• artefacts;
  • atomic force microscopy;
  • correlative microscopy;
  • DNA imaging;
  • single-molecule microscopy

The ability to interrogate phenomena at the nanoscopic level has led to an increasingly refined and quantitative understanding of structures and mechanisms in cells and materials. Recent advances in tools such as fluorescence and scanning probe microscopy have greatly impacted our ability to visualize the structure and molecular components of a wide range of systems. As any technique, they have their individual strengths and limitations. Scanning probe techniques such as atomic force microscopy (AFM) can provide very high-resolution topographical images, but only near the sample surface, and usually with no chemical or functional information. Fluorescence microscopy, on the other hand, is the method of choice in the biological sciences due to its ability to access deep into a sample and the possibility to use specific molecular labels, but it has a limited spatial resolution of about 250 nm due to light diffraction. The combination of the complementary strengths of these two techniques leads to a more comprehensive characterization and understanding of a (biological) system.1, 2 Although the concept of correlative microscopy is not new, recent advances in the individual microscopy techniques, in particular fluorescence microscopy, promises an increasingly important role of this multi-dimensional imaging tool.1, 3, 4

In the first instance, an important strength of correlative microscopy is its use as a validation tool for newly developed microscopy techniques. In particular, it has been crucial in the establishment of super-resolution fluorescence microscopy, which encompasses a new group of techniques capable of imaging with a spatial resolution well below that of standard optical techniques.5 Correlative electron and fluorescence microscopy helped to establish photoactivated localization microscopy (PALM) in 2006,6 and later to provide a subcellular context in the localization of protein clusters with stimulated emission depletion microscopy (STED).7 As super-resolution fluorescence microscopy dramatically spreads across biological, biomedical and materials research, there is an increasing need to identify and exclude artefacts arising from poor labelling, dye photobleaching or other phenomena. Answering to this need, STED and AFM have been recently combined.8 Super-resolution imaging based on single-molecule localization (PALM-like methods) and AFM of protein aggregates have also been achieved on the same area of a sample, however using separate microscopes and aligning AFM and fluorescence images with the help of fiduciary markers.9

In this Communication we present a correlative microscopy tool that combines in situ super-resolution fluorescence microscopy based on single-molecule localization and AFM. We discuss the technical aspects of the correlative microscope, including image alignment and sample preparation requirements. We show how this novel tool is able to reveal typical artefacts in super-resolution imaging related to labelling and image reconstruction. The latter point is particularly relevant to localization-based super-resolution methods.

The correlative microscope is an adaptation of a commercially available platform that integrates AFM on a standard inverted optical microscope. To allow efficient single-molecule fluorescence detection, lamp excitation in the commercial setup was replaced by laser excitation, and a very sensitive electron multiplying charge coupled device (EMCCD) camera was added (see Experimental Section for a full description). For alignment of fluorescence and AFM images, the system integrates dedicated software (DirectOverlayTM, JPK Instruments), which compensates potential distortions of the optical image due to aberrations.8 The image alignment process involves scanning the AFM cantilever on a defined area (typically 30×30 μm2) and optically mapping the actual position of the cantilever during the scanning process (Figure 1). We use special AFM cantilevers that are specifically designed for the combination with optical microscopy and that allow focusing the optical image on the AFM tip rather than on the cantilever (Top Visual, NT-MDT). Alternatively, fiduciary markers like quantum dots or small fluorescent beads, which give contrast in fluorescence and topography, may be used for image registration.1013

thumbnail image

Figure 1. Alignment of AFM and optical images is achieved by optically imaging the AFM tip at several scan positions.

Download figure to PowerPoint

The next aspect to consider is which substrate to use for sample preparation. The substrate should have good optical properties to allow efficient single fluorophore detection and precise localization, as well as smoothness and flatness for optimal AFM imaging. Mica, which is an appropriate substrate for AFM, has poor optical properties, namely light absorption and birefringence, causing loss of fluorescence photons and distortions of the fluorescence signals. However, the negative impact of its optical properties can be minimized for very thin layers of mica.12 We obtained best results by using a very thin layer of mica glued on a glass coverslip (see Experimental Section).10, 13, 14 We also tried to perform AFM on glass, which should be flat enough for imaging large objects,8 but it was very challenging for small objects.

We tested the correlative setup with stretched λ-DNA labelled with the intercalating cyanine dye YOYO-1. In the presence of a suitable “switching buffer” that incorporates an enzymatic oxygen scavenging system and a reducing thiol compound (see Experimental section), YOYO-1 can blink in the appropriate timescale for localization-based super-resolution imaging.15 We first quantified the effect of the additional layer of mica and glue on the ability to localize single molecules of YOYO-1, and we obtained an experimental localization error of 21 nm, which is only slightly worse than for bare glass (18 nm). Correlative imaging of λ-DNA was then performed. Panel a in Figure 2 shows an AFM image acquired in air. After AFM imaging, the AFM head was carefully removed, the switching buffer was added (as the presence of enzymes in high concentrations precludes AFM imaging in the switching buffer) and super-resolution imaging was performed. It is worth noting that the need for anaerobic conditions (very common in super-resolution imaging with organic dyes) requires special attention in these correlative experiments. We verified that even though the sample holder used was open to air to simplify the experiment, the oxygen scavenging system was able to keep the dissolved oxygen concentration low enough to achieve reasonably good blinking and low photobleaching. Panels b and c show the super-resolution reconstructed image and the standard fluorescence image, respectively, of the same area as panel a. From the cross-sections of the super-resolution reconstructed image at different areas of the sample, we estimate a spatial resolution of about 40–50 nm, only slightly worse than previously reported values,15, 16 and features that cannot be resolved in the standard fluorescence image (panel c, yellow arrow) are resolved in the super-resolution image (panel b). We note that the use of non-enzymatic oxygen scavenging systems17 in combination with a closed AFM chamber may provide a useful alternative when super-resolution demands anaerobic conditions.

thumbnail image

Figure 2. In situ correlative AFM and super-resolution fluorescence imaging. a) AFM, b) super-resolution and c) standard fluorescence image of stretched λ-DNA labelled with YOYO-1. The green arrow marks a DNA section that is visible in AFM but not visible in fluorescence. The yellow arrow shows two DNA fragments that are close to each other and that can be resolved in the AFM and super-resolution image but not in the standard fluorescence image. Scale bar is 1 μm.

Download figure to PowerPoint

Detailed inspection of Figure 2 reveals that although most of the λ-DNA seen in the AFM image appears also in the super-resolution image, there are some patchy sections due to, for example, incomplete labelling, rapid photobleaching of YOYO-1 in that region, poor signal-to-noise ratio leading to exclusion during image analysis and so forth. This highlights the power of the correlative setup to reveal labelling and image reconstruction artefacts. There is also a DNA strand that does not appear in the fluorescence images (green arrow), which may have detached from the surface upon addition of the switching buffer after AFM imaging.

The correlative capabilities of the novel microscope are indeed ideally suited to compare different super-resolution image reconstruction methods and different analysis parameters in a robust way. Figure 3 shows how having an AFM image as a reference can be extremely helpful to optimize analysis parameters such as threshold sensitivity or point spread function (PSF) width. For example, the two triangles within the green circle in the AFM image (panel a) can be recognized in panels b and c (which only differ in PSF width), although the images are rather sparse (especially in panel b). Lowering the detection threshold of the image reconstruction software to increase localization density (panel d) blurs the smaller triangle. On the other hand, the lower threshold in panel d does make visible some structures that do not appear in panel b, for example that marked by the blue arrow. The AFM image provides verification that it is indeed a DNA fragment and not background due to, for example, unspecific attachment of YOYO-1 on the surface. Therefore, correlative imaging helps finding the best balance between spatial resolution and localization density.

thumbnail image

Figure 3. Correlative microscopy is a useful tool to compare different parameters for super-resolution image reconstruction of λ-DNA labelled with YOYO-1. a) AFM image. Super-resolution reconstructed image18 with b) detection sensitivity 25 and PSF fixed to 2.5 pixels; c) detection sensitivity 25 and PSF 3.5 pixels; d) detection sensitivity 10 and PSF 2.5 pixels. Scale bar is 1 μm.

Download figure to PowerPoint

In conclusion, we have presented a novel correlative microscopy tool that allows testing the veracity of the picture that localization-based super-resolution methods are painting. Correlative microscopy can help to prevent potential pitfalls in image interpretation, and to provide important opportunities to improve labelling and analysis methods in super-resolution microscopy. In the context of fluorescence labelling, a potentially useful application of correlative microscopy is to identify undesired morphology changes upon labelling, as has been shown recently for amyloid fibrils in separate super-resolution and AFM experiments.19 Besides being a validation tool, this novel setup has the potential to answer new questions about biological systems. Indeed, correlative AFM and fluorescence microscopy has helped to gain insight into a variety of biological processes (reviewed in Ref. [1]), with particular emphasis on the characterization of nanodomains in cell membranes and supported lipid bilayers.20, 21 Another field with potential application is the study of chromatin higher order structures, where correlative imaging will be ideal to corroborate the models of chromosome substructures proposed on the basis of separate AFM22 and super-resolution imaging studies.23 In addition to high resolution topography, AFM can provide information about elasticity,24, 25 interactions between (bio)molecules26 or friction,27 which may be a very valuable complement to super-resolution fluorescence images.1 1

Experimental Section

  1. Top of page
  2. Experimental Section
  3. Acknowledgements

Samples were prepared as previously described,15 with minor modifications. Briefly, λ-DNA (New England Biolabs) labeled with YOYO-1 (Life Technologies) at 1:17 dye/bp was spin-coated on a polylysine-coated substrate to achieve stretching. The substrate was prepared by detaching a thin layer of mica with tape and subsequently gluing it (still attached to the tape) to a glass coverslip with UV-curable adhesive (NOA63, Norland). Finally the tape was peeled off obtaining an even thinner layer of mica.

The correlative microscope is an adaptation of a commercially available platform that integrates AFM (Nanowizard II, JPK Instruments) on a standard inverted optical microscope (Nikon Eclipse Ti), and is placed on an active vibration isolation optical table (Thorlabs). Lamp excitation in the commercial setup was replaced by laser excitation (488 nm, 0.75 kW cm−2 at the sample, Luxx, Omicron). Wide-field illumination was achieved by focusing the expanded and collimated laser beam onto the back focal plane of the objective (TIRF, 60x, 1.49 NA, oil immersion, Nikon). Emission was collected by the same objective and imaged by a EMCCD camera (iXon Ultra 897, Andor Technology) after passing through a dichroic mirror (z488rdc, Chroma Technology) and additional spectral filters (HQ500 LP, HQ530/50, Chroma Technology). Additional lenses resulted in a final magnification of 219x, equivalent to a pixel size of about 73 nm. For super-resolution acquisition, integration time per frame was 100 ms, and the total number of frames collected was typically 3000–4000.

AFM imaging was performed in tapping mode with Top Visual cantilevers (force constant ∼50 N m−1, NT-MDT) under dry conditions. During AFM scanning, the objective was slightly retracted to reduce coupled vibrations.

Super-resolution imaging was performed in “switching buffer”: phosphate-buffered saline (pH 7.4, Sigma) with an oxygen scavenger [0.5 mg mL−1 glucose oxidase (Sigma), 40 µg mL−1 catalase (Sigma) and 10 % w/v glucose (Fischer Scientific)] and 50 mM β-mercaptoethylamine (Fluka).15 The chamber was open to air, but the scavenging system could achieve a low enough oxygen concentration for the observation of adequate fluorophore blinking.

Super-resolution images were reconstructed using the Localizer software.18 Localization errors are given as median values. The detection sensitivity values provided in Figure 3 refer to “GRLT sensitivity” in the Localizer software. AFM images were treated with the software JPK Data Processing.


  1. Top of page
  2. Experimental Section
  3. Acknowledgements

This work has been financed by the Spanish Ministerio de Economía y Competitividad (RyC2011–07637, MAT2012–34487, PTA2011–6702-I) and the European Commission Marie Curie Actions (FP7-PEOPLE-2011-CIG n° 303620). We thank Dr. Peter Dedecker (Katholieke Universiteit Leuven, Belgium) for the Localizer software.

  • 1

    Note: as this Communication was under review, we became aware of a related paper on correlative AFM and stochastic optical reconstruction microscopy for characterization of the cell cytoskeleton.28