DNA-enabled rational design of fluorescence-Raman bimodal nanoprobes for cancer imaging and therapy

Recently, surface-enhanced Raman scattering nanoprobes have shown tremendous potential in oncological imaging owing to the high sensitivity and specificity of their fingerprint-like spectra. As current Raman scanners rely on a slow, point-by-point spectrum acquisition, there is an unmet need for faster imaging to cover a clinically relevant area in real-time. Herein, we report the rational design and optimization of fluorescence-Raman bimodal nanoparticles (FRNPs) that synergistically combine the specificity of Raman spectroscopy with the versatility and speed of fluorescence imaging. DNA-enabled molecular engineering allows the rational design of FRNPs with a detection limit as low as 5 × 10−15 M. FRNPs selectively accumulate in tumor tissue mouse cancer models and enable real-time fluorescence imaging for tumor detection, resection, and subsequent Raman-based verification of clean margins. Furthermore, FRNPs enable highly efficient image-guided photothermal ablation of tumors, widening the scope of the NPs into the therapeutic realm.


Synthesis of 60 nm AuNP core
The AuNP core was synthesized using a modified protocol using a seed mediated growth method. In a typical synthesis procedure, we first synthesized ~ 15 nm AuNP cores. To 99 ml deionized water, 1 ml 25 mM HAuCl4 was added and the solution was heated on a heating plate to boil. To the boiling solution 1 ml of 3.3% sodium citrate solution was added. After 15 minutes the color of the solution had changed to red confirming the formation of 15 nm AuNP cores. In the next step, to 100 ml deionized water 125 μL of 200 mM HAuCl4, 30 μL 500 mM trisodium citrate and 700 μL 15 nm AuNP cores were added under stirring. Further, 250 μL of 1M hydroxylamine hydrochloride solution was mixed. Within few seconds the color of the solution turned deep red confirming the synthesis of 60 nm AuNP.

Synthesis of 40 nm-10 nm AuNR core
The synthesis of AuNRs was carried out using the silver-assisted growth procedure from previous report. 1 a. AuNP Seed synthesis: 60 μL of 10 mM ice cold sodium borohydride solution was mixed to 1 mL of 2.5 mM Hydrogen tetrachloroaurate(III) hydrate solution in 100 mM Cetrimonium bromide (CTAB) and under vigorous mixing. The solution color immediately changed to yellowish brown confirming the formation of small sized seeds that act as nucleation seed for AuNR growth.
b. AuNR synthesis: To 100 mL of 100 mM CTAB solution, 80 μL of 100 mM silver nitrate and 500 μL of 100 mM Hydrogen tetrachloroaurate(III) hydrate was mixed. After gentle mixing, 1200 μL of 100 mM ascorbic acid solution was added and mixed thoroughly. To this mixture, 600 μL of the previously prepared AuNP seed solution was added. The mixture was kept undisturbed at least one hour. The solution turned purple confirming formation of AuNP and verified using UV-Vis spectroscopy and TEM imaging.

Molecular dynamic simulation methodology
Gold nanoparticles are known to form face-centered cubic (fcc) lattice with a truncated octahedral motif. 2,3 We have also chosen a truncated octahedral gold nanoparticle composed of 249 gold atoms for this study. The starting structure for gold nanoparticle, Au-55 core (1.2nm), was taken directly from Online JSmol Resources (https://chemistry.beloit.edu/edetc/pmks/pages/gold.html). Further, we developed two single stranded DNAs: (i) ss-DNA (PODNA) and (ii) ss-Phosphorothioate DNA (PSDNA). Both of the single stranded oligonucleotides are composed of six adenine nucleotides, linked to the surface of the gold nanoparticle by a six-carbon alkyl thiolate linker and tagged with the Lumiprope Cy family dye through linker Lysine.
The initial helical parameters of the PODNA were adapted from the structure of an oligo(dA).oligo(dT) tract. 4 The force field parameters for PODNA are taken from AMBER-ff99bsc0, while the parameters for the alkyl thiol are taken from the work of Hautman and Klein. 5,6,7 The force field parameters for the dye and associated linker (Cy7+Lys) were taken from Graen et. al., and are compatible with AMBER force fields. 8 The interactions between the gold atoms are described by Lennard-Jones potentials where the parameters σ = 2.569 Å and ε = 0.458 eV are taken from the literature. 9 The starting structure of ss-DNA connected with alkyl thiolate linker was developed using MOLDEN software and the PSDNA structure was obtained by replacing O1P-atom (i.e., one of the O-atoms of backbone phosphate group) of PODNA with a S-atom. 10 The initial structure of the Cy7+Lys (dye) system was again taken from Graen et. al.
In each of the simulations, the gold nanoparticle linked and dye-tagged PODNA/PSDNA was explicitly solvated with TIP3P water molecules in a rectangular periodic box whose dimensions were at least 1.5 nm larger than the size of the corresponding solute molecules; and subsequently charge neutralized by adding 4 Na+ ions. The entire set-up was generated using GROMACS 5.0.7. 11 Each system with a cubic box size of 9.084 nm, consisted of ~24363 water molecules and a total of ~73643 particles. The initial round of simulation with explicit solvent and ions involved 50000 steps of steepest descent energy minimization to remove the high-energy contacts. The position of the gold atoms was fixed with harmonic constraints. Thereafter, a 100 ns molecular dynamics simulation at 300 K with a NVT ensemble was performed while keeping the coordinates of gold atoms frozen using 0.001 ps as time step. The particle mesh Ewald (PME) summation method was used to treat long-range electrostatic interactions (with fourier spacing of 0.12 nm and interpolation order 4) and force-switch method was applied for non-bonded interactions (van der Waals) with a cutoff of 1.0 nm. 12 The real-space cut-off was set to 1.0 nm. The verlet cut-off scheme was implemented. We used V-rescale for maintaining the average temperature of 300 K. 13 All the hydrogen atoms bonded to the heavy atoms are constrained using LINCS algorithm to their respective equilibrium bond-length. 14 All the water molecules were simulated as rigid molecule using SETTLE. 15 Both the systems (PODNA and PSDNA) were individually simulated multiple times by varying initial velocity distributions of the systems. Atomic coordinates were saved every 0.5 ps for the trajectory analysis.
We have also performed control simulations with the sequence 5'-TCGCGC-3'. The initial helical parameters of this PODNA were adapted from the structure of Dickerson Drew Dodecamer and the corresponding PSDNA structure was obtained by replacing O1P-atom of PODNA with a S-atom. 16 The simulation protocol was exactly the same as described here in the method section for the A6-sequence. The PSDNA and PODNA systems were each simulated multiple times independently with different initial seeds, reaching a cumulative of 1 µs long MD-run.
All the trajectories were analyzed with GROMACS 5.0.7 and the movies are prepared with VMD. 17

Electromagnetic simulation methodology
We calculated the electric field near metal nanoparticles using the discrete dipole approximation method.

Supplementary Figures
Supplementary Figure 1. (a) UV-Vis spectra of 60 nm AuNP-core based FRNPs before (black) and after (red) silication. We observed ~9 nm bathochromic shift of the absorption maximum. (b) Fluorescence background subtracted Raman spectra of same concentration of FRNPs before (black) and after (red) silication. After silication the Raman peak at 796 cm -1 10 times higher due to decreased surface-fluorophore distance and change in dielectric environment.  . DCLS identifies the presence of a particular spectral signature. In order to generate the DCLS model, the reference spectra were preprocessed by baseline subtraction using a Whittaker filter (with width λ = 200 cm −1 ) and subjected to L1-norm (normalization by the area), followed by a Savitzky−Golay derivative filter (second-degree polynomial fit, first-order derivative, width = 15 steps). (b) The limit of detection of OFRNPs in the tissue phantom was determined to be between 3-10 fM. (c) The limit of detection of OFRNPs in blood was determined to be ~5 fM.