Highly Luminescent Positively Charged Quantum Dots Interacting with Proteins and Cells^†

Preparation and characterization of MUTAB-QDs

We prepared water-soluble MUTAB-QDs by replacing the organic shell of commercial, lipophilic CdSe/ZnS QDs with (11-mercaptoundecyl)-N,N,N-trimethylammonium bromide (MUTAB) in a one-step solvent phase transfer (Figure 1a), during which the MUTAB ligands attach to the QDs surface via covalent Zn-thiolate bonds. Thus, the hydrodynamic radius, R_H = 5.5 ± 0.2 nm, determined by dynamic light scattering (DLS) in phosphate buffered saline (PBS), pH 7.4, was significantly greater than the CdSe/ZnS core radius, R_c = 2.8 ± 0.4 nm, measured by transmission electron microscopy (TEM) (Figures 1b, c). A similar size increase was reported for MUTAB adsorption onto gold NPs.^[29] The absorption band was centered on 565 nm, and the luminescence peak was at ~600 nm in PBS (Figure 1d). Using Rhodamine 6G as a reference, the PL quantum efficiency was determined as 25 ± 2%. The high ζ-potential of +36 ± 2 mV due to the quaternary amine groups of MUTAB ensures that these QDs are electrostatically well stabilized in water; no aggregation was observed after storage for several months. Even in solvents with high ionic strength such as PBS, these QDs were stable over many hours, as indicated by the constancy of the hydrodynamic radius (Figure 1e).

Protonation-induced quenching of MUTAB-QDs

Thiolates are common ligands for coating QD surfaces to render them soluble and colloidally stable in hydrophobic or hydrophilic solvents.^{[30, 31]} As thiolate is the conjugate base of the weak thiol acid, it can easily be protonated, which destabilizes the covalent bond.^[32] This in turn promotes ligand desorption and generates QD surface defects. Trapping of charge carriers at defect sites is a well-known mechanism of QD luminescence quenching. Depending on pH, MUTAB carries a thiol or thiolate on one end; the quaternary amine group on the other end carries a positive charge regardless of pH. To analyze the effect of thiolate protonation, we investigated the PL and colloidal stability (via size determination) of MUTAB-QDs as a function of pH.

Images of vials containing MUTAB-QDs in 40 mM potassium phosphate/citrate buffer adjusted to pH 3—8, taken under 365-nm illumination, show that the emission intensity of MUTAB-QDs is high under basic conditions but sharply drops near pH 7 and is practically zero below (Figure 2a, inset). For quantification, we measured the pH dependence of the emission spectra after incubating the QD samples in buffer for 30 min (Figure 2b). Importantly, the brightness loss observed at lower pH cannot be recovered by raising the pH again (Figure 2b, dotted spectrum), indicating that low pH modifies the QDs irreversibly. The pH dependence of the (normalized) intensity, I(pH), shown in Figure 2a, was fitted with a Hill equation, I(pH) = (1–I₀)[1 + 10^n(pKa–pH)]^–1 + I₀, with (1–I₀) and I₀ representing the overall intensity step and the pH-independent background, respectively. The fit yielded pK_a = 7.2 ± 0.1 as the midpoint of the sigmoid curve. The Hill coefficient, n = 3.7 ± 0.1, which controls the steepness of the transition, is much greater than one, implying that a highly cooperative protonation reaction is responsible for the loss of luminescence. Data taken after 3 h show only a minimal reduction of the intensity (red symbols in Figure 2a). Thus, the MUTAB-QDs are stably luminescent for several hours at pH ≥7.

To examine if pH affects the colloidal stability of MUTAB-QDs, we measured the hydrodynamic radius, R_H, of the diffusing entities by DLS. R_H stayed constant at 5.5 nm between pH 6 and 8, i.e., in the pH range where the luminescence quenching step was observed. Thus, aggregation does not play a role in this process. Below pH 5, however, R_H started to increase enormously and reached values >400 nm at pH 3—4 (Figure 2c), revealing formation of huge QD aggregates. The pH dependence of R_H can be fitted with a Hill equation as well, R_H(pH) = ΔR_H[1 + 10^n(pH–pKa)]^–1 + R_H,0, with ΔR_H = 451 ± 1 nm and R_H,0 = 5.5 ± 0.1 nm representing the overall size increase due to aggregation and the bare MUTAB-QD size, respectively. The fit yielded a transition midpoint of the sigmoid curve, pK_a = 4.2 ± 0.1, and a cooperativity parameter, n = 2.9 ± 0.1. As the quaternary amine of MUTAB does not protonate, it ensures colloidal stability as long as there is a ligand shell around the QDs. Thus, we conclude that low pH causes depletion of the MUTAB layer and renders the QDs colloidally unstable.

Interactions of MUTAB-QDs with serum proteins

Protein adsorption and luminescence quenching. For many years, we have carried out elaborate in-vitro studies of the adsorption of model proteins and blood serum onto luminescent NPs using FCS, which allows us to monitor protein adsorption precisely via changes of the hydrodynamic radius, R_H.^{[24, 33-40]} In these experiments, the thickness of the adsorption layer was always found to vary with the protein concentration in a sigmoidal fashion, as expected from the Hill equation, and the adsorption process was complete within the time required for the measurement (ca. 3 min from sample mixing and loading till the end of data acquisition). By contrast, HSA adsorption onto MUTAB-QDs showed a surprisingly slow R_H increase in the range 1.56 μM ≤ [HSA] ≤ 50 μM (Figure 3a). Within the experimental error, the temporal evolution did not depend on [HSA], only the overall adlayer thickness, ΔR_H, grew with increasing [HSA]. Thus, we fitted the entire data set with exponential curves, taking the lifetime parameter of HSA adsorption, τ_ads, as a shared (global) fit variable, to obtain τ_ads = 1,460 ± 140 s (Figure 3a). The R_H change with concentration due to the HSA adsorption layer can be described with a relation based on a Langmuir-Hill binding equilibrium (Figure 3b), which has been widely applied to FCS and DLS data on protein corona formation.^{[24, 41, 42]} The fit yielded an overall radius increase of 3.7 ± 0.3 nm, an apparent equilibrium dissociation coefficient, K_D1 = 9.8 ± 1.7 μM, and a cooperativity parameter, n = 1.2 ± 0.2.

In the FCS experiments, we observed a pronounced decrease of the MUTAB-QD PL at greater [HSA], limiting the FCS measurements to below 50 μM. By using optical fluorescence spectroscopy with 470-nm excitation, we measured the temporal decay of the emission as a function of [HSA] (Figure 3c). With increasing [HSA], the quenching process accelerated and the residual fluorescence at long times decreased (Figure 2c). Notably, up to ~50 μM HSA, quenching did not proceed to completion. The decay curves are well described by single exponentials plus offset. The saturation levels follow a Langmuir-Hill isotherm, shown in Figure 3d as a red line, using the same parameters as those obtained from fitting the R_H change with concentration in Figure 3b. This implies that the extent of PL quenching depends on the degree of surface saturation with HSA. For [HSA] < 50 μM, the pH changes by less than 0.05 upon adding HSA to the solution (Figure S1), and thus does not affect the observed quenching behavior.

Close inspection of the data in Figure 3d indicates that the long-time residual PL (offsets) above 50 μM deviate from the Hill isotherm, which could hint at a different quenching behavior at higher [HSA]. Thus, we further measured the temporal decay of the MUTAB-QD luminescence up to 1,400 μM HSA (Figure 3e). Above 200 μM, we found essentially complete quenching on timescales shorter than τ_ads, the timescale of formation of a strongly bound HSA adlayer with K_D1 ≈ 10 μM. Thus, in this concentration regime, quenching precedes the HSA association process seen in the FCS data (Figure 3a). We were able to fit the kinetic data with single exponential decay functions; the obtained rate coefficients are plotted in Figure 3e (inset). Their dependence on [HSA] can be modeled by the expression k([HSA]) = k_q[1/(1 + (K_D2/[HSA])ⁿ)], which assumes that the observed kinetics are governed by an intrinsic, [HSA]-independent quenching rate coefficient times the degree of saturation of the MUTAB-QDs with HSA, as described by a Langmuir-Hill equation. Figure 3e (inset) shows that a fit of the data with this expression works very well, yielding a lifetime, τ_q = k_q^–1 = 150 ± 25 s, an apparent equilibrium dissociation coefficient, K_D2 = 700 ± 250 μM, and a cooperativity parameter, n = 1.0 ± 0.1. In Figure 3f, the degree of saturation with HSA is plotted against [HSA] together with the corresponding binding isotherm. Thus, at high [HSA], the quenching process reveals a 70-fold decreased HSA affinity. Although HSA binding is weak, it is still effective for PL quenching. We note that, at the highest [HSA] values, slight changes of the solution pH (Figure S1) may also affect quenching.

Dynamics of the protein corona following concentration jumps. As we have shown, HSA adsorption on MUTAB-QDs is a slow process for [HSA] ≤ 50 μM, with a lifetime of about 30 min (Figure 3a). The concentration dependence of the adsorbed HSA follows a Langmuir-Hill isotherm (Figure 3b), which is theoretically based on an underlying dynamic binding equilibrium. However, in earlier studies, we found that protein binding can nevertheless be irreversible, and the saturation curve reflects an equilibrium behavior before the corona turns irreversible due to structural dynamics at the NP-protein binding interface.^{[38, 39, 42]} To address these issues for MUTAB-QDs, we performed concentration jump experiments, in which [HSA] was quickly changed by a factor of four, and the ensuing response of R_H and the luminescence intensity were recorded.

Using FCS, we monitored the size increase due to HSA adsorption at 3.125 μM HSA over 4 h (Figure 4a). Subsequently, we quickly raised [HSA] to 12.5 μM and continued with the measurement. Comparison with the curve measured with 12.5 μM HSA from the beginning shows the same final corona thickness. This implies that the protein layer formed at 3.125 μM HSA does not passivate the QD surface, so that it can be loaded with additional protein molecules at higher [HSA]. Note the slow growth response in all steps. In another experiment, we first formed a protein corona at 6.25 μM HSA and then jumped [HSA] down to 1.56 μM. Surprisingly, R_H dropped to 5.5 ± 0.2 nm, which is less than the value measured for a protein corona formed at 1.56 μM and close to R_H of bare MUTAB-QDs (Figure 4b). This experiment strongly suggests that HSA adsorption in this concentration range led to removal of the MUTAB layer from the QDs, so that HSA desorption leads to a smaller size than expected.

Next, we performed the corresponding concentration jump experiments with monitoring of the PL intensity. We note that this procedure entails mixing of the MUTAB-QD solution at a certain [HSA] with another solution of different [HSA]. Therefore, the intensity must be rescaled to account for the dilution. For jumping the concentration up by a factor of four, the final extent of quenching was essentially identical, regardless of adsorption in one or two steps (Figure 4c). This observation suggests that the additional HSA binding inferred from the R_H increase gives rise to additional quenching, and that the final state depends on the HSA occupancy but not on the details of the HSA adlayer formation process, i.e., in one or two steps. For a four-fold reduction of the HSA concentration from 12.5 μM to 3.125 μM by PBS addition, the emission intensity of the MUTAB-QDs dropped immediately upon reducing [HSA] by ~25% and then stayed constant for several hours (Figure 4d). The fact that it did not at all recover to the values corresponding to the lower [HSA] in the mixture further supports our conclusion from the FCS experiment (Figure 4b) that the MUTAB-coating of the QD surface is irreversibly damaged upon tight (K_D1 ≈ 10 μM) HSA binding. There is even more quenching after HSA dilution than expected from the experiment at constant [HSA], suggesting a protective effect of HSA adsorption on the surface, which counteracts PL quenching.

Mechanistic interpretation of protein-MUTAB-QD interactions

The MUTAB layer around our QDs is primarily stabilized by Zn-thiolate covalent bonds and, additionally, by van-der-Waals interactions between the undecyl chains. The dense grafting of the ligands on the surface implies that the quaternary amine groups have a destabilizing effect due to charge repulsion. Unlike many other capping layers featuring carboxyl, amine, PEG or sugar moieties,^[43-45] quaternary amines expose hydrophobic methyl groups at the surface that cannot form hydrogen bonds at the interface. Thus, proteins can only bind to the QD via (weak) van-der-Waals and charge interactions, with charge stabilization relying on a net attractive interaction at the binding interface. Note that protein charges that are more distant from the binding interface are not relevant, as the high ionic strength of the protein solvent (PBS) ensures effective charge screening (Debye length < 1 nm).^{[27, 35, 46]} Taken together, it is not surprising that HSA binds only weakly to pristine MUTAB-QDs (K_D2 = 700 μM, Figure 3f), despite its net negative charge near neutral pH (isoelectric point, pI = 4.7).

We have inferred weak HSA binding from the quenching kinetics of the QD PL at high [HSA], which depends on the occupancy of the QDs with HSA. Such quenching has also been observed upon adsorption of polyacrylic acid (PAA) polymer rather than proteins (Figure S2a).^[10] HSA is known to provide an acidic environment^[47] and, therefore, we propose that the PL quenching effect originates mainly from a high local density of carboxylic acid sidechains on HSA; minor changes of the solution pH at high [HSA] may further amplify the effect (Figure S1). This hypothesis is supported by quenching studies with charge-modified proteins (Figures S2b—d). We have rendered carboxyl groups on HSA to amines by reacting them with ethylenediamine in the presence of the catalyst 1-ethyl-3-(3-dimethylamino-propyl)-carbodiimide (EDC). As expected, the quenching effect of aminated HSA (HSAam) was less than that of HSA (Figure S2b). By reacting the positively charged ε-amino groups of lysine sidechains of Tf with succinic anhydride,^[48] we turn them into negatively charged carboxyl moieties, so that their surface density is markedly increased. Indeed, unlike native Tf, succinylated Tf (Tfsuc) shows a pronounced quenching effect (Figure S2c). Thus, taken together, these experiments support the notion that luminescence quenching depends on the density of carboxylic acid groups on the protein. This raises the question if there are proteins other than HSA that show such a behavior. To address this question, we compared the quenching effect of HSA and human blood serum, which contains ca. 3700 different polypeptides including HSA as the most abundant protein,^[49] with a concentration of ~500 μM. Figure S2d shows that the decay kinetics look very similar for HSA solutions and serum diluted such that it has the same [HSA]. Consequently, other serum proteins appear to have only a minor if any quenching effect in comparison with the HSA component. Of course, the concentrations among proteins in serum differ greatly, and we cannot exclude that similar effects arise with proteins that remain unnoticed due to their low abundance.

The initial effect of HSA upon adsorption onto MUTAB-QDs likely involves proton injection into the MUTAB layer by HSA, converting thiolate groups to less stably bound thiols.^[32] In this process, characterized by a low HSA affinity (K_D2 ≈ 700 μM), surface defects are created that give rise to the observed quenching (Figure 3e). A tightly bound HSA corona develops only slowly (within roughly 30 min), however, as measured by the size increase due to HSA adsorption (Figure 3a), characterized by a greatly enhanced affinity (K_D1 ≈ 10 μM) for HSA binding and the accompanying quenching effect (Figures 3b, d). Thus, we suggest that, after proton injection into the MUTAB layer by HSA, less stably bound thiols slowly detach from the surface. This process prepares the ground for HSA binding with high affinity. The modified, less densely grafted MUTAB layer may still provide attractive interactions to carboxylate sidechains and reduce repulsion by positively charged amine groups. Possibly, the presence of bare QD surface patches can also enhance the HSA binding affinity. Experimental support for MUTAB removal from the capping layer comes from the concentration jump experiments (Figure 4). We note that, at [HSA] < 50 μM, quenching appears to saturate according to the HSA occupancy, suggesting that HSA binding protects against further quenching. Presumably, bound HSA prevents or at least hinders free HSA from further proton transfer. However, we cannot exclude that further quenching proceeds slowly at [HSA] < 50 μM but remains unnoticed on our timescale of several hours.

MUTAB-QDs interactions with live HeLa cells

The surface charge of nanomaterials is one of the key factors controlling their interaction with cells.^[50] Cell membranes are negatively charged and thus interact strongly with positively charged NPs.^[51] Accordingly, high cellular uptake efficiency is seen for positively charged NPs.^[52] Such cationic NPs often elicit cytotoxicity due to physical damage to the plasma membrane, e.g., by perforation, or they may interact with membrane proteins such as Ca²⁺ channels, causing an increase of intracellular Ca²⁺ levels.^[53] Therefore, we assessed the biocompatibility of MUTAB-QDs and their internalization by live HeLa cells, a popular cervical cancer cell line that is frequently used in this context.

A MTT assay showed no obvious adverse effects of MUTAB- QDs on the metabolic activity of HeLa cells for concentrations up to 10 nM (Figure S3). Beyond that level, however, the overall cell viability started to drop markedly. To visualize possible effects of the MUTAB-QDs on the integrity of the cell membrane, we loaded HeLa cells with the cell membrane-permeant calcium indicator Fluo-4 AM before adding the QD solutions. The emission intensity of this dye is a measure of the cytosolic Ca²⁺ concentration, which is low under normal conditions (typically ca. 100 nM^[54]). When the MUTAB-QDs were suspended in PBS with 0.9 mM Ca²⁺ (PBS/Ca²⁺), the Fluo-4 emission from the HeLa cells was observed to increase in a stepwise fashion after several minutes, indicative of heavy Ca²⁺ influx into the cytosol as a result of severe membrane damage (Figures 5a—c). Movie S1 and Figure S4 show both the green Fluo-4 and the red QD emission as a function of time. Frequently, we observed a subsequent intensity (and thus Ca²⁺ concentration) decrease, as cells responded with membrane repair and the action of Ca²⁺ pumps until the next major Ca²⁺ influx event occurred (Figure 5b). After 20 min, however, all cells were brightly fluorescent and therefore dead. Smaller and slower increases of the fluorescence were observed for MUTAB-QDs suspended in PBS without Ca²⁺, presumably due to the release of Ca²⁺ ions from internal stores of the cell in response to MUTAB- QD-induced damaging of the plasma membrane (Figure 5a—c). Notably, it is well known that elevated intracellular Ca²⁺ levels are the key trigger for processes involved in plasma membrane repair.^{[55, 56]} With plain PBS solvent, bright cellular fluorescence indicating cell damage was not observed (Figures 5a—c). We also tested dihydrolipoic acid-functionalized QDs (DHLA-QDs,^[57] 20 nM in PBS/Ca²⁺), which bind to QD surfaces in a bidentate fashion and expose negatively charged carboxylate moieties on the surface, but we found only a decrease of the background intensity due to photobleaching but no indication of cell damage (Figures 5a—c).

Next, we assessed the capacity of MUTAB-QDs to damage cell membranes in a setting that is closer to the physiological situation, for instance, injection of an NP-based diagnostic or therapeutic agent into the blood stream.^[18] Thus, we pre-incubated MUTAB- QDs with solutions of varying [HSA] for 10 min before adding the mixture to the HeLa cells. In the light of our in-vitro experiments, a significant fraction of HSA will coat the QD surfaces, and the reduced density of positive charges on the MUTAB-QD surface should eventually relieve cytotoxic effects.^[51] Figure 5d shows confocal images from an experiment over 40 min, in which HeLa cells were exposed to MUTAB-QDs incubated with 50 μM [HSA]. Unlike in Figure 5a, there were no bright and thus damaged cells after 10 or 30 min. At longer times, however, the fraction of bright cells increased. We analyzed the fraction of cells that displayed a sudden fluorescence intensity increase due to membrane damage within the first 30 min upon pre-incubation with [HSA] = 10, 50 and 100 μM (Figure 5e, Figure S5). These data clearly prove that HSA coating has a pronounced protective effect on the cells, and further confirm that surface-exposed, positively charged trimethylamine moieties of the MUTAB ligands are responsible for plasma membrane damage.