Open Access
Add to BibSonomy
Abstract
The achievable image quality in fluorescence microscopy and nanoscopy is usually limited by photobleaching. Reducing the light dose imposed on the sample is thus a challenge for all these imaging techniques. Various approaches like CLEM, RESCue, MINFIELD, DyMIN and smart RESOLFT have been presented in the last years and have proven to significantly reduce the required light dose in diffraction-limited as well as super-resolution imaging, thus resulting in less photobleaching and phototoxicity. None of these methods has so far been able to transfer the light dose reduction into a faster recording at pixel dwell times of a few ten microseconds. By implementing a scan system with low latency and large field of view we could directly convert the light dose reduction of RESCue into a shorter acquisition time for STED nanoscopy. In this way, FastRESCue speeds up the acquisition locally up to 10-fold and allows overall for a 5 times faster acquisition at only 20% of the light dose in biological samples.
© 2020 Optical Society of America under the terms of the OSA Open Access Publishing Agreement
1. Introduction
Fluorescence microscopy [1] allows for highly specific imaging of tissue, cells, organelles, macromolecular assemblies and even individual molecules. Therefore, it has become an integral part of modern biomedical research [2]. Especially for capturing fast dynamics in living cells, both the achievable temporal resolution and the image quality are of great importance. The image quality is usually limited by photobleaching [3,4] of the dyes and under live-cell imaging conditions, in particular by phototoxicity [5,6] resulting therefrom. A major challenge for imaging living cells is thus finding the right balance between image quality and cell viability [7]. A good signal-to-noise ratio requires a relatively high light dose, i.e. a rather high product of the incident intensity of the excitation light and the (local) exposure time. For cell viability, on the other hand, a low light dose is of utmost importance.
When imaging with a resolution below the diffraction limit, this trade-off is even more pronounced. In optical nanoscopy [8] the fluorescence signals from closely spaced features within a diffraction-limited spot are recorded sequentially in time by switching fluorophores between a fluorescent (bright) and a non-fluorescent (dark) state [9]. If, for example, the resolution of a point scanning microscope is increased by a factor of $m$, the image acquisition time needed for a 2D recording increases hence by a factor of $m^2$ due to the Nyquist sampling criterion, thereby increasing the light dose by the same amount. As switching of fluorophores usually requires additional irradiation with light of another wavelength, the light dose to which the sample is exposed typically increases even more.
An immediate approach to avoid photobleaching is the use of anti-bleaching agents [10]. Yet, they are often incompatible with live-cell imaging [11] and strongly depend on the dyes employed. Therefore methods which reduce photobleaching by means of advanced image acquisition and illumination strategies are used more and more frequently in scanning based microscopy and nanoscopy. Methods such as resonant scanning [12], bunched pulsed excitation [13] or the use of pulsed lasers at low repetition rate [14] do not directly reduce the light dose, but avoid the transition of dye molecules into states other than the ground and the first electronically excited state. This applies in particular to higher excited singlet states as well as the long-lived triplet state, which are all involved in major molecular photobleaching pathways [14]. Moreover, non-bleached molecules that are in the latter state typically no longer fluoresce during the remaining image acquisition. Although very effective, these approaches also have drawbacks: In resonant scanning, the field of view has to be scanned multiple times [15] in order to obtain a decent signal-to-noise ratio. This is at the expense of the time span needed for the full acquisition of a pixel. When using lower repetition rates, this local time resolution does not decrease that much, but the overall time needed to record an image increases 5- to 10-fold [16].
Other techniques, such as CLEM [17], RESCue [18], MINFIELD [19] or DyMIN [20], reduce the light dose directly by analyzing specific properties of the sample during image acquisition and by locally adjusting the irradiance used for imaging. These direct methods allow to speed up the image acquisition, as instead of or in addition to the adaptation of the intensity, the pixel dwell time could in principle also be adapted locally. This has recently been demonstrated for RESOLFT nanoscopy [21]. However, RESOLFT [22] requires comparatively long pixel dwell times in the order of several hundreds of microseconds. For much faster techniques like confocal [23] or STED [24,25] imaging, a local adjustment of the pixel dwell time is much more demanding, since the next pixel has to be addressed at any point in time within a few microseconds. For decent fields of view this is not possible so far with available scanners. They either have short response times, but can only address small fields of view, or they address a large scan range such as galvanometer scanners do. However, the latter are inert, because relatively high loads have to be accelerated and decelerated.
Exemplarily, we demonstrate for STED nanoscopy that the combination of a scanner with a large scan range and a scanner with a short response time merges the advantages of both systems and thus allows for pixel hopping. Hence, a local reduction of the light dose by RESCue can be almost completely converted into a speedup of the image acquisition. Our implementation which we term FastRESCue results in a local reduction of the acquisition time by up to a factor of 10 and achieves a reduction of the global acquisition time by a factor of 4-5 when imaging cellular structures.
2. Material and methods
2.1 Setup
As depicted in Fig. 1(a), a STED laser system (Onefive GmbH, Switzerland) provides $600$ ps pulses at a repetition rate of $20$ MHz and a wavelength of $775$ nm and triggers a laser diode (PicoQuant, Germany) providing $100$ ps excitation pulses at a wavelength of $640$ nm. The delay between excitation and STED pulses is controlled electronically by a home-built delay line. The use of acousto-optic modulators (not shown, AA Opto Electronic, France) allows for an extremely fast (in the order of some hundreds of nanoseconds) control of the sample illumination for both lasers. The initially Gaussian STED beam is phase-modulated in order to generate a doughnut-shaped intensity distribution employed as depletion pattern for 2D STED microscopy. Excitation and STED beam are superposed by a dichroic mirror before entering the scan system, yielding the same deflection for both beams.
Fig. 1. Implementation of FastRESCue-STED imaging. (a) The optical setup includes two EODs as fast scanning devices additionally to the galvanometer scanner. (Exc: excitation unit; APD: avalanche photo diode; STED: STED laser unit; $\lambda /2$, $\lambda /4$: half and quarter wave plate, respectively; TC1, TC2: telescopes; GT: Glan-Thompson-prism; GS: galvanometer scanner; OL: objective lens; S: sample). (b) Sketch of the combined scan system’s control. Two voltages for the scanner are output by the FPGA I/O device. They serve as inputs to the circuit board (red) and are subsequently linearly transformed. The monitoring output voltages of the scanner are used as additional inputs to derive the voltages for the EOD drivers.
For realizing short scanner latencies, two electro-optic deflectors (EODs) are employed, for which a customized EOD model (Conoptics Inc., USA) in combination with a standard voltage driver (Trek Inc., now part of Advanced Energy, USA) is chosen, as also described in [26,27]. One EOD is utilized for $x$- and the other one for $y$-scanning, implying that their deflection axes are rotated by $90^{\circ}$ with respect to each other. Since the electro-optic effect is only exhibited for one polarization direction of the incident beam, the polarization needs to be rotated accordingly, as realized by a half-wave plate in-between the EODs as shown in Fig. 1(a). After the EOD pair, the polarization of all beams is set circular by a combination of a half- and a quarter-wave plate before they pass the galvanometer scanner (Abberior Instruments GmbH, Germany) and are focused on the sample via an oil-immersion objective lens ($NA = 1.4$; Olympus, Japan). Two telescopes in the beam path as well as the two lenses before and after the galvanometer scanner propagate the objective-conjugated plane to ensure that the beam is only tilting, but not translating in the back-focal plane of the objective lens.
The fluorescence signal is collected by the same objective lens and is de-scanned and separated from excitation and STED light by a dichroic mirror. It is subsequently coupled into an avalanche photodiode (Perkin Elmer, USA) via a multimode fiber, with all detected photons, irrespective of their arrival time, being used for image generation and signal analysis. It is worth noting that the emitted fluorescence signal cannot be completely de-scanned for confocal detection due to the polarization-dependence of the electro-optic effect. Hence, a Glan-Thompson-prism in the beam path before the EODs (Fig. 1(a)) filters the fluorescence signal accordingly. Assuming an isotropic fluorescence emission due to a random distribution of fluorophores’ dipole orientations, this implies a loss of half of the fluorescence signal compared to a setup without de-scanning by the EODs.
2.2 Control
The setup is operated via a self-written LabVIEW (National Instruments corp., USA) routine, which is employed for programming a Field-Programmable Gate Array (FPGA) on a multifunction reconfigurable I/O device (National Instruments corp., USA) in order to allow for a fast control of the scanning and its synchronization with the detection. This program operates the scan system via analog voltages and provides a digital modulation signal for laser control via the acousto-optic modulators. The detector signal is read via a digital input and directly analyzed for the optimization of the pixel dwell time. The FPGA program runs time-independently of the host program, providing accurate operational timing, which is necessary to correctly determine and set the pixel dwell time.
2.3 Combined scan system
For the scan system, a combination of a galvanometer scanner and EODs is realized, as proposed in [28]. To exploit the full advantage of both scanning device types, the EODs are utilized to compensate for the positioning error of the galvanometer scanner, while the latter is employed for scanning large fields of view. As electro-optic deflection is a function of the refractive index dispersion, which is virtually constant over the wavelength range of operation, there is no relative movement between the excitation and the STED focus in the sample plane. Figure 1(b) shows a systematic sketch of the control of the combined scan system. A home-built circuit board, highlighted in red, is utilized to read one voltage per scan direction from the I/O device and to output all necessary voltages to both the galvanometer scanner and the EODs. For the control of the galvanometer scanner, which needs four input voltages, the voltages $U_{\textrm {X1}}$ and $U_{\textrm {Y1}}$ are read from the I/O device and passed on to the scanner together with the linearly related voltages $U_{\textrm {X2}}$ and $U_{\textrm {Y2}}$. The actual position can be accessed via the monitoring output $U_{\textrm {X1, moni}}$, $U_{\textrm {Y1, moni}}$ of the scanner driver and is used as an additional input to the circuit board. The differences between $U_{\textrm {X1}}$ and $U_{\textrm {X1, moni}}$ and between $U_{\textrm {Y1}}$ and $U_{\textrm {Y1, moni}}$ are appropriately scaled and rerouted to the voltage amplifiers of the EODs. In this way, the EODs compensate for the spatial displacement introduced by the slow response time of the galvanometer scanner.
2.4 RESCue and FastRESCue implementation
RESCue [18], or CLEM [17] in case of confocal imaging, allows to reduce the light dose imposed on a sample by switching both the excitation and STED laser off for the residual pixel dwell time if either enough signal has been collected (adjustable upper threshold) or no signal arises within the first e.g. 10% of the pixel dwell time (adjustable lower threshold). The upper threshold $uTh$ defines the number of counts at which the lasers are switched off and is adjusted in accordance with the targeted signal-to-noise ratio. If $uTh$ is met, the acquired signal is respectively scaled to compensate for the pixel dwell time reduction [18].
In this study, we used a slightly modified RESCue implementation: To avoid illumination in the absence of structure, the lasers are switched off, if not at least $lTh_1$ (first lower threshold) counts are recorded within time $dT_1$, where the latter is in the following given in percent of the pixel dwell time $pT$. $lTh_1$ and $dT_1$ have to be chosen depending on the expected signal as well as on the noise level. For a too conservative lower threshold, also background pixel without structure experience illumination, whereas for a too restrictively chosen threshold, dim structures may be overlooked [18]. To improve the accuracy of the decision, a second lower threshold has been introduced: analogously to the first, at least $lTh_2$ counts are required after the time $dT_2$ in order to continue the acquisition, with $dT_2>dT_1$ and $lTh_2>lTh_1$. The upper threshold was implemented in the same way as for conventional RESCue.
For FastRESCue, we used the same threshold parameters as described above. However, when the switching-off criterion was met, the data acquisition was immediately continued at the next pixel. This results in an effective pixel dwell time smaller than or equal to $pT$.
We quantified the similarity between the FastRESCue image and the standard STED image by means of their normalized cross-correlation [29]. Therefore, we calculated $\rho _{\textrm {cc}}$ as the maximum of the Matlab (MathWorks, USA) function normxcorr2 in order to be insensitive to displacements between images. For identical images $\rho _{\textrm {cc}}$ equals 1, however it is reduced for statistically independent images due to noise.
2.5 Sample preparation
As test samples, fluorescent microspheres (FluoSpheres carboxylate-modified microspheres, $48$ nm diameter, crimson fluorescent (625/645); Life Technologies, USA) are diluted $1:5,000$ with purified water and subsequently incubated on a coverslip. To guarantee adhesion, the latter is coated with Poly-L-lysine ($0.1\%$(w/v) in $H_2O$; Sigma-Aldrich). The sample is then embedded in self-prepared Mowiol.
Cultured cells are fixed with either methanol or paraformaldehyde. Immunostaining of the structures of interest is performed with a suitable primary antibody and the fluorophore Abberior STAR 635P (Abberior GmbH, Germany) or Abberior STAR RED (Abberior GmbH, Germany) conjugated to a secondary antibody. Mowiol is again used as embedding medium.
3. Results
3.1 Characterization of the scan system
As outlined previously and demonstrated in the following, a galvanometer scanner, like most of the common scanning devices summarized in [30], does not meet the necessary requirements for the response time in order to allow a pixel-specific variation of the scan pattern within a few microseconds. Due to the inertia of the scan mirrors, the monitoring voltage on the scanner driver’s output has an offset compared to its input, reflecting an offset between the nominal and the actual scan position. For a quantitative analysis of this displacement, the voltage offset is measured while scanning a field of view of size $3{\mu \textrm{m}}\times 3{\mu \textrm{m}}$ with a pixel size of $20{\;\textrm{nm}}\times 20$ nm for varying pixel dwell times. The scan position displacement is constant over the field of view, depends on the chosen pixel dwell time, as shown in Fig. 2(a), and can amount to more than $2{\mu \textrm{m}}$ for a pixel dwell time of $1{\mu \textrm{s}}$. The observed functional dependence can be explained by solving the basic differential equations for the motion of a galvanometer for an input voltage with a constant slope, as it is the case during a line acquisition. Assuming the system to be critically damped, the nominal and actual scan position have, in the limit of large times, a constant offset in time, which depends on the characteristics of the galvanometer, but not on the acquisition parameters. Taking into consideration the slope of the input voltage, which depends on the pixel dwell time, yields the hyperbolic behavior underlined by the respective fit shown in Fig. 2(a).
Fig. 2. Performance of the scan system and impact on image quality. (a) Scan position displacement of the galvanometer scanner as a function of the pixel dwell time for a field of view of $3{\mu \textrm{m}}\times 3{\mu \textrm{m}}$ and $20\;\textrm{nm}\times 20$ nm pixel size, averaged over a line scan. The error bars denote ten times the standard error of the mean. For a pixel dwell time of $1{\mu \textrm{s}}$, the scan position displacement is more than $2{\mu \textrm{m}}$. (b-d) Imaging of fluorescent microspheres using only the galvanometer scanner with (b) STED and (c) RESCue-STED shows equivalent image quality, while the corresponding FastRESCue-STED image (d) exhibits clear distortions. (e) Only with the EODs as additional scanning devices for FastRESCue-STED, the scan position displacement is compensated and the image quality is restored. Note that for a better visualization, the resolution is only increased in the direction of the fast scan axis. The same color table is chosen for all images and the scale bar indicates a length of $500$ nm in (b-e).
The impact of the offset between the nominal and the actual scan position on the characteristics of scanning fluorescent microscopes and nanoscopes is investigated by imaging fluorescent microspheres in STED mode. For better visualization, the STED beam is phase-modulated for these measurements such that the resolution is only increased along a single direction [31], with the high-resolution axis pointing in the direction of the fast scan axis. This facilitates the identification of distortions in the elongated images of fluorescent microspheres. For a scan with a constant pixel dwell time, the scan position displacement is constant and thus results in a constant offset over the whole field of view. Hence, no distortions are caused in the image, as shown in Fig. 2(b) for a scan using the galvanometer scanner only. For a RESCue scan, the lasers are switched off according to the defined parameters and thus the pixel dwell time is not affected. Therefore, the same offset can be observed as in the scan without RESCue, as depicted in Fig. 2(c). Refer to Table 1 in the appendix for the applied RESCue parameters. An analysis of the image similarity between standard STED and RESCue-STED acquisition yields a cross-correlation value of $\rho _{\textrm {cc}}=0.87$, which lies in the range of what one expects for two statistically independent images with the present signal-to-noise ratio.
Contrarily, the image exhibits clear distortions, being reflected in a reduced cross-correlation value of $\rho _{\textrm {cc}}=0.70$, if the next scan position is directly addressed in the FastRESCue imaging scheme instead of switching off the lasers for the residual pixel dwell time, as shown in Fig. 2(d). Because of the resulting variation in the effective pixel dwell time, the offset between actual and nominal scan position of the galvanometer scanner also varies over the field of view, which has to be compensated for. This can be accomplished by combining the galvanometer scanner with EODs as additional scanning devices. Since EODs are driven by an analog voltage, yet have no mechanical parts, they deflect the beam almost instantaneously. A FastRESCue scan of the same field of view utilizing the EODs for the displacement compensation is shown in Fig. 2(e). Additionally to the global offset, the distortions have completely disappeared, as indicated by a cross-correlation value of $\rho _{\textrm {cc}}=0.87$ and $0.86$ in comparison to the standard STED and the conventional RESCue acquisition, respectively. The image quality as well as the signal level of this scan is therefore equivalent to both, the scan without RESCue and the scan with conventional RESCue. Therefore, the EOD correction has proven to be strictly necessary for implementing the FastRESCue imaging scheme.
3.2 Analysis of global and local light dose savings and acquisition time
To evaluate the potential of FastRESCue to translate a reduction in light dose into recording speed, we imaged fluorescent microspheres. Figures 3(a) and 3(b) depict the same area within the sample using standard confocal and STED microscopy, respectively. The corresponding FastRESCue acquisition in STED mode is shown in Fig. 3(c) and exhibits the same image quality ($\rho _{\textrm {cc}}=0.89$) and resolution as the reference in Fig. 3(b), which confirms the correct setting of the RESCue parameters (see Table 1 in the appendix). As the light dose is reduced by RESCue by more than a factor of six to $15.9\%$, by means of pixel hopping the FastRESCue acquisition could be performed in $17.8\%$ of the time needed otherwise. The effective pixel dwell times are shown in Fig. 3(d). As can be seen, the use of FastRESCue speeds up imaging in areas without structure as well as in areas with high signal. In addition, only a few background pixels are detected as false positive, indicating that the lower threshold is set not too restrictively. This is important to avoid overlooking of dim structures.
Fig. 3. FastRESCue reduces light dose and frame time. (a-d) Images of fluorescent microspheres of $48$ nm diameter for (a) confocal, (b) STED and (c) FastRESCue-STED acquisition (scale bar $1{\mu \textrm{m}}$). The chosen RESCue parameters result in a light dose of $15.9\%$ and a frame time of $17.8\%$ compared to the standard STED acquisition. (d) shows the effective pixel dwell times in ${\mu \textrm{s}}$, illustrating the potential to speed up acquisition both in areas without structure as well as in areas of high signal. (e) The FastRESCue acquisition is repeated for varying RESCue parameters (see Table 2 in the appendix) at otherwise identical acquisition parameters, leading to a varying relative FastRESCue light dose. The resultant frame time compared to the standard STED acquisition is depicted as a function of this light dose, showing a direct translation of light dose saving into a faster acquisition. The red curve visualizes equal light dose and frame time saving.
As indicated by this example, the saving in light dose is not entirely translated into a saving in acquisition time. This results from an overhead within the LabVIEW routine as well as from the line fly-back time needed in unidirectional scanning: For fields of view larger than the maximum scan range of the EODs, a wait time is implemented at the end of each line, which adds to the total overhead and which allows the scan system to address the beginning of the next line. As the routine’s overhead as well as the wait time are a constant for both the RESCue and the FastRESCue implementation, the saving in frame time is slightly less than the saving in light dose. This dependence is analyzed in more detail by imaging the same sample with varying RESCue parameters (see Table 2 in the appendix), yielding different light dose and frame time savings. Figure 3(e) depicts the FastRESCue frame time as a function of the FastRESCue light dose, with both quantities given relative to the respective quantities for the standard STED acquisition. In the following, these are denoted as relative FastRESCue frame time and relative FastRESCue light dose. It is apparent that the relative FastRESCue frame time is always somewhat higher than but very close to the bisectrix, which is shown in red. Although the constant total overhead of the program is more dominant at short pixel dwell times, it can be deduced that the advantage of FastRESCue can be instantly used in all applications where RESCue is beneficial.
Since RESCue is typically employed for imaging biological structures, we also analyzed the time savings by FastRESCue for imaging Vero cells stained for $\alpha$-tubulin (Fig. 4). The FastRESCue-STED image in Fig. 4(b) exhibits the same image quality ($\rho _{\textrm {cc}}=0.91$) as the standard STED acquisition Fig. 4(a), at only $22.7\%$ of the frame time, and the detailed distribution of pixel dwell times in Fig. 4(c) corresponds well with areas of high or very low signal (i.e. stained structures or background). In order to analyze the advantage of FastRESCue on different spatial scales, the image shown in Fig. 4(b) is divided in disjoint regions of interest (ROIs). For each ROI, the local relative acquisition time is calculated, which is the FastRESCue acquisition time for this ROI relative to the standard STED acquisition time for the same ROI. The size of the ROIs is varied and the local relative acquisition time is displayed in a histogram as shown in Fig. 4(d) for four different ROI sizes. Here, intervals of $5\%$ are considered for the analysis, i.e. events between $0$ and $5\%$ local relative acquisition time are summed up and displayed at $2.5\%$. The variance of this distribution decreases for increasing ROI sizes and is displayed as an inset in Fig. 4(d). The distribution itself exhibits two peaks at $2.5\%$ and $97.5\%$ for a ROI size of one pixel, i.e. $20{\;\textrm{nm}}\times 20$ nm, corresponding to the first lower threshold and the maximum pixel dwell time, respectively. The events at $7.5\%$ correspond to the second lower threshold, and all other events reflect the pixels where the acquisition was stopped due to the upper threshold. For increasing ROI size, the outer peaks vanish due to the natural inhomogeneities in the structure’s distribution, as seen from Fig. 4(b). Nevertheless, also for large ROI sizes, the main peak remains below $10\%$, implying that, while a global reduction in image acquisition time by a factor of $5$ can be successfully realized, FastRESCue actually speeds up the acquisition locally by up to a factor of $10$.
Fig. 4. Advantage of FastRESCue on different spatial scales. Images of $\alpha$-tubulin in Vero cells (dye: Abberior STAR RED) recorded with (a) STED and (b) FastRESCue-STED (scale bar $1{\mu \textrm{m}}$). The global FastRESCue light dose and frame time is reduced to $20.3\%$ and $22.7\%$, respectively. The distribution of the effective pixel dwell time (in ${\mu \textrm{s}}$) is visualized in (c). Histograms of local relative acquisition times for different ROI sizes are displayed in (d), with the variance in dependence on the ROI size being depicted in the inset.
3.3 FastRESCue imaging of different biological samples
In order to demonstrate the potential of FastRESCue for biological applications, we imaged representative structures, labeled with different fluorescent dyes, within different cell types (Fig. 5). For each sample, a confocal overview image was acquired first and the marked area was then imaged using FastRESCue, whereby the image acquisition could be accelerated by a factor of four to five as given by the percentage values in the respective image. Figures 5(a) and 5(b) show another example of $\alpha$-tubulin stained in Vero cells. Although the RESCue parameters have been slightly modified as compared to the recording depicted in Fig. 4 and found to be optimal as $uTh=40$, $dT_1=4\%$ and $dT_2=7\%$, the image quality is comparable. Figures 5(c) and 5(d) present membrane-stained peroxisomes in Vero cells. Due to the huge variation in structure size and thus signal, the lower thresholds have been chosen much higher ($dT_1=10\%$ and $dT_2=16\%$) as compared to the previous example. This yields an optimal trade-off between imaging speed and quality and therefore also dim peroxisomes are correctly detected and imaged. In Figs. 5(e) and 5(f), vimentin within a fibroblast is displayed. The RESCue parameters have been chosen similarly to those of the recording shown in Figs. 5(a) and 5(b) and also dimmer structures visible in the background can be correctly discerned. Finally, Figs. 5(g) and 5(h) depict $\alpha$-tubulin stained in fibroblasts. The FastRESCue image exhibits no artifacts or distortions and the RESCue parameters were set to $uTh=40$, $dT_1=10\%$ and $dT_2=15\%$. For all images the color table is chosen such that any structure is clearly visible. A complete list of all RESCue parameters and the resulting relative FastRESCue light dose is summarized in Table 1 in the appendix.
Fig. 5. FastRESCue-STED imaging of various cellular structures. Confocal overviews (a, c, e, g) and FastRESCue-STED images (b, d, f, h) of $\alpha$-tubulin in Vero cells (a,b), peroxisomes (PMP70) in Vero cells (c,d), vimentin in fibroblasts (e,f) and $\alpha$-tubulin in fibroblasts (g,h). For fluorescent labeling, the fluorophores Abberior STAR RED (a,b) and Abberior STAR 635P (c-h) were used. The length of the scale bar is $5{\mu \textrm{m}}$ (a,c,e,g) or $1{\mu \textrm{m}}$ (b,d,f,h). The percentage values in the upper right corner of the FastRESCue images denote the respective relative FastRESCue frame time.
4. Conclusion
By using an ultrafast scanner that compensates for the latency-induced displacement characteristic for scanners utilized for imaging large fields of view, it is possible to hop from one pixel to another within a few microseconds. This allows to implement sophisticated spatial scan sequences. We have implemented such a scan system by combining EODs with a galvanometer scanner and exemplarily applied this system to RESCue-STED in order to directly translate a reduction of light dose into imaging speed. FastRESCue-STED measurements on fluorescent beads as well as on various cellular structures have shown a reduction of the acquisition time by a factor of $5$ or more as compared to (RESCue-)STED imaging. As this gain depends on the sparsity of the imaged structure and on the targeted signal-to-noise ratio, a statistical analysis of sub-areas of an exemplary image proved a light dose as well as a time reduction even by a factor of $10$ for substantial parts of the image.
Within this work, the EODs’ control was performed by means of a circuit board, which provides sufficiently fast response times in a fraction of the effective pixel dwell time. Depending on the sampling rate of the analog input and output channels of the FPGA I/O device, this closed-loop scanning control can also be implemented directly on the FPGA, a facilitative approach we did not pursue due to the limited number of analog output channels on the FPGA board employed.
Currently, the polarization-dependence of the electro-optic effect necessitates to reject half of the fluorescence signal for a complete de-scanned detection. This could be avoided by replacing the EODs by a sufficiently fast polarization- and wavelength-independent scan technology. As, to our knowledge, such a technology is not available yet, one possibility could be to only partly de-scan the fluorescence signal by the galvanometer scanner, but not by the EODs. If compensation of the constant displacement of the galvanometer scanner would be omitted by means of an adapted circuit board, only small displacements at the position of the detector would arise, which can most likely be compensated by an adjustment of the pinhole size. However, the effect on the image quality by the therefrom resulting reduced optical sectioning capability and the higher sensitivity to any background signal from out-of-focus planes has to be weighed against the signal gain, which is therefore subject to further investigations.
Quantifying the reduction in photobleaching and phototoxicity when applying FastRESCue to live cell imaging is an important further step. Since FastRESCue reduces the light dose whenever RESCue does, we expect a similar advantage for live cell imaging conditions as for the latter technique. However, the image acquisition time for FastRESCue is shorter, and the resulting effect thereof on the sample should be investigated.
As the scan system allows for pixel-based decisions on the pixel dwell time and hence can yield an improvement in acquisition speed whenever there are wait times due to the shut-off of lasers, it is applicable to all techniques which reduce the light dose by adjusting the irradiance to the properties of the sample. This is, for example, the case for DyMIN, which is even more sample-conserving with respect to the number of necessary on/off transitions than RESCue, making this combination particularly attractive. A further reduction of both light dose and acquisition time could be achieved by utilizing a-priori knowledge about the fluorophores’ positions, as e.g. gained by a confocal pre-scan, and completely skipping empty regions.
Due to the accurate positioning control provided by the combined scan system, applications beyond the speed-up of current microscopy and nanoscopy imaging techniques are conceivable, including particle tracking or tracing of filaments and other structures by continuously adapting the scanned region during image acquisition. The speed of such algorithms is then only limited by the available fluorescence signal. While the here presented system enhances the scan flexibility in the focal plane, an extension to the third dimension seems to be readily accessible by using an electro-optically actuated varifocal lens for fast axial scanning. This will be particularly relevant for the observation of fast dynamics in living cells. Hence, pixel hopping will enable completely new applications and will therefore have a significant impact on how microscopy and nanoscopy based experiments will be carried out in the future.
Appendix A. Acquisition parameters
Table 2. RESCue parameters for the FastRESCue acquisitions analyzed in Fig. 3(e).
Funding
Funded by the Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) under Germany’s Excellence Strategy - EXC 2067/1- 390729940, EXC 171 and FZT 103.
Acknowledgments
The authors thank M. Lübbecke for the realization of the circuit board for the scan system’s control as well as J. Rehman (Abberior GmbH, Göttingen) and K. Soliman for supplying fixed cells.
Disclosures
AE: Abberior Instruments GmbH (I,C).
References
1. C. A. Combs, “Fluorescence Microscopy: A Concise Guide to Current Imaging Methods,” Curr. Protoc. Neurosci. 50(1), 2.1.1–2.1.14 (2010). [CrossRef]
2. Y. Sun and A. Periasamy, “Fluorescence Microscopy Imaging in Biomedical Sciences,” in Biomedical Optical Imaging Technologies, R. Liang, ed. (Springer, 2013).
3. L. Song, C. A. G. O. Varma, J. W. Verhoeven, and H. J. Tanke, “Influence of the Triplet Excited State on the Photobleaching Kinetics of Fluorescein in Microscopy,” Biophys. J. 70(6), 2959–2968 (1996). [CrossRef]
4. T. Bernas, M. Zarȩbski, R. R. Cook, and J. W. Dobrucki, “Minimizing photobleaching during confocal microscopy of fluorescent probes bound to chromatin: role of anoxia and photon flux,” J. Microsc. 215(3), 281–296 (2004). [CrossRef]
5. D. J. Stephens and V. J. Allan, “Light Microscopy Techniques for Live Cell Imaging,” Science 300(5616), 82–86 (2003). [CrossRef]
6. C. L. Rieder and A. Khodjakov, “Mitosis Through the Microscope: Advances in Seeing Inside Live Dividing Cells,” Science 300(5616), 91–96 (2003). [CrossRef]
7. H. Schneckenburger, P. Weber, M. Wagner, S. Schickinger, V. Richter, T. Bruns, W. S. L. Strauss, and R. Wittig, “Light exposure and cell viability in fluorescence microscopy,” J. Microsc. 245(3), 311–318 (2012). [CrossRef]
8. S. J. Sahl, S. W. Hell, and S. Jakobs, “Fluorescence nanoscopy in cell biology,” Nat. Rev. Mol. Cell Biol. 18(11), 685–701 (2017). [CrossRef]
9. S. W. Hell, “Far-Field Optical Nanoscopy,” Science 316(5828), 1153–1158 (2007). [CrossRef]
10. J. Vogelsang, R. Kasper, C. Steinhauer, B. Person, M. Heilemann, M. Sauer, and P. Tinnefeld, “A Reducing and Oxidizing System Minimizes Photobleaching and Blinking of Fluorescent Dyes,” Angew. Chem., Int. Ed. 47(29), 5465–5469 (2008). [CrossRef]
11. C. Boudreau, T.-L. E. Wee, Y.-R. S. Duh, M. P. Couto, K. H. Ardakani, and C. M. Brown, “Excitation Light Dose Engineering to Reduce Photo-bleaching and Photo-toxicity,” Sci. Rep. 6(1), 30892 (2016). [CrossRef]
12. R. T. Borlinghaus, “MRT letter: High Speed Scanning Has the Potential to Increase Fluorescence Yield and to Reduce Photobleaching,” Microsc. Res. Tech. 69(9), 689–692 (2006). [CrossRef]
13. G. Donnert, C. Eggeling, and S. W. Hell, “Triplet-relaxation microscopy with bunched pulsed excitation,” Photochem. Photobiol. Sci. 8(4), 481–485 (2009). [CrossRef]
14. G. Donnert, C. Eggeling, and S. W. Hell, “Major signal increase in fluorescence microscopy through dark-state relaxation,” Nat. Methods 4(1), 81–86 (2007). [CrossRef]
15. J. Jonkman and C. M. Brown, “Any Way You Slice It - A Comparison of Confocal Microscopy Techniques,” J. Biomol. Tech. 26(2), 54–65 (2015). [CrossRef]
16. G. Donnert, J. Keller, R. Medda, M. A. Andrei, S. O. Rizzoli, R. Lührmann, R. Jahn, C. Eggeling, and S. W. Hell, “Macromolecular-scale resolution in biological fluorescence microscopy,” Proc. Natl. Acad. Sci. U. S. A. 103(31), 11440–11445 (2006). [CrossRef]
17. R. A. Hoebe, C. H. Van Oven, T. W. J. Gadella Jr., P. B. Dhonukshe, C. J. F. Van Noorden, and E. M. M. Manders, “Controlled light-exposure microscopy reduces photobleaching and phototoxicity in fluorescence live-cell imaging,” Nat. Biotechnol. 25(2), 249–253 (2007). [CrossRef]
18. T. Staudt, A. Engler, E. Rittweger, B. Harke, J. Engelhardt, and S. W. Hell, “Far-field optical nanoscopy with reduced number of state transition cycles,” Opt. Express 19(6), 5644–5657 (2011). [CrossRef]
19. F. Göttfert, T. Pleiner, J. Heine, V. Westphal, D. Görlich, S. J. Sahl, and S. W. Hell, “Strong signal increase in STED fluorescence microscopy by imaging regions of subdiffraction extent,” Proc. Natl. Acad. Sci. U. S. A. 114(9), 2125–2130 (2017). [CrossRef]
20. J. Heine, M. Reuss, B. Harke, E. D’Este, S. J. Sahl, and S. W. Hell, “Adaptive-illumination STED nanoscopy,” Proc. Natl. Acad. Sci. U. S. A. 114(37), 9797–9802 (2017). [CrossRef]
21. J. Dreier, M. Castello, G. Coceano, R. Cáceres, J. Plastino, G. Vicidomini, and I. Testa, “Smart scanning for low-illumination and fast RESOLFT nanoscopy in vivo,” Nat. Commun. 10(1), 556 (2019). [CrossRef]
22. S. W. Hell, S. Jakobs, and L. Kastrup, “Imaging and writing at the nanoscale with focused visible light through saturable optical transitions,” Appl. Phys. A 77(7), 859–860 (2003). [CrossRef]
23. J. B. Pawley, Handbook Of Biological Confocal Microscopy (Springer, 2006).
24. S. W. Hell and J. Wichmann, “Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy,” Opt. Lett. 19(11), 780–782 (1994). [CrossRef]
25. T. A. Klar, S. Jakobs, M. Dyba, A. Egner, and S. W. Hell, “Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission,” Proc. Natl. Acad. Sci. U. S. A. 97(15), 8206–8210 (2000). [CrossRef]
26. F. Göttfert, “STED microscopy with scanning fields below the diffraction limit,” PhD thesis (Georg-August-Universität Göttingen, Germany, 2015).
27. F. Balzarotti, Y. Eilers, K. C. Gwosch, A. H. Gynnå, V. Westphal, F. D. Stefani, J. Elf, and S. W. Hell, “Nanometer resolution imaging and tracking of fluorescent molecules with minimal photon fluxes,” Science 355(6325), 606–612 (2017). [CrossRef]
28. G. R. B. E. Römer and P. Bechtold, “Electro-optic and acousto-optic laser beam scanners,” Phys. Procedia 56, 29–39 (2014). [CrossRef]
29. C. Ledig and D. Rueckert, “Chapter 14 - Semantic Parsing of Brain MR Images,” in Medical Image Recognition, Segmentation and ParsingS. K. Zhou, ed. (Academic, 2016).
30. G. F. Marshall and G. E. Stutz, Handbook of Optical and Laser Scanning (CRC, New York, 2004).
31. T. Klar, E. Engel, and S. W. Hell, “Breaking Abbe’s diffraction resolution limit in fluorescence microscopy with stimulated emission depletion beams of various shapes,” Phys. Rev. E 64(6), 066613 (2001). [CrossRef]
