Total internal reflection fluorescence (TIRF) microscopy [1] is a popular and useful tool for studying surface features of biological specimens [2] and imaging single molecules [3,4] with a high signal-to-noise ratio. When an excitation beam impinges on the glass-sample interface with an incidence angle that is greater than the critical angle, the illumination beam generates an evanescent field that selectively excites fluorescently labeled biomolecules near the surface with a penetration depth of 50–200 nm [5].

There are a variety of methods to generate TIRF excitation [6,7], with the most common in cellular imaging being objective TIRF. Typically, objective TIRF is achieved by tightly focusing a single laser excitation beam to the periphery of the back focal plane (BFP) of an objective with a numerical aperture (NA) of 1.4 or greater [Fig. 1(a)] [8]. However, this method can cause uneven illumination due to interference fringes from the coherent laser beam and shadowing caused by obscuring objects in cells [9]. These issues have been mitigated via spinning TIRF where the excitation spot is rapidly rotated to various positions around the TIRF annulus of the objective within a single camera exposure [Fig. 1(b)] using refractive optics [10], piezo/galvo mirrors [11–13], acousto-optic deflectors [14], or a digital micromirror device [15]. The result is an incoherent superimposition of the excitation intensity from each azimuthal angle.

Alternatively, an annular mask can be placed at a plane conjugated to the BFP, so that only the TIRF annulus is illuminated and a uniform TIRF excitation field is instantly generated [Fig. 1(c)] [8,16,17]. This implementation is appealing due to the simplified setup with no moving parts and its compatibility with incoherent light sources; however, the use of a mask incurs significant power losses, which can be greater than 99% in some cases [17]. Whereas uniform TIRF illumination with a higher efficiency has been proposed by generating annular beams with axicon optics [18,19], these approaches have a limited field of view (FOV), suffer from a strong zero-order spot that prevents quantitative analysis, or are difficult to implement in a commercial imaging system.

In this Letter, we demonstrate a novel method of generating instantaneous, uniform, and efficient TIRF by coupling an excitation source into a tailored fiber bundle. The basic concepts of our approach are outlined as follows: (i) The individual fibers in the bundle are arranged in a ring at the output end, which is focused and re-imaged within the TIRF annulus of the BFP of the objective, and (ii) the beam exiting from each fiber is spatially incoherent such that they are incoherently summed at the image plane. Figure 1 shows its working principle in comparison with single-spot and spinning TIRF and illustrations of the input and output ends of the fiber bundle.

We designed the annular fiber bundle to be compatible with a 60×/NA1.45 objective (PLAPON60XOTIRFM, Olympus). Our fiber bundle consisted of 137 individual multimode fibers that were close-packed at the input end and arranged in a single ring around a spacer at the output end [Fig. 1(d)]. Considering the size of the BFP of our objective, we designed the bundle such that when the output was magnified ${3} \times$ it would be relayed to the outer annular region of the BFP that generates TIRF illumination. A detailed description of the design process is given in Supplement 1.

 

Fig. 1. Illustrations of (a) single-spot, (b) variable-angle, and (c) annular TIRF described at the back focal plane (BFP) and imaging plane (IP). (d) Illustrations of fiber bundle input and output facets, showing the outer (${{ D }_o}$), inner (${{ D }_i}$), and multimode fiber (${{ D }_{{\rm MMF}}}$) diameters.

Download Full Size | PDF

 

Fig. 2. Experimental setup. ${{\rm FB}_{i/o}}$, Fiber bundle input/output; SMF, Single-mode fiber; MMF, multimode fiber; L, Lens; TL, Tube lens; M, Mirror; FM, Flip mirror; FC, Filter cube. Lower-left inset: Image of fiber bundle beam at the BFP. Scale bar, 2 mm.

Download Full Size | PDF

 

Fig. 3. Single-molecule imaging. Images of the same field of view taken in the presence of 10 nM fluorescent background using (a) TIRF and (b) epi illumination. (c) TIRF image without fluorescent background in a different field of view than in (a) and (b). (d) Intensity distributions generated from 20 images comparing annular TIRF and single-spot TIRF illumination. Inset: illustration of excitation penetration depth while imaging in high background. Scale bars, 5 µm.

Download Full Size | PDF

 

Fig. 4. Beam characterization. Dye layer images recorded in TIRF using (a) a diode laser with shaker motor, (b) a diode laser without shaker motor, and (c) an LED. (d) Line profiles taken along diagonal as indicated by solid lines in (a), (c). Scale bars, 50 µm.

Download Full Size | PDF

The fiber bundle was fed into a custom-made TIRF microscope built around an Olympus IX73 body (Fig. 2). The bundle output was collimated by a lens ${\rm L}_1$ (${f_1} = 100\;{\rm mm}$) and focused to the BFP of the objective by a lens ${\rm L}_2$ (${f_2} = 300\;{\rm mm}$), which was mounted on a manual $xy$ and motorized $ z $ translation stage. By translating ${\rm L}_2$ one inch along the optic axis, we can change the size of the beam at the BFP to achieve TIRF or weakly focused epi illumination. The beam was reflected by a filter cube (TRF89901v2, Chroma) and the fluorescence signal was detected by a scientific complementary metal oxide semiconductor camera (sCMOS, Zyla 4.2 PLUS, Andor) or an electron multiplying charged-coupled device (EMCCD, iXon Ultra 897, Andor). The sCMOS was used to demonstrate large FOV (${222} \times 222\;\unicode{x00B5}{\rm m}^2$) TIRF, but the EMCCD was used in all other experiments. An additional ${1.66} \times$ magnification system was installed prior to the EMCCD for a total image magnification of ${100} \times$ to satisfy the Nyquist criterion for single-molecule imaging. An image of the fiber bundle output taken at the BFP is shown in an inset of Fig. 2.

Two lasers—488 nm and 638 nm (06-MLD, Cobolt) as well as a 470 nm light emitting diode (LED, M470F3, Thorlabs), which was directly coupled via SMA connectors to the fiber bundle, were the light sources used with the fiber bundle to demonstrate multicolor imaging and TIRF with coherent or incoherent sources. The diode lasers were first coupled into a 400 µm core multi-mode fiber (MMF, M28L01, Thorlabs) that was attached to a shaker motor (JRF370-18260, ASLONG) to degrade the coherence of the beam [20] before coupling into the fiber bundle input. For comparison with single-spot TIRF, a 491 nm or 640 nm laser (04-01 Calypso, 05-01 Bolero, Cobolt) was coupled to a single-mode fiber (P5-488PM-FC-1, Thorlabs) and collimated by a lens ($f = 300\;{\rm mm}$) and directed to the microscope by a flip mirror. We observed a total power efficiency of $\sim{30}\%$ when using the 638 nm laser, with 76% of the total loss occurring at the coupling of the MMF into the fiber bundle. This was expected, as $\sim{50}\%$ of the fiber bundle input is void and the MMF was roughly butt-coupled to the fiber bundle input, with an intentional gap between the MMF and fiber bundle to compensate for the difference in their core sizes. A further optimized design will greatly mitigate the power losses.

We first demonstrated the shallow excitation depth of our TIRF field using the 638 nm diode laser coupled into the annular fiber bundle. Single-molecule images were recorded of surface-immobilized IgG antibodies labeled with Alexa Fluor 647 (AF647) at a degree of labeling of $\sim{1.1}$ [21] in the presence of 10 nM STAR635 diluted in an imaging buffer as fluorescent background. Our result shows that the single molecules are able to be resolved with an average signal-to-background ratio of 2.3 ($n = 20$) when imaged in TIRF [Fig. 3(a)], whereas the background overwhelms the single-molecule signal when imaged under epi illumination [Fig. 3(b)]. Note that at this concentration of fluorescent background, even a slight degree of far-field excitation due to partial TIRF could substantially increase the background level [20], indicating that our annular fiber bundle produces clean TIRF excitation. The effective penetration depth was roughly 235 nm, which was estimated by measuring the diameter of 1 µm fluorescent beads (F8816, ThermoFisher) [22]. In addition, we obtained single-molecule images without fluorescent background [Fig. 3(c)] and analyzed the intensity of each spot, comparing annular and single-spot TIRF illumination. The annular TIRF intensity distribution has narrower peaks and two observable populations, while the single-spot TIRF distribution is broader and less defined [Fig. 3(d)]. This demonstrates that our annular TIRF excitation field enabled the distinction between antibodies labeled with one or two fluorophores, which is only possible with uniform excitation [20].

 

Fig. 5. TIRF imaging of actin in U2OS cells using (a) LED annular TIRF, (b) laser diode annular TIRF, and (c) single-spot objective TIRF. Arrow indicates direction of single-spot TIRF excitation. Scale bars, 10 µm.

Download Full Size | PDF

We further examined the uniformity of our illumination by measuring beam profiles taken by exciting a $\sim{5}\;\unicode{x00B5}{\rm m}$ thick dye layer (Atto488 or STAR635) sandwiched between a microscope slide and coverslip and imaged with the sCMOS detector for a ${222} \times 222\;\unicode{x00B5}{\rm m}^2$ FOV (Fig. 4). The uniformity was characterized by the root mean square (RMS) of the line profile intensity [20]. Detector noise from the sCMOS degraded the RMS values of the diode laser [Fig. 4(a)] and LED [Fig. 4(c)] to 0.79 each. But after smoothing via adjacent averaging, the RMS values were 0.84 and 0.86, respectively. In the central ${82} \times 82\;\unicode{x00B5}{\rm m}^2$ region representing the EMCCD FOV, the respective smoothed RMS values were 0.91 and 0.92. Without shaking, the diode laser had an RMS of 0.70 before and 0.77 after smoothing. Raw and smoothed line profiles are plotted in Fig. 4(d) for the LED and diode laser with shaking.

Artifacts from single-spot TIRF illumination are often more severe when imaging subcellular structures in cells. To demonstrate the homogeneity of the TIRF excitation generated by our fiber bundle, we imaged U2OS cells that were stained with Alexa Fluor 488 phalloidin (A12379, ThermoFisher) to label filamentous actin. Images taken with the 470 nm LED or 488 nm diode laser coupled with our fiber bundle are compared with single-spot TIRF. Figure 5 shows that our annular fiber bundle generates uniform TIRF illumination for artifact-free imaging regardless of the light source, whereas single-spot TIRF results in strong interference fringes from the excitation source, scattering and/or shadowing artifacts from the unidirectional excitation.

Finally, we demonstrated high-throughput stitched imaging with our annular fiber bundle using a 15% image overlap on the phalloidin stained U2OS cells to record a ${550} \times 550\;\unicode{x00B5}{\rm m}^2$ area. To demonstrate the ease of switching between illumination modes, Fig. 6 shows a comparison of the same FOV taken in TIRF and epi illumination using the LED light source, where the inset shows a magnified view of the boxed region. Surface features such as focal adhesions are clearly resolved under TIRF excitation, whereas these features are less visible under epi illumination due to elevated background fluorescence from non-surface actin structures.

In this Letter we demonstrated a method of instantly achieving shadowless TIRF excitation using an annular fiber bundle. We showed that this method is suitable for multicolor imaging and generates a uniform and shallow excitation field. It is possible to use other popular TIRF objectives such as a 100×/NA1.49 objective if one designs a new fiber bundle and modifies the imaging system slightly. Our annular fiber bundle was designed to be suitable with both a laser or LED; however, LED excitation has a very limited power throughput [23] and thus is not suitable for imaging weakly fluorescent samples such as single molecules. If one only uses a laser as an excitation source, a more optimized design—for example, utilizing a shorter focal length lens ${\rm L}_1$ and fewer MMF fibers—is likely to increase the power throughput. Versatile control of the incidence angle is possible via calibration of the motorized translation stage on ${\rm L}_2$, which will be useful for depth-matched multi-color TIRF illumination [12] and 3D reconstruction by multi-angle TIRF [24]. Polarization-based TIRF experiments [25] may be feasible by generating radially or azimuthally polarized light using a segmented half waveplate [26]. With no moving parts, our method is compatible with video-rate live-cell TIRF imaging. We expect our method will make quantitative TIRF imaging systems more accessible.

 

Fig. 6. High-throughput stitched imaging of actin stained U2OS cells imaged under (a) TIRF and (b) epi illumination using the LED source and fiber bundle. Upper-right inset: Zoomed view of boxed region. Scale bars, 50 µm.

Download Full Size | PDF