Super‐beacons: Open‐source probes with spontaneous tuneable blinking compatible with live‐cell super‐resolution microscopy

Abstract Localization‐based super‐resolution microscopy relies on the detection of individual molecules cycling between fluorescent and non‐fluorescent states. These transitions are commonly regulated by high‐intensity illumination, imposing constrains to imaging hardware and producing sample photodamage. Here, we propose single‐molecule self‐quenching as a mechanism to generate spontaneous photoswitching. To demonstrate this principle, we developed a new class of DNA‐based open‐source super‐resolution probes named super‐beacons, with photoswitching kinetics that can be tuned structurally, thermally and chemically. The potential of these probes for live‐cell compatible super‐resolution microscopy without high‐illumination or toxic imaging buffers is revealed by imaging interferon inducible transmembrane proteins (IFITMs) at sub‐100 nm resolutions.


| INTRODUCTION
Super-resolution microscopy (SRM) encapsulates optical imaging methods capable of bypassing the~250 nm resolution limit imposed by diffraction. 1 Their resolving power approaches that of electron microscopy 2 while keeping the benefits of fluorescence imaging, such as molecular-specific labelling and potential for live-cell imaging. 3 Single-molecule localization microscopy (SMLM) is a well-establish set of SRM approaches, particularly popular because of their capacity to achieve near-molecular scale resolution (<50 nm) in relatively simple imaging equipment. 4,5 SMLM achieves this nanoscale resolution by exploiting fluorophore photoswitching [6][7][8][9] or, more recently, through emission fluctuation in a new family of techniques derived from SMLM. [10][11][12] An SMLM acquisition aims to capture individualised fluorophores transitioning between non-emitting and emitting states.
The analysis of a sequence of images containing this information then allows the creation of a super-resolution image where the presence and location of these fluorophores can be better-discriminated. 13 To achieve a considerable resolution increase (10-fold increase when compared to diffraction-limited approaches), these methods rely on specialised labels whose switching kinetics are frequently modulated by intense illumination. 14 However, this requirement constrains these approaches to microscopes capable of high-intensity illumination and limits compatibility with live-cell imaging due to phototoxicity. 3,15,16 Exceptions exist, such as genetically encoded fluorophores that photoswitch at low-intensity illumination, these include Dreiklang, 17 Skylan 18 and SPOON. 19 SMLM based on Point Accumulation for Imaging in Nanoscale Topography (PAINT) exploits an alternative mode of blinking, accomplished through transient binding of fluorescent probes to target molecules. [20][21][22] PAINT has the benefit of not requiring high-intensity illumination, as blinking arises from binding kinetics. However, it is often limited to TIRF or Spinning-Disk microscopy, due to the presence of considerable background fluorescence from unbound freely diffusing probes. 23 Recent studies have started to address this issue. 24 We propose a new class of super-resolution fluorescent probes for SMLM, dubbed super-beacons ( Figure 1). With these molecules, the transition between emitting and non-emitting states is stochastic and independent of illumination. This feature eliminates the need for photodamaging illumination or photoswitching inducing buffers ( Figure 1A). At the molecular level this is achieved through a DNA scaffold resembling molecular-beacons ( Figure 1B), 25 Figure 1A). 26 The transition between these conformations thus influences the lifetime of emitting and non-emitting states. We designed a new class of photoswitchable probes using these general principles.
We chose a DNA-hairpin scaffold due to its flexibility and easy synthesis, a structure resembling the well-characterised molecularbeacons ( Figure 1B). [27][28][29][30] Our probe dubbed a super-beacon (SB), is a single-stranded DNA (ssDNA) oligo-nucleotide with a fluorophore and quencher covalently bound to opposing ends ( Figure 1B). Short, complementary, terminal sequences promote self-hybridization resulting in the stable formation of the hairpin-shaped secondary structure, hereafter called the closed-state. This state, dominant in thermal equilibrium, is non-emitting due to resonant energy transfer, collision and contact with a dark quencher. 31,32 The probe can reversibly transition to a short-lived fluorescent "open" state in a stochastic manner ( Figure 1C). The rate of transition between states is modulated by temperature, chemical environment and choice of structure (oligonucleotide sequence) 33 ( Figure 1C). To demonstrate the principles of SBs, we designed a short probe with the sequence 5'-ATTO550 ACG A naming convention that relates to it being a super-Beacon (B3S9L)

| Super-beacons in vitro characterisation
To characterise SBs as photoswitching probes, we imaged their photophysical behaviour while immobilised on a coverslip using Total Internal Reflection Fluorescence microscopy (TIRF) ( Figure 1D,E).
B3S9L-ATTO550-BHQ2 SBs were linked to a coverslip surface at high dilution, to achieve sparse spatial distribution and negligible detectable molecular spatial overlap. In SMLM, it is expected that probes switch stochastically between a non-fluorescent (off) state and fluorescent (on) state. The on-state lifetime, τ ON , should, ideally, be similar to the acquisition rate (10 s of milliseconds) and ratio of onstate to off-state lifetimes, r = τ ON /τ OFF , at least 1/50. 33 Following the protocols described in the Methods section, we measured the switching state lifetimes of our SB structure ( Figure 2A). In these conditions, r = 1/50 is achieved at 250 W/cm 2 , demonstrating desirable photoswitching at 5-fold less illumination than for our control probe.
Interestingly, we observed an illumination dependence on the on-to off-state lifetimes ratio. This can be due to local thermal effects or due to the photophysics of the ATTO555 fluorophore, namely the transitions to triplet states at these illumination regimes.  Figure S1A), an effect that we could recover by imaging this SB in OxEA switching buffer ( Figure S1A). 34 Additionally, the molecular design of SBs favours the closed-state conformation, an advantage over organic fluorophores for SMLM, since a high proportion are observed to be in the off-state at the beginning of acquisition ( Figure 2B). We then disrupted the hairpin

| Super-beacons as super-resolution probes
To evaluate the potential of SBs as imaging probes, we took advantage of the biotin-modified thymine in the SB loop to attach it to a streptavidin-conjugated antibody ( Figure 2D,E). Using these antibodies we labelled and imaged β-tubulin in fixed NIH3T3 cells (MetOH fixed as previously described 11 Figure S2). This was confirmed by measuring the distance between two filaments ( Figure 2D,E, suggesting a resolution better than 80 nm). This experiment demonstrates that for similar low-illumination imaging conditions, using a non-toxic buffer, SBs can achieve SRM images of higher perceived quality than those obtained with the control probe ( Figure 2D).

| Capacity for live-cell SRM
To test the potential of SBs for live-cell SRM imaging, we used cell proteins. IFITM proteins are a family of broad-spectrum inhibitors of virus replication that act primarily against enveloped viruses. 37 IFITM genes have been found in numerous vertebrate species. In humans, IFITM-1, -2 and -3 encode proteins that are thought to act primarily, though perhaps not exclusively, 38 by inhibiting viral fusion. 37,39 When expressed alone in A549 cells, IFITM-1 localises predominantly to the plasma membrane, while IFITM-2 and -3 localise preferentially in late and early endosomes, respectively. 39 Thus, for human IFITM proteins at least, the distributions cover the main cellular portals through which enveloped viruses enter cells. We next investigated the potential of SBs to provide detailed information on the distribution and dynamics of C-terminally HA-tagged IFITM-1 in live-cell SMLM experiments. As IFITM-1 is primarily localised to the plasma membrane, and its C terminus is accessible on the surface of intact cells, 39  photoswitching, there were negligible visual differences in IFITM1 distribution at the cell surface, for both imaging modalities ( Figure 3A).
Quantitative analysis of image quality through the SQUIRREL algorithm 36,40 further shows that both conditions achieve similar quality ( Figure S3). In all illumination regimes, there was a high fidelity between the SMLM reconstructions and the raw data (as seen by SQUIRREL error maps). As expected, this was not the case for AF647 that relies on photo-induced processes ( Figure S3A,B). SMLM provides not only high-resolution structural information but also quantitative information about the subcellular distribution and molecular organisation. 41 Hence, we asked if the high fidelity in protein distribution across different illumination intensity regimes was translated into the ability to extract quantitative data. To access this, we analysed IFITM1 distribution across a range of illumination intensities (from~0.05 tõ 2.5 kW/cm 2 ) and analysed the cluster diameter with SR-Tesseler. 41 When using anti-HA-AF647 labelled antibody, we saw a significant difference (P < .01) between the high-illumination regime ( Thin trails of localizations between these regions were also seen in few instances. Single-molecule tracking was performed using the single-particle tracking plugin for ImageJ, TrackMate 40 ( Figure 4B). The ability to achieve this in a physiological buffer at low-illumination intensities demonstrates the merits of using SBs in live-cell settings.
Furthermore, combining SB with techniques such as single-particle tracking has the potential to allow experiment design using SB imag- ing. This flexibility suggests a model-based design principle could be adopted to reach an optimal structure for a specific application in order to maximally make use of the rapid imaging technologies available.

| Super-beacon probe design
The SB structure was designed to be short to reduce linker and crowding errors 49,50 and to have a melting temperature above room temperature (~33 C-using mFOLD 46

| Surface passivation for single-molecule studies
In order to determine the photoswitching characteristics of SBs, bio-

| Illumination intensity determination
The reported illumination intensities are the peak intensities (I peak ) at the centre of the field of view and were calculated from the percentage of maximum laser power provided by the microscope acquisition software, p. For each wavelength used, a conversion factor, c, was regularly calculated relating the software provided power to the measured power in the back aperture of the objective, P = cp. Over the course of the experiments a maximum of 10% error was measured.
The transmission rate, T, of the objective was confirmed by measurement to be 89% ± 1.5%. For each wavelength used, the SD, σ I , of the Gaussian distributed illumination intensity in the sample plane was measured by fitting the intensity of line profiles taken from illuminated empty coverslips. The calculation for I peak , assuming a Gaussian distributed illumination, I(x,y), of the sample, can be derived as follows, The maximum variation observed of the illumination intensity over the 5.2 μm imaged area was less than 5%.

| Single-molecule imaging of super-beacon probes
Characterisation was performed in phosphate buffer saline (PBS, MiliQ water or OxEA-the latter as previously described 34  continuously for a period of 540 seconds (30 000 frames).

| Single-molecule time-trace analysis
Data obtained were analysed by extracting intensity over time traces (eg, Figure 1D) using a custom analysis pipeline as follows: Raw data was loaded with ImageJ/Fiji 53 and single molecule localization performed using the ThunderSTORM plugin. 35

| Super-beacon for fixed cell imaging of microtubules
Fixed NIH3T3 cells were stained for β-tubulin (as previously analysis of the data was performed using ThunderSTORM 35 and to access the fidelity between the SMLM reconstructions and the raw data we used SQUIRREL. 36 To analyse IFITM1 distribution across a range of illumination intensities (from~0.05 to~2.5 kW/cm 2 ) we analysed the cluster diameter with SR-Tesseler. 41 Localization tables from Thunder-STORM were imported and Voronoi diagrams were created. Regions of interest were selected with a density factor δ of 2. Within individual objects, clusters were identified with δ = 2. Statistical analysis was performed using GraphPad Prism 7 (Prism Software). Significance was calculated using unpaired t-tests: *P < .01.
4.9 | Super-beacon SR single particle tracking of IFITM1 in live cells