Combining FISH with localisation microscopy: Super-resolution imaging of nuclear genome nanostructures
- First Online:
- Cite this article as:
- Weiland, Y., Lemmer, P. & Cremer, C. Chromosome Res (2011) 19: 5. doi:10.1007/s10577-010-9171-6
- 1k Downloads
The optical resolution of conventional far field fluorescence light microscopy is restricted to about 200 nm laterally and 600 nm axially and has been thought for many decades to be an insurmountable barrier for the quantitative spatial analysis of cellular and hence also nuclear constituents. Novel approaches in light microscopy have now overcome this barrier. Here, we report on a special method of localisation microscopy, spectral precision distance/position determination microscopy and its combination with fluorescence in situ hybridization to analyse the spatial distribution of specific DNA sequences in human cell nuclei at the macromolecular optical resolution level. As an example, repetitive DNA sequence DYZ2 located within the heterochromatin region on human chromosome Yq12 was labelled with clone pHY2.1. Between 300 and 700 single-probe molecules were resolved in individual chromatin domains, corresponding to a detected molecule density around 500/μm2, i.e., many times higher than resolvable by conventional fluorescence microscopy. A mean localisation accuracy of about 20 nm indicated a mean optical resolution in the 50 nm range. Beyond new perspectives for light microscopic studies of specific chromatin nanostructures, this may open a new avenue towards the general analysis of copy number of specific DNA sequences in small regions of individual interphase nuclei.
KeywordsFISH Chromatin nanostructure Optical resolution Super-resolution Localisation microscopy Spectral precision distance microscopy (SPDM)
Fluorescence in situ hybridisation
Foetal calf serum
Full width at half maximum
Point spread function
Spectral precision distance/position determination microscopy
Stimulated emission depletion
In recent years, nuclear genome structure has emerged as a central theme of cell biology (for reviews, see Lamond and Earnshaw 1998; Cremer and Cremer 2001; O’Brien et al. 2003; Misteli 2005, 2007; Rouquette et al. 2010). To study chromatin nanostructure, methods of ultrastructure analysis using ionising radiation, in particular electron microscopy, have proven to be a valuable tool for imaging beyond the conventional optical resolution of light microscopy, until recently thought to pose an absolute limit of about half a wavelength or 200 nm in the object plane and around 600 nm in the direction of the optical axis (Abbe 1873; Rayleigh 1896). In the recent past, however, a variety of laser-optical far-field microscopy techniques based on fluorescence excitation has been developed to overcome the ‘Abbe-limit’ of 200 nm.
Some well-known methods are confocal 4Pi-Laser Scanning Microscopy (Cremer and Cremer 1978; Hell et al. 1994, 2003; Hänninen et al. 1995; Egner et al. 2002; Bewersdorf et al. 2006; Baddeley et al. 2006; Lang et al. 2010), structured/patterned illumination microscopy (Heintzmann and Cremer 1999; Gustafsson 2000; Baddeley et al. 2007; Schermelleh et al. 2008), STED microscopy (Hell and Wichmann 1994; Schrader et al. 1995; Hell 2007; Schmidt et al. 2008) or localisation microscopy techniques using far-field fluorescence microscopy (Cremer et al. 1996, 2010; Bornfleth et al. 1998; van Oijen et al. 1998; Edelmann et al. 1999, Edelmann and Cremer 2000; Esa et al. 2000; Lacoste et al. 2000; Schmidt et al. 2000; Heilemann et al. 2002; Betzig et al. 2006; Hess et al. 2006; Egner et al. 2007; Reymann et al. 2008; Lemmer et al. 2008). Using these techniques, an effective optical resolution in the 10–20 nm regime has been obtained. For the first time, this enabled ‘nanoimaging’ of intracellular biostructures at macromolecular resolution using fluorescence excitation by visible light.
The basis of localisation microscopy as a far-field fluorescence microscopy ‘nanoimaging’ technique is the independent localisation of ‘point like’ objects excited to fluorescence emission by a focused laser beam or by non-focused illumination via ‘optical isolation’; i.e. the localisation is assigned by appropriate spectral features (“signatures”), allowing focused, structured or homogeneous illumination schemes (Fig. 1). Furthermore, a large variety of spectral signature modes can be used, including simple spectral absorption/emission characteristics, fluorescence life times and time-dependent phosphorescence/luminescence. This method, designated ‘spectral precision microscopy’ or spectral precision distance/position determination microscopy (SPDM) was conceived and realised in proof-of-principle experiments already in the 1990s (Cremer et al. 1996, 1999; Bornfleth et al. 1998; Rauch et al. 2000; Esa et al. 2000, 2001). Early ‘proof-of-principle’ experiments using confocal laser scanning fluorescence microscopy at room temperature to determine the positions (xyz) and mutual Euclidean distances of small adjacent sites on the same large DNA molecule (translocation chromosome t(9;22) in human cell nuclei) labelled with three different spectral signatures yielded a lateral effective resolution of about 30 nm (ca. 1/16th of the wavelength used) and ca. 50 nm axial effective resolution (Esa et al. 2000): Sites with a 3D distance of only 50 nm were still discriminated; their (xyz) positions were determined independently from each other with an error of a few tens of nanometers, including the correction for optical aberrations.
Compared with related ideas (Burns et al. 1985; Betzig 1995), in the SPDM approach, the focus of application was (1) to perform super-resolution analysis by the direct evaluation of the positions of the spectrally separated objects; (2) to adapt the method to far field fluorescence microscopy at temperatures in the 300 K range; (3) to specify spectral signatures to include all kinds of fluorescent emission parameters suitable, from absorption/emission spectra to fluorescence life times (Heilemann et al. 2002) or any other method allowing ‘optical isolation’ (Cremer et al. 2001).
We demonstrated recently that reversible photobleaching can be used for super-resolution imaging (SPDM) of cellular nanostructures labelled with conventional fluorochromes such as Alexa dyes (Reymann et al. 2008; Baddeley et al. 2009) or fluorescent proteins (Lemmer et al. 2008, 2009; Gunkel et al. 2009). This effect may be obtained by appropriate physical conditions, such as an exciting illumination intensity in the 10 kW/cm2 to several 100 kW/cm2 range; only one exciting wavelength is necessary both for activation and read out of a given molecule type. Under these conditions, a large number of images or frames is taken rapidly one after the other (typical frame rates around 20 fr/s or more, up to a total of 1,000 to 4,000 frames); in each frame, under appropriate conditions the distance between the diffraction images of individual fluorescent molecules is typically larger than the ‘Abbe limit’, i.e., larger than the FWHM of the PSF of the optical system used. Recently, a related technique (Fölling et al. 2008) has been described for several other dyes like Atto532, Citrine or PhiYFP called ground state depletion microscopy followed by individual molecule return.
In the present study, we combined SPDM with fluorescence in situ hybridization (FISH)-labelling to quantitatively analyse the spatial distribution of specifically FISH-labelled DNA sequences in small regions of human cell nuclei. Heterochromatin region Yq12 was labelled with a cloned DYZ2 fragment (pHY2.1, Cooke et al. 1982) in human fibroblast nuclei. Repetitive DNA sequence DYZ2 is a major component of the heterochromatic region on human Yq12 and comprises approximately 20% of the Y chromosomal DNA (Cooke 1976). Arranged in tandem arrays, the 2.47 kb repeat is distributed in about 2,000 copies over the entire length of Yq12 (Schmid et al. 1990).
Materials and methods
Cell culture and specimen preparation
Human diploid fibroblast cells VH7 (kindly provided by Prof. Dr. Beauchamp from DKFZ, Heidelberg) were routinely cultivated in Dulbecco’s Modified Eagle medium supplemented with 10% FCS and penicillin/streptomycin at 37°C in a humidified CO2 incubator. For specimen preparation, cells were seeded onto cover slips, grown until 70–80% confluency and subsequently fixed with 4% formaldehyde in PBS. Permeabilisation steps included treatment in 0.5% Triton-X in PBS, 20% glycerol in PBS, repeated freezing–thawing in liquid nitrogen and incubation in 0.1 N HCl, according to standard protocols for the preparation of 3D preserved nuclei (Bridger and Lichter 1999; Solovei et al. 2002a). Cover slips were stored in 50% formaldehyde in 2×SSC (pH 7) at 4°C until usage.
Clone pHY2.1 (Cooke et al. 1982), composed of the 2.1 kb HaeIII fragment inserted into plasmid pBR328 (a kind gift from Dr. G. Rappold, Institute of Human Genetics, University of Heidelberg) was labelled by standard nick translation using AlexaFluor®568 (Invitrogen, Germany) as fluorescent dye. Labelled DNA was co-precipitated with salmon sperm DNA serving as carrier DNA and resuspended in hybridisation buffer consisting of 50% deionised formamide, 10% dextran sulphate in 4×SSC.
FISH and immunocytochemistry
The setup is equipped with three laser sources providing laser light at λ = 488 nm, λ = 568 nm and λ = 647 nm (Lexel Laser, USA). Lenses L2 and L3 (Linos Photonics, Germany) expand the laser beam to a diameter of approximately 20 mm. A 50:50 beam splitter (Edmund Optics, Germany) evenly distributes the beams onto the two interferometer arms. The left optical path provides wide-field illumination: The beam is directed to the objective (O1), passing through an achromatic lens (L4) and a mirror (S2), resulting in collimated illumination of the sample, positioned between the two objectives. Currently in use are 100× oil objectives with an NA of 1.4 (Leica, Germany). Movement of the sample is provided by a 3-axis stepper motor stage and a 2-axis piezo stage (Physik Instrumente, Germany). The right optical path is utilised for localisation microscopy: In addition to an achromatic lens (L5) focusing the beam into the back focal plane of the objective (O2), a second lens (L3) provides the higher laser illumination intensities needed for this type of localisation microscopy. For the illumination of dyes, intensities were in the order of 10 kW/cm2. The emission light from the sample is collected by the right hand objective lens (O2), passes through the dichroic mirror D4 (AHF, Germany) reflecting the laser lines, and is focused by a ×1 tube lens (Leica, Germany) onto the CCD chip of a highly sensitive camera (PCO Imaging, Germany). Remaining laser light and out of band fluorescence are blocked by blocking filter BF (AHF, Germany). A typical time series encompassing 1,000 2D images (frames) was acquired with a repetition rate between 10 and 16 Hz.
In localisation microscopy, the essential concept to surpass the resolution limit of conventional light microscopy is the optical isolation of single fluorescent molecules by using appropriate spectral signatures, which may include photo-physical or photo-chemical properties of the fluorescent dye molecules (e.g., photoswitching or photoconversion). This means that at any given time, the distance between two molecules of the same spectral signature has to be larger than the conventional optical resolution (FWHM of the PSF). In conventional microscopy, on the contrary, the image of the fluorescence-labelled object emerges from the simultaneous detection of all fluorescent molecules, i.e., also from molecules with a distance smaller than the conventional resolution: in this case, due to diffraction, the object appears blurred and the resulting loss of structural information cannot be reversed. In localisation microscopy, however, a time series of only a few single molecules per image (i.e., with distances larger than the conventional resolution) is acquired. The result is a large number of single images or frames of sparsely distributed molecules with non-overlapping diffraction patterns. This allows to determine the positions of the individual fluorophores with a precision mainly depending on the detected photon numbers (Bornfleth et al. 1998; Edelmann and Cremer 2000; Thompson et al. 2002). All positions obtained from the individual frames are assigned to the same ‘localisation map’. Thus, an image of the measured object (positions of all molecules detected) can be reconstructed that is not limited by diffraction.
Segmentation of the signal within the wide-field data
Fitting of an appropriate model function to the data to determine the individual molecule positions.
For the detection and segmentation of the signals, established algorithms such as CLEAN (Högbom 1974) can be used.
The middle panel of Fig. 3 shows the effect of detected photon numbers per signal on localisation accuracy, assuming equal background levels in each case. For the PSF, the function described above was used. In (d), the photon number Nphot of the fluorescence signal of active molecules averages to 15, in (e) to 100 and in (f) to 1,500 photons. For each molecule, a variation in detected photon number, expressed as standard deviation of a normal distribution, of 30% of the mean was assumed. In (d), it is not possible to differentiate between background and signal. In (e), signals become recognisable but the precise number of molecules cannot be determined. In (f), the number and the position of the individual signals can be readily identified.
The effect of localisation accuracy on the attainable resolution is illustrated in the lower panel. For clarity, the corresponding images of the middle and lower row are simulated with consistent average photon numbers, i.e., in (g), it averages to 15, in (i) to 100 and in (h) to 1,500 photons. In (g), a localisation accuracy of 50 nm along with the detection of a few tens of photons provided no additional structural information about the object, when compared to the conventional microscope image shown in (b). By the detection of around 100 photons/molecule and a resulting localisation accuracy of 20 nm, substructures become distinguishable (h) which are certainly not observable in the conventional wide-field image. The detection of more than 1,000 photons (i) is sufficient for a localisation accuracy better than 10 nm. In combination with an adequate number of detected molecules, this allows a good reconstruction of the given nanostructure.
The area of the labelled DNA region (b, d) is extracted from the wide-field images (a, c) by an operation from the field of mathematical morphology termed morphological closing (Matheron 1975; Serra 1982). As structuring element, a circular disc in the plane was utilised. Basically, this results in the elimination of small holes and the filling of gaps in contours. In addition, a certain amount of smoothing on an object contour is effected.
Without interpolation, the area of the segmented region is calculated to 0.942 μm2 and with interpolation to 1.049 μm2. This means the area is 11.3% larger in the interpolated image. It is easily observable that the non-interpolated images (Fig.4 a, b) appear rather pixelated, whereas the images subjected to interpolation are smoother to some extent, and the edges are shown in far more detail, thus allowing the extraction of considerably more structural information. For this, however, a reliable discrimination of object and background and thus a good signal-to-noise ratio is important.
Both histograms display several maxima at various positions. The analysis of the area yielded values ranging from 0.50 to 1.65 μm2 and a mean value of 0.86 ± 0.28 μm2. The number of signals was found to range from 281 to 713 with an average of 443 ± 114. The comparison of FISH labelled area and cell cycle stage (as determined from the pKi67 pattern) indicates that all nuclei with areas smaller than 1.1 μm2 are in G1/S phase.
In Fig. 10(a), the number of detected signals was plotted against the area of the chromatin regions. The basic trend of the graph shows that the number of signals increases linearly with the area. Figure 10(b) displaying the anisotropy of the chromatin region reveals a strong linear relationship with a slope of 0.972. This suggests that on average the DNA regions do not show a shape deviating considerably from a sphere.
In contrast to Fig. 9, displaying the distribution of segmented areas for all 25 nuclei, the distributions of the standard deviations in Fig. 12 appeared to be more homogeneous, suggesting a bimodal distribution of nuclei with standard deviations smaller than 300 nm and of nuclei with standard deviations larger than 300 nm. It may be noted that the nuclei with a FISH-labelled Yq12 area larger than 1.1 μm2 (Fig. 9) are represented by standard deviations larger than 325 nm in both directions.
In a previous study, we have shown that localisation microscopy (SPDM) can be accomplished with conventional fluorochromes such as Alexa dyes or fluorescein derivatives (Reymann et al. 2008; Lemmer et al. 2009; Cremer et al. 2010). Besides their application in immunocytochemical staining methods, these dyes are widely used to specifically label chromatin structures by fluorescence in situ hybridisation (for review, see Cremer and Cremer 2001). Many of these structures, however, have a dimension too small for conventional light microscopical techniques. Examples for such structures in nuclear cell biology are small gene domains which may change their conformation due to differentiation or regulatory mechanisms. In this study, we have assessed and demonstrated the capability of the combination of fluorescence in situ hybridisation with localisation microscopy. Heterochromatin region Yq12 was labelled with a probe for DYZ2, a tandemly repeated sequence with a length of 2.47 kb, which is distributed in about 2,000 copies over the entire length of band Yq12. In addition, we described a data processing procedure that minimises the loss of structural information present in the original data and hence produces an accurate representation of the measured structure. Several structural parameters were extracted, describing heterochromatin region Yq12 in far more detail than can be gained by conventional microscopical techniques. In particular, the extent of this region was determined directly from the positions of individual hybridised molecules; the number of the molecules hybridised and detected was determined for each individual FISH-labelled region, despite a mean molecule distance much smaller than the conventional optical resolution dconv of approximately 200 nm: a density of 500 molecules/μm2 (Fig. 10a) corresponds to a mean distance of about 45 nm, i.e. about five times smaller than dconv. From the mean localisation accuracy σ of about 20 nm, an average optical resolution of about 2.35 σ corresponding to about 50 nm can be estimated. This means that so far the data do not allow drawing detailed conclusions on the substructure of the Yq12 domain. It is anticipated that this problem can be overcome by technical improvements of the SPDM microscopy technology presently in development (Pertsinidis et al. 2010). Another current limitation is the number of detected molecules: From a FISH labelled Yq12 region containing a total number of 2000 repeats, only about 300–700 signals are obtained. Some reasons for this may be the labelling efficiency of the probes, the inherent optical sectioning capacity of the technique: a ‘slice’ of about 600 nm thickness was analysed while the entire Yq12 domain had a diameter of about 1 μm; or the efficiency of detection of bound probe molecules. Despite these limitations, the method allows currently a large variety of applications where the determination of relative molecule numbers is sufficent.
This paper presents the first detailed description of the application of localisation microscopy (SPDM) for the nanostructural analysis of specific FISH-labelled chromatin domains. This approach may be combined with other SPDM techniques where fluorescent proteins were applied to study general features of nuclear nanostructure (Kaufmann et al. 2009; Gunkel et al. 2009; Bohn et al. 2010). For example, one could analyse specific domains of active/silenced genes or of gene deserts (Shopland et al. 2006) in the context of the histone distribution. Another interesting field of application is the analysis of break point regions induced e.g. by ionising radiation. In contrast to early attempts to use the SPDM methods (Esa et al. 2000, 2001) where only a very few spectral signatures were available, the present FISH-SPDM technique described here would allow positioning multiple molecules in a nuclear area as small as 200 nm diameter. Another attractive field of application is to study condensation variabilities on the macromolecular resolution level, by artificially changing the degree of compaction of the chromatin. This can be done with chemicals such as Trichostatin A, an inhibitor of histone deactelylase, inducing decondensation of heterochromatin regions (Görisch et al. 2005), or Berenil, an AT-specific DNA ligand, inhibiting the condensation of the heterochromatin block on Yq12 (Haaf et al. 1989). The quantitative data obtained this way may be used to test numerical models of chromatin nanostructure (Odenheimer et al. 2009; Bohn et al. 2010) A further appealing perspective is to count copy numbers of specific DNA sequences on a single cell nucleus basis. So far, copy numbers are mainly assessed by PCR and array-based techniques, which have the drawback that they typically show an average over a large number of cells. Changes in copy numbers attract increasing interest, as it becomes more and more apparent that these contribute to several diseases, including cancer (Dear 2009).
The FISH localisation microscopy approach presented here is particularly suited for the analysis of small gene regions in intact cell nuclei. Many of these domains are thought to change their conformation due to activation but so far incontrovertible evidence is still elusive. Furthermore, extending the SPDM approach to two or more colours, each allowing the localization of a large number of molecules, will make possible to study on the nanoscale the interaction of specific chromatin structures such as the widely discussed problem of intermingling of chromosome territories (Branco and Pombo 2006; Rouquette et al. 2010).
The major concerns in the application of FISH-based SPDM to study nuclear nanostructure are changes induced by the fixation and denaturation process (Solovei et al. 2002b). This problem might be overcome in various ways, e.g., by using COMBO or PNA FISH with non-denaturating conditions (Hausmann et al. 2003; Schmitt et al. 2010) or by using thin cryosections (Branco and Pombo 2006).
We gratefully acknowledge the support of the Deutsche Forschungsgemeinschaft (SPP1128) and of the European Union (In vivo molecular imaging Consortium, www.molimg.gr; 3D Genome project). Paul Lemmer is a fellow of the Hartmut Hoffmann-Berling International Graduate School of Molecular and Cellular Biology of the University Heidelberg and a member of the Excellence Cluster Cell Networks of the University of Heidelberg. We would like to thank our colleagues Patrick Müller, Rainer Kaufmann and Manuel Gunkel for stimulating discussions and great support.