Ultrasmall, Bright, and Photostable Fluorescent Core–Shell Aluminosilicate Nanoparticles for Live-Cell Optical Super-Resolution Microscopy

Abstract

Stochastic optical reconstruction microscopy (STORM) is an optical super-resolution microscopy (SRM) technique that traditionally requires toxic and non-physiological imaging buffers and setups that are not conducive to live-cell studies. It is observed that ultrasmall (<10 nm) fluorescent core–shell aluminosilicate nanoparticles (aC’ dots) covalently encapsulating organic fluorophores enable STORM with a single excitation source and in a regular (non-toxic) imaging buffer. It is shown that fourfold coordinated aluminum is responsible for dye blinking, likely via photoinduced redox processes. It is demonstrated that this phenomenon is observed across different dye families leading to probes brighter and more photostable than the parent free dyes. Functionalization of aC’ dots with antibodies allows targeted fixed cell STORM imaging. Finally, aC’ dots enable live-cell STORM imaging providing quantitative measures of the size of intracellular vesicles and the number of particles per vesicle. The results suggest the emergence of a powerful ultrasmall, bright, and photostable optical SRM particle platform with characteristics relevant to clinical translation for the quantitative assessment of cellular structures and processes from live-cell imaging.

Single molecule localization based optical super-resolution microscopy (SRM) techniques, and in particular stochastic optical reconstruction microscopy (STORM), are powerful imaging tools to resolve structures below the diffraction limit of light, deepening our understanding of nanoscale interactions in chemical and biological systems.^[1–3] Imaging via STORM rests on stochastic blinking of organic fluorophores attached to their target structures such that only a subset of fluorophores is imaged with each frame of a time-lapse diffraction-limited fluorescence movie. Precise localizations of these sparsely emitting fluorophores in all frames are then reconstructed into a super-resolved STORM image. Optimal STORM probes include those with high brightness, high photostability, and low on-off duty cycle (ratio of “on” over “off” time, vide infra) in cases of high labeling density, as localization precision is proportional to the square root of brightness and low duty cycle stable probes are less likely to have overlapping point spread functions (PSFs).^[3–6] In practice, dye blinking is conventionally achieved by exciting the fluorophores with two different light sources in a complex STORM imaging buffer cocktail, consisting of, e.g., thiol compound beta-mercaptoethanol (βME) and an oxygen scavenging (OS) system.^[7] The OS system is typically a combination of glucose, glucose oxidase, and oxygen catalase. Fluorophore blinking is caused by the formation of reversible long-lived dark states through dye interactions, e.g., with a primary thiol, which can be recovered to its ground state when aided by a UV light source.^[8] The typical imaging medium uses a non-neutral buffer and thiols that are toxic to many cell types, however, which limits the application of STORM in long-term live-cell imaging.^[9,10] Moreover, the conventional requirement of two light sources/lasers for single-color STORM imaging puts an additional burden on the experiments and often poses technical challenges for non-experts. Nevertheless, STORM is considered an attractive SRM technique relative to other methods such as stimulated emission depletion (STED) as it does not require specialized microscopy systems and provides high resolving power.^[11]

Previous work on STORM systems has produced simpler and less toxic protocols, typically involving self-healing dyes as well as heavy-atom-containing quantum dots.^[12–14] Additionally, it has been shown that certain probes can be used for live-cell STORM, but only in specific organelles, such as the mitochondria, that have increased levels of thiol compound glutathione.^[15] The Tinnefeld group described an experimental strategy to induce blinking with only one excitation light source by using an imaging buffer that contains both oxidizing and reducing agents.^[16–21] Through stochastic photoinduced oxidation and/or reduction of the fluorophore, it is forced into a radical cationic or radical anionic dark state. The fluorophore can then be recovered back to its ground state via another photoinduced redox interaction. While these and other methods are marked improvements upon the classic STORM setup, and some have enabled live-cell super-resolution imaging, they still do not avert all of the complex light source and/or imaging buffer requirements as well as cytotoxicity issues that render live-cell imaging challenging.^[22,23]

We previously introduced a new class of amorphous quantum nanomaterials, in which the incorporation of different amounts of foreign atoms into the amorphous inorganic aluminosilicate glass of an ultrasmall aluminosilicate core and poly(ethylene glycol) (PEG) shell core–shell nanoparticle (aC’ dot) alters the quantum behavior of fluorescent dyes covalently encapsulated in the glassy core, leading to substantially altered macroscopic optical behavior.^[24,25] After introducing iodine as foreign atoms (iaC’ dots), illumination led to increased intersystem crossing (ISC) rates to triplet excited dye states via spin-orbit coupling, in turn resulting in improved reactive oxygen species (ROS) generation useful, e.g., in photodynamic therapy (PDT). When introducing thiol groups as foreign atoms (srC’ dots), illumination led to dye blinking, which enabled STORM based optical SRM. While the latter approach removed the requirement for βME in the imaging buffer, a buffer system with an OS system was still employed, as was a two-laser excitation system, all complicating applications in live-cell imaging.^[6]

Here we describe the surprising discovery that aC’ dots, without incorporation of thiol groups and in the complete absence of any particular imaging buffer and under simple illumination with a single light source, exhibit dye blinking enabling STORM based SRM. As we show, the blinking mechanism of aC’ dots is very different from that in previous thiol-containing srC’ dots. In the latter particles, under illumination thiol-dye adducts disrupt dye π-electron delocalization leading to dark states and associated blinking in the presence of an OS system, whereas such systems quench aC’ dot blinking (vide infra).^[24] In a range of cross experiments with free dyes and aluminum-free C’ dots we reveal that fourfold coordinated aluminum in the aluminosilicate matrix is responsible for dye blinking, likely via photoinduced redox processes. We demonstrate that dye blinking can be generalized over several dye families covering the visible to near-infrared (NIR) spectral range, leading to brighter and more photostable SRM probes relative to their parent free dyes. We finally show examples of applications of these SRM probes in fixed and live-cell imaging. The latter reveals access to a wealth of information on cellular structure and particle processing that can be extracted from quantitative assessment of live-cell SRM data sets. This includes measures of the size of intracellular vesicles confining nanoprobe motion and the number of particles per vesicle, i.e., detailed information about the way cells process such ultrasmall fluorescent core–shell silica nanoparticles highly relevant for their clinical translation, e.g., in the field of drug delivery in oncology.^[26,27]

NIR dye Cy5 encapsulating and PEGylated ultrasmall fluorescent core–shell silica nanoparticles with and without aluminum (PEG-Cy5-aC’ and PEG-Cy5-C’ dots, respectively) were synthesized in aqueous conditions in the presence or absence of aluminum-tri-sec-butoxide (ASB), respectively, as described previously (see Supporting Information).^[25,28,29] The molar ratio of Si:Al in the aC’ dot core is 92:8, according to energy-dispersive X-ray spectroscopy (EDS) measurements (Figure S1, Supporting Information), i.e., only slightly lower than what is expected from the molar ratio of precursors in the synthesis feed (Si:Al of 87:13). A molecular level rendering of the local environment of such an aluminosilicate core with this molar composition encapsulating Cy5 with its π-electron orbitals is shown in Figure 1a (Si white, O grey, Al blue), together with purified PEG-Cy5-aC’ dot characterization results (see Supporting Information). The Gaussian shaped gel permeation chromatography (GPC) elution profile (Figure 1b) revealed high product purity, while the transmission electron microscopy (TEM) images at different magnifications (Figure S1, Supporting Information) suggested a narrow particle size distribution. The high rigidity of the aluminosilicate core with full encapsulation of the Cy5 dye translated into a larger than 4-fold brightness enhancement over free Cy5 dye (Figure 1c) in UV–Vis absorption and emission spectroscopy analysis. This effect has been quantitatively analyzed in earlier static and dynamic spectroscopy studies and is due to an increase in radiative rate and decrease in non-radiative rate.^[30–32] As shown previously, this brightness enhancement relative to free dye carries over into photon emission enhancements during dye blinking, which is crucial for a potential probe for STORM for which localization precision is proportional to the square root of brightness.^[4–6] Confocal fluorescence correlation spectroscopy (FCS) measurements on the diffusing PEG-Cy5-aC’ dots yielded one-component autocorrelation functions (ACF) suggesting an average particle diameter of 4.6 nm and corroborating the narrow size dispersity reflected in the TEM results (Figure 1d). We determined an average of 2.5 Cy5 dyes per particle from the combination of FCS derived sample concentration and absorption spectroscopy results.^[30,32] Furthermore, photobleaching studies based on previously published methods (see Supporting Information) showed that Cy5 encapsulated in aC’ dots is 2.8 times more photostable than in free form (Figure 1e).^[33] Since we already showed in two separate previous studies that Al is fourfold coordinated in aC’ dots, replacing Si in the silica network, we did not repeat the same solid-state ²⁷Al nuclear magnetic resonance (NMR) spectroscopy experiments on aC’ dot samples here.^[24,25] Zeta potential measurements of aC’ dots showed a negative surface charge with an average of −11.9 mV as seen in Figure S2 in the Supporting Information. Characterization results on PEG-Cy5-C’ dots used as a reference in this study, i.e., without Al in the core, suggested a hydrodynamic diameter of 5.5 nm with 2.4 Cy5 dyes per particle (Figure S3, Supporting Information).

Figure 1.

a) Rendering of internal environment of aC’ dot encapsulating Cy5 (Si, white; O, gray; Al, blue). b) Normalized gel permeation chromatography (GPC) elution profile of PEG-Cy5-aC’ dots. c) Normalized and absorbance-matched absorbance (left) and emission (right) spectra of Cy5 dye and PEG-Cy5-aC’ dots in water. d) Confocal fluorescence correlation spectroscopy (FCS) autocorrelation curve with fit of PEG-Cy5-aC’ dots diffusing in water. e) Photobleaching experiments of Cy5 dye and PEG-Cy5-aC’ dots in water with single exponential fit. BE: brightness enhancement; SE: stability enhancement.

We next compared the optical properties of fluorescent C’ dots and aC’ dots to free parent Cy5 dye by single molecule imaging. We immobilized biotinylated Cy5 dye (Cy5-biotin), PEG-Cy5-C’ dots, and PEG-Cy5-aC’ dots on streptavidin-coated glass dishes as described elsewhere, with Hank’s balanced salt solution (HBSS) as the imaging buffer (see Supporting Information).^[24,34] HBSS is an aqueous buffer routinely used in live-cell imaging. The immobilized single emitters were imaged under different conditions by a total internal reflection fluorescence (TIRF) microscope. A red laser (λ_ex = 640 nm) was used to excite Cy5. In the presence of thiolated buffers (βME + OS cocktail), the red laser leads to Cy5-thiol adduct formation (dark state), while a UV laser (λ_ex = 405 nm) was used to dissociate these adducts as described above.^[35] The temporal fluorescence traces of individual dyes or particles across different conditions (Figure 2) were collected with an electron multiplying charge coupled device (EMCCD) camera for 500 s at 50 ms integration times. In Figure 2, traces with different colors represent different single fluorophore or particle localizations. The numbers depicted in green represent the measured average equilibrium duty cycles, i.e., the ratios of the “on” time over “total” time of the respective fluorescence traces at steady-state in the 200–400 s time window, as described by the Zhuang group.^[3] Probes with low duty cycles are easier to resolve, e.g., in STORM, as it is less likely for their emission PSFs to overlap in cases of high labeling density.

Figure 2.

a–j) Single-molecule fluorescence traces for Cy5-biotin dye (a–d), PEG-Cy5-C’ dots (e–h), and PEG-Cy5-aC’ dots (i,j) under different conditions (50 ms integration times). The lines with different colors represent traces from different emitters. Numbers in green represent the average equilibrium duty cycles. Red laser and UV laser signify a 640 nm and a 405 nm light source, respectively. βME: β-mercaptoethanol. OS: oxygen scavenging system. Al salt: dissolved sodium aluminate. HBSS was used as the imaging buffer in all cases.

Neither Cy5 dye nor PEG-Cy5-C’ dots (i.e., without Al) showed blinking in HBSS buffer when excited with only a red laser (Figure 2a,e). In stark contrast, under the same conditions PEG-Cy5-aC’ dots exhibited dye blinking behavior suitable for STORM (Figure 2i), with exceptional equilibrium duty cycles around 0.0003.^[3] Since the only difference between PEG-Cy5-C’ dots and PEG-Cy5-aC’ dots was the absence or presence, respectively, of aluminum, these results suggested that aluminum in the nanoparticle core plays a key role in encapsulated Cy5 dye blinking. While Cy5 dye or PEG-Cy5-C’ dots still did not blink when exposed to illumination from both red and UV lasers (Figure 2b,f), they did show blinking when excited by both lasers and in the presence of thiols and oxygen scavengers of the standard STORM buffer, albeit with higher duty cycles >0.001 (Figure 2c,g). This verified that both samples are capable of blinking under the right conditions. Corroborating that aluminum is the key factor in what makes Cy5 blink in aC’ dots, we observed that Cy5 dye and PEG-Cy5-C’ dots blink (Figure 2d,h) with duty cycles comparable to that of PEG-Cy5-aC’ dots, vide supra, when excited with only a red laser and the addition of 9 × 10⁻³ M of sodium aluminate into the HBSS buffer. We chose sodium aluminate because aluminum is also in a fourfold coordinated state.^[36] The pH of HBSS at this concentration of sodium aluminate is ≈8.7, but we confirmed that PEG-Cy5-aC’ dot blinking is not strongly dependent on pH or dye charge (Figures S4 and S5, Supporting Information). Furthermore, we exposed Cy5 dye to aluminum chloride in HBSS and saw no blinking with only red laser excitation (Figure S6, Supporting Information), suggesting that only certain coordination levels of aluminum are capable of inducing blinking. While it is possible that coordination states other than fourfold could lead to blinking, these experiments are beyond the scope of this present study.^[37] Finally, PEG-Cy5-aC’ dots did not blink under red laser excitation when oxygen scavengers were added to the HBSS imaging buffer (Figure 2j), suggesting that dissolved oxygen naturally present in aqueous solutions also plays a key part in PEG-Cy5-aC’ dot blinking.^[38] Blinking was also observed when PEG-Cy5-aC’ dots were imaged in ethanol, which also naturally contains dissolved oxygen (Figure S7, Supporting Information), demonstrating that an aqueous environment is not required for blinking to occur.^[39] These results together suggest that aC’ dots are optimal STORM probe candidates, in particular for live-cell imaging as they combine high brightness, high photostability, and low equilibrium duty cycle blinking when excited with a single laser source in a regular buffer, i.e., in the absence of cytotoxic imaging cocktails including thiol compounds and oxygen scavenging systems.^[3]

Interestingly, a small subset (≈5–10%) of aC’ dots exhibited early on-off blinking behavior near the beginning of data collection before fluorescence ceased, as shown in Figure S8a,b in the Supporting Information, likely associated with probes sticking to the substrate. Performing imaging FCS on such fluorescence time traces and fitting to a single exponential decay revealed an average off time (τ_off) of 94 ± 14 ms (Figure S8c, Supporting Information).^[40–42] Note that the particles are immobilized, hence any decay in the autocorrelation function is due to photophysical processes rather than diffusion. To obtain a more accurate τ_off, we switched from 50 to 1 ms integration times and performed imaging FCS on this subset of early-blinking particles. As shown in Figure 3e, we observed the same blinking behavior at this shorter integration time. Improved autocorrelation analysis revealed a similar τ_off of 100 ± 10 ms (Figure 3f). This is significantly longer than the triplet τ_off of Cy5 (≈15 ms), suggesting that PEG-Cy5-aC’ dots may undergo a radical-mediated redox-type blinking.^[17] Using higher time resolution (i.e., shorter integration times) for PEG-Cy5-aC’ dots exhibiting low equilibrium duty cycle blinking (as observed with 50 ms integration times, vide supra) revealed rapid on-off blinking during each effective blink (Figure S9, Supporting Information), similar to that described for the particle subset with early on-off blinking behavior. This suggested that the underlying photophysics was similar across all blinking phenomena observed for PEG-Cy5-aC’ dots.

Figure 3.

a–d) Single-molecule fluorescence traces (50 ms integration time) of Cy5-dye with sodium aluminate (a,b) as well as PEG-Cy5-aC’ dots (c,d) without (a,c) and with (b,d) radical scavenger, tert-butanol, suggesting that blinking occurs via a radical ion mechanism. e,h) Single-molecule fluorescence traces of immobilized PEG-Cy5-aC’ dots (top) and PEG-Cy5-C’ dots (bottom) in HBSS and exposed to 640 nm laser excitation (1 ms integration times). f,i) Imaging FCS autocorrelation of fluorescence traces with fits to single exponential functions. Only the first 10 s of (h) were autocorrelated in (i). Inset in (i) has a rescaled y axis to visualize entire data range. g,j) Suggested Jablonski diagrams depicting relevant dye electronic states and transitions for each system investigated.

After a dye absorbs a photon to an excited singlet state, it may perform one of the following actions: drop back to its singlet ground state via fluorescence, cross-over via intersystem crossing (ISC) to an excited triplet state, undergo redox interactions to a radical ionic dark state (in case of a neutral dye), or photobleach (see Jablonski diagram, Figure 3g).^[43,44] In the case of a redox mechanism, a (neutral) dye can either be oxidized into a radical cation dark state and return to its singlet ground state via a reduction step, or it can be reduced to a radical anion dark state and return to the singlet ground state via an oxidation step. However, oxidized dye radical cations typically represent a much higher energy state than reduced radical anions, and therefore are less thermodynamically favorable in the presence of redox agents.^[17] The fact that dissolved oxygen also appears to play an important role in the blinking behavior (vide supra) led us to hypothesize that the fourfold coordinated aluminum in aC’ dots gives rise to a redox mechanism where the fluorophore Cy5 is first reduced into a dark state (e.g., F^•– in case of a neutral fluorophore), and subsequently is oxidized by dissolved oxygen thereby returning to its ground state (Figure 3g). Please note that dissolved oxygen can reach encapsulated Cy5 dye within the silica core via micropores observed in sol-gel derived silica.^[32] To confirm radical-based redox blinking, we exposed PEG-Cy5-aC’ dots as well as Cy5 dye plus sodium aluminate in HBSS to 10% (v/v) tert-butanol, a known radical scavenger.^[45] As expected, all low duty-cycle blinking at steady-state (vide supra) was suppressed (see representative examples in Figure 3a–d).

Our measured τ_off ≈ 100 ms is somewhat higher than the ≈60 ms reported by the Tinnefeld group for Cy5 immobilized on DNA. They also observed that Cy5 τ_off increases if the concentration of reductant is increased.^[17] Similarly, we saw a small increase in τ_off as the amount of aluminum precursor (ASB) used during the synthesis of PEG-Cy5-aC’ dots was increased (Figure S10, Supporting Information), supporting the hypothesis that aluminum plays a role in the reduction of Cy5. Based on a rough calculation (see Supporting Information, section 16), Cy5 dyes within an aC’ dot experience a local fourfold coordinated aluminum concentration of 3 m. If aluminum is involved in the redox process, the reductant concentration is several orders of magnitude greater than the highest reported reductant concentrations in solution of up to 1 × 10⁻³ M.^[19] This may contribute to the higher τ_off values of the Cy5 embedded in the aC’ dots.

We repeated the 1 ms integration time experiments with immobilized PEG-Cy5-aC’ dots in HBSS and added Trolox, an antioxidant, in order to deplete dissolved oxygen. Under this condition, we did not observe any obvious blinking in the temporal fluorescence trace and no meaningful ACF analysis could be performed (Figure S11, Supporting Information). Similarly, with 50 ms integration time, when we added OS system to free Cy5-dye plus sodium aluminate, we did not observe blinking. However, the blinking was recovered when 1 × 10⁻³ M methyl viologen, a known oxidant, was added to the oxygen-depleted solution (Figure S12, Supporting Information).^[17] These results corroborate that oxygen plays the role of the oxidant and is necessary to observe (redox) blinking of PEG-Cy5-aC’ dots.

The temporal fluorescence behavior of PEG-Cy5-C’ dots differed dramatically from that of PEG-Cy5-aC’ dots. In the absence of Al in the PEG-Cy5-C’ dots, within the first 10 s of collection time, Cy5 irreversibly photobleached into a dark state (Figure 3h), and ACF analysis results (Figure 3i) were similar to that of the camera background.^[46] For PEG-Cy5-C’ dots this is consistent with the absence of redox photophysics, where available electronic states can be described by a conventional Jablonski diagram for fluorescent dyes (Figure 3j).

The fluorescence enhancement as well as blinking capabilities provided by covalent aC’ dot encapsulation are not limited to cyanine NIR dye Cy5. As demonstrated in Figure 4a–e, relative to the respective parent free dyes, PEG-dye-aC’ dots substantially enhance the fluorescence of a variety of encapsulated dyes such as 7-diethylamino-coumarin-3-carboxylic acid (DEAC; brightness enhancement factor (BE): 10.0), fluorescein-5 (F-5; BE: 5.3), Cy3 (BE: 5.5), ATTO647N (BE: 2.0), and Cy7 (BE: 2.9). This list includes a coumarin dye (DEAC), fluorescein (F-5), other cyanines (Cy3 and Cy7), as well as a carbopyronine dye (ATTO647N) with absorption and emission spectra covering almost the entire visible spectrum all the way into the NIR regime (i.e., Em from 473 through 760 nm). Aluminum-containing aC’ dots covalently encapsulating all these dyes exhibited excellent blinking behavior with equilibrium duty cycles <0.001 suitable for STORM (Figure 4f–j) in HBSS buffer and exposed to only one respective excitation source. Furthermore, these dyes encapsulated in PEG-dye-aC’ dots all showed substantially enhanced photostability compared to their free form (Figure 4k–o) with stability enhancement factors (SE) when fit to a single exponential function of: DEAC (SE: 7.8), F-5 (SE: 1.4), Cy3 (SE: 7.7), ATTO647N (SE: 11.2), and Cy7 (SE: 18.6). Other dyes that were tested, and showed enhanced brightness and photostability as well as low equilibrium duty cycles, include Alexa Fluor 647 and Cy5.5 (Figure S13, Supporting Information). Optical characteristics together with the net charge of all dyes encapsulated in PEG-dye-aC’ dots relative to their free form are summarized in Table 1. Overall enhancement effects are substantial, sometimes an order of magnitude or larger. Differences in brightness and photostability enhancements across different dyes in part are due to dye encapsulation efficiency as a function of dye molar mass and net charge. For example, the dye exhibiting the least enhancement effects upon covalent encapsulation, Alexa Fluor 647, was the only dye studied with a net negative charge (of −3). From earlier studies and corroborated by recent high-performance liquid chromatography (HPLC) analyses, dye charge is highly correlated with encapsulation efficiency into particle cores, with high negative net charge being particularly detrimental.^[47–49] These data suggest that dyes such as DEAC, F-5, Cy3, and Cy5.5 that in free form, due to their structure and fluorescence properties, have severe limitations for STORM applications may be converted into viable probes when encapsulated in aC’ dots.^[3] Moreover, dyes previously identified as already optimal for STORM applications such as Alexa Fluor 647 and Cy5 may be further enhanced when encapsulated in PEG-aC’ dots. We previously demonstrated that brightness enhancements over free dye indeed directly carry over into photon output enhancements in dye blinking.^[6,24] Taken together, these data point to the emergence of an ultrasmall, bright, and photostable nanoparticle platform for SRM. Building on recent advances, e.g., in bright and photostable nanoprobes including upconversion as well as advanced polymer nanoparticles, ultrasmall aC’ dots constitute an attractive platform for bioimaging and SRM that does not contain heavy metals, exhibits on-off fluorescence cycling characteristics optimal for STORM, and takes advantage of a wide spectral range that can be tailored by dye choice.^[50–52]

Figure 4.

a–e) Normalized and absorbance-matched absorbance (left) and emission (right) spectra for PEG-DEAC-aC’ dots, PEG-F-5-aC’ dots, PEG-Cy3-aC’ dots, PEG-ATTO647N-aC’ dots and PEG-Cy7-aC’ dots and their unconjugated free parent dyes DEAC, F-5, Cy3, ATTO647N, and Cy7, respectively, in water. f–j) Single-molecule fluorescence traces (50 ms integration time) of immobilized PEG-DEAC-aC’ dots, PEG-F-5-aC’ dots, PEG-Cy3-aC’ dots, PEG-ATTO647N-aC’ dots and PEG-Cy7-aC’ dots; Corresponding equilibrium duty cycles are shown in green. k–o) Photobleaching experiments of free dyes and PEG-dye-aC’ dots in water with single exponential fits. Particles were all in HBSS imaging buffer for single-particle experiments and exposed to their respective excitation light sources: 405 nm (PEG-DEAC-aC’ dots), 488 nm (PEG-F-5-aC’ dots), 561 nm (PEG-Cy3-aC’ dots), and 640 nm (PEG-ATTO647N-aC’ dots and PEG-Cy7-aC’ dots). BE: Brightness Enhancement; SE: Stability Enhancement.

Table 1.

Optical characteristics of dyes encapsulated in PEG-aC’ dots compared to their free form in terms of fluorescence and photostability enhancements in water as well as their equilibrium duty cycle statistics using a single excitation light source in HBSS. Error bars for brightness enhancements (BE) were derived from the standard deviation of three separate measurements. Stability enhancement (SE) errors were the cumulative standard error of the photobleaching fit function parameters, described in detail in section 5 of the Supporting Information.

Dye-aC’ dot	Parent Dye Net Charge	Excitation/Emission	Brightness Enhancement (BE)	Stability Enhancement (SE)	Duty Cycle (DC)
DEAC-aC’ dot	0	430/473	9.98 ± 0.63	7.80 ± 0.8	0.0002
F-5-aC’ dot	0	494/508	5.34 ± 0.08	1.40 ± 0.24	0.0007
Cy3-aC’ dot	+1	551/555	5.49 ± 0.21	7.70 ± 2.69	0.0009
Cy5-aC’ dot	+1	647/660	4.12 ± 0.04	2.83 ± 0.30	0.0003
Alexa Fluor 647-aC’ dot	−3	650/655	1.87 ± 0.15	0.90 ± 0.04	0.0003
ATTO647N-aC’ dot	+1	651/659	2.03 ± 0.19	11.16 ± 4.43	0.0007
Cy5.5-aC’ dot	+1	686/691	5.82 ± 0.25	2.82 ± 0.46	0.0005
Cy7-aC’ dot	+1	754/760	3.20 ± 0.19	18.64 ± 3.02	0.0009

The highly tunable surface of aC’ dots, through post-PEGylation surface modification by insertion (PPSMI) reactions, enables attachment of targeting ligands without compromising particle stability.^[53] Through PPSMI and subsequent click chemistry modifications, similar to techniques previously reported, we were able to covalently attach secondary antibodies (Ab₂) to aC’ dots, e.g., yielding Ab₂-PEG-ATTO647N-aC’ dots (Figure 5a).^[54] Using confocal FCS (Figure S14 and Table S2, Supporting Information) we demonstrated this successful attachment, that these modified aC’ dots bind to primary antibodies (Ab₁), and that two Ab₂-PEG-ATTO647N-aC’ dots can bind to one primary antibody, known as secondary antibody binding amplification, to yield Ab₁-Ab₂-PEG-ATTO647N-aC’ dots.^[55] In Figure 5b,c, HeLa cells were fixed with 4% paraformaldehyde, permeabilized with NP-40 detergent, and exposed to mouse anti-α-tubulin primary antibodies followed by goat anti-mouse IgG bearing Ab₂-PEG-ATTO647N-aC’ dot addition (see Supporting Information). These cells were then imaged in aqueous phosphate buffered saline (PBS) with a red laser. Figure 5b shows a diffraction-limited TIRF microscopy image of a resulting cell with a Hoechst 33342 nuclear stain. Some tubulin-like structures can be identified under TIRF microscopy, but due to their overlapping PSFs separating them into individual tubules is not possible. In contrast, when we reconstructed the STORM image in Figure 5c, the labeled tubulin structures now could be resolved below the diffraction limit and overlapping networks of individual tubules could be identified. Intensity profiles across a selected microtubule in both the TIRF and STORM images to show the improvement in resolution can be seen in Figure 5d.

Figure 5.

a) Illustration of two goat anti-mouse IgG secondary antibody (Ab₂)-PEG-aC’ dot conjugates binding to one mouse anti-α-tubulin primary antibody. b) Diffraction-limited TIRF image of a fixed and permeabilized HeLa cell labeled with mouse anti-α-tubulin primary antibodies followed by Ab₂-PEG-ATTO647N-aC’ dots (red) staining. c) STORM reconstruction of (b). d) Intensity profiles of a line across a tubulin structure, highlighted by the yellow boxes in (b) and (c). Profiles were normalized and fit to a Gaussian function to obtain the data shown in (d), demonstrating the improvement in resolution to below the diffraction limit. Please note that the achievable resolution is limited by the fact that typically two aC’ dots are conjugated to a single primary antibody as shown in (a). Cells were stained with Hoechst 33342 (blue) to visualize the nucleus. Images were taken with only a 640 nm excitation light source in PBS.

A major advantage of aC’ dots over traditional STORM fluorophores is the fact that from numerous in vivo studies with conventional C’ dots, including several human clinical trials with their respective preclinical toxicological investigations, they are expected to show no substantial toxicity.^[26,56] Consistent with that expectation, using a PrestoBlue cell viability assay, there were no significant changes in HeLa and MDA-MB-231 triple negative breast cancer (TNBC) cell viability after exposure to 2 × 10⁻⁶ M aC’ dots for 24 h (Figure S15, Supporting Information). This concentration is 40 times higher than that used for the fixed cell experiments (vide supra), and 8 times higher than that used in live-cell imaging as shown in Figure 6a–c. To obtain these live-cell images, MDA-MB-231 TNBC cells were starved for 24 h in serum-free culture media to induce endocytosis of unfunctionalized PEG-Cy5-aC’ dots (red) suspended in complete media.^[57] As controls, it was shown that 11 × 10⁻³ M glutathione, the maximum concentration found in cells, had no effect on PEG-Cy5-aC’ dot blinking, and that exposure to culture media for 48 h does not cause any aggregation (Figure S16, Supporting Information).^[58] After one-hour incubation, excess particles were washed away and the cells were labelled with Hoechst 33342 (blue) and a membrane stain, CellMask Orange (green). Imaging was performed in HBSS buffer with only one excitation (red) laser (see Supporting Information). The diffraction limited TIRF image (Figure 6a) shows relatively heterogeneous and large red fluorescent blobs likely due to particles associated with intracellular vesicles. In contrast, smaller and more homogeneous red features appeared in the reconstructed STORM images (Figure 6b,c). These features are likely due to particles being confined in intracellular vesicles. For 600 such features/ vesicles from 13 individual cells, the vesicle size and number of particles per vesicle (estimated from the number of emitter localizations) were determined using an algorithm estimating the maximum emitter distance as described in the Supporting Information. Results for three example vesicles are exhibited in Figure 6d–f, while three example kernel density plots of the number of localizations, and number of particles derived therefrom, as a function of vesicle size derived from three example cells are shown in Figure 6g–i. Identification of the smaller fluorescent features shown in the panels of Figure 6d–f with aC’ dots confined in vesicles seemed justified as corroborated by the extracted size information of these features (vide infra) as well as the known mechanisms for particle uptake in living cells.^[59] The kernel density plots in Figure 6g–i for individual cells 1–3 suggest well-defined size regimes for these vesicles with particle numbers going up substantially for vesicle sizes of 400–500 nm. This observation is corroborated when the quantitated information from all 13 cells was combined in Figure 6j. Labelled vesicles ranged from ≈50–500 nm in diameter, similar to intracellular vesicle sizes previously reported.^[60] Overall, in the low particle number range, as vesicle size increased, the number of localizations and number of particles derived therefrom in each vesicle roughly increased in a cubic fashion, as suggested by comparison with a cubic fit (blue curve) in Figure 6j. The largest structures/vesicles observed were around 400–500 nm in size. For these largest vesicles, the number of localizations and particles per vesicle sharply increased, well beyond expectations from simple volume arguments (see deviation from blue fit). These vesicles could be maturing towards a degradation pathway as opposed to a recycling pathway.^[60] While the exact analysis of the biology of particle processing in MDA-MB-231 cells, e.g., via colocalization with known markers for different types of vesicles, is beyond the scope of this paper, these proof-of-principle experiments clearly demonstrate that aC’ dots with characteristics enabling clinical translation (e.g., ultrasmall size, nearly neutral PEG surface coating, etc.) constitute a powerful optical SRM particle platform for quantitative assessment of cellular structures and processes via live-cell imaging. This imaging advancement is highly desirable as most recent high-resolution evaluations of vesicle size were performed in fixed cells rather than live cells.^[60,61]

Figure 6.

a) Live-cell composite TIRF image of MDA-MB-231 TNBC cell stained with Hoechst 33342 (blue) and CellMask Orange (green) containing endocytosed PEG-Cy5-aC’ dots (red) in intracellular vesicles. b) Composite image with red channel STORM reconstruction and TIRF overlay. c) Corresponding red channel STORM reconstruction only. d–f) Regions in (d–f) correspond to location boxes indicated in (c). The upper panels show three examples of quantified regions containing: d) 131, e) 7, and f) 74 nanoparticles. The lower panels show scatter plots of the same data with localization densities of: d) 197, e) 11, and f) 111 localizations per square micrometer. g–i) Kernel density plots of number of localizations/particles in three individual cells 1–3 versus maximum emitter distance estimates of PEG-Cy5-aC’ dot derived localizations. j) Combination of kernel density plots of 13 individual cells indicating that the number of particles increases as object size increases. The line in blue is a fit to cubic behavior (see main text).

In summary, this work describes an ultrasmall, bright, and photostable PEGylated fluorescent core–shell aluminosilicate nanoparticle (aC’ dot) platform for optical super-resolution microscopy, particularly STORM, in both fixed and live-cell specimens. When suspended in simple buffers, such as HBSS or PBS, and exposed to only one excitation light source, aC’ dots exhibit fluorescence blinking likely due to a redox mechanism with duty cycles suitable for STORM. With their sub-10 nm diameter, ability to substantially enhance the per dye fluorescence brightness and photostability over free parent dye, the localization precision using aC’ dots is high, and dyes not typically suitable for STORM can be converted into viable SRM probes. aC’ dot properties reliably address current issues of STORM imaging, which conventionally requires specific fluorophores, cytotoxic imaging media, and multiple lasers. The highly tunable surface of aC’ dots enables functionalization, e.g., with antibodies for enhanced targeting capabilities, as demonstrated in fixed cells. Finally, aC’ dots are non-toxic to HeLa and MDA-MB-231 cells and can be used for live-cell optical super-resolution microscopy studies as exemplified here for nanoparticle trafficking in TNBC cells. Overall, aC’ dots offer opportunities for long-term, multi-color STORM imaging in live cells without the need for expensive and technically demanding instrumental setups and chemically damaging imaging conditions, promising quantitative analysis, e.g., of cellular structures and processes in live cells with sub-diffraction limited resolution. This is particularly noteworthy as the C’ dot particle platform encapsulating Cy5 or Cy5.5 has already been clinically translated in multiple human clinical trials (e.g., see clinicaltrials.gov identifiers: NCT01266096, NCT02106598, and NCT04167969). Optical SRM with aC’ dots will therefore enable the direct study, in particular, of intracellular particle processing details in live cells relevant to questions associated with the clinical translation of diagnostic as well as therapeutic C’ dot applications.^[27,62–65]