Fast live simultaneous multiwavelength four-dimensional optical microscopy

Peter M. Carlton, Jérôme Boulanger, Charles Kervrann, Jean‐Baptiste Sibarita, Jean Salamero, Susannah Gordon-Messer, Debra A. Bressan, James E. Haber, Sebastian Haase, Lin Shao, Lukman Winoto, Atsushi Matsuda, Peter Kner, Satoru Uzawa, Mats Gustafsson, Zvi Kam, David A. Agard, John W. SedatProceedings of the National Academy of Sciences 2010 · DOI

TL;DR

Researchers have developed a new microscopy platform (OMX) capable of capturing four-dimensional (4D) images of live cells at unprecedented speeds (sub-second, multi-wavelength). This allows for the observation of cellular processes like chromosome movement throughout an entire cell cycle without damaging the cells, thanks to a novel denoising algorithm that recovers biological information from extremely low-light, noisy images.

What Problem Does This Paper Address?

Observing dynamic biological processes in living cells using fluorescence microscopy is incredibly powerful, but it's a delicate balancing act. The very light used to illuminate the sample can damage it (phototoxicity) or cause the fluorescent markers to fade (photobleaching). This limits how long and how quickly researchers can image, often forcing a trade-off between image quality and cell health. The goal is to achieve high spatial and temporal resolution, identify multiple components simultaneously, and do it all without perturbing the natural behavior of the cell. This paper tackles the challenge of acquiring fast, multi-dimensional, live-cell imaging data while minimizing photodamage and photobleaching.

Methods

The core of this work is the development and application of a new microscope platform called OMX (Objective-based Multi-view Confocal). This platform is designed for fast, multiwavelength, four-dimensional (4D) optical microscopy.

Key methodological innovations include:

  • High-Speed 4D Acquisition: The OMX platform can acquire 3D image stacks at rates of up to 10 stacks per second, enabling the capture of dynamic events over time.
  • Sub-diffraction Structured Illumination Imaging: The platform also provides access to structured illumination microscopy (SIM), which can achieve resolutions beyond the diffraction limit of light.
  • Image Denoising Algorithm: A crucial component is a sophisticated denoising algorithm that exploits spatial and temporal redundancy in the image sequence. This algorithm analyzes image data for statistical noise signatures and averages similar regions to reduce noise while preserving important image features. This is essential for recovering usable data from the extremely low-light images required to prevent phototoxicity.
  • Phototoxicity and Viability Assays: The researchers rigorously tested cell viability by monitoring cell division after imaging. They used specific yeast strains engineered with fluorescent reporters (e.g., LacI::GFP) to quantify the effects of different excitation light intensities and imaging durations.

Datasets

The primary "datasets" used in this study are derived from live-cell imaging experiments. The researchers imaged:

  • Yeast cells (YDB271 strain): This strain contains a GFP fusion to the Lac repressor, allowing visualization of Lac operator binding sites. This was used extensively to study chromosome movement throughout the cell cycle and to assess phototoxicity and viability.
  • Fluorescent latex beads (100 nm): These were used to quantitatively assess the impact of denoising on image features like peak width and intensity, independent of biological complexity.
  • Drosophila melanogaster larval nuclei: These were used to demonstrate the denoising algorithm's applicability to more complex biological systems with multiple fluorescent labels.
  • Caenorhabditis elegans meiotic nuclei: This system was used to show denoising's effectiveness in recovering features from noisy images of subcellular components.

The paper does not introduce a new benchmark dataset in the traditional sense but rather uses these biological and synthetic samples to demonstrate the capabilities of their imaging and processing techniques.

Key Insights and Ideas

  • Unexpected Photosensitivity: Yeast cells, even under what was previously considered "low-light" microscopy conditions, exhibit significant photosensitivity, leading to cell cycle arrest or death. This highlights the need for much lower excitation levels than commonly assumed for live-cell imaging.
  • Denoising as a Game-Changer: The developed denoising algorithm is not just a cosmetic fix; it's a fundamental enabler. It allows researchers to extract meaningful biological information from images that would otherwise be too noisy to analyze, opening up the possibility of imaging at light levels that preserve cell viability.
  • The "Live Imaging Power Landscape": The paper implicitly defines a trade-off space between excitation intensity, imaging speed, and cell viability. By understanding this landscape, researchers can make informed decisions about imaging parameters.
  • Preservation of Quantitative Information: The denoising algorithm doesn't just make images look better; it preserves quantitative features like peak intensity and width, making the recovered data suitable for accurate analysis.
  • Eliminating Fading: Imaging at extremely low light levels, combined with denoising, effectively eliminates fluorophore fading over extended imaging periods, crucial for capturing complete cellular processes.

Key Results

  • Cell Viability Maintained: By reducing excitation light intensity by 100 to 10,000 times and using the denoising algorithm, researchers could image yeast cells throughout an entire cell cycle (2 hours) at one 3D image stack per second without affecting cell division.
  • Sub-second 4D Acquisition: The OMX platform achieved sub-second acquisition of multiwavelength 4D data, enabling the capture of rapid cellular dynamics.
  • Quantitative Image Recovery: The denoising algorithm successfully recovered biological information (e.g., chromosome positions, nuclear boundaries) from extremely noisy images, with quantitative features like peak intensity and width largely preserved.
  • Negligible Fading: At the lowest excitation levels (I = 10⁻³I₀ and below), the imaging showed essentially no fading over time, a critical factor for long-term live imaging.
  • Demonstrated Applicability: The denoising approach was shown to be effective on various biological samples, including Drosophila and C. elegans nuclei, indicating its general utility.

Limitations and Future Work

  • Assay Specificity: The primary viability assay used was cell division. The authors acknowledge that for cell types that don't divide or take much longer to do so, alternative viability assays would be necessary.
  • Denoising Algorithm Limits: While powerful, the denoising algorithm has limits. The authors note that even with denoising, pushing to extremely low light levels can approach the algorithm's ability to retrieve information, depending on the signal brightness.
  • Future Work: The authors suggest exploring other methods to enhance viability, such as increasing expression of free radical scavenging enzymes or modifying the cellular environment (e.g., removing oxygen). They also imply that further optimization of denoising parameters or development of even more advanced algorithms could push the limits of low-light imaging even further. The generalizability of their findings to other microscopes is also discussed, suggesting that the principles are broadly applicable.

Summary updated

Abstract

Live fluorescence microscopy has the unique capability to probe dynamic processes, linking molecular components and their localization with function. A key goal of microscopy is to increase spatial and temporal resolution while simultaneously permitting identification of multiple specific components. We demonstrate a new microscope platform, OMX, that enables subsecond, multicolor four-dimensional data acquisition and also provides access to subdiffraction structured illumination imaging. Using this platform to image chromosome movement during a complete yeast cell cycle at one 3D image stack per second reveals an unexpected degree of photosensitivity of fluorophore-containing cells. To avoid perturbation of cell division, excitation levels had to be attenuated between 100 and 10,000× below the level normally used for imaging. We show that an image denoising algorithm that exploits redundancy in the image sequence over space and time allows recovery of biological information from the low light level noisy images while maintaining full cell viability with no fading.
BibTeX
@inproceedings{carlton2010fast,
  author = {Peter M. Carlton and Jérôme Boulanger and Charles Kervrann and Jean‐Baptiste Sibarita and Jean Salamero and Susannah Gordon-Messer and Debra A. Bressan and James E. Haber and Sebastian Haase and Lin Shao and Lukman Winoto and Atsushi Matsuda and Peter Kner and Satoru Uzawa and Mats Gustafsson and Zvi Kam and David A. Agard and John W. Sedat},
  title = {{Fast live simultaneous multiwavelength four-dimensional optical microscopy}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2010},
  doi = {10.1073/pnas.1004037107},
}
Figure