A Minimal Load‐and‐Lock RuII Luminescent DNA Probe

Abstract Threading intercalators bind DNA with high affinities. Here, we describe single‐molecule studies on a cell‐permeant luminescent dinuclear ruthenium(II) complex that has been previously shown to thread only into short, unstable duplex structures. Using optical tweezers and confocal microscopy, we show that this complex threads and locks into force‐extended duplex DNA in a two‐step mechanism. Detailed kinetic studies reveal that an individual stereoisomer of the complex exhibits the highest binding affinity reported for such a mono‐intercalator. This stereoisomer better preserves the biophysical properties of DNA than the widely used SYTOX Orange. Interestingly, threading into torsionally constrained DNA decreases dramatically, but is rescued on negatively supercoiled DNA. Given the “light‐switch” properties of this complex on binding DNA, it can be readily used as a long‐lived luminescent label for duplex or negatively supercoiled DNA through a unique “load‐and‐lock” protocol.


Introduction
Owing to its central role in the process of life,n ew methods to visualize DNAa nd its dynamics are constantly being sought. Thea bility to image DNA, DNA-protein interactions,a nd changes in DNAs tructure in vivo and in vitro has provided key insights into fundamental cellular function. [1] In this context, fluorescence microscopy has proven to be particularly versatile and this has led to alarge variety of luminescent small organic molecules being investigated and developed as DNAi maging probes. [2,3] Due to some of the drawbacks of these conventional probes,the use of transition metal complexes in this role has been explored.
Thep roperties of water-soluble salts of d 6 metal centres, especially polypyridyl Ru II cations,h ave proven to be particularly promising;a st hese species typically exhibit photostable [3] metal-to-ligand charge transfer (MLCT) excited states,t hey frequently possess bright long-lived emission with large Stokes shifts. [4][5][6][7][8][9][10][11][12] Due to this attractive combination of properties,complexes with extended aromatic ligands capable of intercalating into DNAwere investigated and key early studies revealed that [Ru II (LL) 2 (dppz)] 2+ (LL = 2,2'bipyridine or 1,10-phenanthroline,d ppz = dipyrido[3,2a:2',3'-cj phenazine), Figure 1, displays a"DNAlight switch" effect. [13] Through hydrogen-bonding interactions with the dppz ligand, the emissive state of this complex is solventquenched until solvent shielding through DNAi ntercalation "switches on" luminescence. [14][15][16][17] Consequently,anumber of dinuclear ruthenium complexes based on linked Ru II (dppz) fragments have been developed that thread through DNA duplexes,see Figure 1for examples. [18,19] In this binding motif, one of the bulky ruthenium centres passes through the DNA duplex to produce af inal structure where am etal centre resides in both DNAgrooves.Such systems exhibit increased affinity and enhanced binding specificity towards particular DNAs tructures. [20,21] These threading intercalators have been extensively studied in optical tweezer experiments in which as ingle molecule of DNAt ethered between two optically trapped beads provides amethod to directly measure the position and force on the beads.Using this technology,the very slow DNA threading kinetics of the complexes were accelerated so that they can be monitored in real time. [22] This facilitated the quantification of binding and dissociation kinetics,w hile simultaneously probing any concomitant DNAs tructural changes.These experiments revealed atwo-step "thread-andlock" mechanism, whereby threading results in as table "locked" complex that is exceptionally slow to dissociate. They also disclosed that ancillary ligands play alarge role on this thread-and-lock mechanism;c omplexes incorporating phen ancillary ligands show locked binding,whilst analogues containing more compact bpy ancillary ligands require no DNAe xtension to de-thread. [23] Again, chirality has ap rofound role in the binding affinity of threading events. [24] Although their threading interaction makes these complexes attractive candidates as optical probes for DNA, their use in such applications is restricted by the fact that they are not intrinsically cell permeant. By contrast, in related work, the Thomas group has demonstrated that certain dinuclear ruthenium complexes of the form [{Ru(L) 2 } 2 -m-tpphz] 4+ (where Li sadipyridyl related ancillary ligand and tpphz = tetrapyrido[3,2-a:2',3'-c:3'',2''-h:2''',3'''-j]phenazine) are taken up by live cells. [25] Several derivatives have been developed as DNAp robes for live cell super-resolution microscopy,a nticancer therapeutics,and antimicrobial theranostics. [26][27][28][29] Consequent structural studies on [{(Ru(bpy) 2 } 2 -m-(tpphz)] 4+ (bpy = 2,2'-bipyridine) and related structures (1 4+ ,F igure 1), have demonstrated they possess structural specificity for the DNAq uadruplex structure such as the human telomeric repeat sequence and chirality dependant loop threading into quadraplexes. [30] Whilst spontaneous intercalation of 1 4+ into stable duplexes is not observed, [31] more recent NMR-based studies revealed that it can bind to as hort B-DNAo ligonucleotide through threading. Furthermore,due to the short and rigid tpphz bridging ligand, insertion of this "minimal threader" is acutely dependent on metal centre chirality. Whilst L,L-1 4+ displays locked threading, duplex-bound D,D-1 4+ equilibrates between ag roove-bound and threaded state. [32] These observations suggest that threading of L,L-1 4+ could occur with longer destabilized duplex sequences.T o investigate this possibility,w ec ombine optical tweezers with laser scanning confocal microscopy (LSCM) to directly investigate the kinetics and binding modes of the three stereoisomers of 1 4+ ,w hile simultaneously probing their luminescent properties.Through these experiments we found that L,L-1 4+ functions as a" lockable" luminescent stain for duplex DNA.

DNA Binding Occurs in TwoPhases
Thei nteraction of the chloride salt of L,L-1 4+ with as ingle molecule of l-DNAw as first investigated. AD NA dumbbell was assembled using am icrofluidic laminar flow cell (Figure 2A,c hannels 1-3), and held at ac onstant force (20-50 pN) in the presence of the complex (2-512 nM) ( Figure 2A,c hannel 4). At these forces,t he DNAr emains double-stranded, however the stability is reduced at higher forces (> 60 pN). Thek inetics of the interaction could be readily monitored through the increase in DNAextension as af unction of time ( Figure 2B and Supplementary Data Figure 1) as probe binding increases the spacing between base pairs.
At the lowest concentration (2 nM), the resulting time traces were fit to as ingle exponential increase [ Figure 2C  phase rate constant increases with concentration of 1 4+ , whereas the slow phase rate constant is concentration independent.
These observations indicate that all three stereoisomers interact with DNAi nam ulti-step process,c onsistent with am inimal two-step kinetic model ( Figure 2E)i nvolving ar apid, concentration-dependentb inding phase and as low, concentration independent unimolecular phase.
Previously published single molecule experiments on threading mono-intercalated Ru II complexes,s uch as [{(Ru-(LL) 2 } 2 (bidppz)] 4+ ,s how that they bind to DNAt hrough as ingle state mechanism. [22] Thee xception is ad inuclear complex containing am ore flexible central bridging ligand, which initially binds to duplex through ac onventional intercalated state before threading occurs. [33] On the other hand, bulk optical titrations and linear dichroism experiments on the original, more rigid, threader are consistent with atwostate mechanism, in which groove binding proceeds threading. [24,34,35] Given these observations,a nd the facts that; 1 4+ contains as horter inflexible tpphz bridging ligand, binds extended stable duplex sequences through groove binding, and binds unstable duplexes through both groove binding and threading, we attribute the rapid binding phase observed in these experiments to groove binding and the slow phase to threading ( Figure 2E).

Ru-bpy Stereoisomer Intercalators Differentially Affect DNA Elasticity
Next, we determined the equilibrium DNAe xtension at asaturating concentration of 1 4+ at various forces (10-60 pN, Figure 2F), and fit those to the extensible wormlike-chain model [Eq. (7)] to obtain the contour length, persistence length and stretch modulus of fully intercalated DNAfor each stereoisomer ( Figure 2F,T able 1). All the stereoisomers induce as imilarly pronounced increase in duplex contour length, which is expected from an intercalative interaction, comparable to the widely-used SYTOX Orange (Table 1, Figure 2F). However, whilst L,D-1 4+ and D,D-1 4+ produce significant decreases in persistence length and stretch mod-  ulus,addition of L,L-1 4+ results in only slight changes in these parameters compared to duplex DNA ( Table 1). Thes ubstantial increase in DNAe lasticity produced by addition of L,D-1 4+ and D,D-1 4+ is consistent with our NMR model which shows that each bound D Ru II centre has adestabilizing effect on DNAs tructure. [32] Interestingly, L,L-1 4+ preserves the DNAs tructural properties far better than SYTOX Orange (Table 1a nd Figure 2F).

Force Stretching DNA Increases Ru-bpy Intercalation Affinity
To determine the force dependence of the intercalator binding affinity,wemeasured the equilibrium extension (L eq ) of each stereoisomer at increasing intercalator concentrations and three constant forces (20, 30 and 50 pN,F igure 2G). The resulting curves were fit to the well-established McGhee-von Hippel, MVH, binding isotherm [Eq. (8)], [36,37] which allows estimation of the binding site size per molecule (n)and force dependent K d ( Table 2). Thef its show that all three stereoisomers bind DNAextremely tightly (low nanomolar K d ), and that force stretching further increases affinity (lower K d ). The binding site size, n,o fa ll three stereoisomers increases with increasing force.F or both the L,L-1 4+ and L,D-1 4+ stereoisomers,the binding site size increases from % 1.5 bp (20 pN) to % 2bp(50 pN), suggesting that threading is increasingly likely at high forces with avery high degree of threading occurring at high force.T he binding site size of D,D-1 4+ is consistently smaller, increasing from % 1.3 bp (20 pN) to only % 1.8 (50 pN).This is consistent with our previous NMR studies showing that duplex-bound D,D-1 4+ is equilibrating between groove binding and threaded states. [32] Parametrisation of the Minimal Two-Step Binding and Threading Kinetic Model To determine the parameters of the minimal kinetic model (k 1 , k À1 and k 2 ,F igure 2D), we took advantage of the fact that at low concentration (2 nM) initial binding is rate limiting ( Figures 2B and C), enabling us to determine the corresponding rate constants (k 1 and k À1 )u nder these  conditions for all three stereoisomers.F irst, the DNA extension was measured for 5minutes at forces ranging 10-50 pN ( Figure 3A,left and SI Figure 3A). Asingle exponential fit yields the observed binding rate constant k f [Eq. (4)].
Thesame DNAmolecule was then moved to abuffer-only channel (Figure 2A,P osition 3) to measure the dissociationinduced DNAshortening for 5minutes ( Figure 3A,right and SI Figure 3). As ingle exponential fit directly yields the dissociation rate constant k À1 ( Figure 2D). Theb inding rate constant k 1 is then calculated as indicated by Equation (4). We then determined the binding and dissociation rate constants as af unction of force (10-50 pN). Thed ata show that increasing force results in faster binding ( Figure 3B, k 1 ), whereas dissociation is largely force independent ( Figure 3B, k À1 ). Similar results were obtained for all stereoisomers (Supplementary Data Figure 3), suggesting that increasing the separation between base pairs at higher forces facilitates binding but not dissociation. Afit to the Bell-Evans equation [Eq. (5)] yields the zero force rate constants and the location of the transition state barrier (Figure 3Band (Table 3). Conversely, reaching the dissociation transition state requires no further deformation of the DNA (Table 3, Dx À1 % 0nm). Thebindinginduced DNAe longation provoked by 1 4+ is comparable to the data [23] reported for D,D-[{(Ru(bpy) 2 } 2 (bidppz)] 4+ showing that the structural perturbation required to thread a[ Ru II -(bpy) 2 ]m oiety is similar in both systems,a sm ight be expected. Interestingly all three stereoisomers have a Dx À1 of % 0. To determine the rate constant for the second binding mode (k 2 ), we performed measurements at high concentration (32 nM) for all stereoisomers ( Figure 3C and Supplementary Data Figure 4). Ad ouble exponential fit of the binding curves yields the two observed association rate constants, k f and k s .A th igh concentrations,t he fast phase becomes too fast to be fit accurately,therefore we fixed k f to the k 1 values obtained previously in the low concentration experiments adjusted to the higher concentration, which enables us to accurately fit for k s ( Figure 3C and Supplementary Data Figure 4a nd Supplementary Data Table 1-2). Interestingly,e ven at high complex concentration (32 nM) dissociation remains single exponential, with fits yielding the same observed dissociation rate constant (k off ), consistent with the low concentration experiments ( Figure 3C and Supplementary Data Figure 4). Furthermore,the dissociation rates for all stereoisomers are comparable at low and high concentrations (k À1 % k off )(Supplementary Data Figure 3and Supplementary Data Figure 4), suggesting that dissociation from the threaded state is very slow [k À2 % 0, Eq. (4)].
Theresulting fits show that increasing the force facilitates both groove binding, as indicated by the increasing k 1 ,a nd also threading,indicated by the increasing k 2 .Comparison of the calculated values of Dx 1 and Dx 2 (Table 3) suggests that the DNAi sd eformed by % 2 in the initial groove binding step and an additional % 1 during the slow threading step. As such, both steps are aided by increased extension of the DNA. Our NMR-based studies, [32] and that of others, [38] have shown that groove binding of this class of compound leads to steric clashes within the minor groove.T hus,weattribute the cause of the groove-binding induced extensions to these interactions.A tl ow concentration (2 nM) the initial groove binding is rate limiting however at higher concentrations (! 8nM) the second step becomes rate limiting with groove binding occurring faster than threading ( Figure 3B).
Based on the dissociation and association rate constants, we calculate the equilibrium constant of the initial binding phase as afunction of force ( Figure 3D,Supplementary Data Figure 3and Table 3). As expected, the dissociation constant decreases with increased force,indicating tighter binding.The calculated position of the transition state barrier (Dx eq ) confirms a % 2.5 distortion for binding.F rom this and the previously determined contour lengths at saturation, L eq ,the site size per bound ligand, n,c an be estimated for each stereoisomer as % 2b ase pairs [Eq. (6)],i na greement with the MVH analysis (Table 2). Significantly,all three complexes display exceptional high binding affinities ( 10 nM) at the extrapolated zero-force values.Notably,the estimated K d for complex L,L-1 4+ is six-fold smaller than that reported for D,D-[{(Ru(phen) 2 } 2 (bidppz)] 4+ and is-as far as we are aware-the lowest reported for any mono-intercalator.T his enhanced binding affinity of L,L-1 4+ compared to [{(Ru-(LL) 2 } 2 (bidppz)] 4+ complexes can be attributed to differences in the dissociation rates of the systems.A sd iscussed above, the DNAassociation rates for all the stereoisomers of 1 4+ are close to,orfaster, than those of [{(Ru(bpy) 2 } 2 (bidppz)] 4+ ,but the dissociation rate of L,L-1 4+ is closer to that of [{(Ru-(phen) 2 } 2 (bidppz)] 4+ .W eh ypothesize that the lower DNA distortion required to thread a[Ru II (bpy) 2 ]moiety along with aclose match between the width and steric demands of duplex DNAa nd the bound minimal threader L,L-1 4+ leads to optimization of association and dissociation rates.H owever, further related compounds would need to be validated to confirm this.

Luminescent Imaging of Intercalated DNA
Despite their light-switching properties upon DNAb inding, the distinct emission properties of the different isomers of  Figure 3A,S upplementary Data Figure 2). After 5-minute incubation, as ingle confocal scan was taken ( Figure 4A,l eft, Supplementary Data Figure 5). Next, the same molecule of DNAwas moved to the buffer only channel (Figure 2A,Position 3.) and held at constant force for 5minutes before taking as econd confocal scan ( Figure 4A,r ight, Supplementary Figure 5). Relative DNAemission intensity per base pair was calculated after the association step, I a ,a nd after the dissociation step, I d ,a nd plotted against force for each stereoisomer ( Figure 4B). A clear force dependent increase in DNAl uminescence is observed, consistent with the force dependent increase in equilibrium extension. As expected, the emission intensity is greatest after the association step,a nd reduced after the 5minute dissociation step.A lthough this is ag eneral trend, there are some very clear differences between stereoisomers. L,L-1 4+ displays the greatest emission intensity at all forces after the initial association step and most interestingly retains its higher intensity after the dissociation step ( Figure 4A,B ). This suggests al arger population of permanently-bound, threaded L,L-1 4 . We next performed the same experiments at much higher intercalator concentration (32 nM). As before the DNAluminescence displayed ac lear force dependence; however,e ven at 10 pN ah igh level of emission was observed ( Figure 4C,D ). Furthermore,a ll three isomers displayed comparable DNAe mission intensities despite the fact that L,L-1 4+ displays ah igher equilibrium extension than L,D-1 4+ and D,D-1 4+ .A lthough we cannot directly distinguish between emission from the groove bound versus threaded state this suggests that asignificant amount of emission is coming from non-intercalating molecules,which is consistent with previous reports showing that the groove bound complex is also emissive. [31] Again, after a5 -minute dissociation step, L,L-1 4+ displays the highest level of emission. This provides further evidence for apopulation of luminescent molecules bound through an on-intercalating mode which rapidly dissociate to leave the stable threaded population.
Based on the observed force dependence of both association and dissociation, we compared the three stereoisomers in a" load and lock" assay of DNAl abelling. First, we incubated the DNAwith al ow concentration of intercalator,1nM, clamped at high force, 50 pN,t om aximize intercalative binding ( Figure 4E,left) and after 5minutes we performed aconfocal scan ( Figure 4F,l eft) to confirm Figure 4. Comparison of the imaging properties of the different stereoisomers. A) Confocal images of L,L-1 4+ intercalator bound DNA held between the two trapped beads. The DNA was imaged after incubation with 2nML,L-1 4+ under aconstant force clamp (10-50 pN) for 5minutes ("Association", left). The same piece of DNA was then imaged after af urther 5minutes held at constant force (10-50 pN) in the buffer only channel ("Dissociation") Scale bar = 1 mm(B) Quantification of emission intensities of labelled DNA after association in the presence of 2nMintercalator (I a ,dark red) and dissociation (I d ,l ight red) with DNA held at each force, for each stereoisomer.( )Confocal imageso f L,L-1 4+ intercalator bound DNA held between the two trapped beads. The DNA was imaged after incubation with 32 nM L,L-1 4+ under aconstant force clamp (10-50 pN) for 5minutes ("Association", left). The same piece of DNA was then imaged after afurther 5minutes held at constant force (10-50 pN) in the buffer only channel ("Dissociation") Scale bar = 1 mm( D)Q uantification of the fluorescentintensities of the DNA after association in the presence of 32 nM intercalator(I a ,d ark green) and dissociation (I d ,light green) with DNA held at each force, for each stereoisomer. E) Average extension-time trajectories (n = 3) comparingthe three stereoisomers for the "bind and lock" protocol. First the DNA is clamped at 50 pN for 5minutes in the presence of 1nMof1 4+ , (left). The DNA is then moved to buffer only channel and held at 5pNfor 5minutes (right). F) Confocal images comparingthe emission of each DNA-bound stereoisomer during the "bind and lock" protocol. Ac onfocal scan is taken after a5minute 50 pN clamp with 1nMintercalator ("Association", left) and another scan after a5minute 5pNclamp in the buffer only channel ("Dissociation", right) Scale bar = 1 mm(G) Average extension-time trajectories (n = 3) for binding and dissociation of L,L-1 4+ to torsionally constrained but not supercoiled DNA (s = 0) and negatively supercoiled DNA (s = À0.2). H) Correspondingc onfocali mages from (G), following association and dissociation of L,L-1 4+ in "bind and lock" protocol.
"loading" of the complex onto DNAb efore moving to the buffer only channel and clamping the DNAa tl ow force (10 pN) ( Figure 4E,right). At this low force we reasoned that the rate of dissociation of any threaded 1 4+ would be negligible compared to its groove bound form. Consequently, after 5minutes in the buffer only channel, we took another confocal scan ( Figure 4F,right). Consistent with our assumptions,a nd the threading capabilities of the stereoisomers,w e found that, after this protocol, little emission remained in samples of DNAe xposed to L,D-1 4+ and D,D-1 4+ ,y et DNA loaded with L,L-1 4+ still remained brightly luminescent, indicating that alarge proportion of this stereoisomer remains "locked" onto the duplex through threading. This is in sharp contrast to the extremely rapid dissociation of SYTOXo bserved following the same protocol (Supplementary Data Figure 6).
To investigate whether ionic strength regulates probe binding,w em easured the association and dissociation rate constants at low and high NaCl concentrations (Supplementary Data Figure 7A). Thed ata show that at 50 pN,h igher ionic strengths (200 and 500 mM NaCl) slow probe binding,as evidenced by a % 3-fold reduced association rate constant (k a ), whereas the dissociation rate constant (k d )r emains unchanged (Supplementary Data Figure 7B). These results are consistent with arole of metal ions in shielding the DNA negative charge,thereby reducing the affinity of the positively charged ruthenium compounds.C onsequently,w ea lso observe ac orresponding decrease in DNAe xtension after probe-binding (Supplementary Data Figure 7C).
To further demonstrate the potential application of these compounds,w em easured the kinetics of L,L-1 4+ binding to torsionally constrained DNA( Figure 4G,H). Binding is dramatically impaired in the presence of torsionally constrained DNA( s = 0, Figure 4G,H). Interestingly,w hen the DNAisnegatively supercoiled, DNAbinding is rescued (s = À0.2, Figure 4G,H). Thed ata from binding and dissociation kinetic experiments show that L,L-1 4+ preferential threads in negatively supercoiled DNAcompared to relaxed torsionally constrained DNA( Figure 4G,H).

Conclusion
This work presents the highest affinity mono-intercalating DNAb inding complex to date,a nd represents the shortest linked Ru II (dppz) complex capable of DNAthreading. Given the length of tpphz and the diameter of duplex DNA, any further reduction in the threading linker length could not be accommodated across the DNAhelix (Figure 1).
Theo bservation that all three stereoisomers of 1 4+ bind duplex DNAintwo steps is consistent with previous studies [33] initial groove bound or unthreaded intercalated states prior to irreversible threading (k À2 % 0). Similar to the Michaelis-Menten mechanism, our calculations of K d are only valid under the assumption of rapid pre-equilibrium, which is supported by our measurements of k f and k s ( Figure 2D). Due to the rigid linker structure of 1 4+ ,w ec an rule out the possibility of unthreaded intercalation and attribute the fast step to groove binding;the fact that this step is concentration and force dependent is intriguing. Previous studies on the interaction of [Ru II (LL) 2 (dppz)] 2+ and derivatives with DNA have frequently shown cooperative binding effects,o ften driven by stacking interactions of ancillary ligands.T hus,i t seems likely that at higher complex loading ratios,c harge neutralization effects involving the electronegative minor groove facilitate similar interactions between individual groove-bound 1 4+ cations.T he strong force dependence of the fast initial binding step and addition force dependence of the slower threading step suggests that majority of DNA deformation occurs during the groove binding step,w ith asmaller additional deformation required for threading.
Thes tudies herein highlight the influence of chirality on threading interactions as it is clear that individual stereoisomers exhibit different binding behaviors into stretched DNA. Due to unfavourable steric interactions,incorporation of each D metal centre into the threader causes increased DNA flexibility,seen as areduction in both DNApersistence length and stretch modulus of bound DNA, and also ar eduction in binding affinity (Table 1). This means that, of the three stereoisomers, L,L-1 4+ displays the optimized thermodynamic and kinetic threading stability,binding to DNAwith highest reported affinity for am ono-intercalator.A sac onsequence, once L,L-1 4+ is loaded on to stretched DNA, on removal of the stretching force it is effectively irreversibly threaded into duplex ( Figure 4E).
As ar esult of this load and lock approach, single DNA molecules can be permanently labelled by exposure to al ow concentration of luminescent small-molecule dye without the demanding,a nd often structurally disruptive,p rotocols required to covalently attach small molecules or large biomacromolecule structures such as fluorescent proteins.A st his dye only becomes brightly luminescent on binding to DNA, and its groove bound form is easily washed away,the protocol we describe facilitates high-contrast single molecule imaging of duplex DNAt hrough optical microscopy.A st his class of dye is compatible with super-resolution SIM and STED techniques,arange of possible imaging applications is apparent, such as the intriguing possibility of developing aprobe to specifically visualize destabilized, transcriptionally active,D NA.T his possibility is further supported by the observation of the dramatic reduction in binding to torsionally constrained DNAand the rescue in binding to negatively supercoiled DNA. These studies will also inform the structural optimization of their cell permeant analogues to facilitate the development of new and novel imaging probes and therapeutic leads.W ork addressing these issues will form the basis of future reports.