Advanced Imaging Techniques Notes

Advanced Imaging Techniques

Learning Goals

Brief refresher on the diffraction limit for ‘regular’ light microscopy, understanding its implications for resolving fine details.
Basic principles for 5 major methods for ”super-resolution microscopy,” focusing on how each bypasses the diffraction limit.
Learn something about how these methods are implemented in practice, including sample preparation and instrumentation.
Be able to discuss main features, pros/cons of these 5 methods:
1. 4 $\pi$ microscopy: Enhancing light collection for improved resolution.
2. Total Internal Reflection Microscopy (TIRFM): Selective illumination of structures near the coverslip.
3. Structured Illumination Microscopy (SIM): Using patterned light to capture high-resolution information.
4. Stimulated Emission Depletion Microscopy (STED): De-exciting fluorophores to sharpen the point spread function.
5. Single Molecule Localisation Microscopy (SMLM) - PALM, STORM and related methods: Precisely locating individual molecules over multiple frames.
Focus on main principles due to time constraints, highlighting key differences and applications.

Nobel Prize

The Nobel Prize in Chemistry 2014 was awarded to Eric Betzig, Stefan W. Hell & William E Moerner for “the development of super-resolved fluorescence microscopy.”

Hell (2000): stimulated emission depletion (STED) microscopy—reducing the size of the fluorescent spot.
Betzig & Moerner (2006): single-molecule microscopy (PALM)—achieving resolution by localizing individual molecules.

Diffraction-Limited Resolution

The problem: diffraction-limited resolution of conventional (light) microscopy limits the ability to resolve structures smaller than approximately half the wavelength of light.

Even at the best focus within the 3D blur structure of the “point spread function” (PSF), an infinitely bright single point source of light will be focused as an imperfect ‘diffraction pattern’ with a central ‘Airy disc,’ leading to blurry images.

The “Airy disc” (central bright part) with fainter diffraction rings. See also https://micro.magnet.fsu.edu/primer/lightandcolor/diffractionintro.html

Rayleigh’s Criterion

Diffraction (blur) sets a lower limit on our ability to differentiate nearby structures, making closely spaced objects appear as one.
Two pinpoint objects can be resolved if their Airy discs in the image are far enough apart that the center of one lies on the first minimum of the other, defining the resolution limit.
Visible light (wavelength =~ 0.5 $\mu$ m); maximum numerical aperture in oil (1.4).

Resolution Limit

The resolution limit for a conventional light microscope cannot be better than ~0.2 $\mu$ m even at the shortest ‘visible’ wavelengths due to the wave nature of light and diffraction.

Practical Limits

Lateral resolution 200-300nm: The ability to distinguish objects side by side.
Axial resolution 500-800nm: The ability to distinguish objects along the optical axis.

Under even optimal imaging conditions the grey area in this diagram illustrates the diffraction limit (Airy disc) on the scale of a single bacterium.

Light from each fluorophore molecule would appear at least this large, complicating precise localization!
Makes localization of specific proteins to a part of the cell very challenging, hindering cell biology studies.
As we saw earlier, we can use deconvolution to give approx. 2x better resolution, computationally reducing blur.

Structures Impossible to Resolve

many important structures are impossible to resolve in sufficient detail with conventional light microscopy (even with deconvolution !), e.g., viruses, small organelles, and protein complexes.

Approaches to Improve Resolution

4 $\pi$ Microscopy

Using more than one objective increases the amount of collected light and improves resolution.

Modified confocal using 2 objectives: uses ‘coherent superposition’ of excitation light to gather more information about the sample, enhancing both excitation and emission light collection. e.g. described in the textbook (Cox)
Sample sandwiched between 2 coverslips to allow both objectives access.
Gives some improvement in lateral (xy) and a large improvement in axial (z) resolution, creating more detailed 3D images.
However, very complicated, expensive, and impractical for most samples, limiting its widespread use.

Total Internal Reflection Microscopy (TIRF)

Introduction to TIRF microscopy. Selectively illuminates fluorophores close to the coverslip.

Snell's law: $n = \frac{\sin i}{\sin r}$ , describes the relationship between the angles of incidence and refraction.

Principle

When light passes from glass to air or water at angles greater than the ‘critical angle’ it is totally internally reflected (exit angle > 90°), preventing light from entering the sample directly.
Still gives rise to an evanescent wave (an electromagnetic field) just above the boundary in the low-refractive index medium, which is key to TIRF's functionality.
The evanescent wave decays exponentially over a short distance, typically less than 100 nm, ensuring only structures very close to the surface are illuminated.
The evanescent wave can excite fluorophores that are very close to the interface (typically 100 nm or less), enabling high signal-to-noise imaging of surface events.
Illumination light (laser beam) is focused at the outer edge of the back focal plane of the objective to achieve a high angle of incoming light, essential for total internal reflection.
The illumination light will reach the surface of the cover slip at such a steep angle that it will be totally internally reflected in the coverslip! This is the core of the TIRF technique.
No light actually enters the sample! Only the evanescent wave interacts with the specimen.
But…the evanescent wave can excite fluorophores very close to the coverslip, allowing for selective visualization.
There will be essentially no background from “out-of-focus” objects, providing exceptionally clear images.

A related method is called HILO (Highly inclined laminated optical sheet) microscopy.

The angle of illumination is lower than for TIRF, allowing some light to penetrate deeper.
Some light passes deeper into the sample, providing a compromise between background reduction and penetration.
However, it gives very low background from objects deeper in the sample compared with conventional fluorescence, useful for thicker samples.

Pros

Excellent axial resolution, since only objects within about 100 nm are excited, making it ideal for studying surface interactions.
Excellent signal-to-noise ratio due to the absence of out-of-focus light.
Sample in water hence live imaging possible (e.g., protein dynamics), allowing real-time observation of biological processes.

Cons

No improvement in lateral resolution (but can be combined with single-molecule methods), which may require additional techniques for comprehensive analysis.
Limited to seeing only objects close to the “coverslip” (could be a pro or con…), restricting its use to surface-related phenomena.
Essentially 2-dimensional, providing limited information about the sample's depth.
Requires objectives with very high NA, increasing equipment costs and complexity.

Applications

Using TIRF microscopy to study protein dynamics in vitro. Observing molecular interactions and behaviors on surfaces.

TIRFM can be used to monitor protein binding or assembly on a surface, e.g., on an artificial membrane surface, providing insights into molecular mechanisms.
In the example below, it was used to study assembly of a protein called septin into filaments in a lipid bilayer, revealing details about cytoskeletal structures.
In vitro reconstitution of the Min system, visualized with TIRFM. Studying protein self-organization on membranes.
A lipid bilayer is formed on the surface of a microscopy cover slip (”a model membrane”). Mimicking the cell membrane environment.
Pure MinD and MinE proteins, mixed with about 20% fluorescently labeled MinD (green) and MinE (red) proteins are added on the membrane surface. Tracking protein behavior.
Fluorescence signals on the membrane surface are monitored by TIRF microscopy. Capturing real-time protein dynamics.
MinD and MinE formed self-organizing wave patterns on the flat membrane surfaces in vitro in the presence of ATP! Demonstrating complex biochemical behaviors.

Structured Illumination Microscopy (SIM)

Sample illuminated using a grid pattern close to the resolution limit (‘diffraction grating’) & rotated to several positions. Exploits Moiré patterns to enhance resolution.

Moiré effects arise in the images, due to interference with sub-resolution patterns in the sample. These patterns contain high-resolution information.
Note: the Moiré patterns are larger than the details in sample that gave rise to them so can be imaged! Making the invisible visible.
Can give information about sub-resolution details in the sample since observed patterns can be resolved. Extracting fine structural data.
Several images captured with the grid pattern tilted and also shifted laterally (e. g. 9-15 images). Capturing comprehensive data from multiple angles.
Image processing, involving Fourier transforms of the data, can generate an image of the sample with resolution in both XY and Z dimensions improved by a factor of about 2. Computationally enhancing resolution.
Method does not require any special fluorophores or sample treatment, allowing for versatile applications.

Pros:

SIM can improve resolution about two-fold (similar to deconvolution combined with regular or confocal imaging), expanding the capabilities of standard microscopes.
It works fine with regular fluorophores, simplifying sample preparation.
relatively fast and may work for live cell imaging, enabling dynamic studies.

Cons:

multiple images are needed to generate the final result, increasing data acquisition time.
Data processing is required to create the final images (but this is mathematically less demanding than e.g. iterative deconvolution methods), necessitating computational resources.
Best for high magnification, shallow (thin) samples, limiting its use for thicker specimens.

Variants

Can be combined with other methods (e.g. TIRF-SIM) to leverage complementary advantages.
Rapid growth area for new instruments and methods, offering continually improving capabilities.
Many, many variants…, each with specific strengths and applications.
(far too many to cover here!)

Stimulated Emission Depletion Microscopy (STED)

The pinhole (in the image plane) only allows the central part of the diffraction pattern through, enhancing image clarity.

But remember that the full 3D PSF (including diffraction pattern) is still formed in the object plane: Understanding the PSF is crucial for interpreting STED images.

A confocal technique that exploits a principle called ‘emission depletion’ of the fluorescence excited near the focal plane. Selectively de-exciting fluorophores to reduce the size of the emitting area.
A second (long wavelength) laser forms a donut shaped focal spot, aligned with the central part of the diffraction pattern of the short wavelength (excitation) laser, creating a smaller effective excitation volume.
Produces a much smaller PSF for the excitation spot, enhancing resolution.

Principle

Excitation laser forms a diffraction-limited spot (as in confocal microscopy). Initial excitation stage.
A second ”depletion beam” is applied to immediately de-excite fluorophores, preventing them from emitting light in the outer regions of the focal spot.
Depletion is ”donut-shaped” in the focal plane & very high intensity, ensuring effective suppression of fluorescence outside the central region.
… but intensity is zero at the center of the donut, leaving tiny region in the middle where fluorophores remain excited and can emit fluorescence! This small region defines the improved resolution.
The donut shape of depletion PSF results from ‘spiral phase plate” (destructive interference at the focal plane). Creating the necessary beam shape for STED.
By increasing the intensity of the depletion beam, the non-depleted spot in the middle can be made very small (<50nm), further enhancing resolution.
Only fluorophores in a very small spot in the center of the depletion beam are allowed to emit photons via regular fluorescence. Selectively detecting light from this small volume.
The detector will only detect signal from this very small (“sub-diffraction size”) spot. Enhancing image sharpness.
Otherwise the beams are scanned over the sample exactly like in a confocal microscope. Maintaining a structured scanning process.
Excellent resolution!

Limitations

Photostable fluorophores are required to withstand the intense light.
Risk for phototoxicity in live cells due to high laser intensities.
Z resolution still limited by the original elongated PSF…, requiring additional strategies for improvement.
With an additional depletion beam, it is possible to modify the PSF also along the Z-axis, and thereby improve also axial resolution, achieving isotropic resolution.
But this is very complex and expensive, limiting its practicality.

Single Molecule Localisation Microscopy: PALM, STORM

Basic principle: Localization of individual molecules with high precision.

If (note the big IF) you know that the signal that you detect as an Airy disc comes from 1 fluorophore molecule, you can estimate the likely localization of this molecule to be the middle of the Airy disc pattern (in XY and Z). Precisely determining molecular positions.
Ensure that only a few and physically separated molecules in the sample emit light at each moment… Reducing signal overlap.
Collect a large number of images of small subsets of fluorophore molecules in the sample. Accumulating data for precise reconstruction.
Determine localizations of individual molecules. Pinpointing molecular coordinates.
Repeat, repeat, repeat…. Iterative process for thorough mapping.
Replace the actual image (blurred by diffraction) with a single bright point (dot) labelling the likely location of a fluorophore. Creating a super-resolution image.
Overlay the dots to form a final image. Assembling the complete picture.
Imaging can be in a wide-field microscope, or using TIRFM, or other setups. Versatile imaging configurations.
Images are captured by sensitive CCD cameras (e.g. EM-CCD). Detecting faint signals.
The objects are labelled with suitable fluorophores. Ensuring visibility.
The fluorophore can switch between a dark non-fluorescent state and, on the other hand, a “bright” state in which it can be excited and emit fluorescence (see D). Dynamic fluorophore control.
Samples are handled or treated in such a way that at each moment, only a small fraction (<1%) of the fluorophore molecules in the sample can emit light. Ensuring single-molecule resolution.
Each observed Airy disc is likely coming from one molecule. Simplifying localization.
Images of such emitting molecules will give pixelated intensity distributions (see A) in the images. Capturing molecular signals.
Fitting a 2D Gaussian model to such distributions, the center of the Airy disc can be determined computationally (see B). Precisely locating molecules.
This is the likely localization of the emitting molecule (see C). Mapping molecular positions.
It is essential that only Airy discs that derive from single molecules are considered. Avoiding artifacts.
Any “airy disc” signals that are too bright may emanate from two or more molecules and must be discarded. Ensuring accuracy.
Crucial to obtain a sufficient density of signals for a reliable representation of the structure or pattern. Complete structural mapping.
Important to have even labelling to avoid bias. Consistent representation.
Requires control over state of the fluorophore (active/inactive) – e.g. switchable fluorophore. Dynamic control.
Limits the kinds of samples that can be used!

STORM & PALM

Rely on ‘switchable’ fluorophore proteins, critical for temporal separation of signals.
At any one time we want only a (small and sparse) subset of molecules to be fluorescing, ensuring minimal overlap of signals.
We image those until we capture enough signal to estimate their location with high precision.
Then switch to a different set of fluorophores in the same sample, cycling through all molecules.
Repeat and repeat until we eventually get a complete image, building a comprehensive super-resolution reconstruction.

How do we switch state?

Difference between photoactivated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM) is historical – i.e. two different names from their respective inventors. Different origins but similar principles.
In practice, PALM and STORM are often used interchangeably or grouped by terms such as “blinking microscopy” , “single molecule localization microscopy” (SMLM), “pointillism microscopy”. Broadly similar techniques.
Several different strategies have been used to switch fluorophores between dark and fluorescent states, enabling different imaging modalities.
Commonly used are photoswitchable or photoconvertible fluorescent proteins (in PALM) or organic fluorophores (in STORM). Different classes of switchable probes.
By using a low intensity of light at the “switching” wavelength, only a small subset of molecules are turned ON at each moment, ensuring sparse activation.
They are then imaged by regular excitation and emission, until they are photobleached or deliberately switched back to dark. Collecting data from the activated subset.
Then a new (random) subset is turned ON, a new image is captured, and the process is repeated a large number of times. Iterative data collection.

PALM

PhotoActivated Localisation Microscopy, a specific implementation of SMLM.

PALM involves photoswitchable or photoconvertible fluorescent proteins. Utilizes genetically encoded probes.
E.g. photoconvertible fluorescent protein EosFP (Original paper). A widely used PALM probe.
Normally, EosFP emits green light (516 nm), upon irradiation (~ 400 nm) EosFP switches to become fluorescent at 581 nm. Undergoing a light-induced conversion.
i.e. short pulse of 400 nm light: small subset of EosFP molecules in a sample can be switched to fluoresce at 581 nm. Spatially and temporally controlled activation.
Often, photoactivatable FPs are used: switched from inactive to the fluorescent state by light of certain wavelength. Another class of light-controlled probes.
“switching” light is adjusted so that only occasional molecules in sample switch. Ensuring sparsity.
These then excited and imaged until bleached – won’t corrupt further images. Preventing signal contamination.
Repeated cycles of switching “on”, imaging, and eventually bleaching are carried out. Accumulating data.

STORM

Stochastic Optical Reconstruction Microscopy, another key SMLM technique.

Most common variety (direct STORM = dSTORM), organic fluorophores exhibit spontaneous (or induced) “blinking” between a dark, non-fluorescent state, and “bright”, fluorescent state. Exploiting probe dynamics.
Buffer conditions (e.g. [ $\text{O}_2$ ]) important to control blinking behavior. Environmental influences on fluorophore behavior.
Suitable (small) fraction of the fluorophore population in ”bright state” at each moment. Maintaining sparsity.
Light of certain wavelengths used to switch fractions between dark and bright states. External control over fluorophore states.

Cons (STORM & PALM)

Need very efficient detection of weak signals due to sparse labeling.
Sensitive to background fluorescence, demanding careful optimization.
Background can be reduced by not illuminating deep into the sample (using HILO, “angle epifluorescence” - see earlier slide on TIRF). Minimizing unwanted signals.
TIRF often combined with STORM and PALM: Provides excellent sensitivity and low background…. Synergistic combination of techniques.
BUT only allows us to see a thin layer of the sample. Limiting imaging depth.