- Top of page
Super-resolution fluorescence microscopy is ideally suited to study the complex organization of cell DNA in the 10–100 nm range. Novel methods to image directly labelled DNA, instead of DNA-associated proteins, are being developed and refined. This minireview provides an update of recent progress in super-resolution fluorescence imaging methods for DNA. These developments should allow a deeper understanding of chromatin structure and widen the scope of biological processes that may be investigated with super-resolution fluorescence microscopy.
The compaction of meters long DNA strands into the tiny volume of a cell nucleus is one of the most astonishing feats in Biology. DNA can be packed into a chromosome with a compaction factor of about 10 000-fold with the help of certain proteins (histones), which interact strongly with DNA mostly through electrostatic forces. This dense mix of DNA and proteins, which is called chromatin, has a very complex structure with organization levels that span several orders of magnitude. The underlying details of chromatin have been elusive to the conventional optical microscope due to the limitations in spatial resolution, and in vitro chromatin imaging has mostly been performed until recently with electron microscopy, with the restrictions associated with challenging and invasive sample preparation (Flors & Earnshaw, 2011).
Recently developed super-resolution fluorescence microscopy techniques combine noninvasiveness, sensitivity and specificity of fluorescence methods, with a subdiffraction spatial resolution of 10–100 nm (Hell, 2009). These techniques therefore address the resolution gap between conventional optical microscopy and electron microscopy, which has so far limited our understanding of chromatin structure in that scale. One of the super-resolution fluorescence methods that is becoming broadly adopted by many laboratories is single-molecule switching and localization microscopy (also called stochastic switching and readout). In this method, the emission of individual fluorophores that are stochastically photoswitching or photoblinking is separated in time, and their position precisely localized. Thousands of consecutive cycles of stochastic switching, detection and localization of single molecules on a wide-field microscope are performed, allowing the construction of a super-resolution image that shows localization density of the fluorophores with a typical spatial resolution of tens of nanometres (Hell, 2009).
Since super-resolution fluorescence microscopy heavily relies on the properties of the fluorophores, choosing the most appropriate labelling strategy for a certain sample or biological question is crucial. If one wants to image chromatin, several options can be considered, all with their advantages and drawbacks. Since the methods to introduce photoswitchable fluorophores in proteins are quite mature, labelling DNA-associated proteins like histones is a possibility. For example, histones can be tagged by genetic fusions to photoswitchable fluorescent proteins, by immunolabelling, or by chemical tagging strategies. On the other hand, strategies for direct labelling of DNA are recently starting to be explored and refined. Some benefits of directly labelling DNA are the possibility to achieve a higher labelling density (using an appropriate stain that does not affect DNA native structure) and simple staining protocols. In addition, direct labelling of DNA may potentially result in a better spatial resolution due to the small size of DNA-associated dyes that are incorporated into the DNA double helix itself.
Previous reviews have focused on early work on super-resolution switching and localization microscopy of DNA labelled with sequence unspecific stains (Flors, 2011), as well as on the new biological insight obtained from different super-resolution methods to characterize chromatin substructures in the 10–100 nm range (Flors & Earnshaw, 2011). This minireview provides an update of the most recent work on super-resolution imaging with direct DNA labelling. First, super-resolution imaging of DNA nanostructures will be discussed in the context of the development of new and better standards for the calibration of super-resolution fluorescence microscopes. In addition, the most recent work focused on chromatin imaging in fixed and live cells will be reviewed. Progress in the related fields of DNA mapping and barcoding with subdiffraction resolution will not be discussed in this minireview, since imaging is not the main purpose of these techniques.
When a new super-resolution method is demonstrated, an image of a cellular feature of subdiffraction dimensions such as microtubule filaments is typically shown. However, cellular features have unknown and uncontrollable structure and their labelling is not reproducible. On the other hand, the ability to program DNA self-assembly and to precisely label defined positions with fluorescent molecules provides structural control of labelling distances and stoichiometry. DNA nanostructures like DNA origami provide ideal calibration targets (Schmied et al., 2012). DNA origami is the folding of DNA to create arbitrary two- and three-dimensional shapes at the nanoscale (Rothemund, 2006). The process involves the self-assembly of a long single strand of viral DNA aided by multiple smaller ‘staple’ strands. First proposed in 2009 as calibration targets for switching and localization super-resolution microscopes, there are currently a number of different standards that allow testing different aspects of the performance of a super-resolution fluorescence microscope (Schmied et al., 2012). The versatility of DNA origami assembly and labelling has allowed devising different standards specifically tailored to different super-resolution microscopy techniques, as well as to different spatial resolution ranges. DNA origami can provide a platform to position fluorophores at controlled distances, and can provide a calibration range from molecular scale (about 6 nm) up to the diffraction limit and above. Moreover, the performance of the super-resolution microscope can be tested in different spectral ranges. DNA origami nanopillars also allow testing the axial resolution in three-dimensional switching and localization-based microscopy (Schmied et al., 2013). Fluorescently labelled DNA origami standards have also been designed for stimulated emission depletion (STED) microscopy, another super-resolution technique (Schmied et al., 2012).
The DNA origami standards described above are usually labelled at certain controlled positions in the DNA sequence, however, to image the overall presence of DNA in a (biological) sample, it is convenient to use a dye that can bind to any position in DNA with little or no sequence specificity. Several strategies to produce fluorescence on-off transitions in DNA labels have been demonstrated. One strategy is based on the induction of dye blinking by transient dark state formation via a photochemical process (which would include, for example, direct stochastic optical reconstruction microscopy, dSTORM, and related techniques; Fig. 1a). Another strategy relies on the fluorogenic properties of DNA-binding dyes (Fig. 1b). As discussed below, the combination of both these strategies can also be useful. Using the first strategy, the photochemical induction of dye blinking, Zessin et al. employ a click chemistry reaction for DNA labelling to introduce a bright photoswitchable dye into DNA (Zessin et al., 2012). Specifically, 5-ethynyl-2’-deoxyuridine (EdU) is incorporated into cell DNA in the place of thymidine, and after cell fixation it can be labelled with an azide derivative of Alexa Fluor 647 via click chemistry. For dSTORM imaging (Fig. 2), blinking of Alexa Fluor 647 was induced with an appropriate switching buffer (enzymatic oxygen scavenging system and 100 mM of a reducing thiol compound). This approach allows high-density labelling of cell DNA (the authors report 106 distinct fluorophores per nuclear section) required for super-resolution imaging, which cannot be achieved by other methods such as immunofluorescence of histone proteins (Zessin et al., 2012).
Figure 1. Strategies to induce on-off transitions in fluorophores that directly label DNA. (a) Photochemical formation of a meta-stable dark state D; (b) Fluorogenic strategy: binding of the dye to DNA greatly enhances its brightness compared to the free dye in solution, which allows its precise localization (image adapted with permission from Schoen et al., 2011).
Download figure to PowerPoint
Figure 2. dSTORM super-resolution image of the nuclear chromatin of a HeLa cell. The cells were grown for 24 h in the presence of EdU, fixed and labelled with Alexa Fluor 647 via click chemistry (scale bar in main image is 2 m). Inset (a): diffraction-limited image of the whole nucleus. Inset (b): magnification of the boxed region in the main dSTORM image, scale bar 500 nm. Adapted with permission from Zessin et al., 2012.
Download figure to PowerPoint
Whereas the above approach covalently incorporates an organic fluorophore into DNA, another strategy is to use organic fluorophores that bind to DNA noncovalently. This method typically simplifies sample preparation, it can be less perturbative to DNA structure (depending on the dye) and can be therefore potentially compatible with live-cell imaging. The cyanine dye YOYO-1, when intercalated into DNA, has been shown to blink in a similar switching buffer as that used for Alexa Fluor 647 and provides a spatial resolution below 40 nm when imaging stretched λ-DNA (Flors, 2011). The minor groove-binding cyanine dye SYTO-13, which is cell-permeant and less cytotoxic than other DNA stains like YOYO-1, has been used in a similar way to image fixed mitotic DT40 chicken cells (Flors, 2011). PicoGreen, another minor groove-binding cyanine dye has also been shown to blink in an appropriate timescale for dSTORM imaging (Benke & Manley, 2012). In comparison to YOYO-1 and SYTO-13, PicoGreen selectively stains dsDNA and not RNA. The imaging buffer used in this case also requires oxygen removal, but the reducing thiol compound was replaced by 1 mM ascorbic acid, which is compatible with live-cell imaging. With this buffer, the recovery of PicoGreen fluorescence after the formation of longer-lived dark states was efficient, allowing time-lapse dSTORM imaging on a living cell with 10 min intervals with a rather constant number of localizations in each dSTORM image (Fig. 3). The authors use mitochondrial DNA to estimate a spatial resolution of 70 nm for their method (Benke & Manley, 2012). Importantly, these approaches that use blinking of DNA-binding dyes like YOYO-1, SYTO and PicoGreen also greatly benefit from fluorogenic properties of these dyes, i.e. large fluorescence enhancements (100- to 1000-fold) upon binding to DNA, which greatly minimizes the fluorescence background.
Figure 3. Live-cell dSTORM imaging of a U2OS cell with its DNA labelled with PicoGreen (scale bar 2.5 m). The inset shows three consecutive dSTORM images of the boxed area with 10 min pause between them, showing local rearrangements of chromatin (scale bar 200 nm). Adapted from Benke & Manley, 2012. Copyright © 2012 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim.
Download figure to PowerPoint
The need to induce fluorophore blinking ‘photochemically’ as in Figure 1(a) can be overcome by using the fluorogenic properties of the dyes described above. Dyes that do not associate as strongly to DNA as YOYO-1, e.g. the monomeric cyanine intercalating dye YO-PRO-1, can dynamically bind and unbind to DNA giving rise to apparent on-off blinking due to the binding-induced fluorescence enhancement (Flors, 2011). A similar approach, termed binding-activated localization microscopy (BALM), has been demonstrated with other dyes (Schoen et al., 2011). The authors find that the use of ‘reducing-oxidizing system’ (ROXS) as an imaging buffer enhances the dye brightness and, in addition, promotes the detachment of YOYO-1 and PicoGreen from DNA, which allows the dynamic binding and unbinding that gives rise to fluorescent on-off transitions. Although not discussed in the original paper, ROXS contains methyl viologen, which can also strongly bind to DNA and therefore compete with YOYO-1 for binding sites. This may partly explain why their best results were obtained in ROXS compared to other tested buffers that do not contain methyl viologen. λ-DNA was imaged with BALM with a spatial resolution of 14 nm and high localization density (about one localization per 2 nm) using YOYO-1. For PicoGreen, the spatial resolution was 27 nm due to its lower brightness. PicoGreen was also used to image the nucleoid organization in fixed Escherichia coli bacteria with BALM (Schoen et al., 2011). Another approach that contains some of the concepts described above relies on the combination of fluorescence activation by DNA binding, and single-molecule high-resolution imaging with photobleaching (Simonson et al., 2011). In this work, the dyes YOYO-1, POPO-3 and SYTO-16 were used and it was found that the addition of bovine serum albumin slowed the diffusion rate of the dyes, which allowed the binding events to be more distributed in time. This strategy was used to image chromosomal DNA in fixed HeLa and HEK 293.
So what have we learned so far from these new methods? The labelling strategies described above have been applied to super-resolution fluorescence imaging of DNA in fixed and live cells and some of this work has provided new insight into chromatin structure. Super-resolution fluorescence images have revealed a rather heterogeneous distribution of chromatin, i.e. the presence of high- and low-density regions, in the nucleus of eukaryotic cells and in the bacterial nucleoid, which were previously inaccessible by conventional light microscopy (Figs. 2 and 3; Benke & Manley, 2012; Zessin et al., 2012) (Schoen et al., 2011). In eukaryotic cells, it has been suggested that this heterogeneity probably represents the borders between hetero- and euchromatin (Zessin et al., 2012). More sophisticated labelling techniques in combination with dSTORM will allow addressing some problems inherent to chromatin, such as dense packing. For example, pulse labelling of nascent DNA with EdU and further labelling with Alexa Fluor 647 has already been instrumental to image individual extended chromatin fibres and quantify their length, as well as the formation of folded chromatin clusters in HeLa nuclei (Zessin et al., 2012).
The methods described here show great potential but are still in early stages of development. New and better strategies to fluorescently label DNA directly with minimal disruption to the structure of chromosomes will be of great interest. In addition, it will be useful to combine direct DNA labelling and histone tagging in multicolour experiments. Sequence-specific labelling of DNA will also be beneficial to many applications. Fluorescence in situ hybridization has already been used in switching and localization-based super-resolution imaging in a proof-of-principle study (Weiland et al., 2011). A repetitive DNA sequence located within the heterochromatin region on human chromosome Yq12 was fluorescently labelled and a detected molecule density around 500 m–2 was achieved, with a spatial resolution in the 50 nm range. From the spatial position information of the single fluorescent molecules, clusters were identified and several structural parameters were extracted to obtain a description of the shape of Yq12 (Weiland et al., 2011).
Advances in labelling geared towards the use of STED microscopy, the most representative of the ‘targeted switching and readout’ super-resolution techniques (Hell, 2009), are also of great interest. STED imaging of λ-DNA labelled with YOYO-1 has been demonstrated with pulsed depletion beams of 568 and 647 nm, achieving a spatial resolution of 42 and 62 nm, respectively (Persson et al., 2011). The extension of this initial STED study to chromosomal DNA will certainly constitute an important addition to the toolbox for super-resolution fluorescence microscopy of DNA.
In conclusion, super-resolution fluorescence microscopy is uniquely placed to expand our understanding of the highly condensed DNA structures inside of the nucleus. Together, all these developments will allow a deeper understanding of chromatin folding models, and will widen the scope of biological processes that may be investigated such as DNA transcription, replication and repair.