Molecular states and spin crossover of hemin studied by DNA origami enabled single-molecule surface-enhanced Raman scattering

The study of biologically relevant molecules and their interaction with external stimuli on a single molecular scale is of high importance due to the availability of distributed rather than averaged information. Surface enhanced Raman scattering (SERS) provides direct chemical information, but is rather challenging on the single molecule (SM) level, where it is often assumed to require a direct contact of analyte molecules with the metal surface. Here, we detect and investigate the molecular states of single hemin by SM-SERS. A DNA aptamer based G-quadruplex mediated recognition of hemin directs its placement in the SERS hot-spot of a DNA Origami Nanofork Antenna (DONA). The configuration of the DONA structure allows the molecule to be trapped at the plasmonic hot-spot preferentially in no-contact configuration with the metal surface. Owing to high field enhancement at the plasmonic hot spot, the detection of a single folded G-quadruplex becomes possible. For the first time, we present a systematic study by SM-SERS where most hemin molecule adopt a high spin and oxidation state (III) that showed state crossover to low spin upon strong-field-ligand binding. The present study therefore, provides a platform for studying biologically relevant molecules and their properties at SM sensitivity along with demonstrating a conceptual advancement towards successful monitoring of single molecular chemical interaction using DNA aptamers.


Atomistic simulations of the G-quadruplex
The atomistic structure of PS2.M was constructed following the proposed folding of ref 1.
(monomeric with anti-parallel strands). The ideal model was then built using the 3D-NuS webserver. 2 Firstly, water 3 (TIP3P model) and ions 4 (0.1 M of KCl, parametrized by Joung and Cheatham) are added to the G-quadruplex (parametrized using the OL15 force-field). 5 The system is energy minimized using a steep descent algorithm. Afterwards, the system is equilibrated at a high temperature (400 K) for a short time (1 ns). We constraint, during equilibration, all the G-quadruplex hydrogen bonds, the torsional angle which impose planarity for the three quartet and the position of the K + ions between the quartet planes. The constraint are imposed using the PLUMED package 6 through harmonic restraints, which are subsequently lowered gradually during a second equilibration phase at 300 K for about 40 ns. The system was then simulated at 300 K for 1 µs. The software used for the simulation is GROMACS. 7 During both equilibration and simulations, we used periodic boundary conditions, constant pressure (1 bar) and long-range electrostatic interactions were treated by the particle mesh Ewald algorithm. 8 Configurations were sampled every 100 ps.
The groove width is measured as the backbone to backbone distance between the C4ʹ atoms of the central quartet (shown as blue spheres) in Fig. 2A, manuscript with label 1,2,3,4 corresponding to the C4ʹ atoms of G-nucleotides 4,9,13, and 18 using the sequence GTG GGT AGG GCG GGT TGG as reference for numbering). The average values are 1.46 nm, 1.35 nm, 1.47 nm, 1.38 nm, respectively for 1-2, 2-3, 3-4 and 4-1 C4ʹ distances (distributions and values as a function of time are in Fig. S7A-B). Fig. S7C shows the distribution of the vertical Gquadruplex length, defined as the maximum distance perpendicular to the G-quartet planes between all backbone atoms (see Fig. 2A, manuscript). The most probable value calculated for the vertical length is 1.99 nm, which is within the theoretical hot-spot gap distance in the DONAs.

Coarse-grained simulations of the nanofork -G-quadruplex complex.
The setup for the oxDNA2 coarse-grained simulation was prepared in the following way. The caDNAno template for the nanofork and the atomistic G-quadruplex DNA were converted to the oxDNA representation using the tacoxDNA package. 9 They were then merged and linked using oxView 10 , and subsequently relaxed with the protocol described in ref 8. Simulations were carried out with the LAMMPS software implementation. 11 The temperature was set to T= 300 K and the monovalent salt concentration to 1 M (typically large in order to have stable origami structures). Note that the G-quadruplex structure is not stable with the oxDNA model, due to the lack of Hoogsteen-Watson Crick pairing and of explicit counterions. Thus the Gquadruplex was constrained to its initial predicted position through a harmonic restraint on the root mean square deviation. The system was simulated for about 5×10 6 simulation time units, and configurations sampled every 10 3 time units.
In order to measure the orientation of the G-quadruplex planes with respect to the nanofork, we used the nanofork to define a fixed cartesian coordinate system (x-axis is the bridge, z-axis the pillars as represented), see the image below. This coordinate system is used to compute the relative orientation P of the G-quadruplex over the nanofork, where P is a normalized vector shown in the enlargement (red circle), connecting the lower to upper G-quartet centres and expressed in spherical coordinates (polar and azimuthal angles (θ,ϕ), while the radial distance is fixed to one). The histogram of (θ,ϕ) (not shown) is compatible with a random orientation of the P vector.

FRET experiment to demonstrate the successful folding of hemin aptamer to Gquadruplex:
To ensure the conformational change induced G-quadruplex formation, we carried out a (Förster resonance energy transfer) FRET measurement employing coumarin (C343) and fluorescein (6-FAM) as donor-acceptor dye pair. 21 The aptamer sequence was modified with FAM dye at the 5ʹ-end and C343 dye at the 3ʹ-end, which is connected to the DNA origami fork via running bridge staple sequence separated by two T-spacer sequence. (Fig. S6A and

B)
Emission spectra recorded for the aptamer sequence at the bridge of the DNA origami fork showed dominant emission spectra of C343 dye at 498 nm (Fig. S6C) before K + addition.
Successful folding of the aptamer sequence upon addition of K + into a G-quadruplex leads to the close up pairing of the dyes resulting in efficient FRET as depicted in Fig. S6B. This is reflected in Fig. S6C (blue curve) with an additional shoulder appearing at 515 nm from FAM due to successful FRET. (blue dots, Fig. 2A; manuscript) distances (distributions and values as a function of time are in Fig. S7A-B). On the other hand, the most probable value calculated for the vertical length of the G-quadruplex is 1.99 nm (Fig. S7C). Discussion: Based on the MD simulation, the G-quadruplex attached to the bridge can attain random orientation, but the absence ( Figure 2E(i),(ii) and 2F(i), manuscript) or weak occurrence ( Figure 2F (ii)) of the ring breathing mode 643 cm -1 indicates that there is a higher probability to adopt a parallel orientation to the nanoparticle surface.

Fig. S9
Exemplary TEM images of Au DONAs. Because of the different material contrast the actual DONA structure could not be seen against nanoparticle although negative staining was used. The Au DONAs consists of 60 nm particles size. The gap distance calculated in each of the dimers are highlighted in yellow text. The average gap distance was calculated to be 1.7±0.13 nm. To be mentioned that DONAs with gap distance in the range of 1.1-1.3 nm were also found which were not taken into account in the statistical analysis. We infer such DONAs to be devoid of the G-quadruplex with hemin moiety.   In few situations, SERS signal was observed in Au aggregates (an exemplary situation shown in Fig S13 (iii)). This could arise if hemin remains intercalated to the nanofork body due to inefficient washing step, which might produce some SERS signal from aggregated dimer body.

Time series measurement on single Ag DONAs in dark field mode
To further understand the SM behaviour, the time evolution SERS spectra from single Ag DONA was examined (Fig. 3E)  cm -1 and 1527-1541 cm -1 corresponding to δ(CmH), ν20(B2g) and ν38(Eu) modes of hemin (Table   S4, ESI), which was not detected in correlated SM-SERS spectra reported above. The characteristic peak of G-quadruplex around 1476 cm -1 could be observed during initial 18s which however remains off for prolonged period ( Figure 3E) The high spin state marker band ν10 clearly exhibits peak wandering in the range 1602-1616 cm -1 (indicated in white boxes; Fig.   3E) accompanied by intermittent blinking behaviour. The behaviour includes shorter bright onperiod (53 s-56 s) followed by total peak disappearance (57 s -72 s) which then gradually shows up as weak peak intensity for prolonged time to finally reappear as bright signal at 85 s to 97 s of the acquisition time (acquisition time was 0.5s). The total peak disappearance could be an indication that the hemin moiety could flip out of the hot-spot volume due to thermal vibration. This is in tandem with blinking pattern observed for vibronically active bands (corresponds to non-totally symmetric mode) that typically shows erratic blinking behaviour with switch-on behaviour for brief time periods and off with weak intensity for longer period, however exhibiting the brightest intermittent signal while on. 23    Discussion: Although hemin ideally should be in a non-contact configuration in the plasmonic hot-spot, the fluctuation interaction (mentioned above) might result in redox state selfexchange and hence the observed distribution of Fe (II)/Fe (III) supported by a study by Wang et al. 20 where the fluctuation trajectories observed for ν4 mode were ascribed to charge transfer dynamics that exist on Ag-hemin interface. Further they report that under no potential applied the single molecule of hemin tends to fluctuate between oxidized and reduced state passing through possible intermediate states consequently giving rise to one dominant distribution.
Ligand binding: The hemin Fe (III) unit which predominantly exists in high spin state should ideally switch to low spin Fe (III) state upon successful binding of the respective strong field ligands, resulting in a shift of the ν10 mode. [26][27]