Graphene setting the stage: tracking DNA hybridization with nanoscale resolution

In this study we use nanophotonic effects of graphene to study DNA hybridization: the z−4 nanoscale distance-dependence of the fluorescence lifetime for fluorophores located in the vicinity of graphene is for the first time used to track a DNA hybridization reaction with nanoscale resolution in real time. First, a nanostaircase with ≈2 nm steps from 0 to a total height of 48 nm is used as a nanoruler to confirm the distance dependence law. We find that the axial sensitivity is suited to determine the nanoscale surface roughness of these samples. The proof-of-concept DNA experiments in aqueous medium involve the hybridization of fluorescently labelled DNA beacons attached to CVD grown graphene with complementary (target) DNA added in solution. We track the conformational changes of the beacons statistically by determining the fluorescence lifetimes of the labelling dye and converting them into nanoscale distances from the graphene. In this way, we are able to monitor the vertical displacement of the label during DNA-beacon unfolding with an axial resolution reaching down to 1 nm. The measured distance increase during the DNA hybridization reaction of about 10 nm matches the length of the target DNA strand. Furthermore, the width of the fluorescence lifetime distributions could be used to estimate the molecular tilt angle of the hybridized ds-DNA configuration. The achieved nanoscale sensitivity opens innovation opportunities in material engineering, genetics, biochemistry and medicine.


Introduction
In the last two decades, several optical microscopy techniques were developed that allowed significant improvement in the lateral resolution of acquired images. In 2014 Hell, Betzig and Moerner were awarded the Nobel Prize in Chemistry for their pioneering work to overcome the optical diffraction limit, creating the field of 'Super-resolution microscopy'. Prominent techniques include stimulated emission depletion (STED), which reaches down to about 20 nm in lateral resolution, and single molecule localization microscopy (SMLM) techniques, such as photoactivated localization microscopy (PALM), stochastic optical reconstruction microscopy (STORM) and related techniques, which more recently reached similar resolutions [1]. However, the resolution in the axial (vertical) direction has not been improved accordingly. Typically, the axial resolution only reaches down to 170 nm in STED [2] and with the use of additional optics (typically decreasing the signal detection yield) e.g. via the insertion of a cylindrical lens in SMLM, axial resolutions down to 30 nm were achieved [3].
The ability to achieve high resolution in axial direction on the range of few nanometers, reaching the extension of biological molecules such as proteins or DNA strands is not yet well explored. Such achievement will be of utmost importance for the study of nanometric features present in biological systems, such as cell membrane topography, organelles, protein complexes and other nanosized biomolecules, such as DNA strands. A recent review article summarizes the accomplishments involving axial super-resolution techniques based on near-field interactions between fluorophores and surfaces and highlights their versatility for a variety of future applications [4].
Substrates used to establish near-field distancedependent contrast so far include metals and more recently, graphene [5]. We propose to use graphene, as a very promising carbon-based biocompatible mat erial, for the development of a biosensor with nanoscale resolution.
Biomolecular sensing is crucial for areas like Biomedicine, Food safety, Environmental monitoring and other areas where the detection of traces of biomolecules or DNA can be used for early-stage diagnosis, food quality and authenticity control, or contaminant detection [6]. Biosensing of hybridization chain reactions are key tools in drug development and engineering of complex DNA architectures, such as DNA origami [7] with applications in nanodrug delivery. Previously reported graphene-based biosensors include electrochemical sensors [8,9], field-effect transistors (FETs) [10][11][12], as well as fluorescence quenching sensors [13][14][15]. Most of these sensors rely on the use of graphene oxide colloidal suspensions [16] rather than large-scale graphene sheets, and are characterized by a binary 'onoff' operation.
To overcome this limitation and go beyond 'on' and 'off' state sensitivity we explore for the first time the use of graphene induced near-field effects to distinguish DNA conformations at the nanometer scale, using a functionalized graphene surface grown by chemical vapor deposition (CVD)-. The honeycomb lattice of graphene allows the use of linker molecules with aromatic headgroups that stably anchor to the graphene surface via π-π stacking.
Near-field effects that change the optical properties of emitters in the vicinity of functionalized surfaces have been reported for metallic thin films [17][18][19], transparent conductive oxides [20], and graphene [5], where the resonant electromagnetic coupling between the electric dipole of an emitter (energy donor) and the occupied electronic states on such surfaces (energy acceptor) enables nanoscale distance-dependent nonradiative resonance energy transfer (RET) from the former to the latter [21,22]. Shorter distances between acceptor and donor lead to increased RET rates, reduced fluorescence intensity, as well as shortened fluorescence lifetimes.
The nanoscale distance-dependent fluorescence quenching by graphene was first experimentally quanti fied using nanometer-thin dielectric spacing layers that separated rhodamine dyes from graphene [5]. The obtained relation matches the Chance, Prock and Silbey (CPS) model [23], which formulates the scaling effect of fluorescence lifetime with a z −4 depend ence at near-field axial distances from graphene [24,25]. So far, the effect has been exper imentally tested for selected emitters, including organic dyes [26], quant um dots [27], and semiconductor nanostructures [28], but sensing applications on relevant systems with impact beyond academic interest are lacking. Applications of other near-field effects have been theoretically predicted to produce plasmonic signal enhancement [29] and were applied to a NEMS device [30].
In this work we take the first step in using this effect for the study of biomolecular dynamics. In figure 1 we show the design of the genosensing platform which involves large CVD-grown graphene sheets, functionalized with fluorescently labeled hairpin DNA beacons (hp-DNA). In its initial state, the single-stranded DNA beacon is expected to assume a folded configuration in which the fluorescent labelling dye is placed very close to the graphene-at a distance approximately equal to Figure 1. Concept of the proposed Atto488@hp-DNA-Graphene genosensing platform. Initially, before the hybridization with target DNA the beacons are in the folded configuration. Upon the DNA hybridization, the labeling dye is displaced away from the graphene, and its fluorescence is restored in dependence of the nanoscale distance to graphene. the length of the linker molecule used to bind the beacon to the graphene.
The DNA hybridization-probing strategy consists of monitoring at short intervals the fluorescence lifetime of the labeling dye after the addition of the complementary (target) DNA. During the DNA hybridization, the DNA beacons are expected to unfold into a double-stranded extended (vertical) configuration [31], thereby displacing the dye away from the graphene. The measured fluorescence lifetime values are then used to analyze the degree of near-field coupling, which scales with the fourth power of the inverse of the distance from the graphene surface (z −4 dependence). The distance travelled by the dye should approximately correspond to the molecular length of the DNA strand used. In this work, we perform the proof-of-concept experiments with a 29 nucleotides long DNA strand from a local Port wine grape.
While previous works have demonstrated the graphene-induced nanoscale resolution on nanofabricated static samples we propose for the first time to use the near-field quenching effect in the statistical study of the dynamics of a biomolecular system in aqueous environ ment. This proof-of-concept system is of relevance not only for the study of DNA hybridization kinetics, but indeed for the dynamic observation of conformational change in a vast variety of biomolecules. Such sensing platforms may find applications in various areas of societal relevance-health, security, environment and food authenticity and safety.

Graphene fabrication
The graphene used in this work was grown by CVD (EasyTube 3000, FirstNano) on a copper sheet and transferred onto #1.5 microscope glass coverslips. The transfer was performed by removing the graphene from one of the sides of the copper sheet by plasma etching, spin-coating of a poly(methyl methacrylate) (PMMA) support layer on the graphene (16 000 RPM; soft-baking temperature 80 °C), dissolution of the Cu in 0.48 M FeCl 3 at 35 °C for 30 min, followed by rinsing in 3 cycles of chemical baths (per cycle: 5 min in water; 30 min in 2% HCl, v/v), manual transfer onto microscope glass coverslip and dissolution of PMMA layer (sample dipped in acetone overnight, followed by a 2 h, 300 °C bake). Further details of the graphene growth and transfer can be found in [9] by Campos et al. Dark-field microscopy and Raman confocal microscopy were used to characterize the quality of the sample and results can be found in figure S1 (stacks. iop.org/TDM/6/045056/mmedia) of the supporting information (SI).

Design and fabrication of a nanostructured verification sample
SiO 2 spacer layers were deposited using Plasma Enhanced-CVD (PECVD) (MPX CVD, SPTS) in a stair-like configuration on top of graphene as shown in figure 2(a). Then, 8 successive PECVD depositions were performed to create the SiO 2 steps by using a manually displaced shadow-mask between depositions. Prior to the depositions, the deposition rate of the PECVD was calibrated on auxiliary Si wafer samples using optical reflectometry measurements.
Fluorescence lifetime imaging microscopy (FLIM) images were measured on the Atto 488-functionalized sample. The arrangement of the SiO 2 spacers in steps allows an analysis of the distance-dependent RET process on graphene. FLIM images were taken with 25 × 25 pixels over areas of 50 µm × 50 µm.

Graphene bio-functionalization with Atto488@ hp-DNA
The binding of hp-DNA molecular beacons to the graphene surface was performed using 1-pyrenebutyric acid N-hydroxysuccinimide ester (PBSE) as linker. The pyrene group of the PBSE binds to graphene non-covalently, but irreversibly, and the reactive succinimide ester conjugates with the aminolabeled oligonucleotides of the DNA molecule [32,33]. When the succidimine ester reacts with the aminolabeled oligonucleotides of the DNA molecule, the N-hydroxysuccinimide (NHS) ester group is dropped as a subproduct of the reaction and the 5′-Amino-Modifier C6 binds in its place.
After the transfer of CVD-grown graphene on 18 × 18 mm coverslips, graphene was functionalized with 10 µl of 10 mM PBSE in dimethylformamide (DMF). After 2 h the surface was rinsed with DMF, H 2 O and dried with N 2 stream. Following this, 10 µl of 10 µM probe DNA in TSC1 buffer solution were dropcasted at the same location of the previous drops and left to react overnight via NHS reaction with PBSE, at 4 °C. The DNA-modified surface was rinsed with TSC2 buffer solution and passivated with 10 µl of 100 nM of ethanolamine (C 2 H 7 NO), which were allowed to react for 30 min with the PBSE molecules that had not reacted with the amino group of the probe DNA. For the target DNA solution, a concentration of 10 µM in TSC2 was used to ensure that an excess was used when compared to the probe DNA on the surface.

FLIM setup and experimental parameters
For all FLIM measurements in this study, we use a custom-built FLIM microscope based on timecorrelated single photon counting (TCSPC), with a 467 nm picosecond laser (PLP-10, Hamamatsu) at variable repetition rate-typically operated at 20 MHz, coupled to an inverted microscope setup (RM21, MadCityLabs), equipped with an high NA oil immersion objective (CFI plan APO 100×, NA: 1.45, 0.17 WD 0.13, Nikon) with a combined microand piezo closed loop controlled nanoscanner (NanoLPS200, MadCityLabs), and an avalanche photodiode (APD) (PD50CTD, MPD), which is placed in quasi-confocal configuration with an active surface of 50 µm in diameter.
The TCSPC is implemented using a single-photon counting card (SPC130, Becker&Hickl) to which the sync signal of the pulsed laser controller and the APD detector output are fed. The laser passes a spatial filter, composed of two lenses and a 100 µm pinhole and a 470.0 nm ± 25 nm clean-up filter (470FS10-25, Thorlabs) to block potential background luminescence of the laser source.
A beam expander is integrated into a cage system at the entry to the inverted microscope. A 480LP Dichroic (F38-482, AHF Analysentechnik) is used to guide the excitation signal to the sample and the fluorescence emission signal towards the detector. Additionally, to block laser excitation from reaching the detector a 472 nm long-pass filter (F76-472, AHF) is placed in the detection path before a lens focusing on the detector. A stage top incubator (UNO-T-H-PREMIXED, Okolab) is used to keep the temperature at 21 °C and increase the stability of the sample inside the wet chamber.

FLIM image analysis
A 'fast fitting of multi-exponential decay curves' algorithm developed by Enderlein et al [36] was integrated in a custom-made data analysis platform, with a graphical user interface in MATLAB for userfriendly data analysis of FLIM data. The curvefitting routine deconvolves the experimentallymeasured instrument response function (IRF) from the TCSPC spectrum measured at each scanning position, performs a multi-exponential curve fitting and returns an array of exponential coefficients and weighting factors, which we weight-average and colormap to form a FLIM image.

Experimental verification of nanosensitivity
The integrated TCSPC curves obtained from the scans at each spacer layer are plotted in figure 2(b) (the respective FLIM images can be found in figure S3 of the SI). The fluorescence lifetime distribution at each spacer layer (figure 2(c)) was obtained by calculating the histograms of the fluorescence lifetimes determined per FLIM image.
Using Gaussian curve-fittings we calculate the mean and standard deviation of each Gaussian distribution and assign them to the characteristic fluorescence lifetime and associated error of each spacer layer, respectively.
In figure 2(d), the average fluorescence lifetime is plotted as a function of the distance of the Atto488doped layer to graphene and compared with the z −4 distance-dependence RET model developed for graphene by Gaudreau et al [5]. Details regarding the model and the parameters used are presented in section S3 of the SI. A monotonous distance-dependent reduction of fluorescence lifetime is observed towards shorter distances, which is in excellent agreement with the model (as can be seen in figure 2(d) and the log-log plot of the same datapoints in figure S4, SI). Interestingly, the widths of the fluorescence lifetime distributions shown in figure 2(c) can be correlated with nanoscale distance distributions which match the surface roughness observed using AFM (see section S5 of the SI).
Henceforth in this document we use this model (see equation (7) of the SI) to convert the fluorescence lifetime of the Atto 488 fluorophores into nanoscale distance from the graphene surface.  and nanoscale distance from the graphene surface: z (c). The latter is calculated from (b) using a theoretical RET model for graphene [5]. The scale bar is 2 µm long, and the same for all images. The displacement of the Atto 488 dyes from the graphene surface, Δz, (figure 3(j)) is calculated by the difference between the nanoscale distance images of the initial and final states. Regions of displacement of the fluorophores where, presumably, the DNA hybridization reaction occurred are clearly distinguishable. We calculate the Δz distribution (figure 3(k)) and use Gaussian curve-fitting to determine the average molecular distance change.

'On-Off
The obtained distribution is dominated by a contrib ution at Δz around 10.6 ± 0.5 nm and a minor contribution around 8 ± 2 nm. The latter stems from the rims of the emergent feature, as a result of an averaging edge effect due to the finite lateral resolution of the measurements, and can thus be disregarded.
We highlight the substantial agreement firstly between the distribution characteristic of the initial state (1.7 ± 0.3 nm, figure 3(i)) and the estimated 1.7 nm length of PBSE, and secondly between the displacement of the Atto-488 dyes (Δz = 10.6 ± 0.5 nm, figure 3(k)) and the estimated 10.5 nm of molecular length of the DNA strand. The inset graphics of figure 3(i) illustrate DNA beacons in folded and stretched configurations, which can be associated with the initial and final distributions, respectively. Repetitions of the experiments can be found in the SI, section 7.

Dynamic operation of the genosensing platform
We perform time-lapse FLIM imaging to dynamically track the molecular displacement of the Atto 488 dyes during DNA hybridization, upon the addition of target DNA to the functionalized graphene surface. In this experiment, 9 measurements were performed, with time intervals of ≈13 min (see FLIM images in figure S7, SI). As before, nanoscale distance images and the respective fluorescence lifetimes and nanoscale distance distributions were obtained from the FLIM images [5]. The empirical distributions were fitted using three Gaussian distributions-as can be seen in representative plots associated to three different time points of the series ( figure 4(a)). We label the 3 Gaussian distributions as G1, G2 and G3 in correspondence with the initial, intermediate and final distances observed. The resulting curves fitted for of each point of the time-series are plotted in figure 4(b), firstly as line plots (top panel), secondly in color coded 2D representation in dependence of the reaction time t r (center panel), and thirdly as 2D representation of the three Gaussian distributions (G1, G2 and G3 color coded in blue, green and red, respectively) used in the fittings (bottom panel).
In the initial state, before the addition of target DNA, a single narrow peak is observed, t r = 0 min (figure 3(i)). However, in the first measurement after the addition of target DNA at t r = 3 min (figure 4(a)) two distributions can be distinguished, one of them originating in the narrow peak observed at the 'off' state (1.9 ± 0.2 nm) and a second broadened distribution around 5 ± 2 nm. Over time, the sharp peak at 1.9 ± 0.2 nm decreases in intensity before vanishing entirely after 92 min, while an initially non-existent peak centered around 10 nm appears and becomes dominant towards the end of the hybridization reaction (total displacement Δz of up to 9 ± 1 nm).
Meanwhile, the broadened G2 peak initially observed around 5 ± 2 nm (green contributions in figure 4(b), bottom panel) gradually shifts to a final position at 9 ± 2 nm (t r = 105 min).
The hybridization kinetics are assessed by plotting the mean and mode of the nanoscale distance distributions over time-inset plot in figure 4(b), top panel. An abrupt jump in the value of the mode (from 2 to 10 nm) together with a change of slope in the values of the mean as a function of time (from a higher to a lower slope) is observed at reaction time t r = 40 min, indicating the conclusion of the hybridization reaction.
In figure 4(c) we summarize our findings in an illustrative way based on the molecular configurations potentially associated to each Gaussian peak used for the dynamic analysis-see Discussion section.
Reproducibility of the ability to observe dynamic changes in time-lapse FLIM images recorded during DNA hybridization reaction, was confirmed (see supplementary time-lapse results in figure S10, SI).

Discussion
The results obtained from the nano-staircase verification experiment show an excellent agreement with the theoretical model formulated by Gaudreau et al [5]. We take this as evidence of the sensitivity of the system and suitability of the method to probe near-field interaction changes between the Atto 488 fluorophores attached to DNA molecular beacons and the CVD-grown graphene.
In the static 'on-off' experiments (figure 3), the strong variation from very short to longer fluorescence lifetimes upon addition of target DNA, in the central area of the sample, can be associated to initially folded Atto488@hp-DNA that hybridizes and outspreads to an upright extended configuration. During the DNA hybridization, the displacement of the dye away from the graphene is responsible for the partial restoration of the fluorescence lifetime, due to the reduction of the near-field coupling RET efficiency. This interpretation is strongly supported experimentally since the measured displacement of the Atto 488 dyes during DNA hybridization is in agreement with the theoretical predictions. This agreement is verified both for the initial state: 1.7 ± 0.3 nm versus the estimated length of PBSE (≈1.7 nm); and for the final state: Δz of about 10.6 ± 0.5 nm versus the estimated length of the selected DNA strand (≈10.5 nm).
We highlight the extremely small uncertainty values obtained in the converted nanoscale distance values. While we cannot rigorously guarantee subnanometer resolution across the full distance range of this near-field sensing method, it is nonetheless remarkable to observe a precision of around 1 nm in the distance range of 1-12 nm, as demonstrated by the narrow distance distributions associated with the folded and stretched DNA beacons.
In the analysis of the results, we assume that the stretched DNA configuration is vertically aligned on the graphene plane. However, previous works have shown that 5′ bound double-stranded DNA (ds-DNA) may have tilting angles up to 45° [37]. It has also been reported that the vertical alignment of the Atto488@hp-DNA is affected by the density of ds-DNA [8], where higher molecular packing leads to more vertically aligned molecules. In a recent work, we have shown that, under similar conditions, the surface density of this DNA strand on a graphene surface is 1.3 ± 0.1 × 10 13 molecules cm −2 (maximum theoretical value of 4.8 × 10 14 cm −2 ) [12], which can be considered high molecular packing conditions, thus restricting the possible angular variance. It has also been reported that the elasticity of the ds-DNA depends on the strand contour length (L) [37,38]. In long DNA strands (L > 50 nm) each base can be seen as a randomly oriented segment, and the elasticity is said to be of entropic nature, and thus affected by temper ature. However, short DNA chains (≈10.5 nm in our case) can be considered as elastic rods whose bending can be affected by electric potentials, but where temper ature effects are negligible [12]. As discussed in [37] if L l p , with l p being the persistence length that can be determined from l p = λL 4 (λ = 30) [38], the duplex behaves like a rigid rod; in our case l p ≈ 71 nm and L = 10.5 nm, which confirms the L l p condition.
Based on the widths of the initial (FWHM = 0.6 nm) and broader final (FWHM = 0.8 nm) nanoscale distance distributions, we estimate the maximum tilt angle φ max for the DNA in the stretched configuration. The estimation relies on the assumption that the axial resolution can be characterized by a Gaussian having the same 0.6 nm width as the narrow distribution. We then fit the broader distribution with the minimum number of such Gaussians. This can be done using three Gaussians separated by FHWM/2 = 0.3 nm, centred at 12.1, 12.4 and 12.7 nm, as can be seen in figure 5(a). Considering 12.7 nm to be the total molecular length L T (L T = DNA length + linker length) in the fully stretched configuration, the shorter 12.1 nm distance could be related to a tilt angle φ max by L T cos φ max = 12.1 nm, where φ max ≈ 18°, see figure 5 The obtained total molecular length L T = 12.7 nm is comparable with our initial estimation of 1.7 + 10.5 = 12.2 nm (section S2 of the SI).
In figures 3(c) and (f) we also observe the presence of dyes located at longer distances above the graphene surface (around 12 nm), that remain nearly unchanged from the initial to the final state of the reaction. This could be related to the presence of defects. Localized defects can form directly during the CVD growth and micro-wrinkles in the graphene sheet may be formed during the transfer process (see enhanced D and G bands in the Raman spectra of defective graphene regions in figure S1, SI) [39,40]. Such defects may lead to a decreased near-field coupling strength between the graphene and the dyes, which weakens the fluorescence lifetime quenching yield of the graphene. As a result, longer, unchanging lifetimes may surge as an artifact of longer distances. Another possibility is the presence of misfolded DNA beacons. Such loop disruptions can be triggered by surface interactions with the functionalized graphene surface [41], or electrostatic forces between neighbor probes [42], particularly in case of high probe densities. As a consequence, these alternate conformations lead to the placement of the dyes at a longer distance from the graphene [43,44], which is not the initial configuration required for this study.
In the dynamic analysis, we assign the sharp and fixed peak ( figure 4(a)) at distances of 1.9 ± 0.2 nm to single-stranded (folded) and the broader peak at 10.9 ± 0.8 nm to the double-stranded (stretched) DNA conformations, respectively. We believe that the broad and moving distribution (G2, represented in green in figures 4(a) and (b)) is the result of the dynamic changes occurring during the hybridization process. As the excitation area of the laser is limited and reaches around 1 µm 2 in the lateral directions, multiple DNA molecules in different hybridization states are probed simultaneously. Therefore the collected signal results from the ensemble emission of the molecules, which might include folded, partly hybridized and fully stretched configurations. The peak shift from an average distance of ≈5 nm to ≈9 nm over time can be accounted for by the increasing population of molecules approaching the stretched configuration within the excitation area of the laser. Ultimately, the possible heterogeneous conformational states of the ensemble of DNA molecules within the excitation volume cannot be distinguished at the used densities of labeled DNA molecules. However, the observed dynamic range is much larger than the width of the fluorescence lifetime distributions at any time point of the chemical reaction. Also, the converted nanoscale distance distributions are in line with the expected molecular model of an unfolding DNA strand. We thus show that for such synchronized chemical reaction, here triggered by the addition of target DNA, the described fluorescence lifetime method is able to resolve mean molecular conformational changes. To resolve single DNA strand configurations, ultimately a single-molecule sensitive assay is needed. The absence of collective effects at low molecular densities may alter the double-stranded molecular configuration. To overcome such potential differences, alternative assays e.g. using sparsely labeled hp-DNA probes could be devised.
Previous reports have shown that the hybridization kinetics are highly affected by the concentration of the target DNA, temperature and salts concentration in the buffer solution [45]. In our work, despite of the high target concentration, the observed reaction time of approximately 40 min is consistent with the rather low temperature of 19°C at which the experiments were performed, which significantly decreases the collision frequency of target and probe DNA [46].
To the best of our knowledge, we provide for the first time an axial super-resolution optical image (achieving super-resolution in the vertical direction) of double-stranded DNA on an isolated functionalized graphene flake. We consider this an important sensing achievement with vast applicability in surface bio-functionalization studies and DNA beacon based biosensing. The micrometric lateral resolution enables differentiation between preferential bindings at edges, versus at the flake central region, while the nanometric axial resolution provides topographical information of the DNA strands bound to the graphene.
While previously reported optical super-resolution imaging techniques [47] enable subwavelength lateral resolution, they often provide poor axial resolutionnot smaller than around 40 nm [48]. In contrast, we were able to achieve axial resolution down to 1 nm using the graphene-based approach, thereby providing an outstanding method for nanometer-scale optical topography characterization. Similar near-field interactions between fluorescent dyes and gold thin films have been used by Chizhik et al to achieve axial superresolution in a technique entitled metal-induced energy transfer (MIET)-FLIM. Such technique was used for the study of cell membrane morphology [17] and nuclear membrane architectures [49], with an axial resolution reaching down to 3-4 nm, thus demonstrating the great potential in fundamental cell biology and medical sciences. In the present work, the improved axial resolution of about 1 nm via the use of graphene, thereby reaching sub-molecular length scales, provides a new level of sensitivity for biosensing and imaging applications.
Standard topography imaging techniques, such as AFM have been used to map the distribution of ss-DNA and ds-DNA in functionalized surfaces before and after the hybridization [50], providing information on the conformation of the DNA strands. However, such methods require intrusive probing and are limited to horizontal-lying DNA strands [51,52]. Also, due to the limited penetration capability of the tip, these are rather inadequate for densely packed molecular regimes. Furthermore, such characterization is limited to the comparison between initial and final states of the hybridization reaction, requiring parallel measurements such as x-ray photoelectron spectr oscopy (XPS) [50] and fluorescence intensity [51] to assess the status of the hybridization reaction, and thus offering no information regarding the DNA conformation during the hybridization. Advantageously compared to such techniques, the approach described in this work relies on non-intrusive optical probing which allows assessing the heigth of the biological structures, even during hybridization, excluding the need of a reference.
Other advanced DNA biosensing techniques have shown the ability to probe nanoscale information such as determining rotational dynamics of the DNA-bound dye molecule [53]. We take a step further by implementing a method that enables the mapping of the axial position of such dyes, which can be used to probe DNA interactions involving mechanical displacement of the strands, in this case, hairpin-based DNA hybridization. This could be used for the nanoscale imaging of complex biological organizations such as lipid bilayers, using a much simpler apparatus than other superresolution imaging techniques e.g. scanning near-field optical microscopy (SNOM) [54].

Conclusions
We demonstrate that the near field non-radiative coupling between graphene and fluorescent dyes can be used to achieve nanoscale axial resolution in a range of 0-30 nm using a nanofabricated test sample. The nanoscale sensitivity and analysis of fluorescence lifetime distributions allows the estimation of surface roughness, as demonstrated by comparison with AFM measurements.
Time-dependent measurements allow obtaining real-time DNA hybridization information with axial super-resolution in aqueous solution. The nanoscale distance information enables distinguishing between contributions from folded, stretched and intermediate DNA conformations along the DNA hybridization reaction, which may greatly improve our understanding of the DNA hybridization process.
Our DNA sensing method, employs fluorescence lifetime as the physical quantity, whose value can be used as an absolute measure of the distance of the labelling dye from the graphene. The achieved sensitivity enabled the discussion of possible tilt angles of ds-DNA. This is a great improvement compared to other DNA sensing platforms which are based on relative measures, e.g. of charge-induced local gating in FET based sensing, or fluorescence intensity based sensors that do not carry specific information about molecular reaction states or conformations.
Being a fluorescence-based method, the technology could be converted into an ultrasensitive singlemolecule detection method, enabling identification of specific DNA configurations and real-time reaction kinetics at the single DNA strand level. Compared to near field platforms based on thin gold layers of 20 nm thickness and with reported sensitivities of 3-4 nm in axial direction [17], the 2D material graphene is found to reach record breaking of 1 nm axial resolution.
We envision that this nanoscale near field sensitivity of graphene gives the platform the potential to detect misfoldings of DNA strands due to, e.g. mutations in the DNA sequence.
In the limit of high DNA densities the technique will allow the study of complex DNA contact networks. Finally, the fluorescence lifetime susceptibility of fluorescent dyes located in the near field of graphene, a material that exhibits excellent functionalization possibilities [55], may trigger advances in the fields of medicine, food authenticity and safety as well as other biological fingerprinting applications.