DMD-based super-resolution structured illumination microscopy visualizes live cell dynamics at high speed and low cost

Background

Structured illumination microscopy (SIM) is a widely used super-resolution fluorescence microscopy technique. Its particular strength is the capability to image at high frame rates and with low phototoxicity, which makes it a highly effective tool for live-cell imaging, for visualizing the dynamics of cellular organelles. SIM relies on the creation of an interference pattern at the diffraction limit using the coherent addition of laser beams created by a diffraction pattern.

Opto-mechanic implementations use a diffraction grating that is either mechanically shifted and rotated or steered by galvanometric mirrors to create interfering light beams and, thus, the SIM pattern in the sample plane. However, some limitations are that these systems are complex to build, align and maintain. Ferro-electric light modulators (FLCOS devices) are also conventional options for fast SIM systems, however, drawbacks include the need to constantly switch between positive and inverted images, and that there are limited suppliers and system choices.

Digital micromirror devices (DMD) are a promising option for creating interference-based SIM patterns. They are available in a variety of models, can provide even faster switching times than FLCOS, and can maintain a set pattern for extended durations without the need to switch or refresh the image. Additionally, they are more cost-effective. One of their limitations is the blazed grating effect. This effect arises due to the fixed angles between which the mirrors can be switched, creating a sawtooth arrangement of mirrors and thus leading to a change in the intensity distribution of the diffracted beams. This results in SIM patterns with varying modulation contrast which are prone to reconstruction artifacts. Only if the blazed grating effect is modeled and properly taken into account, can DMDs be effectively employed for structured illumination microscopy.

In their work, Sandmeyer, Lachetta et al studied the blazed grating effects of DMDs by simulations, and identified settings required to generate SIM patterns (1). They used low-cost components to generate a compact SIM system, and tested its performance on fixed and live-cell imaging of biological samples.

Key findings and developments

Simulation of blazed grating effect

An optimal illumination pattern for SIM features high modulation contrast. As the pattern is generated by interfering two beams of coherent light, these beams have to be of the same intensity and, ideally, the same polarization. This even intensity distribution is difficult to achieve with a DMD. If a DMD is used, parameters including mirror dimension and tilt angle, among others, are readily fixed by the device manufacturer, and therefore only the angle of incidence can be adjusted for any given wavelength to fulfill the blaze condition and to find the blaze angle. The authors went on to explore the underlying causes for the blazed grating effect, and performed simulations of the effect. They modeled the DMD, and simulated the diffraction pattern depending on the angle of incidence. In order to do this, they described the DMD mathematically, and calculated the electric field reflected from its surface for different positions and states of the mirrors. As part of this work, the authors created a software package to numerically perform these calculations. Following the modulation of a single mirror, the authors then modeled a two-dimensional array of mirrors.

The authors demonstrated that the SIM pattern in the Fourier plane can be simulated if a DMD is used as the primary device to create interference patterns. Having done this, the authors determined that all possible illumination conditions and their resulting blaze angles needed to be identified in order to further guide the experimental implementation. As this process needs to be repeated for large datasets running through all possible variations of angles, they implemented the simulation on a graphics card to accelerate the calculations.

The authors then proceeded to perform a comparison of simulations and experimental results. To do this, they projected the experimentally obtained intensity pattern diffracted by the DMD onto a camera chip using a single lens. All nine SIM patterns (three illumination angle and three associated phase shifts) were displayed by the DMD. The comparison of distribution of diffraction orders in ON and OFF setups was similar in both cases, but different to that obtained by the simulation. The discrepancy was attributed to either the physical DMD microstructure being deviated from the ideal structure used in the simulation, or to a protective glass plate covering the DMD chip, which was not considered in the simulations, but which could change the light path and the distribution of the diffraction orders.

The authors found that the absolute values of the experimentally found blaze angles were not the same as those obtained by the simulation. They proposed investigating the impact of different tilt angles along the diagonal, on the blaze angle, and found that indeed, small changes in tilt angles leads to very significant shifts of the blaze angle. Ultimately, the authors found a tilt angle and a blaze angle with perfect match between experimental setup and simulations, both in OFF and ON cases.

 

Construction of a compact and cost-effective SIM system

Having determined the correct blaze angle, the authors determined that constructing the remaining components of the DMD-SIM microscope was straight-forward. They discuss that using a DMD is more cost-efficient and allows for a compact design due to the small pixel size of the DMD mirrors. They went on to construct a SIM microscope with a small footprint, and of a total cost ten times lower than commercial solutions.

To test the functionality of the DMD-SIM setup, they imaged TetraSpeck beads (TS) with a diameter of 200 nm and then reconstructed the frame set with fairSIM [2,3]. The SIM reconstruction process also allows the estimation of the pattern modulation depth achieved by the instrument. For the setup hereby presented, this estimate yields reasonable values for a well-aligned 2D SIM system.

Altogether, the estimated modulation depths, the experimentally determined resolution calculated from a Fourier Ring Correlation analysis, and the theoretically expected resolution calculated from the SIM pattern spacing and optical parameters were fully consistent.

 

SIM images of biological samples

The authors then went on to demonstrate the functionality of the DMD-SIM microscope on biological samples. This included visualization of fixed cells stained for actin with Phalloidin Atto532 or labeling fixed transfected cells with mScarlet, for which the SIM mode allowed resolving individual filaments. In both cases, the DMD-SIM microscope had sufficient sensitivity to resolve the actin filaments. The authors also labeled the outer membrane of lysosomes with mScarlet followed by fixation of the cells. Spherical structures are considered a good quality control for SIM microscopes, and in this case, the authors demonstrate that their setup allows resolving small spherical lyososomes.

Having shown the potential of the setup on fixed samples, the authors then went on to test it on live samples. Live samples present further challenges for SR including the refractive index of the medium, and the fact that dynamic processes occur, which may lead to motion-blur in the images. For live-cell experiments, they stained mitochondria using MitoTrackerRed, and using DMD-SIM, demonstrated that they were able to resolve cristae. They also used ER-TrackerRed to stain the ER. They found that the ER network could be easily resolved and the movements of single filaments were easily resolved.

What I like about this preprint

This work bridges a technology gap in a thorough manner, going from simulation to experimental proof. I enjoyed reading it because I like work that improves accessibility to science. The authors found a way to take a relatively complex and expensive setup, and make it accessible to everyone (consistent with the philosophy of democratizing science). I also like that it explains clearly the working of a DMD-SIM microscope- which is very exciting for those interested in microscopy and the principles behind it.

Open questions

1. Did you find specific limitations of your setup that are key to consider?
2. What are the limitations of your setup in terms of speed and resolution?
3. You mention in the live imaging section, the additional hindrances and complexities that might be faced in live samples. Do you have suggestions specific to your DMD-SIM setup when used for live imaging? (for instance, suggestions on media, dyes, etc.?
4. You tested the setup only with dyes and tags around the red wavelengths (500-600). Was there a reason for this? What is your suggestion in terms of number of dyes for simultaneous imaging?

 

References

1. Sandmeyer A, Lachetta M, et al, DMD-based super-resolution structured illumination microscopy visualizes live cell dynamics at high speed and low cost, bioRxiv, (2019).
2. Müller et al, Open-source image reconstruction of super-resolution structured illumination microscopy data in imagej, Nature Communications 7 (2016).
3. Markwirth et al, Video-rate multi-color structured illumination microscopy with simultaneous real-time reconstruction, Nature Communications (2019).

 

doi: https://doi.org/10.1242/prelights.24175

Read preprint (No Ratings Yet)