Open Access
Add to BibSonomyAbstract
We demonstrate a new method for obtaining sub-diffraction resolution in fluorescence microscopy. The technique involves the analysis of the time evolution of fluorescence images in the presence of weak and unstructured (fundamental Gaussian) continuous wave stimulated emission depletion. A reduced point spread functions (PSF) is obtained by the recombination of time segments of the evolving image. A significant reduction in the PSF for 20nm fluorescent beads (ca. 240nm to 125nm) is obtained with an on-sample power of 7.5mW (17MW/cm2) – substantially lower than that required for spatially structured stimulated emission depletion microscopy.
© 2014 Optical Society of America
1. Introduction
Super resolution refers to the ability to resolve objects and/or observe contrast on a distance scale below that afforded by conventional optical imaging (e.g. the confocal microscope [1]). From the Abbé criterion, the 2-dimensional image resolution limit for an objective of numerical aperture (NA) and wavelength λ is given by [2]
In the visible region of the spectrum this is on the order of a quarter to a third of a micron. There is a vast range of biological structures/systems where non-invasive observations below this length scale are desirable. There has been considerable academic and commercial activity aimed at developing techniques that reveal structure below this limit. These fall into four categories namely: stochastic reconstruction techniques (PALM & STORM) [3–5], Structured Illumination (SI) [6], Stimulated Emission Depletion (STED) techniques [7–12] and Ground State Depletion (GSD) microscopy [13,14]. All four approaches have their attendant limitations.PALM and STORM (photo-activated localization microscopy and stochastic optical reconstruction microscopy) require the use of specialised (photo-activatable) fluorescent probes and often long periods of data collection with repeated measurements of the same point to build up an image. In SI, STED and GSD, image resolution is intensity dependent. SI uses structured illumination of the sample and detailed computer analysis of the resulting fringe structure to provide increased resolution. At low powers the technique only affords a factor-of-two improvement in resolution. Higher resolution can be achieved with increased laser powers (saturation effects) at the expense of an increased risk of sample damage [7].
STED creates a sub-resolution fluorescent spot by the overlap of the initial exciting beam (PUMP) with a depletion (DUMP) laser (pulsed or continuous wave) which is ‘shaped’ to provide a toroidal intensity profile at the focal plane. The DUMP removes close to 100% of the fluorescent molecules outside the central minimum through stimulated emission. The PUMP-DUMP combination is scanned over the sample to produce the image. The drawbacks of STED are the complex experimental design and the on-sample DUMP power required to obtain the necessary level of resolution. This corresponds to intensities where the onset of photochemical damage and sample heating becomes a significant risk. GSD microscopy operates on a similar principle to STED. A spatially offset second laser is used to (strongly) drive excited molecules into long lived non-fluorescing triplet states resulting in a reduced fluorescent spot. GSD resolution is degraded by triplet lifetime shortening due to oxygen quenching, requiring the development of customised fluorescent probes and/or the removal of oxygen by specialised mounting media.
2. Super resolution in time resolved imaging
Here we describe a new approach to achieving super resolution in fluorescence microscopy, through analysis of the time evolution of fluorescent images in the presence of continuous wave (CW) stimulated emission depletion. Under these conditions the time evolution (spatial broadening and the in-growth of additional structure) of the point spread function (PSF) of a sub-resolution fluorescent object can be split up into time segments which are recombined to yield a reconstructed PSF whose new width is below the constraints of Eq. (1). The reconstruction can then be applied to the sample as a whole to yield a sub-resolution image.
To demonstrate the principle of the technique, consider the simplified case of a point source of fluorescence in a conventional (e.g. confocal) microscope with one dimensional (line) scanning. In the far field limit light of wavelength λ can only be focused to an area whose radius is given by c.a. λ/2 (Eq. (1)). A scanning microscope will therefore not only excite and detect fluorescence from the region of the sample at the centre of the laser focus, but also from adjacent regions, albeit with a reduced probability. The quantitative description of this effect is encapsulated in the PSF. This causes a point source of fluorescence (i.e. an object much smaller than the focused spot size) located at x0 to appear spread out from this point with an intensity distribution of fluorescence described by PPSF(x). For a high numerical aperture this is well approximated by a Gaussian of the form [2]
where is the full width at half maximum (FWHM) of the Gaussian PSF [15,16] and A is a dimensionless normalisation coefficient. The intensity of detected fluorescence is given by:The dimension x can be in units of length or number of pixels (with appropriate choice of ω) in an image with x0 the location of the point source. The FWHM will depend on the method of excitation, the wavelength of light and the detection optics (e.g. microscope objective and, where present, confocal optics and pinhole). The time dependence of the fluorescence intensity I(t) at time t after pulsed excitation is a function of the fluorescence lifetime tf of the fluorescent probe molecule;I(t0) is the initial fluorescence intensity detected after excitation. The fluorescence lifetime is the same for all identical molecules in equivalent environments. However in the presence of a CW depletion field the rate of population decay from the excited state is increased due to stimulated emission, and the detected fluorescence will show an increased decay rate (a shorter apparent fluorescence lifetime). The intensity of the focused depletion field ID(x) also has a spatial (Gaussian) variation of the formwhere ω′ is a measure of the beam radius as defined for Eq. (2). The stimulated emission rate at position x is linearly dependent on I’D(x):where B is a constant of proportionality. As a consequence probe molecules at different positions within the PSF will display different fluorescence decay rates. For time periods which are comparable to, or longer than, the normal fluorescence lifetime the fluorescence will increasingly arise primarily from molecules located toward the edges of the PSF as these experience lower rates of stimulated emission depletion. In the explanation given here the effects of the orientation of the emission dipole moment of the fluorescent species have been neglected as has any rotational diffusion of the chromophore. The latter assumption is realistic for molecules whose rotational times are long in comparison to their fluorescence lifetimes such as fluorescent proteins or fluorescent probe molecules rigidly attached to a much larger biomolecule. This removes any angular or time dependence of B. This is not a requirement for the technique to work, but greatly simplifies its explanation.For pulsed excitation and CW depletion the fluorescence decay can be modeled as a simple two level system by grouping together the thermally accessible vibrational levels of the first excited electronic state (population NEX) and the uppermost vibrational levels of the ground state that are resonant with the depletion laser wavelength (NGS). The decay of the populations are given by
where kf is the sum of the radiative and non radiative decay rate constants and kVIB is the rate of relaxation of the uppermost ground state vibrational levels. For all practical rates of CW depletion kVIB can be considered so fast that no build up of population in the ground state is produced (NGS = 0). The proportion of fluorescence to the uppermost vibrational levels of the ground state NGS is negligible. The solution to the above rate equations is thus simplyIt is convenient to define the strength of the depletion field in terms of the relative increase in the decay rate for a particular chromophore. The dimensionless quantity Fd describes the proportion of molecules that undergo stimulated emission in the most intense region of the depletion field as defined bySubstitution of Eq. (8) into Eq. (9) and combining terms into a single decay rate gives an effective fluorescence lifetime in the presence of depletion tD(x) [17]Chromophores located at or near the centre of the depletion laser focus will thus have the shortest fluorescence lifetimes, whilst those at the margins (and above and below the focal plane) will exhibit lifetimes only slightly reduced from their intrinsic values. For early times after excitation this variation of lifetime has little effect on the spatial distribution of fluorescence. As time progresses the population of the faster decaying species at the centre of the focal volume reduces to a lower level than that at the periphery, resulting in a local minimum (‘dip’) in the centre of the PSF. After a few fluorescence lifetimes from excitation the principle contribution to the remaining fluorescence signal will come from molecules located at the edges of the PSF. As the lasers scan, chromophores previously distant from the focus (possessing a near to normal lifetime) will subsequently be subject to an increased depletion intensity and thus exhibit a shorter fluorescent lifetime. The time at which a fluorescence photon is emitted now becomes correlated with the scan position of the two lasers. For algebraic clarity it has been assumed thus far that the beam radius of the focused depletion field ω′ = ω (Eqs. (1), (5)), although it should be noted that this is not a necessary requirement. An example of the evolution of the effective PSF in one dimension for a point source is shown in Fig. 1(a).
Fig. 1 Reconstruction of a reduced point spread function with Fd = 0.33. a) The five ‘time window’ distributions used in the reconstruction of the improved PSF in units of tf, plus an example of a later time window (5.0tf) showing a PSF with a significant local minimum. b) Confocal lateral PSF in absence of CW stimulated emission depletion and reconstructed PSF using the linear combination spread of the five time windows as in Eq. (11) with coefficients: c0 = 1, c0.5 = 5.8550, c1 = −10.028, c2 = 11.1066, c3 = −7.3387. Fitted Gaussians yield FWHMs of 6.66 pixels and 1.99 pixels respectively, indicating a 3.3-fold improvement. c) Confocal and reconstructed axial PSFs, with coefficients as before. Fitted Lorentzians yield FWHMs of 5.66 and 1.16 pixels respectively, indicating a 4.9-fold improvement. d) x-z surface plot comparing confocal and reconstructed PSFs.
The key approach to the technique is that the spatially varying time resolved fluorescence intensities are divided into images corresponding to separate temporal slices or ‘time windows’ within the intensity decay. Each time window therefore corresponds to an image recorded with its own effective PSF (Fig. 1(a)). These PSFs form a basis set from which it is possible to construct an optimised PSF (a linear superposition). For example, subtracting some of the 3.0tf fluorescence distribution from the 0tf distribution in Fig. 1(a) would reduce the signal from the edges of the 0tf PSF by a greater degree than at the centre, leading to a narrower combined PSF (albeit at the cost of signal intensity). Each subsequent addition/subtraction of an effective PSF in the correct proportion will further reduce the signal intensity at the margins of the original PSF relative to the centre. An enhancement in resolution beyond the Abbé criterion (Eq. (1)) is possible as although the later time window PSFs are broader than the diffraction-limited case they (and the differences between them) contain spatial variations that are on a shorter scale than the diffraction-limited case (e.g. Figure 1(a), 5.0tf). As the intensity of the fluorescence from non-interacting molecules is additive (linear) a fluorescence image can be reduced to a sum of individual point sources individually broadened by the PSF. Each time window therefore corresponds to the same point sources of fluorescence broadened by different effective PSFs. A linear superposition of these images that leads to the optimum PSF for a single point should therefore lead to the same optimisation for all point sources that constitute the image. The resultant/reconstructed image will therefore be equivalent in terms of resolution to one measured using the optimised/reconstructed PSF. This is quantified and demonstrated for one and two dimensional images in Fig. 1(b) and Fig. 2.
Fig. 2 Simulated 2D image reconstruction. a) Original structure to be imaged. b) ‘Measured’ image simulating the effect of the microscope PSF on (a). c)-e) Simulated images of the effect of the PSF and CW stimulated emission depletion after 1, 2 and 3 fluorescence lifetimes respectively (Fd = 0.33). f) Reconstructed image from time slices of the evolving image. The coefficients in the reconstruction and original PSF FWHM are as in Fig. 1.
For a one dimensional image the reconstructed PSF is therefore given by
where n is the number of time windows, ti is the location of the time window and ci are dimensionless coefficients which determine the contribution of each time window to the reconstructed PSF Pres(x) . Figure 1(b) shows the results of combining 5 such time windows together in the appropriate proportions and the resultant narrowing of the PSF. The coefficients of the superposition can be determined in a number of ways. In this simple case they are set such that the values of the superposition at 4 points at the edge of the PSF are zero (pixels 2, 3, 4 and 6 in Fig. 1(b)) by solving the resulting simultaneous equations. For clarity the width of the time windows used in this model reconstruction are effectively infinitesimal, therefore not requiring integration of the intensity decay over the appropriate time period. Any calculation of coefficients for real image data will however require this relatively simple operation to be performed. Alternatively the coefficients can be optimised using an iterative algorithm - for example to minimise the standard deviation of the PSF. For real images the measured fluorescence intensities at any particular sampling point (pixel) will consist not only of the molecules at that point but also a contribution from nearby/neighbouring points weighted by the PSF evaluated for the distance between these points and the central point. The one dimensional ‘image’ can therefore be represented by a linear superposition of individual single pixel/sampling point PSFs and intensities. The fluorescence signal measured in the nth pixel/sampling point Intot is given bywhere Nx is the detected number of fluorescence events within that pixel. The range of the summation (xmin→xmax) would ideally be the entire range of pixels but can easily be truncated to a point where the contribution to the PSF is negligible. As discussed above, in the presence of the depletion laser the fluorescence lifetime of the probe molecules has a spatial variation dependent on the depletion intensity at that point. The contribution to the total fluorescence intensity at any one pixel from nearby/neighbouring pixels in the above sum will have a time dependence given bywhere tD(x) is calculated as in Eq. (10). The spatial distribution of fluorescence in each time window is thus well described. As stated previously, the key concept in reconstructing a high resolution image is that the measured image can be considered as a linear superposition of single point PSFs and intensities. Therefore the optimal combination of time windows calculated to minimise the PSF of a single point described above will produce the minimum PSF for all such point sources that contribute to the fluorescence image. The image formed from this combination will be identical to that which would have been produced from a conventional instrument with a PSF identical to that of the reconstructed PSF.It is important to note that optimisation of the PSF in the xy plane also leads to improvement in the axial direction (z axis). The intensity of the depletion beam decreases above and below the laser focal plane and hence leads to a longer effective lifetime for molecules emitting in these regions. Later time windows will therefore contain a higher signal from the regions above and below the focal plane as well as from regions that have a larger radial distance from the focal point. A combination of time window images intended to improve the resolution in one of these dimensions should in principle lead to a similar improvement in the other. Considering only the axial direction the PSF, and laser intensity, has a Lorentzian dependence about the beam waist
where the parameter k is the inverse of the square of the Rayleigh range of the focused Gaussian beam [15] and A′ is a constant of proportionality (c.f. Eq. (2)); for algebraic simplicity the PSF and the spatial variation in the depletion intensity are assumed to be the same as above (this is also not necessary). Figure 1(c) shows the resultant PSF in the axial direction by combining the time window one-dimensional images in the combinations used in Fig. 1(b). The result shows a decrease in the PSF that is actually greater than in the radial direction with a trade-off in a reduction of contrast.The approach is easily extended to two-dimensional images. All that has to be done is to describe the normal PSF and the relative depletion intensity as functions of both dimensions.
where n,m are the x and y pixel numbers of the image. Again for clarity it has been assumed the PSF and the spatial intensity profile of the CW depletion laser have similar size and radial symmetry but this is not a fundamental requirement. In non-radial symmetry the PSF and the depletion laser intensity functions are simply the products of two one-dimensional functions that separately describe the intensity variation in each dimension. Figure 2 shows simulations of the process for a specified two-dimensional image consisting of 32 × 32 pixels. Importantly, the time-windows and the coefficients for the reconstruction are identical to those calculated in Fig. 1 for optimization of a single point source. This shows the original structure (Fig. 2(a)), the simulated confocal image (Fig. 2(b)) and three of the five time window images used in the reconstruction (Figs. 2(c)-2(e)), where each time window image depicts the original structure blurred by the corresponding effective PSF from Fig. 1(a). Figure 2(f) shows the reconstructed image resulting from combining the component images in the previously calculated proportions.‘Time gating’ has been deployed in CW STED microscopy to improve the obtainable resolution [18]. In this approach the early time fluorescence is neglected as very little STED has yet taken place. By only looking at fluorescence emitted after some time following excitation, when the toroidal depletion field has reduced more of the signal from the edges of the PSF, a resolution enhancement closer to that of pulsed depletion can be achieved at the expense of a large reduction in fluorescence signal. In contrast here we use the differences in the effective PSFs of different time regions of the decay (which contain structure below the diffraction limit) to engineer a reduced PSF.
The coefficients for the reconstruction depend only on the intrinsic PSF of the instrument, the pixel resolution and the degree of depletion Fd. The first two are independent of the sample and are either constant or experimentally controllable. Fd depends on the intensity of the depletion beam, the particular probe molecule chosen and to some degree the local environment of the probe (e.g. fluorescence lifetime variations, molecular orientation and rotational diffusion rates). If Fd is unchanging between images there is no need to re-evaluate the coefficients for different images. If Fd does change (even between different regions of the sample) this does not prevent the reconstruction of an improved resolution image. However the calculated coefficients will no longer represent the optimum reconstructed PSF in these regions.
Experimental verification of the resolution enhancement was investigated using structures of known size and shape i.e. sub-diffraction fluorescent nanospheres both in inert and live-cell environments.
3. Materials and Methods
The apparatus employs an inverted confocal fluorescence lifetime imaging microscope (FLIM) system (MT200, Picoquant GmbH Berlin). Fluorescence was excited using a 490nm 100ps pulsed diode laser (PicoTA 490, Picoquant), CW depletion was achieved using a 594nm diode laser (Mambo, Cobolt inc. Sweden). Both lasers were coupled into the objective (Olympus UPLSAPO 60x water immersion, NA 1.2) via a single mode polarization maintaining optical fibre. The PSF of the microscope was determined using 20nm fluorescent nanospheres (yellow-green fluospheres Invitrogen inc.) in both agar and live HEK cells (see Fig. 3 and Fig. 4) On sample laser powers were measured using a powermeter (S130C, Thorlabs USA).
Fig. 3 a) A 1D line profile across a 20nm fluorescent bead in agar following 490nm excitation in the presence of a 7.5mW 594nm CW depletion laser, comparing well to theoretical prediction (Fig. 1). b) Reconstruction of a reduced PSF for the bead by a linear combination of the five time windows in a) with coefficients: c1 = 1, c2 = −1.2, c3 = 9.2, c4 = −41, c5 = 99. The PSF with no depletion has a width of 240nm and the width of the reconstructed PSF is approximately 125nm. Image scan time was 17 seconds with a total acquisition time of 26 seconds (3.2x3.2μm scan).
Fig. 4 a) A confocal fluorescence image of a cluster of 20nm nanospheres in a live HEK cell. Scale bar = 1μm. b) The reconstructed image using a linear combination of the time windows as in Fig. 3 with coefficients chosen to minimize the PSF of nanosphere ‘1’ applied to the whole image. A CW depletion power of c.a. 7.5mW at 594nm was employed. c) Reconstructed image from b) smoothed as described in Methods. d) Profiles across nanosphere ‘1’ and fitted Gaussians for confocal (FWHM = 284nm) and raw reconstructed image (FWHM = 147nm). e) Profiles across line in a) showing distinct separation of three closely-spaced nanospheres in the reconstruction. FWHM of nanosphere ‘2’ is 158nm in the raw reconstructed image. Recombination coefficients were: c1 = 1, c2 = −1.4, c3 = 0.5, c4 = −3, c5 = 3.5. Original images in a) and b) were 150x150 pixels with a pixel size of 31.5nm and were subsequently 2x2 binned to give a pixel size of 63nm, shown here. Image scan time was 17 seconds with a total acquisition time of 29 seconds.
3.1 Preparation of nanospheres in agar
Agar blocks containing 20nm diameter nanospheres were prepared according to a method adapted from [19]. A 0.04% (wt%) solution of fluorescent nanospheres was mixed into a 4% agar (Sigma) solution at a 3:7 (nanosphere/agar) ratio. The mixture was pipetted onto a glass coverslip and allowed to set at room temperature. During imaging the agar blocks were immersed in water to prevent drying.
3.2 Preparation of nanospheres in living cells
Uptake of 20nm nanospheres into living HEK cells was achieved using a method adapted from [20]. HEK-293 cells were cultured in advanced DMEM (Invitrogen) supplemented with FBS (10%), penicillin (100 units/ml), streptomycin (100ug/ml) and l-glutamine (2mM), and plated onto glass coverslips at a density of 100,000 cells per coverslip. The following day cells were incubated with a 1:1000 dilution of 2% nanosphere solution in DMEM for 4 hours at 37°C before re-immersion in phenol red-free HEPES-buffered DMEM (Sigma), in which the cells remained immersed during imaging. A trypan blue exclusion test was used to assess cell viability after imaging.
3.3 Image Analysis
For clarity of display, the reconstructed super resolution image presented in Fig. 4(c) was smoothed using a standard routine (Origin 6, Micro Cal Labs). This consists of a compression of the image using adjacent point averaging followed by a re-expansion using a bi-linear interpolation [21]. Although this slightly reduces the attained resolution the reduction in noise significantly improves apparent contrast/clarity. All PSF measurements (Fig. 3 and Fig. 4) are made using unsmoothed data.
4. Results
Figure 3 shows data from a 1-dimensional segment of a 20nm fluorescent nanosphere in agar. In the absence of the depletion laser the PSF is well described by a Gaussian with a FWHM of 240nm. The fluorescence decay of the bead is found to be bi-exponential
with A1 = 0.29 τ1 = 3.88ns, A2 = 0.71 τ2 = 1.68ns. In the presence of an on-sample CW depletion power of ca. 7.5mW at 594nm both lifetime components are shortened with A1 = 0.32 τ1 = 3.15ns, A2 = 0.68 τ2 = 1.43ns. The original decay parameters were recovered on subsequent re-measurement of the sample in the absence of CW depletion. Figure 3(a) shows five time windows (0-1.12ns, 1.12-3.52ns, 3.52-6.72ns, 6.72-9.92ns and 9.92-14.72ns, with corresponding coefficients c1-5) of the 1D line profile of a single nanosphere under these conditions. It can be seen that the PSF evolves as predicted (Eqs. (2-10) and Fig. 1) with an increasing reduction in the relative intensity of the centre of the PSF following excitation. A linear combination of the five time windows yields a reduced PSF of ca. 125 nm (Fig. 3(b)).Figure 4 shows images of a cluster of 20nm nanospheres within a living mammalian (HEK) cell. Figure 4(a) shows the confocal fluorescence image of the cluster where the centre nanosphere ‘1’ has a PSF with a FWHM of 284nm. In the presence of ca. 7.5mW CW depletion power at 594nm and taking the same lifetime windows as in Fig. 3 the PSF of this section of the image was reduced to 147nm. Applying the same linear combination of time windows to the whole image yields the image shown in Fig. 4(b). A similar improvement in resolution for the other nanospheres is achieved. From the reconstructed 1-dimensional PSFs (Fig. 3 and Fig. 4), it is clear that the cost of increased spatial resolution is a decrease in the signal to noise ratio.
The non-destructive nature of the technique was confirmed using the trypan blue viability test [22]. HEK cells containing nanospheres were subjected to the same pump and CW depletion powers as used in the image reconstruction experiments. All cells survived exposure with no apparent morphological changes or loss of membrane integrity.
5. Discussion: time window, Fd and fluorescence lifetime effects
The choice of appropriate time windows and Fd are interrelated with the fluorescence lifetime and the available signal to noise. The aim in choosing these parameters is to have effective PSFs for each time window that are significantly different from each other to make the reconstructed PSF more effective in increasing the resolution and containing minimal artifacts. Fd is a measure of the ratio of the depletion rate to the fluorescence rate. In principle both low (close to 0) and high (close to 1) values can be used with an appropriate choice of time windows. If Fd is low however it will take a greater number of fluorescence lifetimes for the ‘dip’ in the PSF to evolve (Eq. (11), Fig. 1(a) and Fig. 3(a)). This in turn will require wider and later time windows to observe this behaviour. The converse is true of high values of Fd. The depletion laser intensity required for a particular value of Fd depends on the STED cross-section and fluorescence lifetime of the chromophore, as the STED rate has to compete with combined radiative and non-radiative rates to observe a significant change in the fluorescence lifetime. Although in principle the method can be used for any value of the fluorescence lifetime, a chromophore with a longer fluorescence lifetime is advantageous in the sense that it requires a lower laser power to achieve the same Fd (for the same STED cross-section). As stated above, this is particularly useful when imaging photosensitive biological samples. A particularly rapid fluorescence lifetime will not only be more difficult to deplete (requiring a higher depletion intensity), but is also likely to have a low quantum yield and therefore reduced brightness. A high degree of time resolution will be required when using a significant number of time windows. In the simulations the choice of Fd is arbitrary as the time window positions can be varied to produce the same effective PSFs.
Crucially, the obtainable resolution is not limited by the laser intensity, but rather by the number of time windows and spatial sampling points. However, increasing either of these will reduce the signal to noise ratio. In the experimental measurements the positions and number of time windows were chosen as a compromise between achieving good signal to noise (low in late time windows) and observing the ‘dipped’ PSF in the late time windows (Fig. 3(a)). In this case a higher value of Fd has some advantage in that it produces the ‘dipped’ PSF earlier in the fluorescence decay where the signal to noise is higher. This was kept relatively low in the experiments however to minimize the risk of damaging the sample (conventional STED uses 1 to 2 orders of magnitude higher CW depletion powers). Finding the optimum positions of the time windows for a specific Fd in the presence of noise is a complex problem and is the subject of continuing work.
6. Conclusions
We have shown that it is possible to achieve sub-diffraction limited spatial resolution in fluorescence microscopy via the reconstruction of an evolving fluorescent image in the presence of a low-power unstructured (Gaussian) CW depletion laser. The technique permits the non-invasive imaging of live cells whose viability is maintained during the measurement process. A side effect of using much less depletion laser power is a reduction both in scattered light and fluorescence caused by the DUMP. Less PUMP power is therefore required for the PUMP fluorescence signal to be observed above these levels. This reduces the amount of photobleaching of the fluorescent species and heating of the sample. The simplicity of this approach affords the possibility of achieving super resolution by low-cost modifications to standard FLIM systems. A particular advantage is that the degree of super resolution to image contrast can be retrospectively adjusted – i.e. only one modified image needs be taken and subsequently re-analysed for different resolution/contrast trade-offs. A limitation however is the low signal-to-noise ratio in later time windows where the photon count is reduced relative to early times. However all super resolution methods are constrained by the interplay of increased spatial resolution at the cost of reduced image contrast (signal-to-noise ratio). The SPAD detector employed here has significant after-pulsing degrading the signal-to-noise ratio compared to a hybrid-PMT or micro channel plate detector. The acquisition speed of the technique is principally limited to the time taken to acquire sufficient photons given that the photon detection rate for TCSPC is 1% of the laser repetition rate. We are currently investigating the effects of an increased photon collection rate (a departure from Poisson statistics) on the degree of resolution attainable. We are in the process of incorporating a z-scanning capability into the microscope to fully quantify the 3-dimensional resolution that the technique affords.
Acknowledgments
We are grateful to UCL, the BBSRC and the EPSRC for funding this work. We thank Dr Thomas Blacker for preparing the HEK293 cells.
References and Links
1. M. Minsky, “Microscopy Apparatus,” US Patent 3,013,467 (1961).
2. S. G. Lipson, H. Lipson, and D. S. Tannhauser, Optical Physics (Cambridge University, 1998).
3. 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]
4. M. J. Rust, M. Bates, and X. Zhuang, “Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM),” Nat. Methods 3(10), 793–796 (2006). [CrossRef] [PubMed]
5. S. T. Hess, T. P. Girirajan, and M. D. Mason, “Ultra-high resolution imaging by fluorescence photoactivation localization microscopy,” Biophys. J. 91(11), 4258–4272 (2006). [CrossRef] [PubMed]
6. M. G. L. Gustafsson, “Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy,” J. Microsc. 198(2), 82–87 (2000). [CrossRef] [PubMed]
7. R. Heintzmann, T. M. Jovin, and C. Cremer, “Saturated patterned excitation microscopy--a concept for optical resolution improvement,” J. Opt. Soc. Am. A 19(8), 1599–1609 (2002). [CrossRef] [PubMed]
8. 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] [PubMed]
9. T. A. Klar and S. W. Hell, “Subdiffraction resolution in far-field fluorescence microscopy,” Opt. Lett. 24(14), 954–956 (1999). [CrossRef] [PubMed]
10. 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] [PubMed]
11. S. W. Hell and J. Wichmann, “Process and device for optically measuring a point on a sample with high local resolution,” US Patent 5,731,588 (1998).
12. S. W. Hell and K. Willig, “STED-Fluorescent light microscopy with two-photon excitation,” US Patent 20,100,176,307 (2010).
13. S. W. Hell and M. Kroug, “Ground-state depletion fluorescence microscopy, a concept for breaking the diffraction resolution limit,” Appl. Phys. B 60(5), 495–497 (1995). [CrossRef]
14. S. Bretschneider, C. Eggeling, and S. W. Hell, “Breaking the diffraction barrier in fluorescence microscopy by optical shelving,” Phys. Rev. Lett. 98(21), 218103 (2007). [CrossRef] [PubMed]
15. A. E. Siegman, Lasers (University Science Books, 1997).
16. A. Yariv, Quantum Electronics, 3rd ed. (Wiley, 1989).
17. A. J. Bain and R. J. Marsh, “Improvements relating to fluorescence microscopy,” Patent WO 2,013,008,033 (2013).
18. G. Vicidomini, G. Moneron, K. Y. Han, V. Westphal, H. Ta, M. Reuss, J. Engelhardt, C. Eggeling, and S. W. Hell, “Sharper low-power STED nanoscopy by time gating,” Nat. Methods 8(7), 571–573 (2011). [CrossRef] [PubMed]
19. R. Niesner, V. Andresen, J. Neumann, H. Spiecker, and M. Gunzer, “The power of single and multibeam two-photon microscopy for high-resolution and high-speed deep tissue and intravital imaging,” Biophys. J. 93(7), 2519–2529 (2007). [CrossRef] [PubMed]
20. W. Zauner, N. A. Farrow, and A. M. Haines, “In vitro uptake of polystyrene microspheres: effect of particle size, cell line and cell density,” J. Control. Release 71(1), 39–51 (2001). [CrossRef] [PubMed]
21. W. H. Press, B. P. Flannery, S. A. Teukolsky, and W. T. Vetterling, Numerical Recipes in C: The Art of Scientific Computing (Cambridge University, 1988).
22. W. Strober, “Common cell techniques for immunology,” in Short Protocols in Immunology, J. E. Coligan, B. E. Bieher, E. M. Shevach and W. Strober ed. (Wiley, 2005).
