The Effect of Chemical Structure of OEG Ligand Shells with Quaternary Ammonium Moiety on the Colloidal Stabilization, Cellular Uptake and Photothermal Stability of Gold Nanorods

Introduction

Colloidal gold nanoparticles (GNPs) have been studied in a wide range of research areas attributable to their specific optical properties which originate from the interaction of light with their free conduction electrons, known as localized surface plasmon resonance (LSPR).^1–3 Especially gold nanorods (GNRs) have attracted particular attention due to their two distinctive extinction bands that are associated with transverse and longitudinal modes of LSPR. Optical response can be further controlled by tuning the particle size, shape, composition, and dielectric constants of the particle material as well as of the surrounding medium.^4–6 Under light irradiation, the LSRP of GNRs leads to partial conversion of absorbed light into heat. In such a scenario, gold nanoparticles act as nanosized sources of heat.⁷ The described phenomenon attracts attention of scientists for its promising applications in biomedical fields such as plasmonic photothermal cancer therapy,⁸ controlled drug delivery and release,⁹ and microtissue surgery.¹⁰

Gold nanoshells stabilized by polyethylene glycol (PEG) were demonstrated as the first plasmonic nanoparticle-based agents for photothermal therapy of a cancer in clinical trials based on their accumulation in tumor ascribed to the enhanced permeability and retention effect.¹¹ Ethylene glycols are particularly promising ligands of GNPs for their minimal toxicity, increased polarizability, minimal interaction with biomacromolecules and ability to escape opsonization, the latter preventing recognition of GNPs by immune system.^12–17 Moreover, PEG increases the stability and solubility of the NPs under physiological conditions by increasing the steric distance between NPs that prevents their aggregation and increasing hydrophilicity of NPs due to strong surface hydration of ethylene glycol units.^18–20 Although the PEG coatings improve the pharmacokinetics of NPs and increase the blood circulation time,^21,22 they also hinder the desired nanoparticle uptake by cancer cells. It is known that the cellular uptake of neutral NPs, including uncharged PEGylated NPs, is much lower in comparison to NPs with positively and/or negatively charged surfaces.^23–25 Additionally, the rod-like shape of NPs can significantly reduce the uptake of NPs by, for example, macrophages.^26,27 These both limit the utilization of PEGylated gold nanorods in applications that require high incorporation of NPs into cells such as cell transfection^28,29 and photothermal cancer therapy utilizing the tumor tropic properties of stem cells to deliver NPs to the tumor area.^30–32

Depending on the light excitation, one can trigger a variety of photothermally induced effects, namely nanoparticle fragmentation,³³ nanoparticle reshaping,³⁴ microbubble generation,^35,36 or generation of acoustic and shock waves³⁷ that may occur during the photothermal ablation of cancer cells. Several studies reveal underlying mechanisms of the photothermal conversion consisting of a sequence of events, namely the plasmon excitation, electron-phonon thermalization, and heat diffusion.^7,38 Recent advances in the thermoplasmonics propose both theoretically and experimentally that the solvation layer of nanoparticles may considerably influence the heat transfer from the nanoparticle to its surroundings, especially for the case of pulsed illumination.^39–44 Horiguchi et al⁴² showed that the photothermal reshaping of GNRs depends on the type of their surface modification when nanosecond pulsed laser irradiation induced larger spectral changes of PEG-GNRs and phosphatidylcholine (PC)-passivated GNRs than in the case of poly(ethylenesulfonate hydrochloride) (PSS)- and poly(vinylpyrrolidone) (PVP)-passivated GNRs. Similarly, a method of transient absorption spectroscopy revealed that the thermal dissipation from photoexcited GNRs depends on the ligand chemistry and on the ligand ability to exclude water from the surface of the GNRs that makes the phonon transport between the GNRs and solvent less efficient.⁴³ The role of ligand hydrophilicity was confirmed by Wu et al,⁴⁴ who demonstrated that the effective thermal conductance of the GNRs/surfactant layer/water system is higher for GNRs modified by hydrophilic PEG than for GNRs modified by hydrophobic cetyltrimethylammonium bromide (CTAB) with the same length of surfactant layer. Possibly the covalent bonding between Au and thiol-containing PEG and water penetration into the PEG layer are the reasons of enhanced heat dissipation. In addition, Centi et al⁴⁵ found that relatively small thiols (methylbenzenethiol) dramatically enhance the thermal stability of GNRs during annealing in an oven set to 90°C and their photostability during photoacoustic monitoring compared to PEG (MW 5,000 Da).

Recently, we developed the GNRs stabilized by surfactant composed of the short ethylene glycol chain with thiol moiety and quaternary ammonium head (trimethylammonium) as terminal functional group – N,N,N-trimethyl-3,6,9,12,15-pentaoxaheptadecyl-17-sulfanyl-1-ammonium bromide (POSAB).⁴⁶ While the polar ethylene glycol chain strongly reduced the compound cytotoxicity in free state compared to the common alkyl chain, quaternary ammonium salt (QAS) provided GNRs high cellular uptake. In connection with these results here we investigated the role of different length of oligoethylene glycol (OEG) chain and various structure of quaternary ammonium head on the physical–chemical and biological properties of free compound as well as ligand shell of GNRs. Three series of ligands with 3, 4 and 5 ethylene glycol units and six variants of quaternary ammonium heads were rationally designed and synthesized. Their ability to ensure the colloidal stability of GNRs, the rate of GNRs cellular uptake and cytotoxicity of GNRs as well as free ligands were evaluated. Finally, the effect of solvation layer on GNRs thermal stability under the femtosecond (fs) laser irradiation was measured for the representative OEG ligands showing high cellular uptake and good stability and was compared to a known alkyl analogue (16-mercaptohexadecyl)trimethylammonium bromide; MTAB).^47,48

Materials and Methods

Chemicals and Antibodies

3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT), 4ʹ,6-diamidino-2-phenylindole (DAPI, D9542), ascorbic acid, bovine serum albumin (BSA), hydrogen tetrachloroaurate(III) trihydrate (HAuCl₄. 3H₂O), cetyltrimethylammonium bromide (CTAB), silver nitrate (AgNO₃) and sodium borohydride (NaBH₄) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Antibody against lysosome-associated membrane protein 1 (LAMP-1; sc-20011) was from Santa Cruz Biotechnology (Dallas, TX, USA). Alexa Fluor 488 phalloidin (A12379), Alexa Fluor 488-conjugated secondary antibody and ProLong Gold Antifade Mountant were from Invitrogen (Carlsbad, CA, USA). Milli-Q water (18.2 MΩ at 25°C) used for nanoparticle synthesis was from Ardeapharma a. s., Sevetin, Czech Republic.

Cell Cultures

Human cervical carcinoma cell line HeLa obtained from ATCC (Manassas, VA, USA) was cultured in Dulbecco᾿s Modified Eagle᾿s Medium (DMEM; Biochrom AG, Berlin, Germany; containing 4.5 g/L glucose) supplemented with 10% Fetal Bovine Serum (FBS; Gibco; Grand Island, NY, USA). Chinese hamster ovary cell line CHO-K1 from ECACC (Salisbury, UK) was cultured in Nutrient Mixture F-12 Ham (Sigma-Aldrich) growth medium supplemented with 10% FBS (Sigma-Aldrich). Both culture media were supplemented with penicillin (100 U/mL) and streptomycin (100 µg/mL; Sigma-Aldrich). Cells were cultivated at 37°C under 5% CO₂ atmosphere and 95% humidity.

Experimental Evaluation and Calculation of the Hydrophobicity of Prepared Ligands

Hydrophobicity of final compounds expressed as ClogP was calculated in Open Babel, version 2.3.1 (http://openbabel.org, accessed October 2011). The correlation with experimental evaluation of hydrophobicity by high performance liquid chromatography (HPLC) with isocratic elution (logk) is described in the Supplementary Material.

Calculation of Molecular Electrostatic Potential (ESP)

The models of prepared compounds were pre-designed in HyperChem 8.0 software (Hypercube, Gainesville, FL, USA) as free cations and exported as mol files for further calculations. In Spartan 14 (Wavefunction, Irvine, CA, USA), semi-empirical quantum chemistry PM6 method was used for determination of the compound equilibrium geometry in vacuum. Electrostatic potential of the geometrical optimized conformers in the local minimum of potential energy was mapped on the electron isodensity surface of 0.002 e/b³. The atomic partial charges were determined on the same level of theory in Spartan 14 applying CHELP algorithm for a least-square fit of the partial charges to the molecular electrostatic potential (ie, ESP atomic partial charges). Additional computational models obtained by molecular dynamic based conformational analyses with the semi-empirical method RM1 are described in the Supplementary Material.

Cell Viability Assessment of Ligands in Free State

In order to compare the cytotoxic effect of the studied compounds standard MTT assay (3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide) was utilized according to the manufacturer's protocol using the CHO-K1 cells.⁴⁹ The protocol is described in detail in the Supplementary Material.

Preparation of Cetyltrimethylammonium Bromide (CTAB)-Stabilized Gold Nanorods and Their Functionalization by OEG Compounds

Gold nanorods (GNRs) were prepared by a modification of the seeded-growth method in the presence of silver nitrate according to the protocols described previously.^48,50,51 The protocol is described in detail in the Supplementary Material.

To modify GNRs by OEG compounds and by MTAB, CTAB-coated GNRs (5.7 mL of GNRs tuned to 633 nm or 7.0 mL of GNRs tuned to NIR) were purified by two cycles of centrifugation (4,700 ×g for GNRs tuned to 633 nm or 7,400 ×g for GNRs tuned to NIR for 20 minutes), redispersation of sedimented GNRs in the Milli-Q water and concentration to 1 mL (residual CTAB in solution was ≤0.5 mM). After that, 1 mL of 5 mM aqueous solution of MTAB or cationic OEG compounds was directly added to the concentrated GNRs dispersion and kept at room temperature for a minimum of 7 days. Next, the GNRs solution was diluted by Milli-Q water (7 mL) and sonicated at 85 W and 45°C for 2 hours. Afterwards, the residual unconjugated ligand was removed by two cycles of centrifugation in the following 2 days. The final concentration of prepared GNRs ranged from 0.35 mM to 0.45 mM of gold atoms (Au⁰) in the nanoparticle dispersion as calculated by formula Au⁰ [mM] = 0.355 × A_{390 nm}/path length [cm].^48,52 To assess the colloidal stability of GNRs, the UV-VIS-NIR absorption spectra of GNRs dispersions were measured by a DU 730 Life Science UV/VIS Spectrophotometer (Beckman Coulter, Fullerton, CA, USA).

Quantification of Ligand Packing Density by Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES)

To quantify the ligand packing density, the GNRs functionalized by cationic ligands were analyzed by ICP-OES (Spectro Arcos MV, Spectro Analytical Instruments, Kleve, Germany; see the Supplementary Material for detailed measurement conditions). The ligand packing density on GNRs was then calculated from the ratio of the mass fraction of sulfur (S) and gold (Au) as described previously.^53,54 The parameters of GNRs were estimated from FE-SEM size analysis of 300 NPs, the calculation is shown in the Supplementary Material. The packing density σ_calc. was calculated by the equation:

where w_Au and w_S are the Au and S mass fractions determined using ICP-OES, N_A is the Avogadro constant, ρ_GNRs is the density of gold, M_S is the molar mass of S, V_GNRs and S_GNRs are the volume and surface of GNRs, respectively. Data were expressed as a mean of two independent functionalizations of GNRs measured in the triplicate.

Cellular Uptake of GNRs Determined by Confocal Microscopy and Fluorescence-Activated Cell Sorting (FACS)

HeLa cells were seeded onto coverslips or 6-well plates (4×10⁴ cells per coverslip and 3×10⁵ cells per well) and allowed to attach overnight (16 hours). Then the cells were incubated with ^MTABGNRs or ^OEG+GNRs at a concentration 20 µM (Au⁰) diluted into serum-containing cell culture medium for 24 hours. Cells cultured in nanoparticle-free media were used as controls.

For microscopic analysis, the cells were washed 3× with phosphate buffered saline (PBS), fixed with 4% formaldehyde (15 minutes, r.t.) and permeabilized with 0.2% Triton X-100 (10 minutes). Washed cells were blocked with 3% BSA (30 minutes) and incubated with Alexa Fluor 488-conjugated phalloidin for F-actin staining (diluted 1:200 in PBS/3% BSA; 37°C; 1 hour in dark) or with anti-LAMP-1 primary antibody (diluted 1:250 in PBS/0.1% Tween-20; 1 hour), washed and stained by Alexa 488-conjugated secondary antibody (diluted 1:1,000 in PBS/0.1% Tween-20; 1 hour in the dark). After the final wash, the nuclei were stained by 4ʹ,6-diamidino-2-phenylindole (DAPI; 1 µg/mL) and coverslips were mounted with the ProLong mounting medium. The fluorescence and back-scattered laser light from the longitudinal LSPR mode of GNRs was detected with Leica TCS SP8 AOBS confocal microscope (Leica Microsystems, Wetzlar, Germany) using the following excitation and detection wavelengths: DAPI (405/413 – 488 nm), Alexa Fluor 488 (498/510 – 560 nm) and GNRs (633/630 – 636 nm) as described previously.⁴⁸

For FACS analysis, the cells in 6-well plates were washed 3× with PBS, trypsinized with 0.25% trypsin/0.53 mM EDTA and after centrifugation (700 ×g; 3 minutes) fixed with 4% formaldehyde (15 minutes, r.t.). The cells were counted by BD LSR II (BD Biosciences, San Jose, CA, USA). The scattered light of the longitudinal LSPR mode of GNRs was detected using 633 nm laser in the side scatter mode of the cytometer at 635±10 nm, as described previously.⁴⁸ Data were expressed as a mean value and SD of triplicate experiments.

Evaluation of the Plasmonic Photothermal Stability of GNRs

^MTABGNRs or ^OEG+GNRs were deposited on the formvar coated TEM locator grid. The extra thick layer of formvar was resistant to applied fs-laser irradiation, the locator grid allowed to characterize the same area of the sample before and after the laser exposition on FE-SEM. Desired distribution of GNRs on the TEM grid was accomplished by depositing a droplet of GNRs solution of known concentration and subsequent control of the deposition time and air humidity. After the deposition process, TEM grids were carefully washed in ultrapure water and imaged on JSM-7500f JEOL FE-SEM. Laser exposition was performed on an upright two-photon excitation microscope (2PEM, Bruker Ultima IntraVital, Middleton, WI, USA; equipped with Coherent Chameleon Ultra I laser, Santa Clara, CA, USA) using the following layered setup: microscope slide – TEM grid – water droplet – cover glass – water droplet – objective (25× water immersion objective with a 1.1 numerical aperture). Samples were exposed to 140 fs laser pulses at 750 nm with a repetition rate of 80 MHz and peak fluence ranging from 14.8 mJ/cm² to 148.5 mJ/cm². After the laser irradiation, the TEM grid was left to dry out and the selected areas were again characterized on FE-SEM. The statistical data from more than 300 nanoparticles were extracted by software analysis of acquired micrographs.

Statistical Analysis

Student’s t-test was used to determine the statistical significance of arithmetic means’ differences among groups: not significant (ns), P<0.05 (*), P<0.01 (**), and P<0.001 (***).

Results

Organic Synthesis

The preparation of monoquaternary thiol with OEG side chain was based on commercially available symmetrical diols 1–3 with required length of chain (Figure 1A; for details see Supplementary Figure S1). Briefly, the desymmetrization reaction of two chemically equivalent hydroxyl groups was made in the first step by the introduction of the dimethoxytrityl group, which provided monoprotected alcohol intermediates 4–6. The subsequent step was the tosylation of the remaining free hydroxyl with following nucleophilic substitution of the tosyl group with potassium thioacetate to obtain intermediates 10–12. Further, thioacetates 13–15 were prepared by selective deprotection of dimethoxytrityl under acidic condition and consequent brominated thioacetates 16–18 via formation of triphenylphosphonium salt and following nucleophilic substitution. The quaternization of brominated thioacetates 16–18 were made by Menshutkin-like reaction with appropriate amine leading to quaternary thioacetates 19–21a–f. Deprotection of the thiol group was carried out in a mixture of anhydrous solvents (methanol, dichloromethane) by in situ generated hydrogen chloride. Synthesized QAS were obtained in 35–53% overall yields for final products 22–24a–e (Supplementary Table S1) after seven steps. For quinolinium based QAS (22–24f) the overall yields were 15–28%. The ligands were grouped into four series according to their side chain: ligands with ethylene glycol chain with different number of units, namely with three (OEG₃ series; cpd. 22a–f), four (OEG₄ series; cpd. 23a–f), and five (OEG₅ series; cpd. 24a–f) ethylene glycol units and the ligand with alkyl chain (MTAB) (Figure 1B and C).

Figure 1 Retrosynthesis and structure of prepared cationic compounds for functionalization of GNRs (n, number of ethylene glycol units). (A) Retrosynthetic scheme of preparation quaternary ammonium compounds based on oligoethylene glycol chain. (B) Chemical structures of utilized compounds 22–24a–f and MTAB tethered to the surface of GNRs via S-Au bond and of CTAB non-covalently bonded to the GNRs surface. (C) Scheme of CTAB bilayer compared to monolayer of MTAB and OEG-ligands stabilizing the GNRs.

Hydrophobicity and Molecular Electrostatic Potential of Prepared Compounds

As the organic compounds are an important constituent part of the ligand-GNRs complex that can modify nanoparticle-cell interactions, cytotoxicity of NPs and their health-related applications,^55–57 several theoretical and experimental molecular descriptors including the hydrophobicity and molecular electrostatic potential (ESP) of free compounds were assessed. Hydrophobicity supports transmembrane penetration, while the hydrophilicity provides the solubility in aqueous environment.^58–60 In addition, cytotoxicity of QAS is closely related to their hydrophobicity and especially the ability to incorporate the side chain of QAS compounds into phospholipid bilayer membrane.^61,62 The hydrophobicity of final products was expressed as ClogP calculated in Open Babel 2.3.2 software. The relationship between ClogP and cytotoxic potential of compounds is shown in Figure 2A. The calculated data proved that the compound with an alkyl chain (MTAB) and its analogue without thiol moiety (CTAB) are more hydrophobic than compounds with OEG chain (see Supplementary Table S1 for values of ClogP). Moreover, the correlation of ClogP with experimental evaluation of hydrophobicity expressed as logk⁶³ is described in detail in the Supplementary Material (see Supplementary Figure S2A). For OEG compounds, the hydrophobicity slightly increased with the length of OEG chains. Based on different structures of the cation, the compounds with trimethylammonium head (22–24a) followed by pyridinium salts (22–24b) showed the lowest hydrophobicity. Increased and comparable hydrophobicity was found for quinolinium (22–24f), isoquinolinium (22–24e), and benzalkonium (22–24d) salts. The most hydrophobic compounds were the phenylpyridinium salts (22–24c).

Figure 2 Hydrophobicity, models of molecular electrostatic potential and cytotoxicity of prepared compounds in free state. (A) Correlation between cytotoxicity (logIC50) and calculated hydrophobicity (ClogP) of cationic compounds with OEG chain (22–24a–f) and alkyl chain (MTAB, CTAB). (B) Electrostatic potential maps and ESP atomic partial charges on the quaternary nitrogen (NESP) of selected compounds determined in Spartan 14 molecular modeling and computational chemistry software.

The interaction of NPs with proteins, the rate of their cellular uptake as well as their organ and sub-organ distribution strongly depend on the surface charge of NPs.^24,64,65 Therefore, the ESP mapped on the frontier electron isodensity surface (ie, 0.002 e/b³, covering 99.1% of the total electron density) and the ESP atomic partial charge projected on the atomic nucleus on isodensity surface of the quaternary nitrogen were calculated in Spartan 14 software by semi-empirical quantum chemistry PM6 methods (see Figure 2B for selected compounds, Supplementary Table S1 and Supplementary Figure S3A for all final products and Supplementary Figure S3B for additional computational models obtained by conformational analyses on the semi-empirical method RM1).^61,63 As expected, the OEG chain exhibited higher fluctuation of ESP than the alkyl chain due to the presence of oxygen, which caused variation in electron distribution along the chain. However, the disparity between the OEG chain with a different number of ethylene glycol units (22a and 24a) and alkyl chain (MTAB) did not exhibit a significant influence on the charge of quaternary nitrogen. In contrast, the introduction of quaternary ammonium nitrogen in the heteroaromatic system or solely aliphatic substitution exhibited substantial influence on the nitrogen ESP partial charge. Apparently, the nitrogen with the methyl substituent, namely trimethylammonium (24a) and benzalkonium (24d) salt, exhibited the highest partial ESP charge due to the electron-donating effect of the methyl group. In the case of compounds with quaternary nitrogen in the aromatic system, the partial charge was distributed through the aromatic ring.

Functionalization of GNRs by OEG Compounds and Evaluation of Colloidal Stability and Ligand Packing Density of Cationic GNRs

The GNRs coated by novel ligand shell (^OEG+GNRs) were prepared by a ligand exchange of ^CTABGNRs (56.3±4.9 nm in length and 28.2±3.9 nm in width; see Supplementary Figure S4A and C for size distribution of GNRs and their UV-Vis-NIR spectra, respectively) according to our protocol published previously.^46,48 The presence of silver, which is a common reagent used in synthesis of GNRs, in ^OEG+GNRs was measured by elementary analysis (ICP-OES) and was approx. 2% (Ag/Au). The dimensions of GNRs were designed to obtain the nanorods with light absorption in the red region of VIS (633 nm) to enable visualization and determination of their behavior in living cells by commonly available equipment (confocal microscopes and fluorescence-activated cell sorter). The functionalization of GNRs by cationic OEG ligands allowed to prepare stable colloidal GNRs without affecting their shape, as was confirmed by measurements of UV-VIS-NIR absorption spectra. The typical absorption spectrum of GNRs consisted of the transverse (λ_max ~ 520 nm) and longitudinal (λ_max ~ 610–630 nm) absorption peaks (Figure 3A). Nevertheless, the ligand exchange was not successful for ligands 23a, 23c, 23f, and 24f. In this case, UV-VIS-NIR absorption spectra measured after functionalization of GNRs showed that GNRs tend to aggregate together as demonstrated by extensive change of the GNRs absorption spectrum. The decrease in the ratio of longitudinal and transverse absorption peak intensity accompanied by duplication/broadening and shifting of the longitudinal peak to longer wavelengths was observed. ^23a,c,fGNRs and ^24fGNRs were thus excluded from further biological evaluation.

Figure 3 Colloidal stability and ligand packing density of GNRs stabilized by cationic OEG compounds. (A) UV-VIS-NIR spectra of GNRs modified by OEG compounds with 3, 4, and 5 ethylene glycol units (OEG3, OEG4, and OEG5 series) and various quaternary ammonium heads (a–f) after surface ligand exchange in storage solution (water) normalized to 50 µM (Au0) concentration and (B) dispersed in 10% FBS/DMEM at 50 µM (Au0) concentration. (C) Quantification of ligand packing density on GNRs using ICP-OES.

The good colloidal stability of GNRs in biological media containing salts, low molecular compounds and biopolymers is a prerequisite for their further biological utilization. Therefore, the ability of cationic OEG ligands to stabilize GNRs in colloidal solution was assessed in the cell culture medium. The water stable ^OEG+GNRs preparations were diluted to DMEM supplemented with 10% FBS (Figure 3B). Except MTAB analogue ligand 22a that underwent aggregation, the GNRs modified by OEG₃ series showed similar colloidal stability in cell culture media as in water solution. In contrast, the OEG₄ and OEG₅ series provided less stable colloidal nanoparticles as indicated by changes of GNRs absorption spectra in most cases from slight reshaping of absorption peaks up to the fusion of transverse and longitudinal peaks and their extension to longer wavelengths demonstrating extensive GNRs aggregation (namely ligands 23b in OEG₄ and 24a and 24b in OEG₅ series). The colloidal stability of cationic GNRs according to the side chain of compounds decreased in the following order: MTAB alkyl chain > OEG₃ chain > OEG₄/OEG₅ chains (for absorption spectra of ^MTABGNRs, see Supplementary Figure S5A). According to the divergent structure of cationic heads, the best colloidal stability was obtained for the benzalkonium (d) and isoquinolinium (e) ligands that showed in all tested lengths of OEG chain a sufficient stability demonstrated by a stable spectrum in the cell culture medium (OEG₃ and OEG₅ series for quaternary ammonium head d; OEG₃ and OEG₄ series for e) or the altered spectrum still consisting of two distinctive plasmonic peaks (OEG₄ series for d; OEG₅ for e).

The packing density of OEG-coated GNRs was calculated based on the ratio of the mass fraction of sulfur (S) and gold (Au) measured simultaneously by ICP-OES,⁵³ where the gold-to-sulfur ratio depends proportionally on the volume-to-surface area ratio.⁵⁴ Assuming complete removal of free ligand by centrifugation during sample preparation, the ligand coverage was estimated for all ^OEG+GNRs, including unstable colloidal dispersion of GNRs (23a, 23c, 23f, and 24f; Figure 3C). Importantly, the ligands that did not allow us to prepare stable GNRs in colloidal solution after the ligand exchange, reached less than half the ligand packing density compared to MTAB providing the highly stable colloid of GNRs. The different length of OEG chain did not have a significant influence on coverage of GNRs. However, the different packing density was conspicuous in comparison of each terminal cationic group. The more effective GNRs coverage exhibited the OEG ligands with benzalkonium salt (d) that in the combination with five ethylene glycol units (OEG₅; 24d) reached a similar packing density (6.10±1.04 molecules/nm²) to MTAB (6.30±0.72 molecules/nm²).

Assessment of Cytotoxicity of Cationic Ligands in Free State and in Complex with GNRs

Previously, we showed that the cationic alkanethiol-(MTAB)-stabilized GNRs are non-toxic both in vitro and in vivo.⁶⁶ Similarly, no significant decrease in viability of HeLa cells was observed for all ^OEG+GNRs tested at a concentration range from 30–50 µM (Au⁰) (Supplementary Figure S6). However, the possibility of ligand release from the surface of GNRs motivated us to evaluate the cytotoxicity of free compounds by colorimetric MTT assay as a half maximal inhibitory concentration (IC₅₀).⁶⁷ CHO-K1 cells previously established for the evaluation of cellular toxicity of QAS^63,68 were used for the cell viability assay. As we showed recently,⁴⁶ the compound with alkyl chain (MTAB) showed more than two orders of magnitude higher cytotoxicity than its analog with OEG chain of similar length (24a). This correlated with higher hydrophobicity of alkanethiols compared to the OEG compound (see Figure 2A and Supplementary Table S1). Comparing the cytotoxicity of each quaternary ammonium head, compounds containing trimethylammonium group exhibited in the chain with four and five ethylene glycol units (23–24a) almost one order of magnitude lower in cytotoxicity compared to other OEG compounds (Supplementary Table S1). Additionally, the increasing length of chain of the aromatic compounds and the compound with aromatic side chain (b–f) led to slightly higher cytotoxicity (Supplementary Figure S2B), but still nearly two orders of magnitude better to MTAB. This finding leads to the conclusion that compounds with OEG chain are, in general, less toxic and more suitable for usage in living cells than the compounds with alkyl chain.

Estimation of Cellular Uptake of ^OEG+GNRs

The cellular uptake of GNRs modified by different ligands was evaluated in HeLa cells using confocal microscopy at a 20 µM (Au⁰) GNRs concentration. The experiment confirmed that the ^OEG+GNRs are effectively absorbed by cells and localized into the perinuclear space (Figure 4A). Moreover, for all cationic GNRs, transport of GNRs into lysosomal compartment was confirmed by a co-association with lysosomal marker LAMP-1 (Supplementary Figure S7). Next, the GNRs level was semi-quantitatively estimated by fluorescence-activated cell sorting (FACS; Figure 4B). The median of scattered intensity of ^OEG+GNRs taken up by the cell was normalized to the signal of ^MTABGNRs known for their highly efficient cellular uptake.^47,48 In line with microscopic analyses, the FACS results showed a trend to increase the cellular absorption of ^OEG+GNRs depending on the length of the side chain of cationic ligand, where the ligands with four (OEG₄) and five (OEG₅) ethylene glycol units in the chain were taken up more effectively compared to ligands with three ethylene glycol units in the chain (OEG₃). In addition, the level of OEG₄ and OEG₅ series ligands reached or exceeded the intensity of alkyl chain ligand (MTAB). Nevertheless, any correlation of cellular uptake in dependency on the different cationic ligand head was not observed.

Figure 4 Efficiency of cellular uptake and intracellular localization of GNRs stabilized by QAS. (A) Cellular uptake of various cationic GNRs modified by ligands with OEG (OEG3, OEG4, and OEG5 series) chain and different terminal QAS group (a–f) and by QAS-ligand with alkyl chain (MTAB) determined by confocal microscopy in HeLa cells incubated with 20 µM (Au0) GNRs for 24 hours (a back-scattered light from the longitudinal LSPR mode of GNRs was utilized to visualize GNRs; the actin and nuclei were stained by phalloidin and DAPI, respectively; whole cells were scanned as a series of z-stacks) and (B) semi-quantified by FACS in HeLa cells after 24 h-incubation with 20 µM (Au0) GNRs (the median of scattered intensity of OEG+GNRs was normalized to MTABGNRs). Bar, 10 µm; Student’s t-test: P<0.05 (*), P<0.01 (**), and P<0.001 (***).

Estimation of the Photothermal Stability of Differently Coated Cationic GNRs

To assess the effect of the ligand structure on the photothermal properties of GNRs, we used synthesized GNRs tuned to near-infrared (NIR) region (81.5±24.2 nm in length and 27.9±3.4 nm in width; see Supplementary Figure S4B and D for size distribution of GNRs and their UV-Vis-NIR spectra, respectively), where tissues exhibit lower absorption.^4,69 As the results obtained by FACS analysis showed that the uptake of GNRs by cells increases with the length of OEG chain, the ligands from the series with the highest number of ethylene glycol units (OEG₅) were selected for subsequent evaluation of photothermal stability. From OEG₅ series, ligands providing the highest surface coverage density and satisfactory colloidal stability of GNRs both in water and cell culture medium (24c and 24d) were evaluated as the most promising and selected as the representative samples. 24c- and 24d-coated GNRs (see Supplementary Figure S5B for UV-Vis-NIR spectra of ligand-modified GNRs) were deposited on TEM grids and then were irradiated by 140 fs pulses with 80 MHz repetition rate, central wavelength of 750 nm, and peak fluence set to 14.8 mJ/cm², 45.0 mJ/cm², and 142.5 mJ/cm² on two-photon excitation microscope. The sets of samples were characterized prior to and after the laser irradiation on FE-SEM using specialized particle shape recognition software package⁷⁰ (see Supplementary Figures S8 and S9). Morphological stability of metal nanoparticles of ^24cGNR and ^24dGNRs was compared to ^MTABGNRs. As revealed in Table 1, the highest fluence (142.5 mJ/cm²) caused reshaping of all GNRs regardless of their solvation layer due to transient accumulation of heat in or around GNRs. In all cases, reshaping of GNRs resulted in a decrease of rod length and their aspect ratio (Figure 5A). However, for 24d-coated GNRs, the fluence of 14.8 mJ/cm² triggered already morphological changes of GNRs as indicated by a decrease of GNRs aspect ratio from 2.7±0.5 to 2.1±0.4 and a decrease of GNRs length from 28.6±4.0 to 23.8±4.2. Similarly, the photothermal stability of 24c-coated GNRs saturated already at 45.0 mJ/cm² (see Figure 5B), when aspect ratio of GNRs decreased from 2.7±0.5 to 1.6±0.5 and rods length from 29.1±4.0 to 20.8±4.4. Under femtosecond laser irradiation, the reshaping of GNRs is competitive to the thermal diffusion across the shell layer of NPs.⁷¹ Our results thus indicate lower thermal stability and lower thermal transport into the surrounding environment from heated ^OEG+GNRs in comparison to alkanethiol-stabilized ^MTABGNRs.

Table 1 Morphological Stability of Different Types of GNRs after Laser Irradiation (Fpeak, Peak Fluence); Analyzed > 300 NPs/Group

Figure 5 Photothermal stability of GNRs coated by selected ligands. (A) Representative FE-SEM images of MTAB, 24c-and 24d-coated GNRs (tuned to NIR region) before and after two-photon irradiation by 140 fs pulses with 80 MHz repetition rate at wavelength of 750 nm and peak fluence (FPEAK) set to 14.8 mJ/cm2, 45.0 mJ/cm2, and 142.5 mJ/cm2 (bar, 250 nm). (B) Tukey box plot showing median of aspect ratio of GNRs (solid line inside the box) before and after the laser irradiation for MTAB, 24c and 24d and three values of laser peak fluence. Whiskers extend to outliers up to 1.5-times the interquartile range and further outliers are marked as dots above boxes; Student’s t-test: not significant (ns), P<0.01 (**), and P<0.001 (***).

Discussion

Rationale and Design of Quaternary Ammonium Salts with OEG Chain

Due to high application potential of GNPs in clinical medicine,^11,72–74 in this study we prepared various OEG compounds containing the positive charge in the terminal group. Our aim was to improve the cellular uptake of GNRs compared to PEGylated GNRs and to decrease the toxicity of free surfactant compared to alkanethiols. Moreover, it was hypothesized that the ligands with the OEG-chains will provide a less compact passivating layer due to their ability to absorb molecules of water, which could lead to conveniently modified photothermal characteristics.^15,75,76 Vigderman et al⁴⁷ previously demonstrated that cationic alkanethiol ligand MTAB of the same length of alkane chain as a CTAB used to cover the GNRs during synthesis allows the complete ligand exchange and highly increases the colloidal stability of GNRs. Therefore, here we designed the analogues of MTAB (C16) with OEG chain considering the divergent bond lengths, bond angles, and spatial arrangement of atoms in molecules. The OEG/PEG can adopt helical or all-trans conformation and ratio between these two forms is a function of many factors. The attachment of (EG)n via sulfur group on flat gold surface allows both conformation such as for n≥6 helical and for n=4 a mix of helical and all-trans conformations.^77,78 Based on that, OEG compounds with three different lengths of chain (22–24) and with six various quaternary ammonium heads were prepared: QAS with sole aliphatic substitution (quaternary heads a, d) or as a part of heteroaromatic system (quaternary heads b, c, e, f). The various properties of QAS such as the different conformation of OEG chain, ESP charge distribution and hydrophobicity could have an important influence on the ligand cytotoxicity and its packing density on GNRs and thus on the GNRs colloidal stability, cellular uptake, as well as on the photothermal properties of GNRs.

Cytotoxicity of Cationic OEG Ligands

The utilization of thiol-containing compounds as the ligand for GNRs stabilization may also exhibit some disadvantages. Thiol-containing small molecules such as dithiothreitol or amino acids such as methionine and cysteine at physiological concentrations can cause release of the thiol ligand and subsequently bind to vacant adsorption sites on the nanoparticle surface.^15,79,80 In contrast to the formerly accepted view, recent research demonstrating that the gold-sulfur coupling in gold–thiol self-assembled monolayers (SAMs) has a physisorbed rather than chemisorbed (or covalent) character supports the possibility of ligand release.^81,82 Moreover, a ligand desorption can be caused by heating⁸³ and by an intensive fs-laser irradiation of GNPs, leading to their reshaping and melting that trigger the breaking of the S-Au bond.^84,85 In the blood stream, the released free ligands can then cause toxicity or induce tissue pathology.^86–89 Despite the GNRs being capped with cationic alkanethiol (MTAB)^47,66 as well as OEG ligands being evaluated non-cytotoxic, the excessive ligand release during the exalted or repeated GNRs dose or photothermal therapy may increase the probability of free ligand toxicity in vivo. For this reason, the cell viability assay of free compounds was performed on non-cancerous CHO-K1 cells. In general, the cytotoxicity of quaternary ammonium compounds increases with their increasing hydrophobicity, mainly attributable to their efficient incorporation into cell phospholipid membrane and formation of mixed micelles with membrane constituents causing the membrane ruptures and leakage of cytosol.^49,90,91 Our results showed that all synthesized QAS with OEG chain exhibit almost two orders of magnitude lower cytotoxicity compared to conventional QAS with the alkyl chain, which is in agreement with lower hydrophobicity of OEG compounds. Similarly, OEG compounds containing aliphatic trimethylammonium group that showed the lowest hydrophobicity of the QAS heads also exhibited nearly one order of magnitude lower toxicity against cells when combined with the chain of four and five ethylene glycol units (OEG₄ and OEG₅). The introduction of aromatic system led to a slight increase of cytotoxicity when the cytotoxic effect was endorsed by the prolongation of OEG chain, which is also in agreement with the hydrophobic descriptors (Supplementary Table S1 and Supplementary Figure S2B).

Colloidal Stability Studies of GNRs Modified by Cationic OEG Compounds

Agglomeration/aggregation may dramatically affect nanoparticle properties which can mediate their biological effects.^92,93 This underlines the importance of evaluation of colloidal stability in relevant media such as water or cell culture medium. Although most prepared OEG ligands provided stable colloidal dispersion of GNRs in the water, ^OEG+GNRs tended to aggregate in the serum-containing cell culture medium. It is known that the blood serum, a cocktail of proteins, is amenable to unguided attachment onto NPs and is responsible for the colloidal stability or anti-aggregation effect.^48,94 Thus the ability of OEG chain to reduce protein adsorption on the surface of NPs¹⁴ induces likely lower GNRs colloidal stability in the cell culture medium where the high content of salts promotes their aggregation.⁹⁵

In addition, the ligand packing density on NPs plays an important role.^96,97 In agreement with a previous report showing that the grafting densities of high molecular weight PEG on GNP are significantly lower than those of thiols⁸⁰ we observed lower GNRs coverage also by most of OEG compared to ligands with alkyl chain of similar length. Lower number of cationic nitrogen groups bound to the surface of GNRs that provides the electrostatic repulsion between NPs may result in lower colloidal stability of GNRs in aqueous environment.¹⁸ In relation to the different structure of cationic head, the benzalkonium salts (22–24d) that provided good colloidal stability of GNRs in the cell culture medium also showed the highest packing density on GNRs. Surprisingly, the different length of OEG chain did not show any significant differences in ligand coverage of GNRs, while the colloidal stability of GNRs decreased with the number of ethylene glycol units. Likely the structural diversity of OEG compounds itself may lead to different orderliness of organic shell on GNRs surface due to different intermolecular forces and thus to diverse tendency to form aggregates in aqueous environment. The differing structure of quaternary ammonium head led to different inter- and intra-molecular interaction, such as electron-donating/withdrawing effects of substituents, aromatic π–π interactions or cation–π interactions.^97–99 The bulky positively charged head groups can control ligand density via volume of the terminal groups and repulsive forces between them.^47,97,100 For example, ligand without aromatic moiety 24a exhibited almost half coverage of analogous ligand with aromatic ring 24d despite both substituents having similar N_ESP and thus should be presumed to have similar repulsion of cationic parts. Moreover, the ligand 24d can occupy larger space due to the presence of aromatic ring. The introduction of aromatic ring itself can therefore contribute to colloidal stabilization of the ligand packed-GNRs and the flexible linkers can help to avoid a jam of head groups by slight alteration of their conformations.¹⁰⁰

Cellular Uptake and Intracellular Localization of Cationic OEG-Modified GNR

The cellular uptake of GNRs was firstly studied by confocal microscopy including the determination of GNRs localization and subsequently quantified by FACS. Similarly to ^MTABGNRs,^{48 OEG+}GNRs adhered onto the cell plasma membrane were engulfed and trafficked to lysosomes. A semi-quantitative estimation revealed that the cellular uptake of ^OEG+GNRs increases with the length of OEG chain and thus with increasing hydrophobicity of the ligand. This is consistent with a study by Lorenz et al,¹⁰¹ demonstrating that with the increasing hydrophobicity of a linear alkyl side chain the absorption of poly(alkyl methacrylate) nanoparticles by HeLa, primary human bone marrow-derived MSCs, KG1a, and Jurkat cells significantly increases. However, no correlation was observed in relation to differences in cationic heads that also provide a different degree of hydrophobicity. It has been shown that cationic surface charge enhances the cellular absorption of NPs compared to anionic and neutral charge.^24,94 However, the initial surface charge may not be a simple predictor of nanoparticle uptake. Against expectation, the methyl-containing compounds that exhibited the highest EPS partial charge on the quaternary nitrogen did not show the highest level of cellular absorption in all tested series. On the other hand, the cellular uptake of ^OEG+GNRs correlated with decreasing stability of GNRs in the cell culture medium and with increasing formation of nanoparticle aggregates. Previously, we showed that the aggregated nanorods are endocytosed via macropinocytosis, while clathrin-mediated endocytosis (CME) is a preferential mechanism for individual nanorods.⁴⁸ Therefore, the enhanced tendency to form aggregates may increase the amount of internalized GNRs on behalf of macropinocytosis, that is the dominant endocytic pathway for NPs.^48,102 Thus, the affinity to form aggregates was probably a major factor determining the mechanism and level of cellular uptake of ^OEG+GNRs.

Thermal Stability and the Heat Transport from GNRs during Exposure to Femtosecond Laser Irradiation

The thermal therapy by GNPs promises progress in non-invasive cancer treatment.^103,104 Upon interaction of GNPs with the laser pulses, the rate of heat dissipation from the hot metal particle to the surrounding media determines the efficiency of the thermal killing of cancer cells. In case of metal NPs coated with water soluble surfactant, the effective thermal conductance (G_eff) includes the thermal conductance of the metal/surfactant interface, the intrinsic thermal conductance of the surface molecules, and the thermal conductance of the surfactant/water interface.⁴⁴ Previously, the transient absorption measurements as well as computer simulations were applied to study the heat transfer from GNPs to their surroundings.^39,40,43,44 Here, we utilized the method evaluating the photothermal stability of GNRs under the femtosecond laser irradiation. Particle cooling depends on the metal/surfactant interface thermal conductance. The excessive accumulation of heat energy may result in NPs fragmentation or changes in their size and shape.^30,34,71 This approach allows to determine a risk of secondary cytotoxic and genotoxic effects of “reshaped” NPs on cells surviving the laser irradiation and neighbouring/unexposed cells.

According to the surface ligand properties, several studies reported higher thermal conductivity for ligands with higher hydrophilicity that brings associated water molecules near the GNRs surface.^40,43,44 As the direct contact between water molecules and gold provides an additional thermal path, the thermal conductance of the Au/hydrophilic PEG (^~23 ethylene glycol units) interface comprises G_Au-PEG and G_Au-Water components. While the calculated value of G_Au-PEG is comparable to the Au/lipophilic alkanedithiol interfacial thermal conductance,^105,106 the additional thermal path between Au and water (G_Au-Water) increased the overall thermal conductivity of the GNR-PEG system about half.⁴⁴ Besides the hydrophilicity, the type of solvent,¹⁰⁷ bond strength between the Au and surface ligand⁴⁴ as well as ligand packing density^40,108 determine the efficiency of the heat transport from the GNPs. In case of CTAB-stabilized GNRs, the thermal interface conductance decreased with the increasing concentration of free hydrophobic CTAB.¹⁰⁸ In contrast, for hydrophilic molecules, the thermal conductivity of the GNRs polyelectrolyte layer increased with the increasing number of poly(acrylic acid) (PAA) layers and slightly decreased with increasing number of polyallylamine hydrochloride (PAH) layers.⁴⁰ Our results showed that despite less hydrophobic OEG-chain, the cationic OEG-GNRs lost the photothermal stability at lower laser fluence and thus heat transport to the surrounding was less effective than for similar cationic alkanethiol-GNRs. The difference in ligand packing density could lead to a discrepancy in the rate of heat dissipation. However, the OEG ligands used for evaluation of photothermal stability reached nearly analogous ligand packing density as MTAB. This indicates that also other factors that reflect the divergent chemical structures of MTAB and OEG compounds such as different assembly kinetics of thiolated ligands and orderliness of the formed ligand shell may influence the photothermal properties of GNRs.¹⁰⁹

In terms of thermal effectiveness, the OEG-GNRs are thus suitable for photothermal therapy to induce cancer cell death by programmed cell death pathway – apoptosis, with avoidance of necrosis.¹⁰⁴ Upon these settings, the cells are heated only to a temperature around 42–43°C.¹¹⁰ However, lower heat dissipation and higher risk of NPs fragmentation in applications requiring irradiation by high laser intensity such as tissue micro-surgery^111,112 may lead to lower effectiveness and cytotoxic and genotoxic damage of nanoparticle remnants.

Conclusion

In conclusion, the physico-chemical and biological properties of 18 newly-designed and synthesized cationic OEG compounds have been examined as ligand shells of GNRs and compared to each other as well as to their alkyl analogue – MTAB (summarized in Table 2). Our results showed that while the ability of cationic OEG ligands to stabilize the colloidal dispersion of GNRs in water and in serum-containing cell culture medium decreases with length of OEG chain, the cellular uptake of GNRs by cancer cells increases. The most promising QAS group appears to be the benzalkonium moiety that allowed to prepare GNRs with sufficient colloidal stability, high ligand packing density and reasonable cellular uptake in combination with all tested chain lengths. In comparison to the alkyl-QAS ligand, all OEG compounds exhibited nearly two orders of magnitude lower cytotoxicity in free state. Despite of the aggregation in cell culture medium, the series of OEG ligands with 4 and 5 ethylene glycol units in the chain revealed the comparable level of GNRs absorbed by cell as their alkyl analogue MTAB. The effect of ligand shell on the heat transport from GNRs under fs-laser irradiation was determined using a novel approach quantifying the reshaping of more than 300 GNRs by software analysis of FE-SEM micrographs that allows the assessment of biological risk of fragmented GNRs. However, the thermal stability of ^OEG+GNRs and thus the heat dissipation to the surrounding media was lower than for ^MTABGNRs. Based on these divergent properties, ^OEG+GNRs appear to be optimal for clinical applications requiring systemic administration of NPs/NPs treated cells such as drug delivery and photothermal therapy inducing apoptosis, while the alkanethiol-modified GNRs may be more suitable for local application with high intensity laser irradiation such as tissue micro-surgery.

Table 2 Summary of Physico-Chemical and Biological Properties of Prepared Cationic OEG Compounds Compared to Their Alkyl Analog (MTAB)