Cell Painting: The Art of Image Analysis

This blog post was written by ImmuONE, a leading provider of inhaled and immune in vitro assay solutions using single-cell analysis. Their services are available on the Scientist.com marketplace.

What is cell painting?

Cell painting is a high-content imaging assay (HCIA) that utilises a collection of fluorescent dyes to label distinct subcellular compartments such as the nucleus, endoplasmic reticulum and mitochondria (Figure 1). Machine learning/AI models analyse these images to extract multiple morphological features, generating a comprehensive phenotypic fingerprint. By comparing treated versus untreated cells under these conditions, it’s possible to predict mechanisms of action, toxicity and classify compounds based on their effects at the cellular level.

Macrophages and morphology

At ImmuONE, we routinely offer cell painting as a service to clients across sectors. One of the key applications of the assay is the phenotyping of alveolar macrophages (AM). AM are phagocytic immune cells, which serve as the first line of defence in the lower respiratory tract (Figure 2). They protect the respiratory system by engulfing inhaled pathogens and particulate matter. These cells play a crucial role in maintaining homeostasis, mediating disease progression and regulating immune responses, making them essential for inhalation toxicity assessments.

Morphological changes of AM provide key insights into biological pathways (phospholipidosis, apoptosis, particle overload, inflammation). This helps to differentiate between adaptive and adverse cellular responses to compounds. Beyond lung toxicology, insights gained from AM extend to other tissue-resident macrophages, offering a broader understanding of immune responses. Their versatility makes them ideal for investigating inflammatory lung safety while also providing valuable data for extrapolating systemic macrophage behaviour in toxicology and disease research.

Baselines and a reference library

In 2013, we began tackling a critical industry issue — the appearance of foamy macrophages in animal studies, which hindered the development of novel inhaled therapeutics. The challenge: to characterise foamy macrophage responses to inhaled compounds and develop an in vitro assay to serve as a read across for in vivo toxicity. The first HCIA assay was first shown to identify foamy macrophage responses in rat and human cells [3].

Over the years, the method was refined through extensive benchmarking by applying soluble and non-soluble compounds in multiple conditions. Further resolution of morphological characteristics over time [4] and links to macrophage activation were reached [5].

This work validated HCIA reproducibility in detecting cellular morphological changes and its ability to differentiate safe inhalation compounds from those causing adverse effects [6]. As part of this work, a robust reference library of compounds, spanning published safe inhaled compounds, known toxicants, phospholipidosis-inducers and antibody-drug conjugates was compiled. This has allowed us to establish a strong understanding of baseline macrophage morphology, identifying what a “healthy” cell looks like and defining a reliable fingerprint of “safe” responses. Beyond pharmaceuticals, HCIA can be applied across multiple industries facing similar challenges [7].

Assay development and phenotyping: morph_ONE^TM

As highlighted in [3], every cell responds differently to a substance. Traditional bulk analysis methods provide average results, masking critical details about cellular heterogeneity and individual cell adaptations. HCIA for morphology requires analysing shifts in cell population profiles, such as the proportion of cells exhibiting elevated vacuolation or lipid content, rather than relying on population averages, which are often too insensitive to detect meaningful changes.

morph_ONE^TM leverages four fluorescent stains to generate HCIA data (Figure 3). Instead of collecting hundreds of morphological features, we focused on key characteristics that align with in vivo histological concerns. Descriptors pathologists identify as relevant in toxicity studies (changes in cell area, vacuolation and macrophage lipid content) and most indicative of adverse responses were prioritised.

To structure the analysis, four descriptors are selected: mitochondrial activity, cell area, number of vacuoles and the extent of vacuolation. Each descriptor is categorised into three levels, generating 81 distinct phenotypes. Rather than considering individual descriptors in isolation, they are combined to create a phenotypic fingerprint of a cell’s response to a compound [7].

Case study: Unmasking toxicity through individual cell analysis

AM were exposed to a known toxicant [7]. When analysed using traditional average assessment, the compound appeared benign, showing only minor mitochondrial activity reductions (Figure 4A) and membrane permeability (Figure 4B) changes.

When subjected to phenotypic fingerprinting at the single-cell level (Figure 5), it was observed certain phenotypes which exhibited adverse effects would have otherwise gone unnoticed. For example, a cell population exposed to the known toxicant had an increased cell population exhibiting a phenotype associated with decreased mitochondrial activity increased cell area and increased number of vacuoles (Figure 5A). Test item A showed a threefold increase in a phenotype associated with increased cell area and increased number of vacuoles after 48 hours of exposure (Figure 5B). In the phenotype associated with decreased mitochondrial activity and increases in cell area, number of vacuoles and total area of vacuolation, the known toxicant (and to a lesser extent test item A) had an effect on the number of cells in that population (Figure 5C).

Key takeaways

Why cell painting?

Morphological analysis provides essential context and meaning to biochemical markers like cytokines, which alone may not distinguish between adaptive and adverse responses. Although single-cell markers help detect acute cellular toxicity, a more sophisticated approach (beyond traditional bulk measurements) is necessary to differentiate individual cell populations and reveal the mechanisms causing toxicity.

Why this approach?

Developed over a decade of research, HCIA methods are standardised and validated for reproducible outputs. With an extensive reference database, establishing a clear baseline for what constitutes a safe response, confidently identifying deviations indicative of toxicity is achievable. This framework for a biomechanistic assay is now a powerful and reliable tool for inhaled and compound safety assessments across multiple contexts.

Connect with ImmuONE on Scientist.com

References

1. Way GP, et al. Evolution and impact of high content imaging. SLAS Discov. 2023;28(7):292 – 305.
2. Mu X, et al. Tissue-resident macrophages in the control of infection and resolution of inflammation. Shock. 2021 Jan;55(1):14 – 23.
3. Hoffman E, et al. Morphometric characterisation of rat and human alveolar macrophage cell models and their response to amiodarone using high content image analysis. Pharm Res. 2017;34(12):2466 – 76.
4. Hoffman E, et al. Investigating the suitability of high content image analysis as a tool to assess the reversibility of foamy alveolar macrophage phenotypes in vitro. Pharmaceutics. 2020;12(3):262.
5. Hoffman E, et al. High content image analysis as a tool to morphologically distinguish macrophage activation and determine its importance for foamy alveolar macrophage responses. Front Immunol. 2021;12:611280.
6. Hoffman E, et al. Profiling alveolar macrophage responses to inhaled compounds using in vitro high content image analysis. Toxicol Appl Pharmacol. 2023;474:116608.
7. Hutter V, et al. High content analysis of in vitro alveolar macrophage responses can provide mechanistic insight for inhaled product safety assessment. Toxicol In Vitro. 2023;86:105506.