Biophysical Chemistry, Biomolecules, and Biomaterials; Surfactants and Membranes DNA Binding Induces a Nanomechanical Switch in the RRM1 Domain of TDP-43

. Understanding the molecular mechanisms governing protein-nucleic acid interactions is fundamental to many nuclear processes. However, how nucleic acid binding affects the conformation and dynamics of the substrate protein remains poorly understood. Here we use a combination of single molecule force spectroscopy AFM and biochemical assays to show that the binding of TG-rich ssDNA triggers a mechanical switch in the RRM1 domain of TDP-43, toggling between an entropic spring devoid of mechanical stability and a shock absorber bound-form that resists unfolding forces of ∼ 40 pN. The fraction of mechanically resistant proteins correlates with an increasing length of the TG n oligonucleotide, demonstrating that protein mechanical stability is a direct reporter of nucleic acid binding. Steered Molecular Dynamics simulations on related RNA oligonucleotides reveal that the increased mechanical stability fingerprinting the holo-form is likely to stem from a unique scenario whereby the nucleic acid acts as a “mechanical staple” that protects RRM1 from mechanical unfolding. Our approach highlights nucleic acid binding as an effective strategy to control protein nanomechanics.


2
ABSTRACT. Understanding the molecular mechanisms governing protein-nucleic acid interactions is fundamental to many nuclear processes. However, how nucleic acid binding affects the conformation and dynamics of the substrate protein remains poorly understood. Here we use a combination of single molecule force spectroscopy AFM and biochemical assays to show that the binding of TG-rich ssDNA triggers a mechanical switch in the RRM1 domain of TDP-43, toggling between an entropic spring devoid of mechanical stability and a shock absorber bound-form that resists unfolding forces of ∼40 pN. The fraction of mechanically resistant proteins correlates with an increasing length of the TG n oligonucleotide, demonstrating that protein mechanical stability is a direct reporter of nucleic acid binding. Steered Molecular Dynamics simulations on related RNA oligonucleotides reveal that the increased mechanical stability fingerprinting the holo-form is likely to stem from a unique scenario whereby the nucleic acid acts as a "mechanical staple" that protects RRM1 from mechanical unfolding. Our approach highlights nucleic acid binding as an effective strategy to control protein nanomechanics.
The nanomechanical properties of individual proteins regulate a number of major biological processes, including the deformability of the extracellular matrix 1 , mechanotrans duction in focal adhesion adaptors 2 , the elasticity of cardiac titin 3 or the degradation of proteins in the proteasome [4][5] . Several molecular tactics have been convincingly used to modify the mechanical stability of proteins; beyond simple protein unfolding -which converts the mechanically resistant native state into a compliant and extended protein conformation 6 -, the introduction of key point mutations in the so-called "mechanical clamp" 7 and the presence of stiff disulfide bonds [8][9] or covalent organometallic bonds [10][11] have all been shown to have a direct effect on the mechanical stability of the natively folded conformation.
In addition to these more common strategies, ligand binding 12 has recently emerged as an effective orthogonal modulator of protein nanomechanics. For example, DHFR was shown to increase its mechanical stability upon binding nicotinamide adenine dihydrogen phosphate (NADPH), 7,8-dihydrofolate (DHF) or inhibitor methotrexate (MTX), converting a purely elastic protein devoid of mechanical stability into an efficient shock absorber able to withstand stretching forces 13 . Likewise, the enzyme staphylococcal nuclease increases its mechanical resistance upon binding its inhibitor deoxythymidine 3',5'-bisphosphate 14 . Similarly, binding of small sugars (in the case of maltose binding protein [15][16] , the hyperthermophilic adenine diphosphate (ADP)-dependent glucokinase 17 and membrane transporters 18 ) or single amino acids (such as leucine 19 ) can change both the height of the energy barriers and the distribution of 4 unfolding pathways. In addition, metal binding has also revealed as a successful strategy to regulate protein stiffness through calcium [20][21][22][23] or nickel [24][25] binding. Perhaps even more conspicuous are the mechanical consequences of small peptide -or full protein-binding. In this vein, the mechanical properties of protein G are substantially increased upon binding the IgG antibody 26 . Furthermore, SUMO1 increases its mechanical stability upon binding small peptides 27 . Similarly, the attachment of the short hydrophobic APPY polypeptide induces selective increase of the mechanical properties of the domain I of the multidomain DnaJ chaperone 28 . Other recent examples epitomise the importance of mechanically revealing key binding pockets that are otherwise hidden in the folded conformation 29 30-32 .
Collectively, these experiments revealed the large knock-on effects that protein-protein interactions have on protein nanomechanics. This growing experimental evidence makes it tempting to speculate that, given the increasingly large number of identified DNA-and RNAbinding proteins (DRBPs) 33 , nucleic acid binding, further to modifying unfolding pathways 34 , might play an analogous role in regulating the mechanical stability of proteins. Direct testing of this hypothesis has remained elusive, mostly due to the lack of an extensive pool of DNAbinding proteins for which the crystal structure in the apo-and holo-forms has been solved, and especially given the difficulty to obtain these proteins biochemically free of the nucleic acid partner. Here, we investigated how nucleic acids of well-defined sequences regulate the nanomechanical properties of the RRM1 domain of the 43 kDa TAR DNA-binding protein (TDP-43). TDP-43 plays important roles in many essential cellular functions involved in DNA transcription and RNA translation 35 , and it is hence capable of binding both RNA and DNA.
Furthermore, TDP-43 has been associated to several important neurodegenerative disorders,  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60   5 including amyotrophic lateral sclerosis (ALS) and frontotemporal lobar degeneration (FTLD) [36][37][38] . Under physiological conditions, TDP-43 is predominantly localised in the nucleus with low levels in the cytoplasm [39][40] and conducts a multiplex of functionalities, being involved in different steps of RNA processing including transcription, mRNA splicing, transport and translation, and works as a transcription factor as well 41 . Given its multifunctional role, accessing the microscopic insights underpinning the protein-nucleic acid interaction is of capital importance towards establishing a link between function and protein conformation. From the topological perspective, TDP-43 is composed of two tandem RNA recognition motifs (RRM1 and RRM2) flanked between an N-terminal domain (NTD), an NLC segment that has been reported to bind RNA 42 , and the C-terminal glycine-rich domain (GRD, Figure 1a) 43 . The crystal structure of both RRM domains has been solved in complex to different UG-and TG-rich single-stranded RNA and DNA sequences 44 , concluding that RRM1 plays a dominant role in nucleic acid binding whereas RRM2 holds a supporting function 45 . Given its ability to effectively bind DNA and RNA, and thanks to the fact that its binding properties have been characterised both structurally and biochemically, RRM1 emerges as an excellent case study to elucidate how nucleic acid binding has direct effects on protein conformation.
Remarkably, the fraction of trajectories displaying a mechanical peak increased with the number of TG repetitions, with a significant transition towards the bound fraction for TG n >5 (Figure 2c), thus recapitulating the binding titration measured in Figure 2a. Altogether, these experiments suggest that the change in mechanical stability of RRM1 is a direct read-out of DNA binding.  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58   We next set out to probe whether the recovery of mechanical stability through DNA binding is a reversible process intricately linked to mechanical refolding. To this purpose, we conducted force-quench experiments (which afford superior control of the folding dynamics) 51 in the presence of TG 15 , whereby the force was first ramped up to 240 pN at a constant rate of 40pN s -1 to trigger the unfolding of holo-RRM1 (fingerprinted by a step-wise increase of protein length of 19 ± 1 nm, occurring at F = 30 ± 8 pN, Figure S5), followed by the unfolding of the I27 domains, marked by the increase of the protein's contour length in ∼25 nm steps (Figure 3a). The force was subsequently withdrawn for t q = 15 s to trigger protein refolding before the force was ramped up again to test the folding status of the protein. Remarkably, in ∼46 % of the trajectories (n = 13), mechanically re-stretching of the protein mirrored the initial unfolding sequence whereby the mechanically resistant RRM1 was first unfolded, prior to the unfolding of the I27 domains occurring at higher forces. The recovery of mechanical stability for RRM1 suggests that, upon refolding, RRM1 is able to effectively re-bind TG 15 . Similar conclusions were  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 11 qualitatively reached in the case of TG 6 binding ( Figure S6). Two distinct scenarios could mechanistically explain the rebinding process; either DNA was not removed from the protein and kept bound after unfolding, or alternatively DNA was able to re-bind from the solution within the experimental quench time (Figure 3b). Discriminating between both possibilities would eventually require an extensive set of experiments where both the quench time and also the time the mechanically unfolded protein is exposed to the solvent environment are independently varied. Repeating the experiments over a range of protein concentrations could also help elucidate whether or not nucleic acids remain bound after RRM1 mechanical unfolding.
The main discovery of our single molecule experiments is that, under the presence of a stretching force, the apo-form extends in a feature-less manner, demonstrating that the forces required to unfold are lower than ∼ 15 pN, the intrinsic resolution of the AFM working under constant velocity conditions. Further work using e.g. optical or magnetic tweezers, with expanded low-force resultion 53 , could provide further insight into the individual unfolding pathways of the apo-form of RRM1. By contrast, the addition of TG nucleotides results in a remarkable mechanical stabilization of RRM1, giving rise to the characteristic saw-tooth pattern of unfolding. However, the distribution of unfolding trajectories corresponding to the apo-form (and also to the control experiments in the presence of CA) shows that a finite number of unfolding events (∼30 %) exhibiting mechanical resistance is always present. It is possible that after our purification, and despite quantitative DNA removal 46 , there is still a low fraction of DNA-bound proteins. Conversely, in the presence of TG 15 , we also observed a number of unfolding events (∼17 %) devoid of mechanical stability. We attribute these trajectories to the presence of a dynamic binding/unbinding kinetics of the ssDNA oligonucleotides. Furthermore, our titration experiments ( Figure S3) demonstrate that, as expected, the short TG 3 shows a lower binding affinity to RRM1 than TG 4 , and that, in general, the binding affinity increases with the  1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 17 length of the TG oligonucleotide. Notably, the different extent of binding of the TG oligonucleotides of different lengths nicely correlates with the fraction of trajectories showing mechanical stabilization (Figure 2c). Hence, an important conclusion stemming from our experiments is that mechanical stability appears to be a direct reporter of DNA binding.
To obtain insight into the molecular origin of the mechanical stabilization process, we conducted SMD simulations under force clamp conditions. Our simulations revealed a plausible mechanism whereby the oligonucleotide binds directly onto the mechanical clamp of RRM1 (ß1 and ß5) and acts as a 'mechanical lid' that locks the protein, preventing unfolding. This protective mechanism largely differs for example from the 'allosteric' binding mechanism described for protein G, whereby the IgG ligand was found to bind on a position away from the mechanical clamp 26 . Thermal fluctuations coupled with the effects of the pulling force are able to perturb the location of the nucleic acid, eventually slightly displacing it from the protein binding site and thus allowing the detachment of ß5 from ß1, which triggers the subsequent unfolding of the protein. Hence, given that mechanical stabilisation is mostly dictated by the residence time of the oligonucleotide bound to RRM1, it is tempting to speculate that in the presence of oligonucleotides able to bind RRM1 with high affinity (i.e. long TG repeats), the protein would take longer to unfold, thus giving rise to an apparent higher mechanical stability under force extension conditions. Indeed, close inspection to the histograms reporting the experimental unfolding forces for the different TG constructs ( Figure S4 and Table S1) shows an overall significant (p < 0.05) increase of mechanical stability for those constructs with longer TG repeats, shifting from ∼25 pN for TG 3 up to ∼40 pN in the case of TG 15 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60 18 seen whether RNA binding gives rise to a similar mechanical stabilization to that obtained by DNA-related sequences. A further interesting feature observed in our long simulations is that, after being displaced from the binding pocket enabling protein unfolding, the oligonucleotide is kept bound to the mechanically stretched protein (Supplementary video 1). It is hence possible that the relatively fast rebinding that we observed in our force-clamp experiments (Figure 3a) occurs because the oligonucleotide remains bound after mechanical unfolding, hence rendering the rebinding process more efficient.