Biotechnol. J. 2011, 6, 1037–1051
DOI 10.1002/biot.201100031
www.biotechnology-journal.com
Review
Superresolution imaging of biological nanostructures by spectral precision distance microscopy Christoph Cremer1,2,3,8,9, Rainer Kaufmann2, Manuel Gunkel2,3, Sebastian Pres2, Yanina Weiland2, Patrick Müller2, Thomas Ruckelshausen2, Paul Lemmer2, Fania Geiger4, Sven Degenhard5, Christina Wege5, Niels A. W. Lemmermann6, Rafaela Holtappels6, Hilmar Strickfaden7 and Michael Hausmann2 1 Institute
of Molecular Biology, Mainz, Germany Institute for Physics, Heidelberg University, Heidelberg, Germany 3 BioQuant Center, Heidelberg University, Heidelberg, Germany 4 Max Planck Institute for Intelligent Systems, Stuttgart, Germany 5 Institute of Biology, University of Stuttgart, Stuttgart, Germany 6 Institute of Virology, University Medical Center, University of Mainz, Mainz, Germany 7 Biocenter, Ludwig Maximilian University of Munich, Munich, Germany 8 Institute for Pharmacy and Molecular Biotechnology, Heidelberg University, Heidelberg, Germany 9 Institute for Molecular Biophysics/The Jackson Laboratory, Bar Harbor, ME, USA 2 Kirchhoff
For the improved understanding of biological systems on the nanoscale, it is necessary to enhance the resolution of light microscopy in the visible wavelength range beyond the limits of conventional epifluorescence microscopy (optical resolution of about 200 nm laterally, 600 nm axially). Recently, various far-field methods have been developed allowing a substantial increase of resolution (“superresolution microscopy”, or “lightoptical nanoscopy”). This opens an avenue to ‘nano-image’ intact and even living cells, as well as other biostructures like viruses, down to the molecular detail. Thus, it is possible to combine light optical spatial nanoscale information with ultrastructure analyses and the molecular interaction information provided by molecular cell biology. In this review, we describe the principles of spectrally assigned localization microscopy (SALM) of biological nanostructures, focusing on a special SALM approach, spectral precision distance/position determination microscopy (SPDM) with physically modified fluorochromes (SPDMPhymod). Generally, this SPDM method is based on high-precision localization of fluorescent molecules, which can be discriminated using reversibly bleached states of the fluorophores for their optical isolation. A variety of application examples is presented, ranging from superresolution microscopy of membrane and cytoplasmic protein distribution to dual-color SPDM of nuclear proteins. At present, we can achieve an optical resolution of cellular structures down to the 20-nm range, with best values around 5 nm (~1/100 of the exciting wavelength).
Received 28 April 2011 Revised 13 July 2011 Accepted 2 August 2011
Keywords: Localization microscopy · Microscopy · SALM · SPDM · Super-resolution imaging
1 Correspondence: Prof. Christoph Cremer, Institute of Molecular Biology, Ackermannweg 4, 55128 Mainz, Germany E-mails:
[email protected];
[email protected] Abbreviations: CLSM, confocal laser scanning fluorescence microscopy; SALM, spectrally assigned localization microscopy; SMI, spatially modulated illumination; SPDM, spectral precision distance microscopy; YFP, yellow fluorescent protein
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Introduction
In conventional far-field light microscopy [1, 2], the optical resolution is limited by diffraction to R = 0.61 · λ/NA, where R is the smallest resolvable distance, λ the emission wavelength and NA the numerical aperture. The resolution limit according to Rayleigh is shown in Fig. 1.
1037
Biotechnology Journal
Biotechnol. J. 2011, 6, 1037–1051
Figure 1. Illustration of the limits of lacteral optical resolution in conventional fluorescence microscopy. Diffraction pattern (Airy disc) produced by the fluorescence emission of a single molecule (top), and of two overlapping diffraction patterns (Airy discs) produced by two adjacent molecules (bottom). The smallest resolvable distance between the two maxima has been defined as the optical resolution. Reprinted with permission from Physik in unserer Zeit (Wiley-VCH) [3].
In recent years, various techniques for far-field fluorescence microscopy have been developed to overcome this limitation. Besides structured illumination microscopy (SIM) [4, 5] and stimulated emission depletion microscopy (STED) [6], localization microscopy is becoming a widely used subdiffraction-limit method. The fundamental concept of this technique is based on the precise determination of the positions of optically isolated single molecule signals [7–10]. The optical isolation of closely adjacent molecules is achieved by labeling with different ‘spectral signatures’. Various names have been applied to this general principle, such as ‘colocalization microscopy’ [11, 12] ‘superresolution by spectrally selective imaging’ [13], or spectral precision distance/position determination microscopy (SPDM) [10, 14]. At present, the most effective way to achieve optical isolation is to use fluorophores that can be switched between two different ‘spectral states’; in this way, the required optical isolation (i.e., an independent registration of the respective diffraction patterns) of the single molecule signals can be realized even in a densely labeled structure. First concepts to use random labeling to enhance the op-
1038
tical resolution were put forward already in 2003 [15], while first experiments to realize this by stochastic ‘blinking’ were described in 2005 for fluorescent nanodots [16]. Using photoactivatable or photoswitchable fluorophores, techniques like photoactivated localization microscopy (PALM) [17], fluorescence-activated localization microscopy (FPALM) [18] and stochastic optical reconstruction microscopy (STORM) [19] allow biological structures to be imaged with a resolution down to 20 nm. All these methods require special fluorophores (or pairs of fluorophores) that are commonly switched/excited by two laser frequencies. In contrast, the special SPDM protocol presented here (SPDMPhymod) uses the stochastic recovery of conventional fluorophores from a light-induced reversibly bleached state to the fluorescent state for the optical isolation of single molecule signals [20, 21]. No specially designed fluorophores, pairs of fluorophores, or additional laser wavelengths are needed to achieve superresolution with a given fluorophore. The advanced SPDM approach [20–22] is based on reversible photobleaching [23]. Unlike the usu-
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
Biotechnol. J. 2011, 6, 1037–1051
al bleaching effect used in PALM or FPALM, where the structure of fluorescent molecules is irreversibly modified towards a non-fluorescent ‘dark’ spectral signature state (at given excitation/emission conditions), in the SPDM technique described here this effect is a reversible one. The functional connection between the three fundamental states of a molecule under these latter conditions may be described with the transitions k2
⎯⎯ ⎯⎯ → M ⎯⎯→ M Mrbl ← fl ibl k3
k1
where Mrbl is the reversibly bleached state, Mfl is the fluorescent state, and Mibl is the irreversibly bleached molecule. The rate constants of the crossing processes are indicated with ki(i=1, 2, 3), where the processes are assumed to be first order reactions. The ratio between the probabilities for reP versible and irreversible bleaching Prblibl can be significantly affected by physicochemical modifications of the molecule due to its environment, or due to illumination with light of appropriate wavelength and intensity (“physically modified fluorophores”). This effect has been well studied in the context of fluorescent protein derivatives from the jellyfish Aequorea victoria. After starting to illuminate fluorescent molecules (Mfl) with excitation light, a certain amount is bleached instantly k3 ( M fl ⎯⎯→ Mibl ). Another amount is transferred into k2 the reversible dark state ( M fl ⎯⎯→ Mrbl). We envisaged that the statistical recovery of fluorescent molecules (Mfl) from this ‘bright’ state (Mrbl) and transition into an irreversibly bleached ‘dark’ state (Mibl), with a delay time sufficient for single fluorescent molecule registration, would allow an additional possibility for optical isolation of single molecules in the time domain. This would offer another approach to high-resolution localization microscopy of the number and positions of molecules (even of the same type) within a given observation volume. In the original description of SPDM as published since the mid 1990s [10, 14, 15], the concept of ‘spectral signature’ included not only differences in the absorption/emission spectrum and fluorescence lifetimes but any photophysical procedure to obtain optical isolation, including, for example, luminescence, that means very long-lived excited states (i.e., ‘dark’ metastable states compared to short-lived excited states of S0 to S1 transitions). In this concept, the stochastic reversible bleaching of the individual molecules is conceived as a special ‘spectral signature’ based on the lifetime of verylong-lived individual dark states. In this sense, the spectral signature of reversible bleaching for an individual molecule can be defined as the duration of
© 2011 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
www.biotechnology-journal.com
the existence (lifetime) of its metastable dark state, before it ‘decays’ to a fluorescent state. This notion has been used to estimate the average ensemble lifetimes required to obtain an acceptable optical isolation by this method [24]. To stress the fact that in this type of SPDM protocol the stochastic recovery rate critically depends on the physico-chemical conditions used, this technique was also denoted as SPDM with physically modified fluorophores (SPDMPhymod) [22, 25]. A first report [21] (published 8 May 2008) demonstrated SPDM of this type with an average localization accuracy of 15 nm (corresponding to an optical resolution of about twice this value) [22, 24] of nuclear pore protein (p62) distribution on the nuclear envelope in intact human cells immunolabeled with the standard fluorescence dye Alexa488, using one laser wavelength only (λexc =488 nm) for both excitation and switching. Shortly afterwards, related (standard fluorophores, appropriate illumination intensities and switching buffers) but in detail and applications differing methods were published and denominated as ground state depletion microscopy (GSDIM) [26]; as direct stochastic optical reconstruction microscopy (dSTORM) [27]; or as reversible photobleaching microscopy (RPM) [28]. Since then, these highly simplified and efficient localization microscopy methods have been extended to allow the simultaneous superresolution of various types of molecules, e.g., dual-color localization microscopy (2CLM) [29], or d4STORM [30]. From the methodical point of view, these methods appear to be related to each other, the technical main differences being the switching buffer to adjust the ‘blinking’ frequency, and the molecule types and laser frequencies used in the first publications [20, 21, 26, 28].All these methods rely on the use of light-induced reversible transitions between extremely long-lived ‘dark’ and ‘bright’ molecular states; however, in contrast to the GSD/RESOLFT techniques of focused nanoscopy [31], in the case of SPDM/RPM/dSTORM etc., superresolution down to the molecular level is possible even at homogeneous illumination; furthermore, the ‘dark’ to ‘bright’ transition times are several orders of magnitude higher in SPDMPhymod and related methods than in the T1 to S0 transitions discussed in GSD microscopy [32]. The various denominations given to the novel localization microscopy approaches have made it possible to include even standard fluorochomes in a straightforward way, and are not only justified by the many differences in the optical setup, the molecule types, and the physicochemical environments used. In addition, they stress various ele-
1039
Biotechnology Journal
ments of the general concept of “spectrally assigned localization microscopy” (SALM) [20, 22, 24, 33]. This name is proposed as a joint ‘family’ name for the various approaches (PALM/FPALM, STORM/dSTORM, GSDIM, SPDM, etc.) achieving superresolution by precision localization of optical isolated point emitters, using as an assignment-appropriate ‘spectral signatures’ (including absorption/emission spectral, lifetimes, luminescence, dark-bright transitions, etc.). For example, in PALM and FPALM the importance of using photoswitchable molecules is put forward; in STORM, dSTORM and d4STORM the optical isolation by stochastic differences between spectral signatures is highlighted; in GSDIM the focus is put on the quantum physical necessity to realize very long-lived excited states for S0 depletion; in “RPM” the effect of reversible photobleaching is denoted; in 2CLM the possibility of extending SALM to the simultaneous superresolution of multiple molecule types is envisaged; and in SPDMPhymod performing highly precise position/ distance measurements of single molecules under specific physicochemical conditions (e.g., illumination intensity, chemical environment) is highlighted. The exact photophysical mechanisms underlying the ‘blinking’ effects used in 2CLM, GSDIM, RPM, SPDMPhymod, dSTORM/d4STORM etc. are still widely unknown. From the formal point of energy terms, they might include a transition through a T1 state as assumed in GSDIM [26] but the molecular mechanisms to proceed from there to the required very-long dark states are most probably highly complex [23]. They are expected to constitute a promising field for future photophysics of organic molecules. The visualization of localization microscopy data is based on the rendering of all detected single molecule positions in one image, such that the mean effective optical resolution of this image is determined by the localization accuracy and the density of detected molecules. This is commonly done by blurring the position of each molecule with a Gaussian function with a standard deviation corresponding to the individual localization accuracy. However, there are other rendering methods for an improved visualization of localization microscopy data [34]. Recently, the wide scope of SPDM for the investigation of biological nanostructures has been shown. 2-D single-color imaging [20, 22, 25], multicolor imaging and 3-D spatial resolution in the 50-nm range has been realized [20, 29, 35].The single molecule information provided by SPDM can be used for statistical analyses of protein distributions
1040
Biotechnol. J. 2011, 6, 1037–1051
[29, 35, 36]. It was even possible to perform SPDM on completely label-free cells, and gather structural information of autofluorescent molecules within these cells with a resolution far below the Abbe limit [37].
2
Basic principles of SPDM
SPDM [9, 10, 14, 15, 38, 39] is based on precision position measurements of small (‘point like’) objects of appropriate spectral signatures. Figure 2 schematically shows the SPDM principle [10, 22]; the example is based on numerical calculations using scalar wave theory, assuming a numerical aperture NA = n · sin(α) = 1.4 (α = half aperture angle, n = refraction index) and different emission wavelengths λ. Imagine that the first three, point-like self-luminous points (e.g., single molecules) have the same spectral signature, and also assume a next neighbor distance of only 50 nm (Fig. 2a), i.e., four times smaller than the conventional limit of optical resolution (compare Fig. 1). From each of these molecules, an Airy disc-like diffraction pattern is formed in the image plane by the microscope optics, with a full-width-at-half-maximum (FWHM) diameter of about D = 0.51 · λ/ (n · sin(α)) · M = 200 nm · M (numerical aperture: n · sin(α) = 1.4; excitation wavelength: λ = 500 nm; magnification factor, M). With the same spectral signature all signals overlap additively in the image plane (Fig. 2b). An intensity cross-section horizontally through the center of this diffraction pattern is shown in Fig. 2c. In this case it is not possible to determine where the three molecules are precisely located and by what distance they are separated from each other. However, if a different spectral signature B, G, R is assigned to each of the closely spaced molecules (Fig. 2d), localization of the individual molecules and determination of their distances is possible. From the Airy disc center positions (XB;YB) in the image plane, the corresponding molecule localizations (xB = XB/M; yB = YB/M) are determined. It is possible to determine the center (XB; YB) of the diffraction pattern (Airy disc) in the image plane with a positioning error σXY = σ·M, which is much smaller than the diameter D ≈ 200 nm of the Airy disc. With an appropriate number of detected photons and after careful correction of all errors due to optical/chromatic aberrations and mechanical shifts, the position of the corresponding emitters can be determined with a localization accuracy down to σ
