Effect of water-DNA interactions on elastic properties of DNA self-assembled monolayers

DNA-water interactions have revealed as very important actor in DNA mechanics, from the molecular to the macroscopic scale. Given the particularly useful properties of DNA molecules to engineer novel materials through self-assembly and by bridging organic and inorganic materials, the interest in understanding DNA elasticity has crossed the boundaries of life science to reach also materials science and engineering. Here we show that thin films of DNA constructed through the self-assembly of sulfur tethered ssDNA strands demonstrate a Young’s modulus tuning range of about 10 GPa by simply varying the environment relative humidity from 0% up to 70%. We observe that the highest tuning range occurs for ssDNA grafting densities of about 3.5 × 1013 molecules/cm 2, where the distance between the molecules maximizes the water mediated interactions between the strands. Upon hybridization with the complementary strand, the DNA self-assembled monolayers significantly soften by one order of magnitude and their Young’s modulus dependency on the hydration state drastically decreases. The experimental observations are in agreement with molecular dynamics simulations.


S2. Experimental setup
In the optical beam deflection method, a laser is focused onto the free end of a cantilever beam and its reflection is collected by a quadrant photodetector or by a position sensitive detector (PSD). However, it is known that the response of a PSD is not uniform along its whole surface; therefore, a static cantilever bending, which moves the laser spot at the surface of the detector, will induce a non-real shift in the measured resonance frequency. In order to prevent this undesirable measurement artifact, we have introduced two mirrors in the optical path actuated by an automatized motor in a feedback closed loop configuration: the output of the PSD is converted into a voltage input signal to the motors controlling the mirror angles in such a way that the change of the angle maintains the laser spot at the central point of the detector surface all the time, Fig. S2. Therefore, the input signal of the mirror angle is translated into the static deflection of the cantilever, whereas the resonance frequency obtained by the Fast Fourier Transform (FFT) of the signal coming from the spot at the center of the PSD is free from undesirable artifacts. The output signal of the photodetector is split up into two different signals, one is injected into the feedback loop controlling the mirrors and the other one is analyzed by a locking amplifier with a phase locked loop (PLL). In order to demonstrate the low noise level, and reproducibility of the dynamic characterization, the upper chart in Fig. S3 shows a sample of the frequency measurement obtained by the PLL, with a frequency noise of barely 50 mHz. By using the locking amplifier, the signal-to-noise ratio of the resonance peak is enhanced reaching a value of 6 at resonance frequency of 5.1 kHz with a mechanical quality factor of about 20, shown in the lower chart of Fig. S3. S4. X-ray photoelectron spectra (XPS) measurements X-ray photoelectron spectra (XPS) were recorded using a Escalab 200R (VG, UK) electron spectrometer equipped with a hemispherical analyzer, operating in the constant pass energy mode, and a MgKa (hn = 1253.6 eV, 1 eV = 1.603×10 '() ) X-ray source operated at 10 mA and 12 kV. The detection angle of photoelectrons was 60º to the surface of the specimen. The spectrometer was calibrated against Au4f7/2 line at 84.0 eV using a gold sheet and Cu2p3/2 at 932.5 eV from a copper sheet. Charge effects on the samples were removed by taking the C1s line from adventitious carbon at 284.8 eV. In order to estimate the photoelectron peak intensities, the background was subtracted from the measured spectra according to the Shirley method and using a combination of Gaussian and Lorentzian lines (90G-10L). The relative surface atomic ratios were determined from the corresponding peak intensities, corrected with tabulated atomic sensitivity factors. The reproducibility of the results was confirmed several times under the same conditions.

S3. Dynamic Characterization
In order to determine the number of molecules at the cantilever surface we have performed a quantitative characterization of the DNA film by X-ray photoelectron spectroscopy (XPS). The presence of nitrogen atoms is typically used as the experimental indicator of adsorbed DNA; however, since the used buffers in the immobilization and subsequent cleaning process are unspecific sources of nitrogen we have chosen the phosphorous as signature indicator. The signal coming from the gold 4f peak is attenuated as the immobilization time for the ssDNA is increased. From this attenuation, it is possible to calculate the actual thickness of the DNA layer by using the clean Au4f spectrum as reference. Then, the calculated thicknesses are used to correct the measured XPS peak ratios of the N and P atoms for attenuation. In order to do this, we have to calculate the practical effective attenuation length (PEAL, LAu) for electrons in the film using a reference film, whose thickness we have measured by atomic force microscopy.
The relationship between the intensity of the XPS peak, -. , and the thickness, , is given by -. = -. 1 − -. . Table S1 shows the results obtained for the surface coverage by looking at the P 2p peak intensity relative to the Au 4f peak. Although the presence of nitrogen atoms is typically used as the experimental indicator of adsorbed DNA, since the used buffers in the immobilization and subsequent cleaning process are unspecific sources of nitrogen we have used the phosphorous as signature indicator.
Table S1 | Molecular surface density determination. By using the number of P atoms per Au atoms N 8 N 9: and the relative coverage Note the saturation at 6.5x10 13 molecules/cm 2 . Fig. S6 shows the thickness of DNA layers calculated from XPS measurements as blue and red circles for ssDNA and dsDNA, respectively.
As it was previously explained, thickness was obtained by means of the Au 4f peak attenuation (the gold signal exponentially depends on the inverse of the thickness). The larger the immobilization time, the thicker the ssDNA layer. This is due to the molecular interactions between the strands that arise when the molecular surface density grows 1    In order to account for the hydration dynamics, the hydration/dehydration cycles were simulated by linearly sweeping the number of water molecules from zero to the maximum number per molecule described above. The water uptake by DNA as a function of the relative humidity has extensively been studied in samples consisting of DNA fibers obtained by standard desiccation methods 5 . In these studies, firstly established by Falk and collaborators the water adsorption isotherms closely follow the Brunauer-Emmett-Teller (BET) equation 6 ; however, it has been demonstrated that a linear dependency is a good approximation 5 . Since the calculation of the mass density takes into account the number of water molecules, it was also correspondingly changed.
S6. Molecular Dynamics Simulations. DNA layer thickness  On the other hand, as the surface density increases, the thickness also increases due to the intermolecular forces between the DNA strands.  In order to study the dynamics of the SAM formation, we simulate the thickness of the layer as a function of the immobilization time for both the ssDNA and dsDNA at fully hydrated state and at high vacuum, Fig. S10. As it was described above, the thickness of the dsDNA is larger than the ssDNA. From these simulations, it is also possible to see how the dsDNA reaches a saturation value of the thickness in barely two hours, which is indicative of the shielding effect of the double helix. Note also that the thickness of the hydrated layer is larger. This is because of the extra rigidity provided by the water molecules placed around the DNA strands.

Simulated bending experiments
Until now we have shown how increasing packing density of DNA molecules has important effects in the layer thickness. As the distance between the strands decreases, there is an increasing attractive collective force due to hydrogen bonding between bases, base-phosphorous interactions, and base stacking. Thus, we expect that the collective elastic properties present a similar dependency. In order to study the elasticity of the DNA layer, we have simulated bending experiments by using a potential mean force using WHAM method 7 . Fig.   S12 shows the simulated bending experiments for ssDNA (blue circles series ranging from dark to light blue for increasing molecular surface density) and dsDNA (red circles series ranging from dark to light red). The energy cost increases accordingly with the increasing molecular density packing of the ssDNA, being the corresponding layer of 3.86x10 13 molecules/cm 2 six times stiffer than the 3.1x10 12 molecules/cm 2 one; indicating a stiffening effect given by the intermolecular forces. Note that the stiffening of the dsDNA is virtually negligible when compared with the ssDNA.

S11. Thickness Uncertainty
Attending the simulations (commercially available software Schrödinger), the length of a 20-base long ssDNA is 6.4nm, far away from the measured value of 1.2nm, but the measured thickness is still in reasonable agreement with the MD simulations, as can be seen from the following images. The first one corresponds to the length calculation in Schrödinger, and the second one is the fig. S8  At this point, we should take into account the way this thickness was calculated. The final point of the thickness growing curve was measured by AFM at the maximum molecular surface density. We set a DNA step by simply shadowing a gold area and measure the thickness variation.
Please, take a look at one of the performed measurements for the grafting density of 6x10 13 molecules/cm 2 .
Although all the trends are correct, we attribute the difference among measurements and the MD simulations to the attractive force in between the gold and the bases composing the ssDNA layer, note that the simulations are done on an inert surface (graphene). Nevertheless, it seems that is too much material to pack into such a small thickness. We really think that the MD simulations could provide a better estimation for the thickness tan the experimental XPS and AFM measurements presented. Note that the AFM measurements averages the thickness measurement in an area equivalent to the tip surface area, and, as it was observed in previous studies, the DNA forms growing coalescent islands.
Thus, the AFM averages with the bare surface.
We must take into account the error in the thickness estimation by using AFM. As it was explained above, this error comes from the calibration method employed. The calibration of AFM was made by using a calibration sample trenches of height -21 ± 1 nm. The maximum tolerance after AFM calibration is of 1%. Therefore, the average of our DNA thickness measurements retains and propagates that error, giving as a result 1.2 ± 1.1 nm. This uncertainty propagates to the Young's modulus determination to reach an error of 33% in the value determination.

S12. Error Propagation Calculation
The calibration of AFM was made by using a calibration sample made of trenches of height -21 ± 1 nm. The maximum tolerance after AFM calibration is of 1%. Therefore, the average of our DNA thickness measurements retains and propagates that error, giving as a result 1.2 ± 1 nm.
We can neglect the cantilever's thickness error (