Interaction of single- and double-stranded DNA with multilayer MXene by fluorescence spectroscopy and molecular dynamics simulations† †Electronic supplementary information (ESI) available. See DOI: 10.1039/c9sc03049b

MXenes show differential affinity towards single- and double-stranded DNA, with unique kinetics and potential for fluorescent biosensing.


Introduction
The interaction of nucleic acids with micro/nanomaterials has been an extensively studied topic with relevant implications in different elds. In materials science, DNA has been utilized as "biological glue" for achieving programmable and precise assembly at the nanoscale. The resulting DNA-linked structures can generate highly ordered nanoparticle structures with possibility of modulation of their optical, magnetic, and electronic properties. [1][2][3] In molecular biology and bioanalytical chemistry, the adsorptive properties of some materials towards nucleic acids have been exploited for the extraction of large and small DNA fragments from biological uids in sample cleanup. 4,5 DNA physisorption onto 2D materials, for example, has been broadly applied in biosensing and nanopore-based sequencing. [6][7][8][9] Material-based biosensing systems have become very popular over the past decade. [10][11][12] They rely on uorescence resonance energy transfer (FRET), taking advantage of the differential adsorption affinity towards single-and double-stranded DNA (ssDNA and dsDNA, respectively). The detection principle consists of uorescence quenching upon ssDNA adsorption and uorescence recovery aer duplex formation. This enables homogeneous hybridization-based detection assays, which are easy and simple to operate while retaining nanomolar limits of detection and high sensitivity. 13 Although biosensing might just be the most popular application of DNA-material interactions, the role of such basic research in biomedical and bionanotechnology-related applications is also extensive. A few of them include targeted drug and gene delivery, imaging theranostics, phototherapy, in vivo biosensing, and tissue engineering. [14][15][16][17][18][19] The appeal of 2D layered materials is based on their high aspect ratio and unique structural and electronic properties, offering a great degree of tuneability. Their characteristic interaction with nucleic acids has driven researchers to continue looking into new layered materials for capturing, analyzing and delivering of these biomolecules. 20,21 MXenes represent a new family of 2D materials consisting of transition metal carbides and carbonitrides rst introduced in 2011 by Naguib et al. 22 The authors presented this new family as complex, layered structures that offer a wide range of properties owed to their multielement content and tunable composition. As shown in Fig. 1A, they are prepared from the respective MAX phase by selectively etching an A-group element (e.g. aluminium) generating MXenes with terminal groups (-OH, -F) that render hydrophilic surfaces. The resulting etched structures exhibit accordion-like shapes with interlayer spaces that can serve as molecular sieving channels and for hosting ions and organic molecules. [23][24][25] Further delamination can be achieved with different methods, resulting in ultrathin 2D sheets. 26 The applications include but are not limited to: environmental remediation, photocatalysis, electromagnetic shielding, sensing and energy storage. [27][28][29][30][31] Recently, MXenes started to be considered as promising candidates for bio-applications mainly due to their hydrophilicity and biocompatibility, plus their strong absorbance in the near-infrared region and adsorptive properties. [32][33][34][35] Very recently, a nanopore DNA sequencing system has been reported with MXene membranes. 36 Aptamer-based assays have also been recently developed for exosome detection, 37,38 as well as MXene composites with DNA for dopamine detection. 39 Herein, we probe the interaction between single-and double-stranded DNA with Ti 3 C 2 T x , the most studied and widely used MXene, by means of uorescence spectroscopy and molecular dynamics (MD) simulations. We aim at investigating such interaction at a basic level to explore the capabilities of this kind of materials not only as prospective biosensing platforms for sequence-specic DNA detection, but also as potential carriers of nucleic acids serving as structural support and biomolecular reservoirs for biomedical applications.

Results and discussion
Based on the already well-established knowledge of DNAmaterial interaction, we projected that the MXene-DNA interaction (Scheme 1) would be based on different affinities towards ssDNA and dsDNA, the former having higher affinity towards the material than the latter. Noncovalent binding of ssDNA to the surface of nanomaterials is generally based on weak interactions such as van der Waals forces, hydrogen bonds and p stacking, involving the phosphate backbone and/or the nucleobases, respectively. p-p interactions are primarily associated to sp 2 -hybridized systems like graphene, whereas the rst two types are more ubiquitous within a wide range of materials. Adsorption of dsDNA to these surfaces is usually much weaker due to the higher rigidity of the double-helix compared to the single-stranded form. If the adsorption is taking place mainly via p-p stacking, then it is expected that dsDNA is far less likely to interact once the bases are not free to interact with the material surface. Fig. 1 shows the bottom-up synthesis of partially delaminated MXene (Ti 3 C 2 T x ) from the MAX phase precursor (Ti 3 AlC 2 ), together with the discreet top-down characterizations to exhibit the structure of MXene from micro-scale to molecular level (see Experimental section in ESI †). The MAX precursor contains Scheme 1 Simplified representation of (A) MXene; (B and C) projected interaction between MXene with ssDNA and dsDNA, respectively. layers of transition titanium carbides (Ti 3 C 2 ), which are interleaved with layers of Al-element atoms (Fig. 1A). In the next step, HF is used to selectively etch the Al-element from MAX phase to achieve MXene phase. The obtained MXene phase has a mixture of -OH, -O and -F terminations, with the chemical formula Ti 3 C 2 T x , where T represents the surface terminations. 24,27 The further delamination step via ultrasonication in aqueous medium results in an intercalated structure, wherein the water molecules can t in and expand the interlayer spacing and consequently increase the nal specic surface area. 27,40,41 In order to assess the nano/microstructure of MAX precursor and intercalated MXene, their SEM images coupled with EDS elemental map are shown in Fig. 1B and C, respectively. The EDS results conrmed that the Al element was almost removed from the precursor through the etching process, and the new atomic terminations i.e. -OH, -O and -F were introduced to the MXene phase. The high-resolution SEM images show the morphological changes aer etching ( Fig. 1D and E). The intercalation of the multilayer is evident aer the etching process. The small species that appeared on the edges of MXene layers are most likely TiO 2 nanocrystals, for it is well-known by now that TiO 2 nanoparticles form on MXene surfaces due to the fast oxidation of titanium carbide in aqueous and/or oxygen conditions and the process starts on the edges. 27,[42][43][44][45] More SEM images of the multilayer akes together with larger accordionlike structures can be found in Fig. S1 in ESI. † The dispersion was let to sediment and the ne supernatant particles consisting mainly of multilayer akes were retrieved for further characterization and DNA tests. The intercalated MXene was further examined by bulk (XRD) and surface (XPS) characterizations. Fig. 1F shows the X-ray diffractogram of MAX phase together with MXene phases aer etching (Ti 3 C 2 T x powder) and ultrasonication (Ti 3 C 2 T x intercalated) in the small angle (2q) regions, i.e. 8-10 , to explore the interlayer spacing. The crystal interlayer distances (d) were calculated from the Bragg equation (l ¼ 2d sin q), where l and 2q are the wavelength of Cu-Ka source (1.54178Å) and (002) peak-position as related to c lattice parameter of hexagonal close-packed crystal structure. 27,40,47 Aer etching, the (002) peak shied to smaller angles and broadened, which respectively indicates the increase in interlayer spacing and reduction in crystal size according to Bragg and Scherrer equations. 47 The ultrasonication of multilayer MXene resulted in a further shi to smaller XRD angles. It is therefore shown that the etching and intercalation with water molecules increase the interlayer spacing to 10% (i.e. from 0.9 to 1 nm). Full-range X-ray diffractograms of MAX phase and intercalated MXene are shown in Fig. S4. † The majority of the non-basal plane peaks of Ti 3 AlC 2 disappeared aer HF etching/ultrasonication in water. 48 The presence of different TiO 2 phases, i.e. anatase and rutile, in T 3 C 2 T x was discreetly evidenced in the 25-28 angle (2q) region. 49,50 Fig. 1G shows the high-resolution spectrum of Ti 2p photoelectron region for the intercalated MXene. The chemical composition consists of oxygen-rich moieties (56%), titanium carbide (29%) and other terminal functionalities such as -F groups (15%). The peak separation of 5.6 eV for the Ti 2p 3/2 and Ti 2p 1/2 suggests that there is a contribution from TiO 2 together with Ti-O terminal groups. 51 The presence of -OH terminations is furthermore conrmed by high-resolution XPS spectra for O 1s region (Fig. S5 †), which is essential for the adsorption of polar species. 52,53 It was reported that the water molecules cannot be removed by further drying and ordinary degassing steps, 27,40 where its presence is observed in our case (Fig. S5 †). Raman analysis showed the vibrational modes for Ti 3 AlC 2 and Ti 3 C 2 T x MXene obtained aer etching (Fig. 1H). The former exhibited vibrational modes at 183, 203 and 273 cm À1 , which have been assigned to Al atoms. Other modes located between 600 and 700 cm À1 have been associated to C atoms and can be seen in both spectra. 54 As expected, Al-related modes disappear aer etching. The modes for Ti 3 C 2 T x MXene are located at 128, 219, 385, 618 and 711 cm À1 , comparable to previous reports. [55][56][57] The heterogeneity of the surface terminal groups affect the overall spectrum as a result of collaborative vibrations from surface and central Ti atoms, central C atoms, and the terminal groups -O, -OH and -F. The band at 128 cm À1 (E g ) is indicative of -F terminal groups from in-plane vibrations of surface Ti and C atoms. Out-of-plane stretching vibrations of surface Ti and C atoms give rise to the 219 cm À1 band as a result of -OH terminal groups. The 380 cm À1 hump can be associated with heterogeneously distributed -O and -OH terminations, while both -F and -OH groups contribute to the in-plane vibration of the C atoms at 618 cm À1 . 55 The presence of TiO 2 was not evident with Raman spectroscopy, suggesting that this chemical species is not predominant in the system.
A picture of the stable MXene dispersion is shown in Fig. S6 † with its corresponding absorbance spectrum showing characteristic absorption at ca. 760-800 nm. 43 The stability of the nal dispersion was also corroborated with zeta potential measurements at pH $ 7 (z ¼ À29.7 AE 7.4 mV). In order to gain more information about the surface area and the pore structure, the BET non-local density functional theory (NLDFT) pore size distributions at 77 K are shown for both MAX precursor and MXene phase in Fig. S7A and B, † respectively. Aer the etching process, the volume of mesopores was sharply increased, which was already illustrated by SEM local measurement (Fig. 1D and E). The relatively large amount of mesopores show the potential for entrapping small fragments of nucleic acids.
DNA-MXene interaction was assessed with uorescence spectroscopy by rstly incubating the MXene material with FAM-ssDNA (also denoted as ssDNA). The sequences used in this work, which correspond to an apolipoprotein-E-encoding DNA fragment, 13 are illustrated in Fig. S8. † FRET is envisioned to take place due to the proximity of the uorophore, 6-carboxyuorescein (FAM), covalently bound to one end of ssDNA, to the surface of the material, leading to uorescence quenching. Fig. S9 † shows the spectral overlap between the broad absorption spectrum of Ti 3 C 2 T x and the absorption and emission peaks of the FAM dye. Fig. 3A shows that for ssDNA-MXene, the uorescence of FAM decreased by ca. 48% and, in the case of dsDNA, negligible difference can be seen compared to the spectrum of FAM alone, agreeing with the foreseen interaction illustrated in Scheme 1. Further evidence was also provided by MD simulations. In the case of ssDNA, the DNA strand stayed in close contact with MXene surface during 200 ns long MD simulations, as indicated by a broad density maximum of DNA 5.8-9.6Å (estimated as an interquartile range; median 7.7Å) from the surface when simulated without the uorophore, and 7.2-12.6Å (median 9.9Å) with the FAM-labeled ssDNA ( Fig. 2A and B and S12 †). On the other hand, dsDNA and FAM-dsDNA were more distant from the surface with a median of 12.3Å (interquartile range 8.5-16.1Å) and 13.2Å (interquartile range 9.2-17.4Å), respectively, which could be attributed to a lower interaction of dsDNA to the surface compared to ssDNA. In the case of ssDNA labeled by FAM, the uorophore interacted mainly with the surface, whereas in the case of dsDNA the FAM not only interacted with the surface but also with the end of the double helix by stacking. These observations can explain the quenched uorescence of ssDNA which has also been evidenced in recent works, 37,58 and can be assigned to resonant electron transfer given the spectral overlap and the low-range distance. But given the low distance (#10Å), the question of whether Dexter energy transfer has a role also arises. Other contributions can be playing a role too such as metal damping, providing additional non-radiative decay of FAM's excited state. 59 Elucidating the specic mechanisms responsible for the quenching phenomenon observed here is beyond the scope of this work. Inner lter effects can also have implications in the attenuation of uorescence due to the absorption of light at both excitation and emission wavelengths by the MXene. We however kept the concentration of these absorbing species constant throughout most of the experiments, thus the correction of such effects would not impact the observed trends.
We then proceeded to test different amounts of complementary DNA (cDNA) for the dsDNA + MXene incubation, to assess whether a correlation between uorescence intensity and cDNA amount was feasible. Fig. 3B shows the nonlinear response of the system. Repeated measurements systematically showed statistically signicant (p < 0.05) uorescence changes when cDNA was equal or above 5 pmol, i.e. when the ratio of ssDNA to cDNA was at least 1 to 5. Thus, unequivocal detection of 5 pmol of a complementary DNA sequence could be attained with this system. Interestingly, one mismatch gave ca. 50% less response than the fully complementary one, suggesting that this platform has potential for sequence-specic discrimination. Even though this needs to be evaluated further with a variety of mismatched sequences, it is relevant to mention that single-base discrimination with other layered/2D materials such as graphene oxide is generally very low, e.g. ca. 70-80% of the response obtained with fully complementary strand. 13,60 As a conrmatory tool, we performed uorescence anisotropy measurements, as these are commonly used to probe biomolecular interactions and affinities. 61 By placing the uorescent signal on the smaller FAM molecule binding to the much larger material akes, substantial changes in anisotropy can be monitored as this binding will signicantly decrease the rotational diffusion of the uorophore. The uorescence anisotropy of free FAM-ssDNA and the ssDNA-MXene complex was appreciably different: 0.06 and 0.12, respectively (MXene concentration of 50 mg mL À1 ), suggesting that an interaction is  taking place. 62 Fig. 3C shows the increase of anisotropy as a function of MXene concentration. Fig. S10 † shows the effect of different MXene concentrations on uorescence intensity and calibration. The uorescence changes were relevant from 50 mg mL À1 on, plateauing aer 100 mg mL À1 (Fig. S10A †), whereas for calibration we found 50 mg mL À1 to be the optimum value with the highest sensitivity (Fig. S10B †). Higher MXene concentration in the media showed a marked decrease in the recovery of uorescence by increasing cDNA amounts, which can be associated with the high absorbance of the material at these wavelengths. The concentration of MXene was kept constant at 50 mg mL À1 throughout subsequent experiments).
In order to gain more information about the system, we carried out a kinetic assessment of DNA-MXene binding (Fig. 4). This kind of experiments has been typically undertaken in previous reports as follows: (1) ssDNA-uorophore is incubated with the material and uorescence changes are monitored either in time-resolved measurements for the calculation of quenching efficiencies/mechanisms or in larger time scales to withdraw information on binding kinetics; (2) the ssDNAmaterial complex is isolated and puried and then exposed to cDNA in order to record the increase in uorescence as a function of time, due to desorption of ssDNA off the surface to hybridize with cDNA. The second phase of this experimental assessment requires several centrifugation/washing cycles and oen involves the use of centrifugal ltration devices with cutoff molecular weight specically selected to entrap DNA-material complexes and to get rid of unbound DNA strands. The MXene prepared in this work underwent re-stacking aer subsequent centrifugation/washing cycles. The resulting cake could not be redispersed in the ionic strength working conditions, making it difficult to separate ssDNA-MXene from unbound/free ssDNA. This led us to assess the binding kinetics by a distinct approach, i.e. on one hand, ssDNA + MXene incubation was carried out and, on the other hand, a threecomponent system in one-pot reaction was used to assess desorption kinetics: ssDNA + cDNA + MXene. Fig. 4A shows the kinetic proling of free FAM-ssDNA and FAM-ssDNA incubated with MXene over the course of 30 min. The measurement was done in real-time mode, i.e. uorescence emission was recorded every second while the sample was constantly irradiated at 490 nm. This approach needs to consider the unavoidable photobleaching of the uorophore, thus the loss of uorescence, registered in terms of uorescence changes ((F 0 À F)/F 0 ), was ca. 17% for the FAM system alone, and ca. 63% in the presence of the material. Fig. 4B shows that in the three-component system, the desorption of FAM-ssDNA takes place aer 20 min of reaction time, resulting in a uorescence recovery of ca. 6%. Fig. 4B and C show the kinetic proling experiments carried out in a longer time scale (up to 2 h) in a single-point fashion so that the uorescence changes remained minimally affected by photobleaching. As a result of minimizing such contribution, the uorescence decrease and subsequent recovery were lower and higher, respectively. The real-time adsorption/desorption prole ( Fig. 4B and D) can be conceptually explained by the notion that the reaction between nucleic acids and MXene is kinetically favorable, however DNA-DNA hybridization subsequently governs the system as a thermodynamically-controlled reaction. The latter induces the partial desorption of both DNA sequences off the surface of the material, leading to the recovery of uorescence.
Scheme 2 illustrates the plausible kinetic processes taking place in the two-and three-component systems. The processes are: photobleaching of the FAM dye, governed by k 1 ; FAM-ssDNA adsorption onto MXene (k 2 ); DNA hybridization (k 3 ); adsorption of cDNA onto MXene (k 4 ); desorption of FAM-ssDNA off the MXene surface, induced by hybridization with cDNA (k 5 ); desorption of cDNA off the MXene surface, induced by hybridization with FAM-ssDNA (k 6 ).
This unique behavior of uorescence decrease followed by uorescence recovery in one-pot reaction has not been reported with other materials, to the best of our knowledge. Reference and comparable materials, e.g. graphene oxide and TiO 2 , have been tested by our group. Graphene oxide adsorbs DNA mainly via the nucleobases by hydrogen bonding and p-p stacking interactions, 63,64 while TiO 2 nanoparticles are known to bind with their respective fluorescence intensity changes ((F 0 À F)/F 0 , where F 0 is the intensity recorded at 0 s and F is the intensity recorded at 1800 s or 120 min). The fluorescence decays were fitted with double exponential functions. (B and D) FAM-ssDNA (50 nM) + cDNA (50 nM) + MXene displaying fluorescence intensity changes from 0-600 s for adsorption and 1200-1800 s or 10-120 min for desorption ((F À F 0 )/F, where F is the intensity recorded at 1800 s or 120 min and F 0 is the intensity recorded at 1200 s or 10 min). Inlet of (B) shows the region of fluorescence increase (1200-1800 s) fitted with a sigmoidal (Boltzmann) fit. The increase in fluorescence in plot (D) (10-120 min) was also fitted with a sigmoidal (Boltzmann) function. PL intensity was normalized in (A and B) to FAM-ssDNA's at time zero.
DNA mainly via the phosphate backbone at pH 7.4, and also interacting with the bases to a lesser extent. 65 The kinetic proling does not exhibit in either case a real-time adsorption/ desorption behavior under the same reaction conditions used for MXene. Given the TiO 2 -decorated features on the MXene surface, we show the adsorption/desorption prole of DNA onto TiO 2 nanoparticles in Fig. S11, † where there is no recovery of uorescence aer 30 min. In the case of MXene the MD simulation indicated almost no stacking interaction of DNA with the surface and, rarely, hydrogen bond formation between the DNA and MXene molecules. Additionally, ions bound at the surface formed several ion bridges with the DNA molecule (Fig. S13 †). as-prepared. The MXene dispersion displayed negative zeta potential (ca. À30 mV), which further suggests that the interaction with DNA most likely takes place via ion bridges. 58 MXenes constitute a complex scenario brought by their multielement composition, richness in surface ligands and TiO 2 nano-sized species, making it challenging to intuitively elucidate the binding mechanism with nucleic acids. The entrapment/release behavior particularly suggests that the interaction is weaker than in the other systems tested, 58 i.e. graphene oxide and TiO 2 , and thus the displacement occurs spontaneously aer a short time of exposure to the complementary sequence. Why such spontaneous entrapment/ displacement behavior occurs with MXene in this time frame and not with a chemically-related system such as TiO 2 remains an open question. Further studies will be carried out to investigate the variables affecting such phenomenon, as well as the implications in polymorphism differentiation.

Conclusions
We have investigated the interaction between DNA and MXenes via uorescence spectroscopy and molecular dynamics simulations. The multifaceted features of MXenes make up a complex system that is not only capable of catching nucleic acids via ion bridges, but due to the proximity (#10Å) of such interaction, the uorescence of dye-labeled DNA can be quenched, offering a potential biosensing platform with a relevant degree of mismatch discrimination. The weak nature of the interaction allows for a kinetically-dynamic system with interesting adsorption/desorption features, making MXenes unique among other layered/2D materials. These early ndings reveal the versatility and promising properties of this family of materials in the biomedical eld, specically their potential use in controlled delivery of nucleic acids and advanced biosensing systems.

Conflicts of interest
There are no conicts of interest to declare.