only products from loss of neutrall igand L[ Reaction (II)] by PT from [L + H] + to ad eprotonated phosphodiesterm oiety and subsequent dissociation of the [RNA + (mÀ1) LÀn H] nÀ ·L complexes (Schemes 1a nd 2B). Likewise, only products of Reaction (II) were observed for Gpa, aGpa, and Arg at n = 2t o3 , but the PT reactions (Schemes 2A, Cand 3A, C) involved in their formation include both intra-and intermolecularP Tb etween aminium,g uanidinium,c arboxylate, andp hosphodiester moieties. Finally,t he competition between the latter reactions accounts for the energy-dependent branching ratio between products from Reactions (II) and (III) in CAD of RNA complexes with Gpa, aGpa, and Arg at n = 4t o5.

RelativeStabilities of RNA-Ligand Complexes
As illustrated for [RNA + 3GndÀ3H] 3À ions in Figure 1B, some unintended loss of ligand was observed after isolation of the RNA-ligand complex ions, which we attribute to vibrational excitation in the linear quadrupole used for ion isolation. [45] The extent of ligand loss during isolation generally increased with an increasei nt he complex chargea nd number of ligands bound, and was always highest for tmeGnd (Table S2). However,i na ll experiments herein,t he fraction of [RNA + m LÀn H] nÀ complex ions decreased sigmoidally with increasing energy used for CAD ( Figure 3A);s imilar breakdown curves have been observed in CAD of noncovalentc omplexes [46] comprised of DNA and basic amino acids or smallp eptides, [47] phosphopeptides and basic ligands, [17a] duplex DNA and minor groove binders, [48] and RNA [49] and DNA [50] duplexes.
Because [RNAÀ3H] 3À ions were the only products from CAD of [RNA + 1GndÀ3H] 3À ions in the energy range investigated, the sigmoidal breakdown curve for [RNA + 1GndÀ3H] 3À ions is exactly the inverse of the appearance curve for [RNAÀ3H] 3À ions, with ac ommon E 50 value of % 15.86 eV.A si llustrated in Figure 3A, these data can be fitted with as igmoidal function withoutv ertical offset and ad ecay rate r,t hat is, Equation (1): In this case, the sigmoidal functioni s0 %( andt he inverse 100 %) at low energy and plateausa t1 00 %( inverse 0%)a t high energy,and E 50 is the energy value at 50 %yield. However, Equation (1)  However, the E 50 values fort he appearance of [RNAÀ3H] 3À ions, which correspond to the energies required to dissociate all Gnd ligandsb ound to the RNA, increased linearly with increasing m ( Figure 4A). This means that irrespectiveo ft he complexity of the reactionc oordinates for [RNA + m GndÀ3H] 3À ion dissociation, each additional Gnd ligand increased the energy required for dissociation by af ixed amount (E slope )t hat was, within error limits,i ndependent of the total number of ligands initially bound to the RNA except for m = 1. Similar behavior was observed for all other ligandsa nd complex net charges studied, althought he E 50 values generally increased in the order tmeGnd < meGnd < Gnd < Gpa < aGpa < Arg ( Figure 4A). Importantly,t his order of complex stabilityi s inconsistent with the order of PA (1032, 986, and 1051 kJ mol À1 for tmeGnd, Gnd, and Arg, respectively), [28] but insteads hows an almostl inear correlation with the number of ligand atoms  Figure 4B for n = 3 ( Figure S4 shows dataf or n = 2a nd 4). The higher stability of Arg versus aGpa [RNA + m LÀn H] nÀ complexes can be attributed to the longera lkyl chain of Arg that allows it to better adapt to the RNA structure [51] and reachm ore binding sites. [52] This correlation does not exclude the presence of stacking, cation-p,o ro ther noncovalent interactions, but suggests that hydrogen bonds and salt bridges provide the largest contribution to complex stability.
Because the E 50 values are ar elative measureo ft he minimum energy required for complex dissociation,w ec onclude that the contribution of [RNAÀn H] nÀ ·Lb inding energy to complex stability is significantly higher than that of DH PT,complex (Scheme 1), andt hat the differences in the binding energy of the different ligands primarily result from differencesi nt he number of hydrogen-bond and salt-bridge interactions that they can form. Moreover,f or each ligand, the E 50 values systematically decreased as net charge n increased (Figure4A), which is consistentw ith an increasing PA of [RNAÀn H] nÀ ions with increasing n (asa lso indicated by their PT reactivity in the collisionc ell, discussed above)t hat in turn decreases the binding energy of [RNAÀn H] nÀ ·Lc omplexes (Scheme 1).
Althoughe ach additional ligand increased the energy required for dissociation of all ligands by af ixed E slope value (within error limits), the linear-fit functions in Figure 4A did not generallye xtrapolatet o0eV at m = 0b ut showedi ntercept energiesa ss mall as (À4.23 AE 0.09) eV and as largea s( 26.03 AE 2.82) eV.I no ther words, one of the m ligands (including that for m = 1) can bind to the RNA more strongly than all others, that is, when E 50 (m = 1) > E slope or,f or tmeGnda tn = 3f or which E 50 (m = 1) < E slope ,m ore weakly than all others (Figure 4C). This strongly suggests that the 8-nt RNA provides a single, unique binding site to which only one of the m ligands binds preferentially,a long with four other binding sites to which up to four ligandsc an bind. With the exception of tmeGnda tn = 3, bindingt ot he unique site was always stronger than binding to the other four sites by af actor of up to % 4.4 ( Figure 4C).
Ap ossible RNA structuret hat agreesw ith all the experimental data from this study is the hairpin motif illustratedi n Scheme 4, with as tem that consists of only two G-C base pairs and aC UAG loop to provide au nique bindings ite. The CUAG loop has the potential for hydrogen-bonding interactions similar to those of the highly stable UUCG loop, and1 2-nth airpin structures with the former (GGAC-CUAG-GUCC, melting temperature T m = (69.8 AE 1.0) 8C) are only slightly lesss table than hairpin structures with the latter (GGAC-UUCG-GUCC, T m = (72.9 AE 1.0) 8C). [53] Hairpin structures with UUCG loops have a minimum requirement of as tem comprised of two base pairs, with meltingt emperatures of % 24, % 54, and % 55 8Cf or CG-UUCG-CG,C C-UUCG-GG, and GC-UUCG-GC, respectively. [54] Althought heory predictsn os table secondary structure for our GGCUAGCCR NA by itself, ah airpin fold could nevertheless be stabilized by binding of guanidinium ligands. For example, the crystal structure of ah airpin motif for guanidineb inding, -GG-ACGA-CC-, in which guanidine interacts with all three phos- phodiester moieties of the ACGA loop and is stacked upon the guanineb ase on the 5' side of the loop in ac ation-p interaction (Scheme4C), hasb een reported for an 18-nt guanidine-II riboswitch. [55] As imilarh airpin structure for our 8-nt RNA (GG-CUAG-CC), in whicht he guanidine moiety of the ligandsa tm = 1c an interact with all three phosphodiester moieties of the loop (Scheme 4A), explainst he unique binding site, whereas the exposed phosphodiester moieties of the stem can account for the bindingo fu pt of our additional ligands( Scheme 4B). Specifically,t he binding pattern in Scheme 4B is consistent with similar binding strengthsf or the additional ligands that give rise to the E slope values, and the approximately threefold stronger bindingo ft he ligands at m = 1( Figure 4C). Finally,ahairpin structure agrees with the weaker binding of tmeGnd at m = 1b ecause tmeGnd cannotf orm more than one salt-bridge interaction with the phosphodiester moieties of the loop, and with the binding of up to only three instead of five Arg ligands ( Figure 4A) because each additional Arg can bind to two adjacent phosphodiester moieties of the stem (Scheme 3). Any differences in the E 50 values between different ligands can then be attributed to different numbers and strengths of interactions with the phosphodiester moieties, and to additional interactions with adjacent ribose moieties.
The slopes of the linear fit functions in Figure 4A were largely independent of RNA complex ionn et chargef or meGnd, Gnd, and Gpaa tn = 2t o4 ,a nd for tmeGnd, aGpa, and Arg at n = 2t o3( Figure 4D). In these complexes, the stabilization achievedb yh ydrogen-bond and salt-bridge interactions apparently dominates over any effects of the complex ion net charge for m > 1b ut not m = 1( Figure4C), which suggests that specific binding (m = 1) is far more affected by complex net charge than the binding of additional ligands( m > 1). This is again consistentw ith the hairpin structures shown in Scheme 4, in which the charged ensity around the specifically bound ligand is much highert han that around the ligands bound to the phosphodiester moietieso ft he stem region. By contrast, at n = 5, all slope values were significantly smaller than those at n = 4, whichi ndicated that Coulombic repulsion limits overall complex ion stabilitya tn = 5. Moreover, both the E 50 (m = 1) and E slope values for aGpa and Arg at n = 4s tand out, whereas those at n = 2a nd 3f ollow the general trends discusseda bove ( Figure 4C, D). The E 50 (m = 1) and E slope ratio of close to one at n = 4i si nconsistent with au nique binding site and insteads uggests binding of up to three aGpa or Arg ligands to al argely extended RNA structure.

Sequential DissociationofL igands
As discussed by Rodgers and Armentrout [56] and Kitova and Klassen, [57] the potentiale nergy surface for noncovalent bond cleavage has as taircasea ppearance;t hat is, there should be no reverse activation barriers and endothermic noncovalent complex dissociation generally proceeds once the available energy exceeds the thermodynamic threshold. In the above CAD experiments, in whicha ll ligands were dissociated (Figure 4), we thus probedt hermodynamic complex stability even though the energy values obtained from the E 50 analysis are relative rather than absolute. [46] However,the E 50 values for sequential dissociation of individual ligands, summarized for tmeGnd, meGnd, Gnd, Gpa, aGpa, and Arg at n = 2t o4and m = 1t o5in Figure 5, do not generally indicate relative bindinge nergieso fi ndividual ligands. This is evident from the strong effect of the initial number of ligands, m,o nt he DE 50 values ( Figure 5). For example, for aGpa at n = 3and m = 3, dissociation of the first ligand was observed at an DE 50 (1) value of (20.09 AE 0.16) eV,d issociation of the second ligand waso bserveda ta na dditional DE 50 (2) value of (11.37 AE 0.26) eV,a nd dissociation of the third ligand was observed at an additional DE 50  However,b yf ar the most irregular energy differences for DE 50 were found for Gnd at m ! 3( Figure 3B, Figure 5), and for aGpa and Arg at n = 4( Figure 5). Ap ossible rationale for this observation are intricate conformational rearrangementso ft he RNA during sequential [RNA + m LÀn H] nÀ ion dissociation along with ligand scrambling and PT reactions not only between ligandsa nd the RNA (Schemes 2C and 3C) but also between ligands. As ac ase in point, the multidentate [Gnd + H] + ion should be especially prone to scrambling and PT between Gnd ligandsbecause of its high symmetry.S uch structural rearrangements would allow for the dissociation of individual ligands at energies that can be highero rl ower than the aver-age energy required for ligand dissociation, withoutc hanging the total energy required for dissociation of all ligands.

RNA BackboneC leavage at Elevated Energy
CAD of [RNA + m LÀn H] nÀ ions also produced c, y, a,a nd w fragments from RNA backbonec leavage and loss of charged and neutralR NA nucleobases at elevated energy ( Figure 2). For [RNA + m LÀ3H] 3À ions with m = 0t o2 ,w ed etermined E 50 values for the appearance of fragments from RNA backbone cleavage ( Figure 6A). These E backbone values weres ignificantly higher,b yafactor of 1.7 (Arg) to 46.5 (tmeGnd), than those for the dissociation of all ligands( Figure 6B), from which we conclude that ligand dissociation and backbone cleavage are sequentialp rocesses.S urprisingly,t he E backbone values increased in the order Gnd < tmeGnd < Arg, which is also the order of PA (986, 1032, and 1051 kJ mol À1 for Gnd, tmeGnd, and Arg, respectively). Thisi ndicates that PT to the RNA (Schemes 2a nd 3) does affect backbone cleavage, most likely by facilitating nucleophilic attack of 2'-OH groups on adjacent phosphorus atoms. [41] Specifically,t he timing between PT (Schemes 2a nd 3) and nucleophilic attack should dependo nt he value of DH PT,complex and thus ligand PA ;f acilitation of nucleophilic attack can only occur if the proton is transferred before nucleophilic attack.

Conclusions
Our comprehensive study shows that ESI and CAD can be used to obtain detailed information on RNA-ligand binding. For tmeGnd, meGnd, Gnd, Gpa, aGpa,a nd Arg ligandsi nm ixtures with an 8-nt RNA, the ESI data suggestt hat the gaseous [RNA + m LÀn H] nÀ complex ions predominantly originate from association reactions in solution by the formation of intermolecular salt bridges between the ligand guanidinium and RNA phosphodiester moieties. The order of [RNA + m LÀn H] nÀ complex stability,t meGnd < meGnd < Gnd < Gpa < aGpa < Arg, established in the CAD experiments,r evealed that salt bridges and hydrogen bonds providet he largest contribution to complex stability in the gas phase, whereas ligand PA showed some effect only on RNA backbone cleavage at elevated energy.L igand dissociation in CAD of [RNA + m LÀn H] nÀ complex ions was generally accompanied by PT from ligand to RNA, for which we have proposed mechanismst hat also account for the energy-dependent competition between neutral versus deprotonated ligand loss of Gpa, aGpa, and Arg at n = 4. Evidence for ligand scrambling during CAD, particularly for the highly symmetric Gnd, was also found, althoughs crambling did not change the total energy required for ligand dissociation.M oreover,d ata from CAD of [RNA + m LÀn H] nÀ complex ions with m = 1t o5i ndicatea nR NA structure to which one of the m ligandsb inds more strongly than all others;a hairpin motifi sc onsistentw ith this observation. In future experiments, we plan to study ligandso fi ncreased molecular complexity,s uch as diarginine, together with different RNA sequences to gain further insighti nto RNA-ligand binding and complex stabilityi nt he gas phase.