Download the PDF version

 

1. Introduction

Fluorescence microscopy has revolutionized the study of biological samples. Ever since the
invention of fluorescence microscopy towards the beginning of the 20th century, significant
technological advances have enabled elucidation of biological phenomenon at cellular, subcellular
and even at molecular levels. However, the latest incarnation of the modern fluorescence microscope
has led to a paradigm shift. This wave is about breaking the diffractionlimit first proposed in 1873 by Ernst Abbe.
The implications of this development are profound.

This new technology, called super-resolution microscopy, allows for the visualization of cellular
samples with a resolution similar to that of an electron microscope, yet it retains the advantages
of an optical fluorescence microscope. This means it is possible to uniquely visualize desired
molecular species in a cellular environment, even in three dimensions and now in live cells – all
at a scale comparable to the spatial dimensions of the molecules under investigation. This
article provides an overview of some of the key recent developments in super-resolution
microscopy.
2. The problem

So what is wrong with a conventional fluorescence microscope? To understand the answer
to this question, consider a Green Fluorescence Protein (GFP) molecule that is about 3 nm in
diameter and about 4 nm long. When a single GFP molecule is imaged using a 100X objective
the size of the image should be 0.3 to 0.4 microns (100 times the size of the object). Yet the
smallest spot that can be seen at the camera is about 25 microns. Which corresponds to an
object size of about 250 nm. Thus there is a big disconnect between reality (size of the actual
fluorophore) and its perception (image). The image suggests a much larger object size than is
actually present.

Why is there a disconnect? The answer lies in the fundamental limitations imposed by the
optics used to image such molecules. Conventional fluorescence microscopy utilizes a lens to
focus a beam of light onto a spot. For example a lens is used to focus the emission signal onto
a CCD detector. Or consider an illumination beam that is focused to a small spot on the sample,
such as in a confocal scanning microscope. Performance of all such instruments is dictated by
far-field optics (distances >>λ , the wavelength of light). In the far-field regime diffraction plays
a dominant role in image formation and in fact limits the smallest spot size that can be obtained
at the focal point of a lens (Fig. 1). Because of diffraction, a parallel beam of light is focused by
a lens into a three-dimensional region near the focal point. This intensity distribution near the
focal point is referred to as the point-spread-function (PSF) of a microscope (see Fig. 1) and
forms the basis of resolution of a microscope. The full width at half maxima (FWHM) of the PSF
in the lateral (x-y) plane and along the optical (z) axis are given by

respectively, where λ is the wavelength of light, n is the refractive index of the medium in which
light propagates, and α is one half of the angular aperture of the objective lens.

Figure 1: Principle of image formation in a conventional microscope. This schematic depicts the intensity
distribution from a very small object (i.e. smaller than the wavelength of light). Note that due to diffraction
the actual image (spread over a 3D region, violet) is much larger as well as elongated compared to the
expected magnified image (green). The 3D intensity distribution of the actual image is called the Point
Spread Function (PSF) of a microscope. A camera placed in the focal plane of the lens system (which
includes the objective and the tube lens) captures a cross section of this 3D intensity distribution. The
cross section profiles of this intensity distribution along x and z axes are also shown. Also note that due to
circular symmetry of the PSF Δx = Δy . α is one half of the angular aperture of the objective lens.
Different colors of the expected and actual images are used for the purpose of illustration. Schematic is
not to scale.

Due to the non-zero value of the FWHM of the PSF, when two objects are located very
close to each other their images cannot be distinguished. Resolution of a microscope is the
smallest distance between two objects that can be discerned by the microscope. Ernst Abbe
proposed that the lateral resolution (x-y plane) of an optical microscope is given by,

which is based upon the observation of a PSF (see Fig. 2). An implication of limited resolution is
that fine structures cannot be discerned.

For example, consider microtubules which serve as structural scaffolds within cells and are
about 25 nm in diameter. A 250 nm diameter (diffraction-limited) image of a microtubule
revealed by conventional microscopy may actually represent a bundle of several microtubules
that cannot be distinguished from one another. In this case enhanced resolution would provide
additional information about the cellular architecture. Super-resolution also enables the study of
membrane heterogeneity and dynamics of protein assembly – studies that benefit from
observing single molecules.

Figure 2: Resolution of an imaging system is based on the observation of its PSF. The smallest distance
that can be discerned is given by d (see text). Since the PSF is narrower along the lateral direction than
along the axial direction, the lateral resolution is higher than axial resolution. See Fig. 1 for details.
3. Enhancing the resolution

Super-resolution microscopy is about enhancing the diffraction-limited resolution of a
microscope.

Near Field Scanning Optical Microscopy (NSOM, Table 1) is a near-field optics based superresolution
microscopy technique. In this technique the surface of a sample is illuminated using a
very fine tip (diameter < wavelength of light) and this tip is placed very close (<< wavelength of
light, hence the name “near-field”) to the surface such that it collects only the evanescent waves
from the sample and therefore avoids the diffraction related issues of far-field optical
microscopy. Using this technique a spot size of only a few nanometers can be resolved.
However, this technique is limited to the study of surfaces and additionally it is very slow owing
to low signal throughput.

Initial attempts at enhancing the resolution of a microscope, utilizing far-field optics, involved
designing objective lenses of higher numerical aperture, NA, where NAn sin α. The FWHM
of the PSF is reduced (Fig. 1) along both the lateral and the axial directions, thereby enhancing
both lateral and z resolutions, respectively. Examples of super-resolution microscopy
techniques that utilize this approach include “4Pi” and “I5M” microscopy (Table 1). Both of these
techniques enhance the axial resolution by placing the sample at the focal plane of two
opposing lenses. Therefore, the effective numerical aperture of the system is increased (since
α, is increased), thereby improving the resolution (Fig. 3B). However these super-resolution
microscopy techniques were able to enhance the resolution only along the axial direction
compared to the conventional techniques. The lateral resolution remains unchanged.

Changing the PSF is not the only way to enhance the resolution. Fluorescence microscopy
has another dimension, called color. Imagine two PSFs of different colors that are located very
close to each other (Fig. 3C). Since Abbe’s resolution criterion applies to a given color the
resolution of a fluorescence microscope should not be limited when two closely spaced point
sources emit light in different colors. Therefore in theory if all the molecules of a sample could
be labeled with a different color then a resolution better than diffraction-limited imaging could be
achieved [1]. However, there is a practical limitation of the number of colors that can be used in
a given experiment and, more importantly, this method does not allow distinction among the
same molecular species, all of which are typically labeled with the same colored fluorophore.

Figure 3: Strategies for enhancing the resolution of an imaging system. (A) Resolution of a conventional
microscope. (B) Enhancing the resolution by changing the PSF of the microscope. (C) Abbe’s resolution
criterion is not limited by color. (D) Time distinction (see text) is a basis of all the current super-resolution
techniques. t1 & t2 are different instances of time.

Abbe’s resolution criterion also does not impose a limit on the resolution when two closely
spaced point sources are imaged at different times. That is, imagine two PSFs that cannot be
otherwise distinguished but each is observed at a different time (Fig. 3D). This methodology of
“sequential” imaging at different times forms the basis of the most recent and successful superresolution
microscopy techniques. A table summarizing established and emerging superresolution
microscopy techniques is included at the end of this article for reference (Table 1).
4. Super-resolution microscopy techniques

Depending upon how time distinction is achieved, super-resolution microscopy techniques
can be broadly categorized into two main approaches [1]. In the first approach, called “targeted
switching and readout”, the illumination volume in a fluorescent sample is confined to a small
region, which is much smaller than the diffraction-limited spot size. Stimulated emission
depletion (STED) microscopy is based on this approach (see Fig.5 and Table 1). Knowledge
about the illumination spot size is used to generate a super-resolution image. The second
approach termed “stochastic switching and readout” utilizes stochastic variation associated with
switching fluorophore molecules on or off, under carefully designed experimental conditions,
such that a sequence of images acquired at different instances can be used to generate a
super-resolution image. In both of these approaches, fluorophores are turned on and off at
different time instances, and the images acquired at different instances are combined together
to generate a composite image. However, in targeted switching the location of fluorophore
molecules that are turned on and off is not stochastic (see Fig. 5).

The mechanism by which time distinction is achieved can be explained by the electronic
transition states of a fluorophore (Fig. 4). In conventional fluorescence microscopy, a
fluorophore absorbs energy from the excitation light and almost instantaneously releases the
emission signal. In order to achieve time distinction (which ultimately provides spatial
discrimination), nonlinear relationships between excitation and emission of a fluorophore are
exploited. By virtue of these nonlinear relationships, specific fluorophore molecules can be
switched on or off.

Figure 4: Electronic transition states of a fluorophore. Upon absorption of energy, a fluorophore molecule
is excited and almost instantaneously (typically within nanoseconds) releases energy in the form of
photons when it relaxes to the ground state. Due to inter-system crossing or (external) photo-physics, an
electron in the excited state can also move to the triplet state before returning to the ground state. This
nonlinear process can delay (by microseconds to milliseconds) or even quench the fluorescence
emission. Fluorophores can also behave nonlinearly by virtue of (external) photo-chemistry; for example,
a fluorophore molecule can be rendered excitable (phtoactivation) or non-excitable by altering the
molecule into cis and trans isomerization states.

The first super-resolution microscopy technique that utilized this nonlinear relationship is
STED microscopy. In conventional point scanning confocal systems, all the fluorophore
molecules within a diffraction-limited excitation volume are excited and emit the fluorescence
signal simultaneously (Fig. 5). By moving the scanning beam over the sample, a composite
image is constructed. The basic implementation of STED microscopy is similar in terms of pointby-
point scanning in order to generate a single image. However, in STED microscopy the
emission signal is generated from a much smaller volume than that of a conventional confocal
scanner. In STED the laser scanning illumination (excitation pulse) spot is overlapped by
another beam called the STED beam which has an annular (“doughnut”) shape and is of longer
wavelength than the scanning beam (and also matches the emission wavelength of the
fluorophore). An intense pulse from the STED beam depletes the emission of fluorophores in
the annulus region by stimulated emission – in other words, it confines the molecules to the
ground state. Immediately after the STED pulse only fluorophores in the central region (the
doughnut “hole”) are still in the excited state and are thus able to emit fluorescence. Therefore
the effective emission spot size is reduced to well below the diffraction limit (10s of nanometers
compared to a diffraction limited size of 100s of nm). This means that the PSF for a STED
microscope has a much narrower FWHM along the lateral axis (based upon the a priori
knowledge of the illumination spot size and its location), which improves the lateral resolution of
the microscope. The axial resolution remains unchanged however.

Figure 5: An example of targeted switching and readout. In conventional confocal scanning, all the
fluorophore molecules (marked ‘x’) within the diffraction-limited illumination spot simultaneously emit the
signal. By exploiting a nonlinear relationship between excitation and emission of a fluorophore, such as in
STED microscopy (right panel), a doughnut shaped STED beam (red) confines the molecules to the
ground state (which do not fluoresce) and thereby creates a sub-diffraction-limited spot size from which
fluorophores emit (green).

Saturated Structured Illumination Microscopy (SSIM) is another example of targeted
switching based super-resolution microscopy. However, as opposed to STED this is not a pointby-
point scanning method. Instead a periodic illumination pattern is generated on the sample
plane by imaging a phase mask (for example a grating with a finely spaced linear pattern)
placed in the excitation light path. The phase mask is rotated to scan the entire sample. A
sequence of images is acquired in widefield detection mode, each image of the sequence
corresponding to a given position of the phase mask. The super-resolution effect in SSIM is
achieved by illuminating the fluorophores with patterned excitation light of saturating intensity.
This pattern deliberately establishes very narrow regions which contain fluorophores in the off
state. Super-resolution information encoded in this “negative data” (off state region of
fluorophores) is extracted mathematically. Sophisticated mathematical analysis of the acquired
date is used to generate a super-resolution image.

Consider another radically different approach for achieving time distinction. Rather than
controlling the illumination spot, what if it were possible to switch single molecules of
fluorophores on (i.e., to an excitable state that can be imaged) and off (a state that cannot be
imaged)? At a given time, only a small population of fluorophores is turned on for imaging. Then
this subset is turned off after imaging. The premise of this approach is that if the density of the
fluorophores in the on state is very low and provided enough photons can be collected from
each molecule, then the image of each individual molecule can be resolved (Fig. 6, right). If the
molecules that are switched on in any given image are far enough apart so that the diffraction-
limited spots associated with each molecule are fully resolvable, then each spot can be
artificially replaced with a much smaller spot (well below the diffraction limited spot). By
repeating the process of switching on a small population of fluorophore molecules, imaging and
then turning them off, a sequence of images is generated. Finally all these images with
artificially smaller spot sizes are super-imposed to arrive at a super-resolution image. In
contrast, in widefield imaging by default all the fluorophores are in the on state while imaging
and therefore the images of individual molecules cannot be independently resolved (Fig. 6, left).
Photoactivation Localization Microscopy (PALM, also synonymously referred to as F-PALM for
Fluorescence PALM) and Stochastic Optical Reconstruction Microscopy (STORM) were the first
techniques to exploit this principle of stochastic switching of fluorophores to generate super-resolution
images.

Figure 6: A comparison of widefield imaging with the stochastic switching approach of super-resolution
microscopy. The emission signal from a fluorophore (marked ‘x’) generates a diffraction-limited spot
(green circle). In widefield imaging, usually the diffraction-limited imaging spots cannot be distinguished
from each other, thereby producing a blurred image (superposition of all the green circles). In “stochastic
switching and readout” super-resolution imaging however, at a given time only a few fluorophores
fluoresce. The emission signal from each fluorophore molecule still generates a diffraction-limited spot;
however, because the population of the fluorophores is so sparse the spots can be independently
resolved. Such resolvable individual molecules are imaged, localized (using computational tools) and
then switched off (see Table 2). This cycle is repeated to generate a stack of images and the locations of
individual molecules (marked ‘x’) from these images are used to arrive at a rendered super-resolution
image.

Specially designed fluorophores are used in PALM and STORM. These fluorophores
can be switched on and off with specific wavelengths of light. For example, PA-GFP does not
glow with blue illumination light (called the readout beam) unless it is turned on (or activated), by
virtue of photo-chemistry using a UV activation beam (Fig. 4) by a UV beam. Once activated,
PA-GFP is imaged until it photobleaches (switching off). This is one example of how
fluorophores can be turned on and off. Another approach is to use a photoswitchable
fluorescent probe, such as Dronpa (Fig. 7, Table 2) which can be activated (with a UV activation
beam), imaged (with a green readout beam) and then turned off (also with green illumination).
This principle of photoswitching was employed in the development of the STORM technique.

GSDIM (Ground State Depletion followed by Individual Molecule return) is another
example of stochastic switching and readout. This technique does not require specialized
fluorophores. In this technique, ordinary fluorophores are initially turned off, for example by
driving them into the long-lived triplet states (Fig. 4), and then as individual molecules return to
the ground state stochastically, a readout scheme similar to PALM and STORM enables super-
resolution microscopy.

Due to the numerous possibilities of imaging configurations, the requirements for optical
filters in super-resolution imaging are often best met by custom-selecting filters for a given
system. A combination of filters that can be used for photoactivation and imaging of Dronpa is
shown in Fig. 7. A special characteristic of the dichroic beamsplitter shown in this figure is its
wide reflection band that is compatible with both the activation (~405 nm) as well as readout
(~488 nm) lasers. High transmission of the emission filter ensures maximum signal collection
from a limited population of fluorophores imaged at a given time and thereby enhances overall
throughput. An essential step in single-molecule based super-resolution imaging techniques
(stochastic switching and readout) is to accurately “localize” individual fluorophore molecules.
With higher accuracy of localization higher super-resolution is achieved. Since the accuracy of
localization of a given fluorophore increases dramatically with the number of photons acquired
from a given fluorophore molecule, highly efficient optical filters play an increasingly important
role in super-resolution microscopy.

Figure 7: Optical filters for imaging of Dronpa. The wide reflection band of the dichroic ensures
that both the activation light (e.g. 405 laser) and imaging beam, 488 nm laser are efficiently
reflected and rejected in transmission. A compatible emission filter provides high blocking of both
the activation and imaging lasers.

5. Perspective

Excitement about super-resolution imaging techniques is evident from the surprisingly
rapid commercialization of “turn-key” instruments. For example, Leica was first to launch a
commercially available super-resolution microscope utilizing STED and at the time of writing this
article is currently developing a GSDIM system. Zeiss and Nikon are also in the process of
launching super-resolution microscopes that utilize PALM and STORM, respectively, as well as
variants of SIM (Structured Illumination Microscopy). The availability of commercially available
super-resolution instruments will enable widespread research at unprecedented resolution.
Research still continues to further enhance the performance of super-resolution techniques in
terms of overall throughput so that they can be used even for the study of fast cellular dynamics
in live cells. And efforts are underway to enhance the super-resolution not only laterally but also
axially. Rather than having to develop and use specialized fluorophores, many researchers
have demonstrated the advantages of using standard fluorophores (see table 2), which expands
the applicability of super-resolution techniques.
Table 1: Summary of established and emerging super-resolution microscopy techniques. The
methodology indicates whether each technique is based upon imaging of single molecules or imaging of
an ensemble of molecules (which in the limiting case can also image a single molecule).

Notes: A variant of STED includes isoSTED that combines STED and 4Pi. SSIM is also referred to as SPEM
(Saturated Pattern Excitation Microscopy). RESOLFT (Reversible Saturable / Switchable Optically Linear
Fluorescence Transition) is a generalized name for STED or SPEM. F-PALM (Fluorescence PALM) is synonymously
used with PALM. Variants of PALM include TL-PALM (Time Lapse PALM), PALMIRA (PALM with Independently
Running Acquisition), spt-PALM (Single particle tracking PALM), iPALM (Interferometric PALM) and biplane PALM
called (BP) PALM a method that enhances axial super-resolution. Variants of STORM include: direct STORM
(dSTORM) based on conventional fluorophores and 3D-STORM.

Table 2: Fluorophores being used in super-resolution microscopy. Activating light makes a fluorophore
excitable. Quenching light renders the fluorophore un-excitable; i.e., it does not emit. Note that the readout
beam required for imaging a fluorophore is different from the activating beam. Photobleaching is an
example of quenching fluorescence. Pre and post colors refer to the fluorescence emission before and after
photoshifting or photoactivation, respectively. Additional fluorophores that have been imaged using STED
and GSDIM microscopy include several ATTO dyes such as ATTO 532 and ATTO 565. Manganese-doped
quantum dots have also been imaged using RESOLFT (Table-1). Fluorophores marked with an asterisk (*)
must be used in the presence of another fluorophore called an activator. The choice of activator fluorophore
dictates the required activation light. (UV) Ultraviolet. (NA) Not Applicable.

6. References

[1] Special Feature: Method of the year, Nature Methods, 6 (1), January 2009.
[2] M. Fernandez-Suarez, and A. Y. Ting, Fluorescent probes for super-resolution imaging in
live cells, Nature reviews, Molecular Cell Biology, 9: 929-943, December 2009.
Authors

Prashant Prabhat, Ph.D. and Turan Erdogan, Ph.D., Semrock, Inc., A Unit of IDEX Corporation.
E-mail: pprabhat@idexcorp.com; Tel: (585) 594-7064; Fax: (585) 594-7095.