Live Cell Imaging

The Art and Science of Live Cell Imaging

Live cell imaging requires users to carefully balance the imaging conditions – the aim is to collect the minimal essential data, which minimizes perturbance of the model system. While it can be tempting to use more illumination power, longer camera exposures, etc. to obtain higher signal-to-noise ratio (SNR) images, it’s important to understand that phototoxicity can compromise the physiological integrity and continued viability of the system. Furthermore, photobleaching imposes a practical limit to how many photons can be extracted from each fluorophore over the course of the experiment.

Another factor to consider is choice of objective lens. High numerical aperture (NA) oil objective lenses may work well for imaging features near the coverslip in flat cells, but suffer from spherical aberration the deeper one images into the sample. This is due to the lower refractive index (RI) of the cell/culture medium compared to glass/oil. For deeper and 3D imaging applications, it is recommended to use objectives that utilize immersion media more closely matching the RI of the imaging environment, such as water or silicone.

The CFI Apochromat Lambda S Objective Series includes a number of water immersion objective lenses, including the CFI Plan Apochromat IR 60XC WI, which has the highest NA for a water immersion objective at NA = 1.27 that we are aware of. Nikon has more recently introduced its Silicone Immersion Series objective lenses. Silicone has an RI of about 1.4, more closely matching the cellular environment while allowing for higher NAs than water immersion. These objectives are suited to imaging 3D culture systems such as organoids.

Minimizing Phototoxicity and Photobleaching in Live Cells

The intense irradiation required by fluorescence-based imaging techniques is inherently harmful to cell health, especially with higher-energy (blue-shifted) wavelengths. For this reason, the use of red-to-near-infrared fluorophores are recommended, though the continued use of popular green fluorophores such as EGFP is not overly-prohibitive and continues to work well for many multicolor imaging applications.

Most microscopes utilize software triggering of system devices. However, this can be somewhat slow due to the relative inaccuracy of the computer clock and the required checks and callbacks of device states. Conversely, hardware triggering uses the high accuracy camera pixel clock to coordinate devices while bypassing device callbacks, thus maximizing system speed while minimizing photon dose. Triggered devices may include Z piezo stages, laser/LED illuminators, and even custom devices featuring I/O communication capability.

Triggering also allows for more complex illumination patterns. For example, it is possible using the Nikon NIS-Elements software to illuminate the sample only when all pixels of a sCMOS camera are exposing (accounting for the effect of the rolling shutter) – decreasing photon dose and photobleaching compared to continuous illumination and free-running image acquisition. It is also possible to pulse illumination on the micro-second time scale during the exposure with various LED and laser illuminators. Pulsed illumination on this time scale allows fluorophores excited to a triplet state extra time to relax to the ground state, and thus help avoid major photobleaching pathways.

Using Deep Learning-based Artificial Intelligence to Improve Live Cell Imaging

The advent of artificial intelligence (AI) methods for microscopy has opened up new possibilities for live cell imaging with reduced photon dose. Methods based on deep learning are proving suited towards assays based on images of biological specimens. Nikon is committed to the development of stable software solutions based on deep learning for various applications – available as NIS.ai modules for our NIS-Elements software.

Clarify.ai provides automatic removal of out-of-focus blur from standard widefield fluorescence images. This module is pre-trained at our factory and does not require manual user-based adjustment of the final image, removing the opportunity for subjectivity. Clarify.ai thus allows for optical sectioning to be performed on single plane widefield images acquired at the maximum system rate, not requiring data from multiple planes as with 3D iterative deconvolution of widefield data.

Denoise.ai is another pre-trained module and used to remove Poisson noise (shot noise) from confocal images in real time. This allows for reduced exposure times and is advantageous when used in conjunction with resonant scanning images acquired using the AX R confocal microscope, where shot noise is the predominant source of noise.

Nikon also offers the Convert.ai module for predicting image features that are typically detected in a fluorescence channel, with the predicted signal based on a corresponding brightfield image (e.g., prediction of DAPI nuclear staining patterns from DIC image data). This allows for elimination of fluorescence imaging channels, avoiding the intense illumination demanded by fluorescence techniques. Brightfield imaging techniques are inherently less phototoxic.

Enhance.ai can be trained to predict details in low signal-to-noise (SNR) images, effectively improving SNR and allowing for imaging with reduced photon dose and photobleaching.

Which Contrasting Technique is Right for Your Research?

For relatively flat two-dimensional samples, such as adherent cells cultured in vitro, widefield fluorescence imaging may be sufficient. Compatible use-cases and sample types are much broader when combining widefield imaging with deconvolution analysis or automatic deblurring using the Clarify.ai software module.

Confocal imaging becomes necessary when imaging samples that are more than ~20 μm thick but are also useful for high-resolution imaging of samples even just a few μm thick. Spinning disk confocal instruments are fast and suited to samples up to ~50 μm (with the CSU-W1 able to image deeper than other models due to greater inter-pinhole spacing). Both systems are applicable towards imaging different types of 3D cell culture systems, such as organoids and spheroids.

Point-scanning confocal imaging is able to image deeper, up to several hundred μm. The resonant scanner featured on the AX R confocal is able to provide video rate imaging and is intended for combination with Denoise.ai for real-time shot noise removal. Unlike spinning disk systems, the AX R features a continuously adjustable pinhole for optimization of optical sectioning and resolution with any compatible objective lens.

Total internal reflection fluorescence (TIRF) imaging is an optical sectioning technique that allows for exclusive observation of cell features occurring within a few hundred nm of the cell-coverglass interface. This provides a improvement in SNR, allowing for the use of lower laser power to protect cell health and/or maximize imaging speed. Nikon’s CFI Apochromat TIRF Series objective lenses also feature a very high NA of 1.49.

Super-resolution microscopy techniques are for when optical resolution beyond the diffraction limit is specifically required to resolve the necessary details. Live cell super-resolution microscopy has generally been challenging due to the tradeoffs demanded of these varied techniques to improve resolution. The N-SIM S structured illumination system realizes faster illumination pattern modulation than previously using as spatial light modulator. The Yokogawa CSU-W1 SoRa is a super-resolution system implemented in the context of a spinning disk confocal system, allowing for imaging at speeds matching the CSU-W1 alone (assuming sufficient signal).

While fluorescence imaging techniques are powerful, allowing for multiplexed detection of distinct molecular targets, brightfield imaging techniques such as phase contrast and differential interference contrast (DIC) provide detailed cell images with significantly lower photon dose. While such techniques may lack the required molecular-specificity, it is good practice to acquire occasional brightfield images in order to help assess cell health over the course of the experiment.