Structural and Kinetic Profiling of Allosteric Modulation of Duplex DNA Induced by DNA‐Binding Polyamide Analogues

Abstract A combined structural and quantitative biophysical profile of the DNA binding affinity, kinetics and sequence‐selectivity of hairpin polyamide analogues is described. DNA duplexes containing either target polyamide binding sites or mismatch sequences are immobilized on a microelectrode surface. Quantitation of the DNA binding profile of polyamides containing N‐terminal 1‐alkylimidazole (Im) units exhibit picomolar binding affinities for their target sequences, whereas 5‐alkylthiazole (Nt) units are an order of magnitude lower (low nanomolar). Comparative NMR structural analyses of the polyamide series shows that the steric bulk distal to the DNA‐binding face of the hairpin iPr‐Nt polyamide plays an influential role in the allosteric modulation of the overall DNA duplex structure. This combined kinetic and structural study provides a foundation to develop next‐generation hairpin designs where the DNA‐binding profile of polyamides is reconciled with their physicochemical properties.


Introduction
DNA-binding polyamides( PAs) are cell-permeable transcriptional modulators whichf unctionb yi nhibiting RNA polymerase-mediated elongation and/ort ranscription factorb inding to its target double-stranded DNA (dsDNA)c onsensus sequence. [1] Of the variousd esignsr eported, [2] hairpin PAsa re the mostw idely used [1b,c,f, 3] where the primary sequenceo fNmethyl pyrrole( Py) and N-methyl imidazole (Im) heterocyclic amino acids defines the selectivity of dsDNA binding ranging from 7u pt o2 4b ase-pairs in length (e.g., PA1,F igure 1). [1b, 4] At present,a nu nmet challenge in their furtherd evelopment as ag eneral tool to modulate gene-selective transcriptioni sa n in-depthu nderstanding of the interplay between the dsDNA binding profile of PAsd etermined in vitro, with their overall [a] K. Aman, + Dr.G .P adroni, + Dr.J.A.P arkinson, Prof. G physicochemical properties which impactc ell uptake, and ultimately target engagement in vivo. [5] We have recently expanded the heterocyclic repertoire of current Py-Im hairpin PA designs to include N-terminal thiazole-4-carboxylic acid units (Nt). [6] Nt-building blocks (e.g., PA2-3)d irect ah ydrogen-bond-acceptor (N3) atom towards the floor of the minor groove and forms ah ydrogen bond with the exocyclic hydrogen bond donor amine (N2) of G. A key structural difference with the incorporation of an Nt-unit in the N-terminal positiono fahairpin PA is the endocyclic sulfur atom whichc hanges both the geometry and hydrophobicity (logD) of this heterocycle (Figure1). [7] Furthermore, when a bulky isopropyl substituent is installed in the 5-position (i.e., iPr-Nt, PA3), am ore pronounced compression of the major groove is observed relative to the archetypical hairpin PA1·dsDNA complex. [6] These results imply that allosteric modulation of the DNA duplex imparted by PAsi si nfluenced by both the nature of the N-terminal heterocycle pairing with the N2 of G, and the steric bulk of substituents not directly involved in selectivem inor groove recognition. [8] What is unclear at present is how thesec hanges to the N-terminus influence the kinetics of target versusm ismatch binding to dsDNA sequences.
In this manuscript, we reportalabel-free biophysical assay to profile the affinity,s equence-selectivity and binding kinetics of PA·dsDNA interactions where the N-terminal heterocycle is systematically altered( PA1-4). PAsc ontaining N-terminal Im units (i.e., PA1 and PA4)e xhibit enhanced selectivity for their target sequences relative to cognateN tu nits (i.e., PA2-3). Whilst increasing the steric bulk of the iPr-Im unit (PA4)d oes not impact DNA binding affinity for its target sequence, NMR structurala nalysis reveals the larger iPr-Im unit does induce more pronounceds tructural perturbation of the target dsDNA duplex relative to PA1,w hich contains an N-terminal Me-Im unit.

Design and synthesis of hairpinpolyamides (PA1-4)
The heterocyclic core of ak nown hairpin PA sequence (PA1) was chosen as our exemplar scaffold to explore the dsDNA binding profile as af unction of four different N-terminal heterocycles. [4f, 5b, 8a] PA1 has an established high affinity binding profile for the general sequence 5'-WWGWWCW (whereW= A/T), for which we used 5'-ATGTACT as the target sequence in an immobilized DNA duplex (ODN1). [6, 8a, 9] Compounds PA1-4 were prepared using Boc-based solid phase synthesis on a b-Ala PAMr esin via amide coupling of the corresponding heterocyclic carboxylic acid (Scheme S1). [6,10] Polyamides incorporatingN-terminal imidazole units exhibit picomolar bindingaffinity for their target dsDNA sequence As chematic of the experimental setup is shown in Figure 2. DNA duplexes (ODN1-3,T able 1) were immobilizedo nagold surfacea nd contained af luorophore reporter positioned in close proximity to the proposed PA bindings ite. PA binding to an immobilized DNA duplex containing the target binding sequence( ODN1) [11] resultsi nf luorophore quenching, which is then restored upon dissociation. This provides an isothermal reporter of the binding kinetics (i.e., k on and k off )a nd the equilibrium dissociation constant( K D ). [12] The same fluorescencer eporter setup was also used to determine the duplex stabilization profile( i.e., DT m )o fP A ·ODN complexes as af unction of a temperature gradient.
Kinetic analyses of the binding profileo fPA1-4 to ODN1 show all four PAse xhibit high-affinity binding (Table 1). Whilst the Im-containing PAs( PA1 and PA4)e xhibit K D values in the picomolarr ange, the Nt-containing PAs( PA2-3)e xhibited a binding affinity that is approximately an order of magnitude lower (i.e.,i nt he low nanomolar range). Rate maps of PA1-4 targeting ODN1 provided deeper insight into the origin of the differences in the K D values of our PA set ( Figure 3). Although the dissociation rate (k off )f or each PA wass imilar,t he rate of

G-selective dsDNAb indingobserved for all four polyamides
The sequence selectivity profile of PA1-4 was exploredu sing duplexes where the target binding sequence in ODN1 was replaced with mismatched sequences (ODN2-3). Analyses of the binding kinetics show that Im-containing PAs( PA1 and PA4) are more G-selective relative to Nt-containing PAs( PA2-3, Figure 4). Whilst the rates of association (k on )o fPA4 for all sequences ODN1-3 weres imilar,t he dissociation rates (k off )w ere significantly faster form ismatched sequences (ODN2-3). Al ess pronounced k on /k off trend was observed for PA1 binding to ODN2,w hile no interaction was measured with ODN3.
Consistent with our previous DNA-foot-printing data, [6] the most promiscuous dsDNA binding profile observedw as PA2 (Figure 4b)w here the K D was virtually the same for the target (ODN1)a nd the mismatch (ODN2)s equence. Out of the PA series, PA3 displayed the most uniqueb inding profile (Figure 4c). In this case, ad ecrease in both k off and k on was ob-served for the bindingp rofile of PA3 for ODN2,w hile no interaction was observed for ODN3.
This experimental setup was also used to determine duplex stabilization of PA·dsDNA complexes compared to the free DNA duplex melts. Ag lobal Boltzmann fit over three independentr uns was used to determine the mid-points of the meltingt ransitions (T m )f or free ODN1-3 andi nc omplex with 20 nm PA1-4.T he UV/Vis melting profiles of the PA·dsDNA complexes confirm asimilar trend in dsDNA sequence selectivity (i.e.,h igher DT m )o bserved in the fluorescencee xperiments ( Figure S3). Of particular note was the meltings tabilization of PA4,w hich displayed excellent G-selectivity relative to PA1-3. Consistent with our kinetics profiling ( Figure 4) and previous DNA-foot-printing work, [6] PA2 exhibitedl imited sequence selectivity as highlighted by duplex stabilization observedf or all three ODNs. Taken collectively,the kinetic and melting analyses show that the sequence selectivity of Im-containing PAs( i.e., PA1 and PA4)i ss uperior to Nt-containing analogues (i.e. PA2-3). Furthermore, whilst enhancing steric bulk on the 5-position of the Nt-series enhanced G-selectivity,t his had little effect on the Im-series (i.e., PA1/PA4).

NMR structurala nalysis of the PA4·dsDNAcomplex
In order to gain insighti nto the influence of the iPr-Im unit incorporated in PA4 when in complex with its target dsDNA sequence, NMR studies were undertaken using the self-complementaryd odecamer sequence d(CGATGTACATCG) 2 (ODN4). Titration experiments of PA4 into as olutiono fODN4 confirmed the formation of a1 :1 PA4·ODN4 complex. 2D NOESY studies at 4d ifferent mixing times identified as uite of strong NOE cross-correlations from H4 of the iPr-Im building block to G5H1 and the G5N2 exocyclic amine, whichi mplies that the iPr-Im N3 is directed towards the flooro ft he minor groove ( Figure 5; Figure S9). NOE cross-peaks from H4 and H5 of the iPr-Im buildingb lock to Py1 and the b-alanine tail in the PA4·ODN4 complex is indicative of the PA binding to its target sequence in the hairpin conformation. Comparative NMR structural analyses of polyamide·dsDNA complexes Previous NMR structural work highlighted an increased propensity of PA3 to compress the major groove when in complex with its target DNA sequence (PA3·ODN4) relative to PA1·ODN4.Asimilar trend in enhanced major groove compression was also observed with PA4·ODN4 relative to PA1·ODN4 ( Figure 6). However,t he extent of major groove compression was not as pronounceda st hat observed for the PA3·ODN4 complex. The origins of thesed ifferences become apparent when comparing the extent of minor groove penetration of the three complexes (Figure 7). NMR-restrained molecular dynamics of the PA1·ODN4 complexr eveal PA1 penetrating deep within the minor groove,e xemplified by ah ydrogen bond distance of 2.01 between Me-Im N3 and the exocyclic amine G5N2 (Figure 7a). [13] In contrast, the PA3·ODN4 complex shows ar educed level of minor groove penetration with an average distance of 2.36 between the iPr-Nt N3 and the exocyclic amine G5N2 (Figure 7b). [13] The PA4·ODN4 complex on the other hand shows as ignificant level of minor groove penetration (2.10 )r elativet oPA3·ODN4 but it is not as extensive as  that observed for the PA1·ODN4 complex (2.01 ). [6] We therefore concludet hat both the nature of the N-terminal heterocycle and the steric bulk distal to the DNA-binding face of aP A scaffold influences the allosteric modulation of at arget dsDNA sequence.

Discussion
This combined kinetica nd structurals tudy has shown that the type of N-terminal heterocycle and its substituents influences the dsDNA binding profile and the overall structure of the duplex. We discussh ere severalc onclusions that emerged from our results.

N-terminal heterocycle of ahairpinpolyamide influences rate of association to target dsDNAsequence
Firstly,all four PAsexhibit high affinity (low nanomolar-picomolar) for its target dsDNA sequence. However,t he two N-terminal Im-containing PAs( PA1/PA4)s howed ah igher binding affinity relative to the Nt-containing PA2-3 via an increasei nt he rate of association. Although there has not been as tudy dedicated to evaluating the influence of the hairpin PA N-terminus, ap revious SPR-based study by Sugiyama et al. has shown that  the number of Me-Im and their positioning in ah airpin PA scaffold can have ad isproportionate impact on the k a and K D relative to only small changes in the k d. [14] In contrast, replacing internal Py/Im heterocycles with more flexible b-alanine units influences both k a and k d parameters. [15] Extensive work by Dervan et al. has investigated heterocyclic changes to the internal positions of hairpin PA structures. [16] However, our results highlight the N-terminal position can be used as ac onvenient site to tune parameters of dsDNA binding and overall physicochemicalproperties.
The N-terminal heterocycle position of hairpinpolyamides influence DNA structural perturbations Our structurala nd binding analyses show that whilst an increase in the steric bulk of the iPr-Im unit does not impact dsDNA binding affinity to its target binding site (i.e., PA4·ODN4 complex), an improvement in G-selectivity relative to the iPr-Nt unit (i.e., PA3·ODN4 complex) is likely due to a greater level of minor groove penetration (see Figure 7), and in turn improved recognition of the N2 amine of G. However,t he extent of major groove compression of the PA4·ODN4 complex ( Figure 6a) is less than in PA3·ODN4 (Figure 7). This suggests af ine interplay between minor groove penetration versus major groove compression, with enhancedm ajor groove compression occurring if the hydrogen-bond between the N-terminal buildingb lock and the N2 of Gi sw eaker as in PA3·ODN4,t hereby reducingp enetration of the minor groove.

Conclusions
These experiments wered esigned to probe how an increasei n the steric bulk of heterocyclic buildingb locks of PA impacts the binding kinetics and the allosteric distortion of dsDNA containing the target binding sequence. Although what superficially appears to be as ubtle increase in steric bulk at locations within aP As caffold not directly involved in dsDNA base-readout, these data suggest that strategic changes in the Im and Nt substitution pattern can be used to fine tune the sequenceselectivity of dsDNA binding as well as the overall physicochemicalp roperties of PA scaffolds. [17] We envisage that the strategic incorporation of modified heterocyclic building blocks within aP As caffold could be applied more broadly as a strategyt oe nhance cell uptake and potencyo ft ranscriptional modulation in cellulo.