A tube-source X-ray microtomography approach for quantitative 3D microscopy of optically challenging cell-cultured samples

Development and study of cell-cultured constructs, such as tissue-engineering scaffolds or organ-on-a-chip platforms require a comprehensive, representative view on the cells inside the used materials. However, common characteristics of biomedical materials, for example, in porous, fibrous, rough-surfaced, and composite materials, can severely disturb low-energy imaging. In order to image and quantify cell structures in optically challenging samples, we combined labeling, 3D X-ray imaging, and in silico processing into a methodological pipeline. Cell-structure images were acquired by a tube-source X-ray microtomography device and compared to optical references for assessing the visual and quantitative accuracy. The spatial coverage of the X-ray imaging was demonstrated by investigating stem-cell nuclei inside clinically relevant-sized tissue-engineering scaffolds (5x13 mm) that were difficult to examine with the optical methods. Our results highlight the potential of the readily available X-ray microtomography devices that can be used to thoroughly study relative large cell-cultured samples with microscopic 3D accuracy.


nature research | reporting summary
April 2020 Field-specific reporting Please select the one below that is the best fit for your research. If you are not sure, read the appropriate sections before making your selection.

Life sciences Behavioural & social sciences Ecological, evolutionary & environmental sciences
For a reference copy of the document with all sections, see nature.com/documents/nr-reporting-summary-flat.pdf

Life sciences study design
All studies must disclose on these points even when the disclosure is negative.  No preliminar sample-size calculations were made. The exact number of samples, FOVs, and acquired particles in silico are described in the paper.
It was assumed that the raw 3D data will include imperfections with various origins such as imaging noise. Systematic data filtration, explained in detail in the manuscript and supplementary data, was used to remove one-to-a-few-voxel objects (much smaller than the observed median nuclear sizes were), and incomplete objects cut by the edge of the FOV. According to the reported data-quality indicators, other data anomalies were excluded from the analysis as well. Part of these anomalies were under-segmented close nuclei, which were later reprocessed and disintegrated in silico for the second round of analysis. After the disintegration, the statistical analysis applied on the new separate nuclei reproduced similar results in line with the original data. Data-filtration flow and the number of objects rejected at various steps are tabulated in detail in Supplementary Table 1, 2 and 3. Special attention was paid for cleaning the Figure 4 CLSM data to ensure that it represented only true nuclei well-resolved used as reference, the data exclusions described in detail in methods.
Not only the examined sub-cellular phenomenon was captured with the used uCT analysis more efficiently than we expected, but the same results were succesfully reproduced with rather different analytical setups. The same clear nuclear rounding, induced by the cytoskeletondisturbing agent, was observed in both wet and dry imaged samples, and in the in silico disintegrated under-segmented nuclei. All the observations within and across the experimental series were coherent and expected. Furthermore, in Figure 4 the visual and quantitative similarities and dissimilarities of the µCT imaged nuclei with CLSM imaged nuclei in 3D were assessed in detail. These observations help to define the accuracy and reproducibility of the demonstrated uCT imaging and analysis method.
All the samples were prepared and handled in parallel with great control on reproducibility. Half of the scaffold samples were randomly chosen for the cytochalasin-D exposure. After the wet imaging, two randomly chosen samples were needed for optical imaging: one for optimizing the refractive-index matching water-glycerin solution, and one that was directly put into the optimized solution for the actual 3D optical imaging. Thus, the first wet imaging experiment included 3 + 3 samples; and the following dry experiment included 2 + 2 samples. Similarly, for the Figure 4 experiment the flat-substrate samples were randomly chosen for the µCT and CLSM labeling and imaging. The acquired FOVs were randomly positioned on cellular parts of the samples avoiding the overlapping of the FOVs and counting of the same cells multiple times.
To avoid additional layer of complexity and possibility for human error, we did not see the use of blinding to be necessary for our experiments. Many of the data processing steps were automatized with computer algorithms.