A general approach to the design of allosteric, transcription factor-regulated DNAzymes

Here we explore a general strategy for the rational design of nucleic acid catalysts that can be allosterically activated by specific nucleic-acid binding proteins.


Introduction
DNAzymes and ribozymes, naturally occurring or in vitro selected RNAs or DNAs 1 that catalyze specic chemical reactions, 2-11 couple the advantages of enzymes (e.g., high turnover and specicity) with those of nucleic acids (e.g., low cost, ready designability) and thus represent a promising set of tools for use in biosensing, 1,12-17 synthetic biology 18 and bionanotechnology. 1, [19][20][21][22] Catalytic nucleic acids have similarly emerged as a new class of gene silencing agents, 23,24 leading to the development of, for example, new cancer therapies. 25,26 An advantage of nucleic acid catalysts is the ease with which they can be rationally redesigned to introduce allosteric regulation, a mechanism that allows the ne-scale regulation of their activity upon the binding of an effector, 27 an effect that has proven of value in a range of applications, including synthetic biology [28][29][30][31][32] and biosensing. 33 Most commonly, the design of allostery has been achieved via fusion of the catalytic nucleic acid with a regulation domain for the binding of the effector. This could be a specic short oligonucleotide sequence [34][35][36] or a small molecule, [37][38][39][40] peptide, 33 or protein 41 target that binds an aptamer domain. This binding event induces a conformational change that, in turn, regulates the catalytic activity. [28][29][30][31][32][33]42 Here we expand on this theme by demonstrating a new class of allosterically regulated, catalytic nucleic acids that employ nucleic-acid-binding proteins as their effectors.
As our test bed nucleic-acid catalyst we have employed a guanine-rich, horseradish peroxidase (HRP)-mimicking DNAzyme. 41,43,44 This single-strand DNA adopts a G-quadruplex structure that, in the presence of the cofactor hemin, catalyzes Fig. 1 Transcription factor-induced activation of a DNAzyme. Here we demonstrate a DNAzyme allosterically activated by specific transcription factors (TF-regulated DNAzyme). To do so we coupled two functional domains: (i) a catalytic DNAzyme domain (red sequence in the cartoon) and (ii) a double-stranded transcription factor (TF)binding domain (green). A sequence element complementary to the sequence of the DNAzyme stabilizes an alternative conformation (left) that "sequesters" both domains in an inactive (i.e., non-catalytic and non-TF-binding, respectively) state. This off-state is in equilibrium with a second conformation, the on-state, in which both domains are functional. TF binding shifts this equilibrium towards the on-state, activating catalysis. Here we used the HRP-like G-quadruplex DNAzyme as our model catalytic domain. In the presence of hemin and hydrogen peroxide, this domain catalyzes the oxidation of the HRP substrate 3,3 0 ,5,5 0 -tetramethylbenzidine (TMB) to give a coloured product which is detectable by absorbance (l max ¼ 650 nm). the oxidation of HRP-substrates. To re-engineer this DNAzyme to introduce transcription factor (TF)-regulated allostery we combined this catalytic domain (red domain in Fig. 1) with a consensus sequence recognized by a specic TF (green domain in Fig. 1) in such a way that the fusion populates two low-energy conformations. In the more stable of these, termed the off-state, the catalytic domain and the double-stranded, TF-binding region are "sequestered" and thus inactive (Fig. 1, le). In the less stable conformation, termed the on-state, both the domains are in their functional forms. TF binding thus pushes the equilibrium between these conformations from the former, offstate, towards the latter, on-state, 45 activating catalysis (Fig. 1, right).

Results and discussion
As our rst allosteric effector we employed microphthalmiaassociated transcription factor (MITF), a DNA-binding protein associated with melanoma and renal cell carcinoma. 46- 48 We designed four variants of MITF-regulated DNAzymes, each presenting the same consensus binding sequence 46-48 for the TF ( Fig. 2A, green). The four differ, however, in the stability of their off-states (Fig. 2B). Specically, by increasing the number of bases in the doublestranded stem used to stabilize the off-state we obtained variants with estimated 49 free energies ranging from À30.5 to À57.7 kJ mol À1 . The stability of the on-state, in contrast, is effectively identical in all four variants (predicted stability ¼ À26.9 kJ mol À1 ), and thus this approach provides a means of tuning the switching equilibrium constant, K s , between the two states. This, in turn, provides a means of controlling the range of TF concentrations over which the DNAzyme is properly regulated. 45,50 To avoid wasting the (rather expensive) TF, we rst characterized our re-designed DNAzymes using a DNA strand (the "MITF-mimicking strand") that binds to and thus stabilizes the on-state mimicking the target TF ( Fig. 2A). As expected, the range of mimicking-strand concentrations over which activation is observed depends strongly on the switching equilibrium constant (Fig. 2C). Variant #1, for example, which exhibits the least stable off-state (predicted DG ¼ À30.5 kJ mol À1 ), is partially activated even in the absence of its effector. In contrast, variant #4, for which the predicted stability of the off-state conformation is the highest (DG ¼ À57.7 kJ mol À1 ), remains nearly inactive even at the highest effector concentrations we have employed (10 mM). The two variants between these extrema, in contrast, exhibit robustly regulated activation and show the optimal increase in activity upon increasing the effector concentration (Fig. 2D).
Because it is well regulated, we selected variant #2 to further characterize the extent to which the allosterically regulated catalyst is regulated by its specic TF (Fig. 3). For this variant we observe a monotonic increase in activation with increasing TF and an EC 50 (the effector concentration at which 50% activation is observed) of 375 AE 105 nM (Fig. 3, le). The activation fold of the DNAzyme activity in the presence of saturating concentration of the target protein is 1.9 which is in good agreement with previous strategies for the allosteric activation of the same G-quadruplex peroxidase-like DNAzyme. 1, 15,27,41 In contrast, the DNAzyme remains inactive when challenged with other, non-effector proteins. Incubation with 500 nM of the transcription factor TATA binding protein, for example, does not produce any signicant cross-reactivity (Fig. 3, right).
To demonstrate the generality of our approach we designed a DNAzyme that is, instead, regulated by TATA binding protein (TBP), a TF present in virtually all eukaryotic cells. 51 To do so, we engineered three variants differing in K s ( Fig. 4A and B) that, upon switching to the on-state, exhibit the consensus binding sequence of TBP. 50,51 We found that the resultant activation proles (using again a TF-mimicking DNA strand) are consistent with the predicted energy gap between the on-and off-states. From these results we selected variant #2, which exhibits an intermediate K s and thus robust activation, for further characterization. The 2.6-fold, TBPinduced activation of this DNAzyme occurs with an EC 50 of 104 AE 12 nM (Fig. 4C) and, as was true for its MITF-activated counterpart, the DNAzyme is insensitive to other, non-targeted proteins (500 nM) (Fig. 4D).

Conclusions
In nature, the regulation of the activity of an enzyme is usually achieved through allosteric regulation, in which a binding site distal from the active site can recognize a molecule (an allosteric effector) and lead to a conformational switch that affects the enzyme's affinity for the substrate. Through this mechanism the activity of naturally occurring enzymes can be nely regulated by a variety of allosteric effectors in a concentration-dependent fashion. Allostery, also called "the second secret of life" by Perutz, 52 can thus be considered as the optimal strategy to regulate the affinities and activities of biomolecules and bioreceptors. Because of this, engineering allostery into articial systems could greatly improve the functionality of biomolecules employed in biotechnologies. In response, we have demonstrated here a general strategy for the design of nucleic acid catalysts allosterically activated by specic transcription factors.
The development of allosterically regulated DNAzymes or ribozymes through both rational design and combinatorial selection strategies has been the subject of several studies starting from the seminal and fundamental work of Breaker and coworkers regarding the design and development of allosteric ribozymes. 2,3, 8,37,[53][54][55] In the past ten years several efforts have been devoted to the development of nucleic acid catalysts that can be activated or regulated by different effectors including short nucleic acid sequences, 21,34-36 and aptamer targets. 33,[37][38][39][40][41] Allosteric nucleic acid catalysts have thus been demonstrated to be useful tools not only for biosensing purposes 56 but also as controllable therapeutic agents for gene therapy strategies [23][24][25][26] or as molecular tools for controlling gene expression. [30][31][32] Despite this, to the best of our knowledge, there is no report of a rational design of DNAzymes allosterically regulated by transcription factors (TF). We demonstrated here a general strategy to obtain TF-responsive DNAzymes by joining a TF consensus sequence with a G-quadruplex peroxidase-like DNAzyme domain. Our approach is based on the rational design of a conformational switching probe that ips between a nonbinding inactive conformation to a second binding-competent active conformation in the presence of the specic TF. Because of this, we can regulate the TF concentration range at which activation of the DNAzyme is observed. We can thus control in a ne-tuned fashion the activity of our TF-regulated DNAzyme not only using different concentrations of the specic TF, but also using variants with different stabilities of the non-binding conformation, over completely different TF concentration ranges.
Compared to similar strategies reported earlier on the allosteric activation of catalytic nucleic acids, our approach appears very versatile. The consensus site recognized by each TF almost invariably consists of a specic double stranded DNA sequence with a length ranging from 6 to 12 base pairs, rendering the rational design of the necessary conformational switch quite straightforward. We also note that our approach can have important biological and clinical implications. While there is only a limited number of molecular targets recognizable via the use of aptamers, more than 10% of the ca. 25 000 human genes Fig. 3 (Left) Our MITF-regulated DNAzyme (variant #2) is activated by its target TF in a dose-dependent fashion and is insensitive to other, non-targeted proteins (right). For the specificity test (right) we have used a concentration of target MITF and of non-specific proteins of 500 nM. In the y-axis of the right panel the difference (Dabsorbance) between the absorbance value obtained in the presence of the target and that obtained in a blank solution has been used. encode DNA-binding proteins suitable for the regulatory role presented here, the majority of which function as transcription factors thus controlling crucial biological mechanisms. 57 Despite the limited activation fold achieved with our strategy, we note that this is comparable to other examples of allosterically activated DNAzymes based on the use of the peroxidaselike G-quadruplex. 1, 15,27,41 Because other examples of DNAzymes and ribozymes activated by different targets show larger activation folds we believe that this difference might be due to the low efficiency and high background signal characteristics of peroxidase-like DNAzyme in comparison to other DNAzymes. 1, 27,[42][43][44] Finally, our strategy could be applied in the design of RNA/DNA chimeras to produce TF-regulated ribozymes and RNA-binding-protein-regulated DNAzymes that could have potential applications in several elds. Engineering allosteric catalytic nucleic acids that respond to TFs could in fact lead to novel molecular tools for gene expression regulation, gene regulatory mechanisms or novel riboregulators that can allow the development of genetic circuits in the growing elds of synthetic biology and bioengineering. [58][59][60] Experimental section In the sequences above the underlined bases represent the stem portion of the non-binding conformation (off-state), while the italic bases represent the TF recognition element of the binding conformation (on-state) (see Fig. 1).
Microphthalmia-associated transcription factor (MITF) was a gi from Prof. Colin Goding's group at the University of Oxford (UK). It is constituted by residues 198-302 of the human MITF-M (melanocyte specic), which encompasses the DNAbinding domain bHLH-LZ. 48 TATA-binding protein was obtained by expression of recombinant, His-tagged proteins in Escherichia coli, as described previously. 61 All experiments with TFs and mimicking strands were conducted with a DNAzyme concentration of 50 nM in a 25 mM HEPES buffer containing 20 mM KCl, 200 mM NaCl and 1% DMSO at pH 7.4. Absorbance at 650 nm was measured 40 min aer the addition of the DNAzyme substrates, 3,3 0 ,5,5 0 -tetramethylbenzidine (TMB) liquid substrate, supersensitive, for ELISA (ready-to-use solution with H 2 O 2 , purchased from Sigma-Aldrich, St. Louis, Missouri) using either a Bio-Rad Model 550 Microplate Reader or a Tecan Innite M1000 PRO.
Activation curves were measured in microtiter plates. In every well, a concentration of the TF-regulated DNAzyme of 50 nM was used in a total volume of 150 ml and different concentrations of the proper mimicking strand or protein were added and allowed to react for 1 h at 37 C. Aer this, the DNAzyme cofactor hemin was added at a nal concentration of 500 nM and incubated for 1 h at 37 C. Finally, 150 ml of the DNAzyme substrates TMB and H 2 O 2 were added and the absorbance was measured at 650 nm aer 40 min of incubation. The absorbance of the off-state was set relative to 0 for all normalized gures, while the absorbance obtained in the presence of a saturating target concentration was set relative to 1. The curves were tted with the following equation: where A T is the absorbance in the presence of a different concentration of target; A 0 is the background absorbance; [T] is the concentration of the TF or the mimicking strand; A b is the absorbance in the presence of a saturating concentration of target; and EC 50 is the target concentration at which the activation is half of the maximum activation.