Access Highlight at ICFO: Imaging tissue mimics with light sheet microscopy

Tissue mimics such as microtissues, spheroids and organoid cultures have become increasingly important in life science research, as they provide a physiologically more relevant environment for cell growth, tissue morphogenesis and stem cell differentiation. In contrast to cell lines cultured in monolayers, these 3D models can be engineered to display most of the hallmarks of native tissue in terms of architecture, cell heterogeneity and self-renewal properties. They thus have great potential for modelling tissue development and disease progression in the context of cancer, heart, and stem cell research, as well as in neurobiology, drug discovery and toxicity testing.

Over the last decade, Light Sheet Fluorescence Microscopy (LSFM) has emerged as a powerful solution for long-term live 3D imaging of organism models and many other application fields, including tissue mimics. However, due to their thickness, inhomogeneity and high light-scattering properties, observing tissue mimics at high resolution, in-depth (beyond 100 μm) and in real-time remains a major technical challenge.

To address these problems, several image processing tools and hardware have appeared. In a recent study we have presented a unified and objective comparison of different Light Sheet Fluorescence Microscopy techniques in terms of practicality and performance in the specific context of 3D tissue mimics imaging.

For this purpose, Corinne Lorenzo and Pablo Loza-Alvarez have worked in collaboration with two associated groups, the Département Optique Théorique et Appliqué at Onera in Toulouse and the Quantitative Image Analysis Unit at the Paris Pasteur Institute, to formally assess the performance of the different light sheet microscopy modalities to study tissue mimics. All groups involved have complementary skills and expertise in cancer biology, cell imaging, image processing, biophotonics and optics.

We generated four types of tissue mimics representing samples with different cell morphology (volume, density, shape, etc.) and cell population heterogeneity. Our tissue mimics correspond to samples generally used in different fields of investigation such as cancer biology (mammary duct spheres, multicellular tumour sphe - roids), cardiac biology (cardio-spheroids) and neurobiology (neurospheres).

In a first approach, we tested six different light-sheet modalities using a single type of tissue mimics. This allowed us to highlight and select the modalities of choice based on:

1. Resolution: 3D Point Spread Function using fluorescence in the tissue mimic sample;
2. Signal-to-noise: Ratio between the signal and the background;
3. Contrast: The Normalised Contrast Index as the differences between adjacent pixels.

The light sheet was produced using a cylindrical lens or by Digitally Scanning Light Sheet Microscopy (DSLM). With DSLM the used modalities were linear (one-photon) and nonlinear (two-photon) excitation regimes, under Gaussian or Bessel beam excitation profiles. Finally, the system included a sCMOS camera in which the digital scan was synchronized using the ‘rolling shutter’ mode to achieve confocal line detection.All measurements were automatically extracted from the images using a set of robust and user-friendly image analysis and quantification protocols, available through the open-source Icy platform.

We first observed that the transversal resolution was similar for all modalities. However, the axial resolution was twice as large in the one-photon Bessel case, due to the side-lobes of the Bessel beam. In terms of standard deviation, two-photon modalities presented sharper resolutions than their one-photon counterparts, reflecting the good performance of the nonlinear versus the linear regime in terms of reducing scattered and out-of-focus light.

We then plotted the histogram of the overall signal-to-noise obtained for each illumination and observed that one-photon modalities were significantly superior to two-photon modalities in terms of fluorescence emission efficiency. On the other hand, when observing the local contrast measures (given by the Normalised Contrast Index), we saw that non-linear modalities yielded the largest values in comparison to any other linear modality. Finally, taking both metrics into account, it was interesting to observe that the linear modalities provided similar signal-to-noise and Normalised Contrast Index, despite the negative impact of the out-of-focus light from the side-lobes of the Bessel beam.

From this first set of results, we selected the three most representative techniques and we then further characterised the performance of the systems when imaging the different types of tissue mimics.

In the second set of experiments, we considered only the gold standard Gaussian with linear excitation and Bessel modalities with filtering capabilities. Here different tissue mimic models, including multicellular tumour spheroids, mammary duct spheres, cardio-spheroids and neurospheres were used to assess the imaging capabilities of the light sheet illumination modalities. In addition to the quantitative measures used in the first stage of evaluation, here we developed an additional set of quality metrics specific to LSFM: penetration depth and the Contrasted Imaging Volume or CIV, also using the Icy platform.

We observed that tissue mimics differed in term of nuclear size, cell density and level of heterogeneity (Fig. 1). However, the signal-to-noise was relatively stable across tissue mimics and light sheet illumination modalities. Nonetheless, non-linear Bessel excitation generally displayed lower signal-to-noise than linear modalities.

In terms of the Normalised Contrast Index, variability in behaviour was notably greater across tissue mimics, showing a maximum NCI for non-linear Bessel excitation regardless of the tissue mimic imaged. In the case of cardio-spheroids and neurospheres, linear Bessel excitation looked like non-linear excitation, providing a similar level of Normalised Contrast Index.

Figure 2: Schematic representation of the multimodal LSFM setup used. A CW and an ultrashort pulsed laser, both aligned to the same optical path, were used for linear and nonlinear excitation, respectively. A cylindrical lens (1) free beam propaga- tion (2) and an axion lens (3) could be also included in the setup for accessing the different LSFM modalities. The light sheet was generated using a 10x, NA=0.3 objective and the images were taken using a 20x, NA=0.5 objective. (Image taken from Andilla et al., 2017)

When looking at the resolution, the results were consistent with the first stage of experiments and no significant differences in behaviour appeared among the tissue mimics investigated herein. By measuring the penetration depth achieved by the different illumination modalities, we observed that the non-linear Bessel excitation offered better overall penetration depth, independent from the tissue mimic type.

As expected, there is no clear winner among the various techniques tested, and the choice of modality typically depends on the application at hand. Linear Gaussian excitation gives the best signal-to-noise, but lacks on contrast and sectioning capabilities. In comparison, linear Bessel excitation provides fair signal-to-noise, but a larger field of view and better sectioning capabilities. Finally, non-linear Bessel excitation provides the largest field of view and a highly homogeneous x,y,z resolution, but it does not perform as well in terms of signal-to-noise. Overall linear Bessel excitation seems to provide the best compromise among all measured criteria if the uniformity of the sectioning capabilities is not demanding for the measurements.

Pablo Loza-Alvarez

Andilla, J., et al., Imaging tissue-mimic with light sheet microscopy: A comparative guideline, Sci. Rep. 7, 44939 (2017)