Photonic lantern TIRF microscopy for highly efficient, uniform, artifact-free imaging

Abstract

We report a method for generating uniform, artifact-free total internal reflection fluorescence (TIRF) excitation via a photonic lantern. Our tapered waveguide, consisting of a multimode input and nine few-mode outputs, enables single-shot TIRF illumination from nine azimuthal directions simultaneously without the introduction of nonstationary devices. Utilizing the photonic lantern for multi-beam excitation provides a low-loss mechanism that supports a wide range of light sources, including high-coherence lasers and various wavelengths in the visible spectrum. Our excitation system also allows tuning of the TIRF penetration depth. The high-quality excitation produced by photonic lantern TIRF (PL-TIRF) enables unbiased imaging across the entire illumination field-of-view. The simplicity and robustness of our technique provides advantages over other TIRF approaches, which often have complicated setups with scanning devices or other impracticalities. In this paper we discuss the lantern design process, characterize its performance, and demonstrate flat-field super-resolution imaging and shadowless live-cell imaging using PL-TIRF.

1. Introduction

Total internal reflection fluorescence (TIRF) microscopy [1,2] is a powerful imaging approach which allows one to visualize the surface features of biological samples with high signal-to-background ratio. TIRF illumination is generated by producing a highly inclined excitation beam at the interface between the glass coverslip and the sample (usually aqueous medium) such that the incident angle exceeds the critical angle. Consequently, an evanescent wave penetrates the sample, exciting fluorescent molecules residing within a few hundred nanometers of the sample’s surface [3,4].

There are many ways to generate TIRF illumination [1,5–8], but perhaps the simplest method is objective TIRF, which requires an objective lens with a numerical aperture (NA) of 1.4 or greater [9,10]. The TIRF beam is produced by tightly focusing the excitation beam to the outermost edge of the back focal plane (BFP) of the objective. The outer ring of the BFP is referred to as the “TIRF region” of the objective lens [Fig. 1(a) and 1(b)], and the excitation beam can be focused anywhere within the TIRF region for the beam to exit the objective lens at an angle exceeding the critical angle. When total internal reflection occurs, the evanescent field propagates laterally across the sample along the same direction that the incident beam was launched. The beam’s intensity profile and coherence properties heavily influence the uniformity of illumination across the imaging field-of-view (FOV) [11–13].

Fig. 1.

(a) Illustration of the back focal plane (BFP) of an objective lens with the TIRF region (δ) denoted by the dashed line. (b) Objective-type TIRF generating evanescent excitation at the glass-sample interface. (c) Schematic of nine-core photonic lantern. (d) Schematic of microscope setup with integration of photonic lantern for excitation. DM, dichroic mirror; FC, filter cube; L1-L3, lenses; M, mirror; MMF, multimode fiber; Obj, objective lens; sCMOS, scientific CMOS camera; TL, tube lens; VM, vibrational motor.

When imaging biological samples, scattering and shadowing artifacts can appear due to the laterally propagating excitation beam being obstructed by the features of the sample [11]. One way to prevent such artifacts is by exciting the sample from multiple directions so that direction-specific artifacts can be averaged over [14]. This can be accomplished by scanning a single excitation beam around the TIRF region of the objective lens using a scanning device [14,15] such as a galvo mirror [16,17], a digital micromirror device [18], or an acousto-optical deflector [19]. These approaches produce 360-degree azimuthal illumination, but they introduce nonstationary elements to the setup that require sensitive alignment. Other methods have been proposed, such as employing an annular mask [9,20] or axicon optics [21,22], to produce 360-degree annular illumination continuously but at the expense of being extremely lossy, complicated, or yielding a small FOV, respectively. All of the aforementioned drawbacks can be readily overcome by using an annular fiber bundle [23]. However, the inevitable coupling loss associated with fiber bundles, although not prohibitive, is not ideal for super-resolution microscopy, where high-power TIRF illumination is commonly utilized [24].

Conceptually, photonic lanterns [25,26] can be used to generate annular or multi-spot TIRF in a manner similar to the annular fiber bundle approach [23]. For this particular application photonic lanterns are far superior because they can function like a 1 × N fiber splitter [Fig. 1(c)], where light from a single multimode input core is distributed to multiple (N) single-core output fibers with minimal loss [26]. To form a photonic lantern, multiple fibers (N) are fused and tapered down together, gradually reducing their size to form a single multimode core. By precisely engineering the tapered transition between the input and output ends, we can achieve a lossless link between the two ends [27]. In addition to being highly efficient, the ability to customize the orientation [28], sizes [29], and number [30] of the lantern’s outputs is very useful for imaging. To date, photonic lanterns have been used in various imaging applications such as optical coherence tomography [31], computational imaging [32], and coherent imaging [33]. However, they have not yet been utilized in fluorescence microscopy. In this paper, we propose using a photonic lantern for uniform, artifact-free multi-spot TIRF excitation which supports high coupling efficiency. This approach overcomes the aforementioned drawbacks of other methods, thereby enabling single-molecule localization microscopy (SMLM) [34] for super-resolution imaging.

2. Results

2.1. Preliminary design of six-core photonic lantern

To serve as a proof of concept, we first designed and fabricated a prototype photonic lantern consisting of six output cores arranged in a hexagonal orientation as shown in Fig. 2(a). We settled on six cores because, while the merits of annular TIRF have been demonstrated [23], we wanted to investigate the lower limit of angular diversity required to produce high-quality TIRF illumination. The hexagonal arrangement of output fibers was chosen so that the six-core beam profile could be projected to the BFP of the objective lens via a 4f system and all six beams would simultaneously be focused to different spots within the TIRF region of the objective lens, inducing multi-spot TIRF from six equally separated azimuthal directions. The input end of the lantern was a multi-mode fiber with a core diameter of 20 μm and the output fibers were step-index fibers with core diameters of 9 µm and δn of 4 × 10⁻³. The lantern was designed to be compatible with a 100 × oil immersion objective lens (NA = 1.4).

Fig. 2.

(a) Image of output beam profile of prototype six-core photonic lantern. (b) Image of Atto647N dye layer when excited by prototype lantern with 638 nm laser light. Linewidth of light source was 0.1–1 nm. (c) Line profile of diagonal across dye layer in (b). (d) Image of output beam profile of nine-core photonic lantern. (e) Image of Atto647N dye layer when excited by nine-core lantern with high-coherence 640 nm laser light. Bandwidth of light source was < 1 MHz. (f) Line profile of diagonal across dye layer in (e). Scale bars, 0.2 mm (a),(d) and 10 µm (b),(e).

We built a custom microscope featuring the prototype lantern to validate the operating principle of our multi-spot TIRF approach. By exciting a fluorescent dye layer (Atto647N) with a 638 nm diode laser coupled to the photonic lantern, we characterized the TIRF illumination uniformity by imaging the fluorescence intensity distribution. Unexpectedly, our preliminary data revealed some flaws in the lantern design. Although each of the six beams was focused to the TIRF region of the objective lens as intended, the overall uniformity of the multi-spot TIRF illumination was not homogeneous across the imaging FOV as can be seen in Figs. 2(b) and 2(c).

We deduced that the nonuniformities were a result of multiple factors. Firstly, the relatively small core size of the output fibers meant that each fiber only supported a few guided modes at our excitation wavelength. Since the number of effective modes within the fibers was small, the speckle contrast could not easily be reduced [35], resulting in large speckle features appearing in the excitation intensity distribution. Mode-scrambling via vibrational agitation of the lantern fibers helped marginally but ultimately proved ineffective despite the coherence of our light source being intermediate (0.1–1 nm linewidth). Additionally, having an even number of output fibers resulted in an excitation symmetry where opposite beams propagated across the sample towards each other along a common axis. So, although we induced TIRF illumination from six different directions, excitation only occurred along three axes. Finally, having only six excitation spots meant that there was small margin for error and that artifacts were less likely to be covered up or averaged out by illumination from other directions.

2.2. Design and assessment of nine-core photonic lantern

After identifying the limitations of the prototype lantern, we fabricated a new lantern consisting of nine graded-index output fibers with core diameters of 15 µm and δn of 16 × 10⁻³. The output cores were oriented in an annular fashion [Fig. 2(d)], revealing a symmetry that would not propagate excitation beams along common axes like the previous six-core lantern did. In addition, segments of step-index fiber were spliced into each of the output fibers to enable easier mode scrambling. The input fiber core diameter was increased to 50 µm to accommodate for the increase in modes. The increased number of cores, the larger output core size, and the new beam profile symmetry each addressed a different drawback of the prototype lantern.

To evaluate the performance of our new photonic lantern, we constructed a custom microscope [Fig. 1(d)] and measured the uniformity of TIRF illumination as we did with the six-core lantern. We can clearly see that the new photonic lantern generates relatively uniform illumination across a 50 × 50 µm² FOV when excited with either a low-coherence laser [Supplement 1 ^{(2.2MB, pdf)} , Fig. S1] (linewidth of 1–5 nm) or a narrowband 640 nm laser [Fig. 2(e)] (bandwidth < 1 MHz). The accompanying line profile in Fig. 2(f) reveals a nearly flat-top distribution across a 20 µm span in the center of the FOV, with a slightly asymmetric drop in intensity from the center of the FOV to the edges. We are aware of the challenges associated with trying to obtain homogeneous excitation using a high-coherence light source from an optical fiber, but we chose to characterize the excitation uniformity under nonideal circumstances to test the limits of our method. And although the excitation is not as uniform as when using a low-coherence light source, the results are sufficient for biological imaging. To help alleviate the issue of modal interference when using high-coherence light sources, we introduced a fiber optic “de-speckler” which reduces modal noise within the fiber, effectively eliminating laser speckle entirely. The robustness of our approach was confirmed by testing the photonic lantern using a different excitation wavelength (488 nm), shown in Supplement 1 ^{(2.2MB, pdf)} , Fig. S2.

To confirm that interference between the nine excitation beams was not occurring at the imaging plane, we designed an interferometric experiment as shown in Fig. 3(a). The excitation scheme consisted of our high-coherence 640 nm laser coupled to the de-speckler and then the photonic lantern. We used a 4f system to collimate and then refocus the outgoing lantern light to a distant plane. A 50:50 beam splitter was used to split our nine-beam excitation profile along two arms such that a conjugate focal plane (CFP) resided in each of the arms. A mirror with an attached slit was placed at both CFPs where the nine-beam lantern profile was tightly focused and where each beam was well isolated. The slits were then positioned to block all but one beam, such that only one beam from each arm was reflected back to the beam splitter where they were spatially combined and sent to the objective lens of a microscope. The two-beam interferogram was used to excite a dye layer (Atto647N) with epi-illumination, and we imaged the fluorescence intensity distribution to inspect for any interferometric phenomena.

Fig. 3.

(a) Experimental setup of interference experiments. Images of Atto647N dye layer when excited by interferogram of two beams originating from the same lantern output core (b) and from two different lantern output cores (c). Scale bars, 10 µm. PL, photonic lantern; L1-L3, lenses; M, mirror; CFP, conjugate focal plane; BS, beam splitter; SM, slit-mirror.

As expected, interference occurred when we interfered two beams which originated from the same output core of the photonic lantern [Fig. 3(b)]. When attempting to interfere two beams originating from different lantern cores, we observed little to no interference [Fig. 3(c)]. This is likely due to the intriguing nature of the photonic lantern and how modes from the multimode core are converted to “supermodes” at the transition region, which are then expressed as intensity distributions to the different single-core fibers. It was previously observed that photonic lanterns can act as mode scramblers [25,26]. Therefore, due to the reciprocal nature of the photonic lantern, phase randomization can also occur when light from the multimode core is coupled to the individual fibers, thereby introducing incoherence between the individual output fibers. Note that the images shown in Fig. 3 were recorded while integrating the laser de-speckler. The de-speckler caused each of the interfering beams to exhibit higher uniformity but did not appear to affect the interference behavior. Even without the de-speckler, interference between common beams was present while interference between distinct beams was minimal [Supplement 1 ^{(2.2MB, pdf)} , Fig. S3].

2.3. Imaging quality of photonic lantern TIRF

Another key property that we assessed is the degree of penetration/leakage into the sample. To confirm that the photonic lantern generates clean TIRF, we recorded freely diffusing fluorescent beads in water and compared the images when excited with epi-illumination or with photonic lantern TIRF (PL-TIRF). If the TIRF illumination is leaky, meaning that the excitation light partially propagates through the sample as opposed to being totally evanescent, then beads residing beyond the surface of the sample will fluoresce and appear in the image as background signal. Clean TIRF, on the other hand, only excites beads residing close to the surface, resulting in a bright signal from the surface beads and negligible background signal. The images reveal a clear distinction between the two modes of excitation: epi-illumination produces an intense, Gaussian-like background signal which pollutes the image and drowns out the signal of the surface beads [Fig. 4(a)] whereas TIRF excitation primarily excites the beads close to the surface and elicits minimal background fluorescence [Fig. 4(b)]. This comparison demonstrates that the photonic lantern produces clean TIRF with no significant leakage. By measuring the apparent diameter of 1-µm fluorescent beads under TIRF excitation [4], we estimate that the penetration depth of our PL-TIRF is about 350 nm under 640 nm excitation [Supplement 1 ^{(2.2MB, pdf)} , Fig. S4].

Fig. 4.

Images of freely diffusing fluorescent beads (excited with 640 nm) in water when excited with epi-illumination (a) and with PL-TIRF (b). Scale bars, 10 µm.

To test the versatility of our approach, we modified our imaging system to enable leaky TIRF excitation [36]. This was easily achieved by introducing a zoom lens to the excitation system and configuring it to slightly reduce the magnification of the lantern’s beam profile, causing a portion of each of the nine beams to be focused to the “epi region” of the objective lens. We then captured live-cell images of endoplasmic reticulum (ER) in U2OS cells labeled with StayGold [37] and excited with a 488 nm laser. The leaky TIRF marginally increased the excitation penetration depth, allowing us to visualize the entirety of the ER structures rather than just exciting the portions close to the surface. Under this excitation scheme we could clearly observe fast dynamics of the ER [Visualization 1 ^{(9.5MB, avi)} ]. The snapshot in Fig. 5(a) confirms that the quality of illumination was adequate for quantitative high-resolution live-cell imaging. Note that conventional single-spot leaky TIRF often produces shadow artifacts as shown in Fig. 5(b) and 5(c).

Fig. 5.

(a) Live-cell image of ER tagged with StayGold in a U2OS cell by leaky PL-TIRF with 488 nm excitation. Images of ER lumen excited by leaky single-spot TIRF from a single-mode fiber when the sample is excited from the left (b) and from the right (c). Scale bars, 10 µm.

2.4. Super-resolution imaging with photonic lanterns

To demonstrate super-resolution performance we attempted single-molecule localization microscopy [38] using PL-TIRF to image microtubules labeled with Alexa Flour 647 (AF647) in U2OS cells. Using a 640 nm excitation laser and a 405 nm activation LED, we imaged 1,000 frames with an exposure time of 60 ms. The excitation power at the objective lens was 70 mW, corresponding to an intensity of 2.8 kW/cm² which lies within the range of acceptable intensities for SMLM imaging of AF647-labeled samples [39]. The SMLM data was processed using DECODE [40], a deep-learning-based localization tool, and the final image was rendered and filtered using SMAP [41]. As shown in Figs. 6(a) and 6(b), we were able to obtain super-resolution images with substantially improved spatial resolution compared to the diffraction limited image.

Fig. 6.

(a) Fluorescence images of microtubules in U2OS cells stained with AF647 obtained by PL-TIRF. Lower region is the diffraction-limited image and upper region is the super-resolution SMLM image. (b) Zoomed-in image of green boxed region in (a). (c) Three-color distribution plot showing distribution of emitters in relation to localization precision. Green and blue curves represent emitters sampled from the green and blue boxed regions in (a), respectively. Red curve represents emitters from entire SMLM image. Scale bars, 10 µm (a) and 3 µm (b).

Statistical analysis of the SMLM data was conducted to evaluate the quality and uniformity of excitation, which can be inferred by the localization accuracy [42]. By sampling emitters from different regions of the image and comparing the localization precisions of the emitters from each sample [Supplement 1 ^{(2.2MB, pdf)} , Note 1], we could determine if any significant variation in illumination intensity induced a variation in the observed localization precision values. As shown in Fig. 6(c), all three curves, representing emitters from the entire image and the green (FOV 1) and blue (FOV 2) boxed regions in Fig. 6(a), share similar representative features: a sharp peak near 5 nm and a shallow tail that decays towards a precision value of 70 nm. The average localization precision for each population is 18.8 nm, 16.7 nm, and 18.7 nm, respectively. The strong overall resemblance between the three distributions and the relatively similar values of average localization precision suggest that the excitation intensity remained consistent across both small FOVs as well as the entire image, despite the variation in the microtubule density and orientation across the FOV.

3. Discussion

In this paper, we present a novel approach for generating uniform, artifact-free TIRF illumination using a photonic lantern. Our method utilizes a nine-core photonic lantern to induce multi-spot objective TIRF, which can offer comparable illumination quality to spinning-spot TIRF but without introducing scanning devices or other complications to the microscope apparatus. While the intensity distribution of our TIRF illumination is not perfectly flat-field [12], we have demonstrated that it is uniform enough for quantitative live-cell and super-resolution imaging across a reasonably sized FOV. The single-shot nature of our technique enables high speed imaging limited only by the camera frame rate. Unlike other single-shot approaches that are limited in power throughput [11,23], the high coupling efficiency of photonic lanterns (> 91% in our case) is beneficial for applications requiring high excitation intensity. We also showed that our excitation scheme accepts high-coherence light sources and supports multiple wavelengths within the visible spectrum while simultaneously mitigating interference between the various excitation beams.

Our technique allows for tuning of the TIRF penetration depth by adjusting the magnification of the excitation beam profile using the zoom lens. This controls the illumination angle of incidence and can be varied to produce different modes of excitation such as leaky TIRF or epi-illumination. In future adaptations of our work, the variable magnification feature may be exploited to produce multi-spot highly inclined and laminated optical sheet (HILO) illumination [43] with adjustable axial sectioning capabilities and uniform, shadowless excitation.

To enhance the performance of our approach, one may consider making a few changes to the photonic lantern design and microscope setup. For example, adopting a more precise zoom lens or other magnification system would allow for more control over the penetration depth. In addition, the imaging FOV could be increased by increasing the spacing between opposite output fibers [Supplement 1 ^{(2.2MB, pdf)} , Note 2]. This modification would have the added benefit of permitting the use of larger output fibers which could improve the uniformity of illumination. The illumination quality may be further improved by introducing more output fibers to produce 360-degree TIRF, but fabricating such photonic lanterns is substantially more laborious than fabricating a few-core lantern for multi-spot TIRF. The overall uniformity could be improved dramatically if steps are taken to address the darkened edges of the excitation profile [Supplement 1 ^{(2.2MB, pdf)} , Note 3]. For SMLM imaging, we may be able to increase the imaging speed by 2-4-fold to enable live-cell super-resolution imaging. For instance, by utilizing DECODE [40], which is capable of localizing dense SMLM data while simultaneously estimating its own fitting error for each emitter, we were able to significantly increase the imaging speed while still generating fairly accurate super-resolution renderings [Supplement 1 ^{(2.2MB, pdf)} , Fig. S5]. For these renderings, we utilized DECODE’s ability to draw upon temporal information from previous and subsequent frames to more accurately process each frame.

Due to their convenience in beam delivery, single-mode [12] and multi-mode fibers [44,45] have been widely utilized in TIRF microscopy for excitation. Our previous work, which generated annular TIRF via a fiber bundle [23], offered excellent performance while maintaining a simple optical setup which was more practical than most other shadowless TIRF systems. In comparison, the photonic lantern method is just as easy to implement but is more optically efficient, can be used with a wider range of excitation sources, and offers penetration depth tunability. The practicality of our excitation system makes it an attractive option for biological researchers who use TIRF illumination in any capacity. Notably, the technology would be easy to integrate with commercial TIRF microscopes, making our technique scalable for wide adoption. As photonic lanterns become more widely available, we foresee our approach becoming commonplace in subcellular TIRF imaging systems.

4. Materials and Methods

4.1. Fabrication of photonic lanterns

The photonic lanterns were fabricated using a tapering station (3SAE Technologies). To form the photonic lantern, either six or nine fibers were inserted into a fluorine-doped capillary (NA = 0.16), then fused and tapered to form a multimode core. We then packaged the multimode input core and connectorized it using an FC-PC connector. We then used a microstructured template [30] to precisely position the cores in the required geometry, i.e., an annular pattern to form the output end of the photonic lantern. We secured the fibers using a two-part adhesive, packaged the combined structure inside a metal ferrule and polished its end face. This final structure was employed to illuminate the sample under test.

4.2. Characterization of objective lens for lantern design

We characterized the dimensions of the BFP of our objective lens (UPlanSApo 100×/1.4 Oil, Olympus) to determine what the physical dimensions of the photonic lantern should be. Equation (1) was used to estimate the diameter of the BFP:

$$D_{BFP}=2\times f_{obj}\times NA$$ (1)

where $D_{BFP}$ is the diameter of the BFP, $f_{obj}$ is the focal length of the objective lens, and $NA$ is the numerical aperture. Our objective has an estimated diameter of 5.04 mm. Equation (2) was used to estimate the width of the TIRF region ( $\delta$ ):

$$\delta =f_{obj}\times (NA-n_{sample})$$ (2)

where $n_{sample}$ is the refractive index of the sample, which was estimated to be 1.335. The width of the TIRF region was estimated to be ∼120 µm.

4.3. TIRF imaging setup

We built a custom microscope setup based on a two-deck Olympus IX73 body [46]. Excitation light was either coupled directly into the photonic lantern or first passed through a fiber optic “de-speckler” (F-DS-AFS105-FC/PC, Newport) and then into the lantern. The output fibers of the lantern can be optionally vibrated by a DC rotational motor if more mode scrambling is necessary. We collimated the outgoing lantern light with a collimating lens (EFL = 45 mm, UPlanFLN 4×/0.13, Olympus), sent it through a zoom lens (set to 1.12×, ZBE1A) which maintains collimation, then sent it into a focusing lens (AC508-400-A) that focuses the lantern output beams to the BFP of the objective lens with a total magnification of 7.9× [Supplement 1 ^{(2.2MB, pdf)} , Note 4]. The beams exit the objective lens collimated, and nine TIRF spots were generated at the imaging plane. The fluorescence from the sample was collected by the objective lens and passed through a multi-band notch filter (Di03-R405/488/532/635, Semrock), then sent to a camera (Sona-11, Andor). A removable long-pass dichroic mirror (Di03-R488, Semrock) was placed in the excitation path to allow for simultaneous multi-wavelength excitation. Note that the excitation light is sent into the microscope body through the lower deck.

Our custom-built z-drift correction module sent a NIR beam to the objective lens along a secondary excitation path which utilizes the upper deck of our microscope body. The beam was initially passed through a collimating lens (AC254-060-B), then through a focusing lens (AC508-400-B), reflected off a dichroic mirror (FF750-SDi02, Semrock), then sent to the TIRF region of the objective lens. The reflected beam was collected by the objective lens, then traversed backwards through the focusing lens and then through a 50:50 beam splitter where the light was focused to a camera (DMK 23U618, The Imaging Source). Using a custom-made MATLAB script, calibration was done to determine how shifts in the location of the reflected beam correspond to objective lens drift. During image acquisition the camera monitored the location of the NIR beam, and the script sent signals to the piezo z-stage (Z-insert.100, Piezoconcept) when the z-drift correction is needed. Optics were purchased from Thorlabs unless specified otherwise.

4.4. Interferometer setup

The interferometer experiment was conducted using the setup shown in Fig. 3(a). The collimating lens L1 (UPlanFLN 4×/0.13, Olympus) and the focusing lens L2 (AC254-300-A) were placed in a 4f configuration. We used a 50:50 beam splitter (CCM1-BS014) to separate and then recombine the beams which were to be interfered. Since the nine-core beam profile of the photonic lantern was focused to two planes residing along the two interferometer’s arms, two mirrors with slits attached to them were placed at the two conjugate focal planes and were used to selectively reflect the beams of interest while blocking the others. The interferogram exiting the beam splitter was collimated by L3 (AC254-300-A) and then focused to the objective lens by L4 (AC508-400-A). The fluorescence from the Atto647N dye layer was recorded in a manner similar to TIRF imaging.

4.5. Preparation of dye and bead samples

An Atto647N (AD 647N, ATTO-TEC) or Atto488 (AD 488, ATTO-TEC) dye was diluted in 2,2’-thiodiethanol (TDE; 166782, Sigma) to 2 µM concentration. A 4 µL of each solution was placed onto a glass slide, then sandwiched by a glass coverslip. The edges of the sample were sealed using epoxy. We used 200-nm or 1-µm fluorescent beads (F8806, F8816, ThermoFisher) for imaging freely diffusing nanoparticles or measuring the TIRF penetration depth.

4.6. Cell culture and immunostaining

U2OS cells (HTB-96, ATCC) were grown at 37°C with 5% CO₂ in a humidified incubator. The cells were cultured using McCoy’s 5A medium (SH30200, Cytiva), supplemented with 10% fetal bovine serum (F0926, Sigma) and 1% penicillin/streptomycin (15140122, ThermoFisher). Prior to fixation, the U2OS cells were subcultured in a glass-bottom dish (P35G-1.5-14-C, MATTEK) for 24-48 hours. For fixation, we first added 0.6% paraformaldehyde solution (PFA, 15750, Electron Microscopy Sciences), 0.1% glutaraldehyde (16019, Electron Microscopy Sciences), 0.25% Triton X-100 (93443, Sigma) in 1 × phosphate buffered saline (PBS), then incubated the sample under 37°C for 1 minute. We then incubated the cells in 4% PFA solution, 0.2% glutaraldehyde, and 1 × phosphate buffered saline (PBS) at room temperature for 15 minutes. Then the cells were incubated in 0.1% sodium borohydride (213462, Sigma) at room temperature for 10 minutes. After washing three times with 1× PBS, the sample was incubated with the blocking solution made of BSA (37525, ThermoFisher) and Triton X-100 in 1× PBS for 1 hour. Then the sample was incubated with mouse anti-β-tubulin (T5293, Sigma) at 4°C overnight. After washing three times with 1× PBS, the sample was incubated with goat anti-mouse secondary antibodies labeled with AF647 (A21235, ThermoFisher) for two hours at room temperature. Post-fixation was done with 4% PFA in 1× PBS for 15 minutes at room temperature. After three washes with 1× PBS, the cells were stored in 1× PBS at 4°C.

For live-cell endoplasmic reticulum (ER) samples, U2OS cells were cultured as described above except penicillin/streptomycin was not used. The cells were then subcultured in a glass-bottom dish for 24-48 hours until the cell confluency reached 50–70%. Two solutions were prepared and incubated in room temperature for 5 minutes: one solution consisted of 100 µL of Opti-MEM (31985-062, ThermoFisher) and 1.5 µL of Lipofectamine 2000 (11668030, ThermoFisher) and the other solution consisted of 100 µL of Opti-MEM and 1 µg of the purified plasmid DNA with the StayGold gene (186296, Addgene) [37]. The two solutions were mixed and incubated in room temperature for 20 minutes, then mixed with a solution consisting of 10% fetal bovine serum in McCoy’s 5A medium. The mixture was added to the dish of cells and incubated in 37°C with 5% CO₂ in a humidified incubator for 16-24 hours before imaging.

4.7. SMLM imaging

Microtubules in U2OS cells immunostained with AF647-labeled secondary antibodies were imaged in an imaging buffer containing 1% β-mercaptoethanol (BME, 63689, Sigma) and 1% oxygen scavenger (glucose oxidase, catalase and dextrose) in 50 mM Tris (pH 8) and 10 mM NaCl. The 640 nm excitation (Bolero, Cobolt) was applied continuously, and the 405 nm activation (M405FP1, Thorlabs) was applied periodically via modulation from a function generator. The modulation was set to 0.5 Hz with a 20% duty cycle. The 405 nm LED power was set to 60 µW. Our z-drift correction module was used during image acquisition. The SMLM images shown in Figs. 6(a) and 6(b) were constructed using localizations with ≤ 5 nm localization precision.

4.8. Software usage

We simulated our TIRF excitation scheme using a MATLAB ray tracing script to help determine which design parameters should be used for our lantern and our excitation system. MATLAB was also used for z-drift correction. All image acquisition was done in MicroManager [47]. For single-molecule localization, we used Google Colab to train DECODE [40] models and fit our SMLM data. SMAP [41] was used to render and filter the localized SMLM data to construct super-resolution images. OriginPro was used for statistical analysis of localization data. Fiji [48] was used for image and video processing.

Funding

National Aeronautics and Space Administration10.13039/100000104 (80NSSC21K0624); National Institutes of Health10.13039/100000002 (R35GM138039, U01DK127422).

Disclosures

The authors declare no conflicts of interest regarding this article.

Data availability

Data may be obtained from the authors upon reasonable request.

Supplemental document

See Supplement 1 ^{(2.2MB, pdf)} for supporting content.