Open Access
Add to BibSonomy
Abstract
As a wide-field imaging technique, super-resolution localization microscopy (SRLM) is theoretically capable of increasing field-of-view (FOV) without sacrificing either imaging speed or spatial resolution. There are two key factors for realizing large FOV SRLM: one is high-power illumination over the whole FOV with sufficient illumination homogeneity and the other is large FOV signal detection by a camera that has large number of pixels and sufficient detection sensitivity. However nowadays, even though the state-of-art scientific complementary metal-oxide semiconductor (sCMOS) cameras provide single molecule fluorescence signal detection ability over an FOV of more than 200 μm × 200 μm, large FOV SRLM still has not been achieved due to the lack of high-power homogeneous illumination. In this paper, we report large FOV SRLM with a high-power homogeneous illumination system. We demonstrate experimentally that our illumination system, which is based on a newly designed multimode fiber combiner, is capable of providing sufficient illumination intensity (~4.7 kW/cm2 @ 640 nm) and excellent illumination homogeneity. Compared with the reported approaches, our illumination system is advantageous in laser power scaling and square-shape illumination without beam clipping. As a result, our system makes full use of the sensor of a representative Hamamatsu Flash 4.0 V2 sCMOS camera (2048 × 2048 active pixels) and achieves a FOV as large as 221 μm × 221 μm with homogeneous spatial resolution. The flexible solution for realizing large FOV SRLM reported in this paper pushes a significant step toward the development of SRLM.
© 2017 Optical Society of America
1. Introduction
Super-resolution localization microscopy (SRLM), including photoactivated localization microscopy (PALM) [1], stochastic optical reconstruction microscopy (STORM) [2], direct stochastic optical reconstruction microscopy (dSTORM) [3] and many others [4], achieves ultra-high spatial resolution up to 20~30 nanometers by positioning and reconstructing single molecules from thousands or even tens of thousands of raw images. As intrinsically a wide-field imaging technique, SRLM has the advantage of increasing field-of-view (FOV) without sacrificing either imaging speed or spatial resolution.
At first, the most popular detector used in SRLM is Electron Multiplying Charge Coupled Device (EMCCD) camera which was designed to provide ultimate sensitivity at fast imaging speed [5]. However, for the EMCCD cameras used in SRLM, the number of active pixels is typically limited to 512 × 512, which consequently corresponds to a maximum FOV of approximately 50 μm × 50 μm at the sample plane, assuming a typically pixel size of 100 nm at the sample plane [6]. Such an FOV is insufficient for observing many biological phenomena which are best interpreted in large FOV, for example, volumetric mapping of synaptic connectivity at multiple scales [7], spatiotemporal imaging of vesicular trafficking to the immune synapse [8, 9], high-throughput imaging of cellular phenotypic responses [10] or bacterial cell division [11].
In the past decades, complementary metal oxide semiconductor (CMOS) technology received great improvements in the sensor design and fabrication due to the high-volume demand in consumer applications such as cell phone cameras and video cameras [12], and a new generation of scientific-grade CMOS (sCMOS) camera was developed into a promising low-light detector. Shortly after the launch of sCMOS camera in 2009, the potentials of using sCMOS camera as an alternative detector in SRLM was explored theoretically [13] and experimentally [14–16], because sCMOS can provide simultaneously large sensor format (2000 × 2000 pixels or more), high pixel rate (> 400 MHz), and excellent sensibility which is sufficient for single molecule fluorescent imaging. These features open up a promising opportunity for SRLM: an FOV of more than 200 μm × 200 μm is possible when a mainstream sCMOS camera (where the sensor has 2000 × 2000 or more active pixels) is used as the detector, assuming a typically pixel size of 100 nm at the sample plane [6]. However, the large sensor format of the sCMOS cameras has not yet been fully exploited for SRLM, and the largest reported FOV is limited to be 100 μm × 100 μm [17].
The major obstacles which hinder a further increase of FOV are from the illumination intensity and the illumination homogeneity requirements at the sample plane. For the illumination intensity, it is generally recommended to be 1-10 kW/cm2 [17] for most fluorescent probes. Under a typically condition of 5 kW/cm2 and 33% system transmission efficiency [18], the required output power from any excitation lasers in SRLM is estimated to be 6 W to ensure that the full sensor of a typical sCMOS camera (for example, Hamamatsu Flash 4.0 V2 with 2048 × 2048 active pixels) can be used in SRLM. In this case, an FOV of about 200 μm × 200 μm can be achieved. Unfortunately, for the common excitation laser wavelengths used in SRLM (for example, 561 nm and 640 nm), it is not easy to obtain such a high power from a single laser due to either availability or cost.
On the other hand, illumination homogeneity is critical in obtaining high-quality super-resolution images. It is reported that a Gaussian illumination will cause significant field-dependent image resolution across the FOV [17]. Therefore, several efforts have been made to improve illumination homogeneity in SRLM. Manley et al [17] combined a rotating diffuser with a microlens array to achieve efficient homogenous illumination for an FOV of 100 μm × 100 μm. In their approach, a single laser was used. However, since FOV is ultimately limited by laser power, imaging with a larger FOV will require the investment on a more powerful laser source which might not be available in the market or is very expensive. In another report, Ries et al [19] combined a multimode fiber with a laser speckle-reducer to provide efficient homogeneous illumination across the entire FOV of about 30 μm × 30 μm. However, the speckle-reducer used in their system results in significant loss of total laser power, and requires the exposure time to be greater than 5 ms.
Here we report a new method that can provide high-power and excellent homogeneous illumination for SRLM. This method is based on a newly designed multimode fiber combiner, which consists of four circular-core input fibers, a fused biconical taper, a square-core multimode output fiber, and a high-frequency vibration motor. The multimode fiber combiner enables separation of the mechanical vibration from laser sources, delivery of multiple lasers with multi-watt output power, and laser power scaling if necessary. The square-core multimode fiber improves illumination uniformity and provides a square-shape illumination without beam clipping to match the square sensor of typical sCMOS cameras. Agitating the square-core fiber [20] with high-frequency vibration [21, 22] allows speckle reduction with minimal total laser power loss, and enables high-speed imaging with an exposure time down to 0.5 ms. After a careful characterization on the performance of the fiber combiner and the entire illumination system, we verify that our method is capable of fully exploiting the large sensor format of sCMOS camera for SRLM.
2. Methods
2.1 Design of the multimode fiber combiner for high-power laser delivery
The multi-mode fiber combiner was custom-made from Tianjin Junlaser Technology, China. As shown in Fig. 1, this fiber combiner has four input fibers and one output fiber, and consists of three parts. Part 1 includes four identical 3-meter graded-index multimode fibers (Corning), each fiber has a core diameter of 50 μm, a cladding diameter of 125 μm, and an SMA connector at the input end. The numerical aperture (NA) of the input fibers is 0.20. These input fibers were connected to the output fiber via a fused biconical taper (FBT, see Part 2 in Fig. 1). The output fiber is a 6-meter square-core multi-mode fiber with a 200 μm × 200 μm square core, a 420 μm cladding diameter, and an SMA connector at the output end. The NA of the square-core fiber is 0.22. The square-core fiber is bended and curled into a radius of about 8 cm, and bound with a high-frequency vibration motor (JRF370-18260, ALSONG, China). The square-core fiber and the vibration motor assemble into Part 3.
Fig. 1 A schematic diagram (a) and photography (b) of the fiber combiner. The fibers in Part 1 are- the same type of multimode fibers: 50 μm core diameter, 125 μm cladding diameter, and NA = 0.20 (Corning); The fiber in Part 3: 200 μm × 200 μm square core, 420 μm cladding diameter, NA = 0.22 (CeramOptec); FBT: fused biconical taper.
2.2 Optical system for testing the illumination homogeneity at the exit of the fiber combiner
As shown in Fig. 2, a 639 nm laser (Genesis MX-639-2000, Coherent) is guided by two mirrors (M1-M2), two irises, and then focused by a fiber coupler (L1) into one of the input fibers in the fiber combiner. The output beam from the fiber combiner is focused onto a Hamamatsu Flash 4.0 V2 sCMOS camera by lens L2 and L3.
Fig. 2 The optical setup for testing the illumination homogeneity at the exit of the fiber combiner. Laser: 639 nm (Genesis MX-639-2000, Coherent); SH: Electronic shutter (UNIBLITZ VS14, Vincent Associates); ND: Neutral density filter (NDC-50C-2M, Thorlabs); M1-M2: Aluminum mirrors (BB1-E02, Thorlabs); Iris-Iris2: Iris diaphragms (ID25, Thorlabs); L1: Fiber coupler with a focusing lens (f = 11 mm, SMA connector, PAF-SMA-11-A, Thorlabs); L2: Microscope objective (Plan 10X/NA0.25, Olympus); L3: Achromatic doublet lens (f = 100 mm, AC254-100-A, Thorlabs); sCMOS: Flash 4.0 V2 sCMOS camera (Hamamatsu).
2.3 Optical system for testing the illumination homogeneity at the sample plane
As shown in Fig. 3, the optical system is based on an Olympus IX 71 inverted microscope. Each laser beam (Laser 1 or Laser 2) is guided by two aluminum mirrors and then focused by a fiber coupler into the fiber combiner. The electronic shutter is used to control the duration of laser irradiance and the neutral filter is used to control laser intensity. The output beam from the fiber combiner is imaged into the sample by L3-L7 and an Olympus 60 × /NA1.20 water-immersion objective. L5 and L6 are used to build a pair of relay lenses. The fluorescence from AB9 solution (3844-45-9, Sigma-Aldrich) [17] is collected with the same objective, filtered with a dichroic mirror and an emission filter, and then focused onto a Hamamatsu Flash 4.0 V2 sCMOS camera. The pixel size at the sample plane is 108 nm. Laser 1/Laser 2: 405 nm laser from CNILaser (MLL-III-405); 473 nm laser from CNILaser (MLL-III-473); 561 nm laser from Coherent (Genesis CX-561-3000); 639 nm laser from Coherent (Genesis MX-639-2000); 640 nm laser from LaserWave (LWRL640-3W). CNILaser and LaserWave are companies both from China.
Fig. 3 The optical setup for quantifying the illumination homogeneity at the sample plane. M1-M7: Aluminum mirrors (Thorlabs); Iris1-Iris4: Iris diaphragms (ID25, Thorlabs); L1/L2: Fiber coupler with a focusing lens (f = 11 mm, PAF-SMA-11-A, Thorlabs); L3: Microscope objective (10 × /NA0.25, Olympus); L4: Achromatic doublet lens (f = 400 mm, AC508-400-A, Thorlabs); L5: Achromatic doublet lens (f = 35 mm, AC254-35-A, Thorlabs); L6: Achromatic doublet lens (f = 150 mm, AC508-150A-ML, Thorlabs); L7: Achromatic doublet lens (f = 300 mm, AC508-300-A, Thorlabs); TL: Tube lens; Objective: Water-immersion microscope objective (60XW/NA1.2, Olympus); DM: Dichroic mirror (ZT405/488/561/647rpc, Chroma); F: Emission filter (ZET405/473/561/640m, Chroma); sCMOS: sCMOS camera (Flash 4.0 V2, Hamamatsu). Pos 1 – Pos 6 indicate the positions used for measuring laser power. Laser 1/Laser 2: 405 nm laser from CNILaser (MLL-III-405); 473 nm laser from CNILaser (MLL-III-473); 561 nm laser from Coherent (Genesis CX-561-3000); 639 nm laser from Coherent (Genesis MX-639-2000); 640 nm laser from LaserWave (LWRL640-3W).
2.4 Determining beam homogeneity
It is well-known that laser speckle will be introduced when a laser passes through a multi-mode fiber [23], thus producing a non-uniform intensity distribution which is not desirable for SRLM. In order to characterize the illumination homogeneity from our multimode fiber combiner, we adopted a widely used parameter called speckle contrast as a measurement for the speckle pattern [23]:
where I, <I> and σ1 are the intensity, the mean intensity, and the standard deviation of the intensity in an observed speckle pattern, respectively. If the speckle contrast becomes zero, the speckle noise is perfectly eliminated and the intensity profile is uniform.We used the horizontal centerline rather than the whole image from an illumination area for calculating the illumination homogeneity, because the Hamamatsu Flash 4.0 V2 sCMOS camera used in this study outputs a smaller image with decreased exposure time due to insufficient bandwidth. Furthermore, to obtain a speckle contrast threshold (hereafter called homogeneity threshold) for defining homogeneous illumination, we performed bright-field imaging of the field diaphragm in a commercial Olympus IX71 microscope under Kohler illumination. The speckle contrast from the bright-field image was calculated to be 0.128. Therefore, in this study we consider an illumination pattern as homogeneous if the speckle contrast of the pattern is smaller than 0.128.
2.5 Sample preparation
COS-7 cells were grown in Roswell Park Memorial Institute 1640 Medium (RPMI 1640 Medium, Gibco) supplemented with 10% fetal bovine serum (FBS, Gibco) and 1% penicillin (10,000 IU/mL)/streptomycin (10,000 μg/mL) at 37 °C with 5% CO2. Before imaging, cells were seeded on 35-mm glass-bottom dishes overnight. For microtubule labeling, cells were washed three times with PBS prewarmed to 37°C and fixed with 3% paraformaldehyde, 0.05% glutaraldehyde and 0.2% Triton X-100 diluted in PBS for 15 min. After that, cells were washed with PBS for three times and permeabilized and blocked with blocking buffer (3% BSA and 0.05% Triton X-100 in PBS) for 1 h while gently rocking. Cells were incubated with mouse monoclonal anti-α-tubulin antibody (Sigma T5168, 1:500 diluted in blocking buffer) at room temperature for 1.5 h. After washed with blocking buffer for three times, cells were incubated with Alexa Fluor 647 goat anti-mouse IgG (A-21236, Invitrogen) at a concentration of approximately 10 μg/mL for 1 h. Samples were washed three times with PBS and stored in PBS at 4 °C until imaging.
3. Results and discussions
3.1 Transmission efficiency of the fiber combiner
SRLM requires the use of several lasers with different wavelength for single molecule activation and excitation. The laser wavelengths in SRLM include typically 405 nm, 473 nm, 561 nm, and 640 nm. We define the transmission efficiency of the fiber combiner as the ratio of the laser power between position 4 and position 3 (that is, Efficiency = Ipos4/Ipos3, see Fig. 3). The transmission efficiency of the fiber combiner is wavelength dependent and can be optimized by controlling the process of fused biconical taper. Using the data shown in Table 1, the transmission efficiency in the visible light range (400 nm −700 nm) is calculated to be 70% at 405 nm, at 92% 473 nm, at 97% 561 nm, and at 92% 639 nm, respectively. We also measured the laser power values at different positions of the optical system (see Fig. 3 and Table 1). The overall transmission efficiency is higher than 50% for both 561 nm and 640 nm, which is higher than that from a conventional SRLM system (appr. 33%) which uses dichroic mirrors for laser combining [18]. Further improvements in the transmission efficiency may be possible by optimizing the dichroic mirrors, aluminum mirrors and lenses (see the supplementary information in [17]).
Table 1. Transmission efficiency of the lasers with different wavelengths a
As discussed previously, it is required to use a high-power laser with ~6 W in the laser output (or 2 W at the sample plane) to provide sufficient illumination for a full FOV of 200 μm × 200 μm. Since a single laser with such a high power may not be available in the market or is very expensive, especially for some special wavelengths (for example, 561 nm), an alternative solution to do so is combining several relative low-power lasers with the same or close wavelengths. Using the fiber combiner developed in this study, it is practical to couple up to four lasers into the same output fiber (see Fig. 3). As shown in Table 2, by coupling a 2.2 W 639 nm laser and a 3.3 W 640 nm laser into the fiber combiner, we are able to obtain a 2.3 W laser output at the sample plane. If necessary, this fiber combiner is still capable of combining two more lasers to provide an even higher illumination intensity. It is worth noting that LaserWave 640 nm laser has a near TEM00 mode and thus results to a lower transmission efficiency (39%) than that from Coherent 639 nm laser (50% efficiency) with a TEM00 mode.
Table 2. Transmission efficiency of the combined laser.
3.2 Illumination homogeneity at the exit of the fiber combiner
Using the optical setup in Fig. 2, we characterized the relationship between the speed of the vibration motor and the illumination homogeneity. The speed of the vibration motor is adjustable in the range of 0 to 10000 round per minute (RPM), with an interval of 1000 RPM. The images of the output beam were recorded using a Hamamatsu Flash 4.0 V2 sCMOS camera and analyzed using the equation in Section 2.4. As shown in Fig. 4, the speckle contrast value is 0.311 when the vibration is off, but decreases significantly with the vibration motor turned on. A minimum speckle contrast value of 0.076 can be obtained when the rotational speed is between 6000 RPM and 10000 RPM. Note that a smaller speckle contrast value means better illumination homogeneity. Therefore, in the following experiments the rotational speed is fixed to be 8000 RPM.
Fig. 4 (a) The speckle contrast as a function of the rotational speed of the vibration motor. The illumination area was averaged over three data sets with 10 consecutive frames in each data set. The exposure time of the sCMOS camera was set to be 20 ms; (b) The dependence of the speckle contrast on the camera exposure time for several rotational speeds.
We also measured the dependence of the speckle contrast on the camera exposure time for several rotation speeds. As shown in [Fig. 4(b)], for a given rotation speed, the speckle contrast value decreases gradually with the increase of exposure time, especially when the exposure time is less than 2 ms. For the optimal rotational speed of 8000 RMP used in this study, the speckle contrast value becomes almost stable when the exposure time is 15 ms or longer.
Using the optical system in Fig. 2, we investigated the illumination homogeneity at the exit of the fiber combiner. Conspicuous speckle noises can be observed when the vibration motor was turned off, resulting in a speckle contrast value of 0.311 [Fig. 5(a)]. The speckle contrast was obviously improved after turning on the vibration motor [Fig. 5(b)]. The normalized intensity profiles showed in [Fig. 5(c)] clearly demonstrated the effectiveness of turning on the vibration motor.
Fig. 5 The illumination homogeneity at the exit of the fiber combiner. The line intensity profiles in (c) correspond to the green lines in (a-b). The calculation was based on single frames and a 639 nm laser (Genesis MX-639-2000, Coherent) was used. Scale bar: 300 μm.
According to Schuberts et al [20], a square-core multimode fiber provides more possibility for ray mixing during light propagation than a circular-core multimode fiber, thus enabling a more homogeneous output beam. And, the square-shape output beam pattern from the square-core fiber matches better with the square-shape sensor of the sCMOS camera used for SRLM with large FOV. Additionally, it is worth noting that, compared to the illumination homogeneity characterization at the sample plane, the illumination homogeneity measured at the exit of the fiber combiner provides a direct observation onto the performance of the fiber combiner with minimum artifacts from the measurement system.
3.3 Characterizing the illumination homogeneity at the sample plane using fluorescent solution
The output beam from the fiber combiner can be guided to a microscope through the use of proper lenses and mirrors, thus providing homogeneous epi-illumination for SRLM [19]. We used the optical setup shown in Fig. 3 to characterize the illumination homogeneity at the sample plane. A thin layer (~8 μm) of AB9 solution [17] was used to avoid out-of-focus fluorescence which may smoothen the fluorescence image. We tested the illumination homogeneity for three different laser wavelengths: 405 nm, 639 nm, and a combined of 639 nm and 640 nm. Note that the fluorescence image from the fluorescent solution (corresponding to the illumination area) covers approximately 200 μm × 200 μm, and that the combination of a 639 nm laser and a 640 nm laser is necessary to enable an illumination intensity of ~5 kW/cm2. The results are presented in Fig. 6. It is clear that the illumination homogeneity is improved after turning on the vibration motor [Figs. 6(a)-6(d)]. And, the illumination homogeneity for all of the three laser wavelengths becomes satisfying after turning on the vibration motor, evidenced by the images and the line profiles shown in [Figs. 6(c)-6(h)]. Specifically, the speckle contrast is 0.118 when the vibrator is turned off [Fig. 6(a), 405 nm]. After turning on the vibration motor, the speckle contrast values are optimized to be: 0.087 for 405 nm, 0.079 for 639 nm, and 0.073 for the combined laser. Additionally, due to the coupling of the fiber, the intensity distribution has a higher intensity on one side. Since the speckle contrast values are all smaller than the homogeneity threshold (0.128, see Section 2.4), homogeneous illumination is achieved for all of the three cases.
Fig. 6 Characterizing the illumination homogeneity at the sample plane using fluorescent solution. (a-d) 405 nm; (e-f) 639 nm; (g-h) the combined laser (639 nm and 640 nm); The vibration motor is turned off for (a-b), and turned on for (c-h). The line intensity profiles in the bottom correspond to the images on the top, horizontally along the center of the illumination area. The calculation was based on single frames. Scale bar: 50 μm.
It is worth mentioning that the speckle contrast was calculated from the light intensity variations across the FOV or in time. However, such intensity variations can be originated not only from the illumination homogeneity, but also from camera noises and/or shot noise (here we called detection noise). The speckle contrast from the detection noise can be calculated from a separate measurement called PTC method [15]. The PTC method is a well-established technique for characterizing the noise performance of the sCMOS camera used in this study. From the PTC results, we can obtain the total detection noise at any given signal level. Since the mean illumination intensity in [Fig. 6(e)] was calculated to be about 4400 photon/pixel, the corresponding detection noise (including shot noise and camera noises) is about 63 electron. Therefore, the speckle contrast is: 63/(4400*0.65) = 0.022. Here 0.65 is the Quantum Efficiency of the sCMOS camera at 680 nm. This indicates that the detection noise plays a non-negligible role to the light intensity variations (and thus speckle contrast).
We concluded that a homogeneous illumination of about 200 μm × 200 μm was achieved for all the lasers used in this study. Furthermore, we found out that it is easy to change the size of the illumination area by adjusting the focal length of L5 in Fig. 3 with no or small loss of total laser power. It is worth noting that the illumination homogeneity in this section was analyzed using an area smaller than the full sensor of the sCMOS camera, so that the performance of the entire illumination area can be explored. Later we changed the lens of L5 to enlarge the illumination area, so that the full sensor of the sCMOS camera can be used (see Section 3.5 for experimental demonstration).
We also investigated whether the illumination homogeneity is acceptable for high speed imaging. We found out that the illumination homogeneity decreases with shorter camera exposure time [Fig. 7(a)], but the speckle contrast value is still less than the homogeneity threshold (0.128) for an exposure time of 0.5 ms which is short enough to realize SRLM at video-rate [24].
Fig. 7 The distributions of the speckle contrast for different camera exposure times. All pixels in the horizontal centerline of the illumination area are used for statistics. The statistics were from 500 successive frames of AB9 solution. The excitation laser was 639 nm (Genesis MX-639-2000, Coherent).
3.4 Characterizing the illumination homogeneity at the sample plane using fluorescent beads
We further used Coherent 639 nm laser as a representative laser source to characterize the illumination homogeneity at the sample plane. Fluorescent beads (F8807, FluoSpheres, Molecular Probes) were sealed between two cover slips, mounted onto a translation stage, moved to five different positions shown in [Fig. 8(a)], and illuminated by the 639 nm laser. 500 successive fluorescence images from the same beads [see Fig. 8(b) for an example image] were captured and analyzed the fluorescence intensity using the following steps (here we take bead 1 as an example). Step 1: we used a maximum likelihood estimator called Maliang [25] to analyze a data set containing 500 successive fluorescence images from bead 1 position 1, and obtained the total photons from the bead in each image. Step 2: we calculated the averaged total photons from the data set. Step 3: we repeated the analysis to obtain the averaged total photons for bead 1 in all of the five positions, and then normalized them according to the value in position 1. Step 4: we calculated the mean and std of the averaged total photons from the five positions.
Fig. 8 Characterizing the illumination homogeneity at the sample plane using fluorescent beads. (a) A schematic diagram showing the positions of the fluorescent beads; (b) A representative fluorescence image for the fluorescent beads in position 1; (c) Normalized fluorescence intensities according to the intensity in position 1. The fluorescence intensities are summed over all the signal area for each beads; (d) Localization precision of the same beads in different positions. The total fluorescence intensities and the localization precision were averaged over 500 successive fluorescence images. Scale bar: 50 μm.
We repeated the fluorescence intensity analysis for all of the three beads shown in Fig. 8(b) and obtained the following results: the mean and std for bead 1 is 0.96 and 0.053, the mean and std for bead 2 is 1.04 and 0.047, the mean and std for bead 3 is 1.01 and 0.040, respectively. We calculated the heterogeneity by dividing the std by the mean, and obtained ~5% heterogeneity. Therefore, we concluded that the intensity homogeneity of the output beam on the sample plane is within approximately 95%. Additionally, we performed molecule localization on the fluorescence images using MaLiang [25] and obtained the localization precision results shown in [Fig. 8(d)]. Based on the localization precision data for the five positions, the statistical results are: mean = 5.12 nm and std = 0.23 nm for bead 1, mean = 5.35 nm and std = 0.24 nm for bead 2, mean = 5.54 nm and std = 0.33 nm for bead 3. The localization results demonstrate a field-independent image resolution over a large FOV (198 μm × 198 μm).
3.5 Super-resolution localization microscopy with the full sensor of the sCMOS camera
To illuminate the full sensor of the sCMOS camera (2048 × 2048 pixels), we changed the focal length of Lens 5 from 35 mm to 30 mm (Fig. 3). In this way, the FOV increases to 221 μm × 221 μm, which is approximately 5 times larger than the previous work [17]. We performed super-resolution imaging on COS-7 cells in which microtubules were labelled with Alexa Fluor 647. A large FOV super-resolution image [Fig. 9(a)] was obtained from a set of 10000 raw image frames with 20 ms exposure time for each frame.
Fig. 9 Homogeneous super-resolution imaging of COS-7 cells using the full sensor of the sCMOS camera. (a) Super-resolution image of COS-7 cells where microtubules were labelled with Alexa Fluor 647; The FOV is 221 μm × 221 μm; (b) Magnified views of the regions with color boxes marked in (a); (c) Pixel intensity profiles through the marked lines in (b) with rectangular box through a microtubule. Red curves are fits to two Gaussian curves. (d) Pixel intensity profiles through the marked lines in (b) with circular box through two adjacent structures. Scale bar: (a) 50 μm, (b) 3 μm.
We estimated the spatial resolution of the super-resolution image via two different methods. The first method is based on the structure feature of single microtubule. As shown in [Fig. 9(c)], two maxima in the line intensity profiles of single microtubule can be found. The distance between the peaks of the two maxima, corresponding to the measured width of the microtubule, is used to present the resolution of the image [17, 24]. Using this method, we confirmed a nearly field-independent image resolution of about 40 nm across the full FOV [Fig. 9(c)]. This resolution value is consistent with that in [17, 24].
The second method uses the separation of two adjacent structures as the standard for spatial resolution estimation. This method is inspired by the Rayleigh criterion, but depends on the intensity distribution from adjacent structures, rather than the Airy disks which are not available from [Fig. 9(a)]. Practically, we used an ImageJ plugin to extract the pixel values for the tilted dash lines in [Fig. 9(d)]. Then, we used Matlab to fit the pixel values to two Gaussian curves and obtained the distance between two maxima. This distance is reported as the separation distance of two adjacent structures. Based on this method, we confirmed a nearly field-independent image resolution of about 71 nm across the full FOV [Fig. 9(d)]. However, it is worth noting that there is no well-recognized standard that can be used to estimate the spatial resolution of a super-resolution image in SRLM, and that the aim of this paper is to provide a homogeneous illumination system for SRLM with large field of view, rather than a super-resolution microscope system with optimal resolution.
Finally, we noticed that localization precision is a key factor for determining image resolution [26], and that localization precision is reversely proportional to the quadratic root of the fluorescence intensity of single molecule [13]. Therefore, we calculated the fluorescence intensity distributions from individual Alexa Fluor 647 molecules [Fig. 10(c)]. We observed similar fluorescence intensity distribution inside several small areas [Fig. 10(d)]. Since the imaging area has been expanded to the full sensor of the sCMOS camera, it is necessary to characterize again the illumination homogeneity at the sample plane using fluorescent solution. The results were presented in [Figs. 10(a) and 10(b)], and the speckle contrast value was calculated to be 0.036, indicating a homogeneous illumination. Importantly, for the sensor with 2048 × 2048 active pixels, a total number of 3974809 pixels exhibit fluorescence intensity larger than 80% of the peak intensity, indicating a high-efficiency use (95%) of the full sensor area. Importantly, such an efficiency is achieved without beam clipping. If a circular core multimode fiber rather than a square core multimode fiber is used in the output end of the fiber combiner, beam clipping would be required to match the square shape of the sensor, resulting in a significant loss of total laser power. Otherwise, for a circular illumination without clipping, more than 20% of the active pixels would capture no signal due to the mismatch between the circular shape of the illumination and the square shape of the sensor.
Fig. 10 Characterizing the illumination homogeneity across a FOV of 221 μm × 221 μm. (a-b) Fluorescence intensity image and distribution of AB9 solution. The image and the statistics are both from a single image frame; (c) Single molecule fluorescence intensity distribution of Alexa Fluor 647 in COS-7 cells. A maximum likelihood estimator called MaLiang [25] was used to calculate the fluorescence intensities from individual Alexa Fluor 647 molecules, and a total of 10000 successive raw images were used; (d) Normalized single molecule fluorescence intensity distribution inside several small area marked in (c). Scale bar: 50 μm.
4. Conclusion
We presented a high-power and excellent homogeneous illumination system for fully exploiting the large sensor of a representative Hamamatsu Flash 4.0 V2 sCMOS camera. This illumination system is based on a newly designed multi-mode fiber combiner. We verified experimentally that this illumination system is capable of providing nearly field-independent image resolution for SRLM with FOV of up to 221 μm × 221 μm. Unlike other reported approaches for homogenous illumination, our illumination system is advantageous in laser power scaling by combining up to four relative low power lasers using the multi-mode fiber combiner, and provides a square homogenous illumination without beam clipping. Unfortunately, oblique-angle or TIRF illumination is not possible from the current objective-type illumination system because the illumination beam is required to be focused onto the back focus plane of a high NA objective, which will bring damage to the anti-reflection coating inside the objective. To achieve oblique-angle or TIRF illumination for SRLM with large FOV, prism-type illumination approach may be a good choice. Nevertheless, we believe that the high-power homogeneous illumination system, especially the multi-mode fiber combiner developed in this study, has great potentials to push the realization and application of SRLM with large FOV.
Funding
National Natural Science Foundation of China (Grant Nos. 81427801, 91332103, 61205196); National Basic Research Program of China (Grant No. 2015CB352003); Science Fund for Creative Research Group of China (Grant No. 61421064).
Acknowledgments
We thank Mr. Sheng-Jun Wang from Tianjin Junlaser Technology for technical support in the multimode fiber combiner, Dr. Qing-lei Hu from Britton Chance Center for Biomedical Photonics for helpful discussions, and Dr. Yina Wang from University of California, San Francisco for English language editing. We also appreciate Mr. Matthias Schulze and Dr. Tao Liang from Coherent for trial use and technical support in the high-power lasers.
References and links
1. E. Betzig, G. H. Patterson, R. Sougrat, O. W. Lindwasser, S. Olenych, J. S. Bonifacino, M. W. Davidson, J. Lippincott-Schwartz, and H. F. Hess, “Imaging intracellular fluorescent proteins at nanometer resolution,” Science 313(5793), 1642–1645 (2006). [CrossRef] [PubMed]
2. M. J. Rust, M. Bates, and X. Zhuang, “Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM),” Nat. Methods 3(10), 793–795 (2006). [CrossRef] [PubMed]
3. M. Heilemann, S. van de Linde, M. Schüttpelz, R. Kasper, B. Seefeldt, A. Mukherjee, P. Tinnefeld, and M. Sauer, “Subdiffraction-resolution fluorescence imaging with conventional fluorescent probes,” Angew. Chem. Int. Ed. Engl. 47(33), 6172–6176 (2008). [CrossRef] [PubMed]
4. T. Klein, S. Proppert, and M. Sauer, “Eight years of single-molecule localization microscopy,” Histochem. Cell Biol. 141(6), 561–575 (2014). [CrossRef] [PubMed]
5. P. Seitz and A. J. Theuwissen, Single-Photon Imaging (Springer).
6. S. van de Linde, A. Löschberger, T. Klein, M. Heidbreder, S. Wolter, M. Heilemann, and M. Sauer, “Direct stochastic optical reconstruction microscopy with standard fluorescent probes,” Nat. Protoc. 6(7), 991–1009 (2011). [CrossRef] [PubMed]
7. Y. M. Sigal, C. M. Speer, H. P. Babcock, and X. Zhuang, “Mapping synaptic input fields of neurons with super-resolution imaging,” Cell 163(2), 493–505 (2015). [CrossRef] [PubMed]
8. J. Rossy, S. V. Pageon, D. M. Davis, and K. Gaus, “Super-resolution microscopy of the immunological synapse,” Curr. Opin. Immunol. 25(3), 307–312 (2013). [CrossRef] [PubMed]
9. M. de la Roche, Y. Asano, and G. M. Griffiths, “Origins of the cytolytic synapse,” Nat. Rev. Immunol. 16(7), 421–432 (2016). [CrossRef] [PubMed]
10. M. Gunkel, B. Flottmann, M. Heilemann, J. Reymann, and H. Erfle, “Integrated and correlative high-throughput and super-resolution microscopy,” Histochem. Cell Biol. 141(6), 597–603 (2014). [CrossRef] [PubMed]
11. S. J. Holden, T. Pengo, K. L. Meibom, C. Fernandez Fernandez, J. Collier, and S. Manley, “High throughput 3D super-resolution microscopy reveals Caulobacter crescentus in vivo Z-ring organization,” Proc. Natl. Acad. Sci. U.S.A. 111(12), 4566–4571 (2014). [CrossRef] [PubMed]
12. M. Bigas, E. Cabruja, J. Forest, and J. Salvi, “Review of CMOS image sensors,” Microelectronics J. 37(5), 433–451 (2006). [CrossRef]
13. T. Quan, S. Zeng, and Z. L. Huang, “Localization capability and limitation of electron-multiplying charge-coupled, scientific complementary metal-oxide semiconductor, and charge-coupled devices for superresolution imaging,” J. Biomed. Opt. 15(6), 066005 (2010). [CrossRef] [PubMed]
14. S. Saurabh, S. Maji, and M. P. Bruchez, “Evaluation of sCMOS cameras for detection and localization of single Cy5 molecules,” Opt. Express 20(7), 7338–7349 (2012). [CrossRef] [PubMed]
15. F. Long, S. Zeng, and Z. L. Huang, “Localization-based super-resolution microscopy with an sCMOS camera part II: experimental methodology for comparing sCMOS with EMCCD cameras,” Opt. Express 20(16), 17741–17759 (2012). [CrossRef] [PubMed]
16. Z. L. Huang, H. Zhu, F. Long, H. Ma, L. Qin, Y. Liu, J. Ding, Z. Zhang, Q. Luo, and S. Zeng, “Localization-based super-resolution microscopy with an sCMOS camera,” Opt. Express 19(20), 19156–19168 (2011). [CrossRef] [PubMed]
17. K. M. Douglass, C. Sieben, A. Archetti, A. Lambert, and S. Manley, “Super-resolution imaging of multiple cells by optimised flat-field epi-illumination,” Nat. Photonics 10(11), 705–708 (2016). [CrossRef] [PubMed]
18. G. T. Dempsey, “A user’s guide to localization-based super-resolution fluorescence imaging,” Methods Cell Biol. 114, 561–592 (2013). [CrossRef] [PubMed]
19. J. Deschamps, A. Rowald, and J. Ries, “Efficient homogeneous illumination and optical sectioning for quantitative single-molecule localization microscopy,” Opt. Express 24(24), 28080–28090 (2016). [CrossRef] [PubMed]
20. F. Schuberts, A. Hoben, K. Bakhshpour, and C. Provost, “Square fibers solve multiple application challenges,” Photon. Spectra 45, 38–41 (2011).
21. D. S. Mehta, D. N. Naik, R. K. Singh, and M. Takeda, “Laser speckle reduction by multimode optical fiber bundle with combined temporal, spatial, and angular diversity,” Appl. Opt. 51(12), 1894–1904 (2012). [CrossRef] [PubMed]
22. W. Ha, S. Lee, Y. Jung, J. K. Kim, and K. Oh, “Acousto-optic control of speckle contrast in multimode fibers with a cylindrical piezoelectric transducer oscillating in the radial direction,” Opt. Express 17(20), 17536–17546 (2009). [CrossRef] [PubMed]
23. J. W. Goodman, Speckle Phenomena in Optics: Theory and Applications (Ben Roberts & Company).
24. F. Huang, T. M. P. Hartwich, F. E. Rivera-Molina, Y. Lin, W. C. Duim, J. J. Long, P. D. Uchil, J. R. Myers, M. A. Baird, W. Mothes, M. W. Davidson, D. Toomre, and J. Bewersdorf, “Video-rate nanoscopy using sCMOS camera-specific single-molecule localization algorithms,” Nat. Methods 10(7), 653–658 (2013). [CrossRef] [PubMed]
25. T. Quan, P. Li, F. Long, S. Zeng, Q. Luo, P. N. Hedde, G. U. Nienhaus, and Z. L. Huang, “Ultra-fast, high-precision image analysis for localization-based super resolution microscopy,” Opt. Express 18(11), 11867–11876 (2010). [CrossRef] [PubMed]
26. R. E. Thompson, D. R. Larson, and W. W. Webb, “Precise nanometer localization analysis for individual fluorescent probes,” Biophys. J. 82(5), 2775–2783 (2002). [CrossRef] [PubMed]
