This Review covers the steps required to create high-quality image-based profiles from high-throughput microscopy images.

Imaging flow cytometry combines the high-throughput capabilities of conventional flow cytometry with single-cell imaging. Here we demonstrate label-free prediction of DNA content and quantification of the mitotic cell cycle phases by applying supervised machine learning to morphological features extracted from brightfield and the typically ignored darkfield images of cells from an imaging flow cytometer. This method facilitates non-destructive monitoring of cells avoiding potentially confounding effects of fluorescent stains while maximizing available fluorescence channels. The method is effective in cell cycle analysis for mammalian cells, both fixed and live, and accurately assesses the impact of a cell cycle mitotic phase blocking agent. As the same method is effective in predicting the DNA content of fission yeast, it is likely to have a broad application to other cell types.

We show that deep convolutional neural networks combined with nonlinear dimension reduction enable reconstructing biological processes based on raw image data. We demonstrate this by reconstructing the cell cycle of Jurkat cells and disease progression in diabetic retinopathy. In further analysis of Jurkat cells, we detect and separate a subpopulation of dead cells in an unsupervised manner and, in classifying discrete cell cycle stages, we reach a sixfold reduction in error rate compared to a recent approach based on boosting on image features. In contrast to previous methods, deep learning based predictions are fast enough for on-the-fly analysis in an imaging flow cytometer.

The delivery of nanoparticles into cells is important in therapeutic applications and in nanotoxicology. Nanoparticles are generally targeted to receptors on the surfaces of cells and internalized into endosomes by endocytosis, but the kinetics of the process and the way in which cell division redistributes the particles remain unclear. Here we show that the chance of success or failure of nanoparticle uptake and inheritance is random. Statistical analysis of nanoparticle-loaded endosomes indicates that particle capture is described by an over-dispersed Poisson probability distribution that is consistent with heterogeneous adsorption and internalization. Partitioning of nanoparticles in cell division is random and asymmetric, following a binomial distribution with mean probability of 0.52-0.72. These results show that cellular targeting of nanoparticles is inherently imprecise due to the randomness of nature at the molecular scale, and the statistical framework offers a way to predict nanoparticle dosage for therapy and for the study of nanotoxins.

Understanding nanoparticle uptake by biological cells is fundamentally important to wide-ranging fields from nanotoxicology to drug delivery. It is now accepted that the arrival of nanoparticles at the cell is an extremely complicated process, shaped by many factors including unique nanoparticle physico-chemical characteristics, protein-particle interactions and subsequent agglomeration, diffusion and sedimentation. Sequentially, the nanoparticle internalisation process itself is also complex, and controlled by multiple aspects of a cell’s state. Despite this multitude of factors, here we demonstrate that the statistical distribution of the nanoparticle dose per endosome is independent of the initial administered dose and exposure duration. Rather, it is the number of nanoparticle containing endosomes that are dependent on these initial dosing conditions. These observations explain the heterogeneity of nanoparticle delivery at the cellular level and allow the derivation of simple, yet powerful probabilistic distributions that accurately predict the nanoparticle dose delivered to individual cells across a population.

We show that deep convolutional neural networks combined with nonlinear dimension reduction enable reconstructing biological processes based on raw image data. We demonstrate this by reconstructing the cell cycle of Jurkat cells and disease progression in diabetic retinopathy. In further analysis of Jurkat cells, we detect and separate a subpopulation of dead cells in an unsupervised manner and, in classifying discrete cell cycle stages, we reach a sixfold reduction in error rate compared to a recent approach based on boosting on image features. In contrast to previous methods, deep learning based predictions are fast enough for on-the-fly analysis in an imaging flow cytometer.

Assessing dose in nanoparticle–cell interactions is inherently difficult due to a complex multiplicity of possible mechanisms and metrics controlling particle uptake. The fundamental unit of nanoparticle dose is the number of particles internalized per cell; we show that this can be obtained for large cell populations that internalize fluorescent nanoparticles by endocytosis, through calibration of cytometry measurements to transmission electron microscopy data. Low-throughput, high-resolution electron imaging of quantum dots in U-2 OS cells is quantified and correlated with high-throughput, low-resolution optical imaging of the nanoparticle-loaded cells. From the correlated data, we obtain probability distribution functions of vesicles per cell and nanoparticles per vesicle. Sampling of these distributions and comparison to fluorescence intensity histograms from flow cytometry provide the calibration factor required to transform the cytometry metric to total particle dose per cell, the mean value of which is 2.4 million. Use of the probability distribution functions to analyze particle partitioning during cell division indicates that, while vesicle inheritance is near symmetric, highly variable vesicle loading leads to a highly asymmetric particle dose within the daughter cells.

No abstract