Skip to main content

Thank you for visiting You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Chemical safety without animals

Primary human cells provide mechanistic information for toxicology studies.

New chemicals are the lifeblood of the pharmaceutical and consumer-product industries, but how best to assess their safety remains controversial. Conventional animal testing has several drawbacks, including high cost, low throughput, uncertainty about the relevance of results to humans, and increasing societal opposition to the use of animals for this purpose. In this issue, Kleinstreuer et al.1 report results from one of several ongoing initiatives aimed at modernizing chemical safety assessment. Their approach is among the first to use primary human cells on a large scale to investigate the biological pathways underlying the potential toxicity of a diverse collection of chemicals. Notably, they find that primary human cells can be used to reliably identify toxicity targets and mechanisms of value in predicting adverse outcomes for consumers, workers and environments exposed to chemicals.

Safety assessors in industry and regulatory agencies usually derive acceptable levels of exposure to new chemicals by combining toxicology data generated in mammals, fish and invertebrates with predicted levels of exposure in humans and environmental species. But this approach is increasingly questioned as public attitudes turn against the use of animals in safety testing. For example, recent revisions to European regulations include a marketing ban on certain products containing ingredients tested on animals2. Some progress has been made in the development of alternatives, such as in vitro tests for skin irritation, but ensuring the safety of chemicals without animal testing remains a significant scientific challenge3.

The shift away from animal testing gained momentum with a landmark 2007 report from the US National Research Council, 'Toxicity testing in the twenty-first century: a vision and a strategy', which called for animal testing to be replaced by “in vitro methods that evaluate changes in biologic processes using cells, cell lines, or cellular components, preferably of human origin”4. It argued that toxicology should be a science based on mechanistic understanding of key biological pathways—'toxicity pathways'—which, if sufficiently perturbed by a chemical, produce an adverse outcome in the exposed individual or environment. Knowledge of these pathways could enable safety assessments based on a mechanistic understanding of human biology rather than by extrapolation from high-dose effects in rodents.

More recently, the Organization for Economic Cooperation and Development (OECD), whose formal guidelines for the conduct of toxicology studies are adopted by safety assessors and regulators worldwide, proposed the use of 'adverse outcome pathways', which provide a more complete description than 'toxicity pathways' of the consequences of perturbing pathways. An adverse outcome pathway begins with a 'molecular initiating event' in which a chemical interacts with a biological target, leading to a sequence of events across different levels of biological organization (for example, cells, tissues, organs, whole organisms, populations or ecological communities) that produce an adverse outcome relevant to a given risk assessment context. The OECD has started to formalize a universal chemicals-risk assessment framework based on adverse outcome pathways to capture and peer review the mechanistic understanding of specific toxic effects, and to evaluate nonanimal methods that aim to predict key events in these pathways5.

The study from Kleinstreuer et al.1 was conducted in accordance with this framework. It builds on previous work identifying primary human cells that can provide readouts of whether a chemical disrupts specific, physiologically important pathways. For example, measurement of 13 endpoints in venular endothelial cells stimulated with the cytokines interleukin-1β, tumor necrosis factor-α and interferon-γ are used to assay chemical effects on vascular biology and chronic inflammation. Unlike many other approaches to in vitro toxicology that rely on studying cell death, this approach uses batteries of well-characterized primary human cells that have been designed to reflect specific human pharmacology and pathology through understanding of disease-related biomarkers.

Kleinstreuer et al.1 report one of the first public, large-scale data sets generated using such cell-based assays6. A total of 776 compounds, including environmental chemicals and 135 failed pharmaceuticals donated by companies, were evaluated using eight culture systems of primary human cell types. Protein biomarkers—87 readouts per compound across all eight cell types—were measured in cells treated with at least four concentrations of compounds, yielding 306,240 measurements in the complete data set (Fig. 1).

Figure 1: Toward improved drug and chemical safety assessment without animal testing.

Debbie Maizels/Nature Publishing Group

In work conducted as part of the US Environmental Protection Agency's ToxCast program, Kleinstreuer et al.1 used a panel of primary human cell types to assess the toxicity of 776 environmental chemicals and failed drugs donated by pharmaceutical companies. The resulting data set of 306,240 measurements was computationally analyzed to predict pharmacological effects, off-target adverse effects and mechanisms of chemical toxicity. This work is part of a global effort to modernize human and environmental safety assessment through improved mechanistic understanding of how compounds affect human biology, in contrast to traditional methods that extrapolate from high-dose effects seen in animals.

The authors analyzed the data with several computational algorithms to infer mechanisms of toxicity for each chemical. Importantly, the large number of drugs evaluated allowed the use of computational clustering approaches to generate predictions of both pharmacological activity and potential off-target effects. Because the complete toxicity profiles of most of the compounds studied were not known, it is not possible to comprehensively evaluate the correctness of the predictions, but the authors highlight many predictions that are consistent with published information on the chemicals' clinical or preclinical safety. Taken together, these results demonstrate that studying a large number of diverse chemicals in a relatively small, well-characterized panel of human primary cell types can produce consistent patterns of bioactivity that can be used to predict intended pharmacological effects of drugs as well as off-target adverse effects seen in the clinic.

Despite the encouraging results of Kleinstreuer et al.1, we are still a long way from robustly predicting human toxicity from in vitro data alone. Additional assays are needed to measure effects on specific organs and specialized cell types, and there is a pressing need to consider xenobiotic metabolism owing to the key role biotransformation can play in the activation of toxins. This work also demonstrates the importance of bioinformatics and mathematical approaches to modeling complex biological pathways. Future decisions on human safety will require methods that integrate pathway networks and are capable of distinguishing normal adaptive responses from those that perturb pathways and give rise to adverse effects.

At the regulatory level, guidance is needed in the use of in vitro data and associated mathematical models for safety decisions on specific chemicals. In particular, one must demonstrate how in vitro dose-response data identifying safe concentrations of ingredients can be combined with exposure information in humans or target organisms. Decisions on human and environmental safety of chemicals and products should be risk-based and driven by an understanding of actual exposure levels and calculation from information about consumer habits and practice6. New approaches to risk assessment that are pathway based require better understanding of the biologically relevant exposure to an ingredient and the development of many new tools, such as pharmacokinetic models to estimate systemic exposure. Examples in which pathway-based in vitro data are brought together with such exposure data are emerging7.

Kleinstreuer et al.1 generated their data as part of the US Environmental Protection Agency's Toxicity Forecast (ToxCast) program ( This work and other related projects, such as the Safety Evaluation Ultimately Replacing Animal Testing (SEURAT-1) program in the European Union (, demonstrate that the National Research Council's 2007 vision is becoming a practical reality. Scientists from diverse disciplines—mathematical modeling, tissue engineering, mechanistic chemistry, high-content imaging, stem cell biology and signaling pathways—are now working together to bring increased mechanistic understanding to decisions on human and environmental safety.


  1. 1

    Kleinstreuer, N.C. et al. Nat. Biotechnol. 32, 583–591 (2014).

    CAS  Article  Google Scholar 

  2. 2

    European Commission. Off. J. Eur. Union L 66, 26–35 (2003).

  3. 3

    Adler, S. et al. Arch. Toxicol. 85, 367–485 (2011).

    CAS  Article  Google Scholar 

  4. 4

    Krewski, D. et al. J. Toxicol. Environ. Health B Crit. Rev. 13, 51–138 (2010).

    CAS  Article  Google Scholar 

  5. 5

    Organization for economic cooperation and development. Report of the workshop on using mechanistic information in forming chemical categories. (no. 138, series on testing and assessment) (OECD, Crystal City, 2011).

  6. 6

    Kavlock, R., et al. Chem. Res. Toxicol. 25, 1287–1302 (2012).

    CAS  Article  Google Scholar 

  7. 7

    Hall, B. et al. Food Chem. Toxicol. 45, 2097–2108 (2007).

    CAS  Article  Google Scholar 

  8. 8

    Adeleye, Y. et al. Toxicology available online, doi:10.1016/j.tox.2014.02.007 (25 February 2014).

Download references

Author information



Corresponding authors

Correspondence to Carl Westmoreland or Paul L Carmichael.

Ethics declarations

Competing interests

Both Dr. C. Westmoreland and Prof. P.L. Carmichael are employed by Unilever.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Westmoreland, C., Carmichael, P. Chemical safety without animals. Nat Biotechnol 32, 541–543 (2014).

Download citation

Further reading


Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing