Identification of the gene was a crucial advance that was hailed as the last piece of the puzzle in the search for a cure for CF. However, as is customary for modelling human genetic diseases, the identification of a specific, causative gene leads to the creation of multiple “artificial” animal models, since animals do not naturally have CF. Identification of the gene associated with CF was confirmed in 1989 and the first animal model was created a couple of years later (Snouwaert et al., 1992). Animal models of CF initially used mice, but since these manipulated mice failed to fully recapitulate the human condition, and the animals did not have symptoms of the lung disease, this expanded to other animals including rats, ferrets, sheep and pigs. There are now more than 750 different genetic animal models of CF, of which more than 690 use mice (Leenaars et al., 2020). Since animals do not have CF naturally, there has to be some manipulation of the animal in order to generate symptoms of the disease and this requires either genetic mutation techniques, used for the genetic models, or non-genetic approaches, including the use of drugs, deliberate infection with pathogens, or grafting human tissues into the animals’ lungs. There are more than 220 non-genetic animal models for CF, again the majority of these use mice, but there are also rats, pigs, monkeys, and rabbits (Leenaars et al., 2021).

The Canadian database which logs the different mutations in CFTR that give rise to CF currently stands at 2,114 (http://www.genet.sickkids.on.ca/StatisticsPage.html). Even before considering the ethics and scientific relevance of this exercise, the costs and time required to develop an animal model for each of these mutations would be exorbitant. One commercial service offering development of genetically modified animals charges a basic fee of around $4,000 for a single mutation and this does not include technical support (such as animal care and surgery) or consumables (e.g., surgical equipment) (https://brcf.medicine.umich.edu/cores/transgenic-animal-model/fees/; https://www.umassmed.edu/globalassets/transgenic-animal-modeling-core/documents/whycostsomuch2015.pdf). To create one animal version of each CF mutation found in people would therefore cost more than 8 million US dollars. This is a very conservative estimate since costs would be greater for larger animals (rabbits costs at least twenty times more than mice and mini pigs are over one hundred times more costly than mice https://minipigs.dk/products-services/enquiries) and it is more expensive and time consuming to create the complex models which would more accurately represent the patient population (almost half of the UK CF population have two different mutations https://www.cysticfibrosis.org.uk/sites/default/files/2020-12/2019%20Registry%20Annual%20Data%20report_Sep%202020.pdf). Developing pig models of the spectrum of CF mutations found in humans would cost upwards of 190 million US dollars.

However, CF is a great example of how human-centred interventions (as opposed to animal-based research) have played a great role in improving understanding of CF such that the bleak prognosis for the child with the salty brow has improved over time. In the 1960s and ‘70s, increased survival was associated with the creation of dedicated CF centers, these were crucial for sharing best practices for nutrition and physiotherapy and for fast, aggressive treatment of infections. With these interventions and the knowledge sharing offered by the dedicated networks of CF physicians, parents and families, the median survival age for babies born with CF reached 11 years in the 70s and is still on the increase (Matthews et al., 1964; Doershuk et al., 1965) (Figure 1).

Traditionally, animals are used in drug development (Figure 2) to demonstrate the safety of the preparation (for a comprehensive review of the drug development process, see the paper by Singh et al. in this edition (Singh et al., 2023)). As stated by Dr van Norman in her 2019 article about animal use for drug safety testing, “[t]here is no doubt that the use of animals in science and medicine has significantly benefitted human beings.” However, despite an historical reliance on animals for medical advances, drug attrition rates of more than 90% (Sun et al., 2022) indicate that animals are not predictive for humans and a drug that is safe for animals is not always safe (or effective) in people (Van Norman, 2019; Van Norman, 2020). Unfortunately, it is not possible to carry out a stringent comparison of the possible failure rates if drugs were tested using the non-animal approaches. The data submitted by the pharmaceutical companies as part of the regulatory approval are not publicly available and whilst the current legislative requirements do not demand animal data, many of the drug safety guidelines to which pharmaceutical companies refer make reference to submission of preclinical pharmacology data from animals. It is therefore not possible to calculate the likely “failure rate” of a drug tested solely on human based methods or using human data. Despite a mounting body of evidence that the human based tools are (and will be) more predictive than animals, we cannot simply assume that the non-animal tests will always accurately predict human responses. However, as data from human relevant tools accumulate, this may change. Indeed, the case study of CF presented here indicates that patient-derived samples and developing tests using human cells offer insight into the biological effects of potential drugs and can revolutionise treatments.

The new approach methodologies (NAMs) are innovative methods that no longer rely on the use of live animals, these include human cells, tissues and organs, organ-chips or microphysiological systems (MPS), use of human data or human volunteers, and computer modelling. When considering the uniquely human nature of a disease, these methods offer a more relevant and physiologically accurate approach to understanding the disease features and therefore developing an effective treatment. They also provide more confidence in the predictivity of the data obtained—recent research has shown that MPS can recreate species-specific effects (Jang et al., 2019) and that MPS using human liver cells are better able to predict toxic drugs than animal testing (Ewart et al., 2022). In fact, analysis of possible savings that could be realised if human liver MPS were included for safety testing within the drug development pipeline revealed that these might reach around 3 billion USD in drug development costs (Ewart et al., 2022).

For drug safety testing, there are national and international guidance documents that have to be adhered to, and these indicate that animal studies should be undertaken (e.g., the International Council on Harmonisation guidance M3 (R2) for “Non-clinical safety studies for the Conduct of Human Clinical Trials and Marketing Authorization for Pharmaceuticals” states that “The development of a pharmaceutical is a stepwise process involving an evaluation of both animal and human efficacy and safety information” and goes on to detail what number and species of animals should be used throughout the process U. S. Food and Drug Administration). However, many of the regulatory agencies claim that they do not require data from animals and that they would be willing to examine data from the non-animal methods and the UK government recently declared that [“t]here is no United Kingdom legislation that mandates animal testing” (UK Parliament, 2023). Indeed, the Food and Drug Administration Modernization Act 2.0, which was signed into US law in 2022, permits drug developers to make use of “certain alternatives to animal testing, including cell-based assays and computer models, to obtain an exemption from the Food and Drug Administration to investigate the safety and effectiveness of a drug” (One hundred and seventeenth Congress of the United States, 2023). Likewise, for drug efficacy testing, (these are studies undertaken to evaluate whether a treatment will be effective), the regulatory agencies do not insist on animal data although animal models of disease are often used here, and so it seems that the door is open for non-animal data to be submitted to support product registration.

In 2006, researchers at Vertex first used human cell cultures with bronchial epithelial cells (taken from the upper airways of people) to develop a method to measure the activity of the CFTR channel (Van Goor et al., 2006). They used this human-relevant tool to screen hundreds of thousands of different chemicals and identify compounds that would increase CFTR activity. This assay revealed many hundreds of compounds with the desired activity, and the researchers then got rid of the ones that were unlikely to be successful drugs based on chemistry (through comparing structures to drugs that have been recognised as toxic). Fifty-three compounds were selected for further investigation using intestinal organoid models—note that organoids are tiny, cell-based models that retain the structure and function of the “parent” organ (Dekkers et al., 2013). Accessing bronchial epithelial cells is quite difficult and can require sedation, making this a complicated procedure to carry out in very young children (the intended patient population), but the cells to create intestinal organoids are more easily accessible and patients have reported limited discomfort with this technique (Servidoni et al., 2013). Thus, intestinal organoids offer a robust method for screening hundreds (or even thousands) of potential drugs. Using this approach, several likely compounds can be rapidly and efficiently assessed using the intestinal organoids, and the activity can be verified with the airways model so that the most promising candidate(s) can quickly move to the clinic.

Another advantage to using intestinal organoids from people with CF is that this offers a simple system with which to reveal patient specific treatments. Organoids are three dimensional structures, so that when CFTR is active, fluid influx leads to swelling of the organoid and measuring the volume of the organoid is a simple way to know whether the drug applied to the organoid is activating CFTR (where increased volume indicates active CFTR). Using this approach, patient-specific theratyping is possible, where organoids from patients can be screened against multiple compounds, or combinations of compounds, to find the optimal dose and identity of drugs to give each individual patient (Clancy et al., 2019; Conti et al., 2022). The major advantage is that the organoid retains the exact same genetic mutation in CFTR as the person, and so “treating” the organoid reflects what will happen in that individual. This testing is just not possible using animal models, or at least it would require that an animal model was available for every single individual with CF, which would take time and money and thousands of animal lives (as described above), and still does not guarantee success, given that animals do not exhibit the lung disease that we are trying to treat! Thus, adoption of the human cell-based intestinal organoid models as a screening platform makes this a very efficient way of screening individual people with CF, to ensure that the medication given will be effective, and therefore can be used to personalise treatment.

The example of curcumin demonstrates another way in which in vitro, human cell-based tools can be used to help advance drug discovery, or at least to prevent false hope. Curcumin is a derivative of the spice turmeric; this was force-fed to genetically modified CF mice (who bore the mutation most frequently found in people with CF) for 3 days, before the electrical activity of their airways and intestines was measured, as an indicator of CFTR activity. The study also used cultures of hamster kidney cells with mutated CFTR inserted in them and tracked the processing of the CFTR protein, showing that curcumin enhanced the insertion of CFTR in the appropriate cellular localisation (Egan et al., 2004). These results suggested that curcumin could correct the CFTR-dependent deficit but unfortunately, this was not recapitulated for people with CF. When the studies were repeated with airways epithelial cells isolated from someone with CF, curcumin did not increase the electrical activity (Song et al., 2004). These data are evidence of the issue with translational failures–where data in animals are not recapitulated in people and indicate that caution is needed in interpreting data from non-human model systems. We suggest that the incorporation of more human-relevant tools, such as the airways models or intestinal organoids, in the drug discovery pipeline, would help reduce these failures.

The non-animal, human relevant tools can be used to reveal insights into the association between physiology and disease symptoms, that are not possible in animals, and that allow researchers to estimate the likely clinical effect of a drug. For example, clinical data comparing the electrical potential difference across the airways in people with or without CF indicates that increasing the amount of CFTR in the cell membrane by around 15%–30% would help to “normalise” this electrical activity (which is associated with ion movement, and therefore CFTR activity) in CF. Researchers can then test potential drugs, existing drugs and combinations of these, in human airways cell models to find those compounds that increase membrane CFTR and enhance CFTR activity. An additional advantage of using human cell-based systems is the ability to carry out direct comparisons. Models developed using CF cells can be compared with non-CF cultures, but importantly models from people with CF who have different mutations in their CFTR can be compared. As we have already mentioned above, research with genetically engineered animal models is almost prohibitively expensive, time consuming, and cannot accurately model human responses, whereas the human cell-based approach allows researchers to clarify the relationship between the mutation in CFTR and its activity, and therefore to make intelligent decisions regarding which CFTR modifiers may be helpful for individual people with CF. Although there is one specific mutation that affects the majority of people with CF (Kerem et al., 1989), as described above, there are over 2000 different mutations. In vitro, cell-based methods offer a more efficient manner with which to test the effects of drugs on different mutations and combinations of mutations (since people with CF may have two different mutations). It is very gratifying to see this potential utility of non-animal models reflected at some level of regulations, as the FDA granted a “label extension” for one of the CFTR modifiers, based purely on data from human cell models (Ratner, 2017), enabling the broader use of this drug for many more people with CF.

The realisation that patient-derived intestinal organoids could be used to filter hundreds or compounds to detect possible CFTR modifiers for treating CF led to another important advance - that of the application of human relevant tools in drug repurposing, an approach suitable for applications beyond CF. Drug repurposing is an efficient way of finding new treatments as it employs existing, approved drugs for a purpose other than that for which they were approved, and so this bypasses the lengthy, expensive safety testing that is needed for entirely new compounds.

This extends beyond CF: there are other examples of drug repurposing where the efficacy data were obtained using human cell-based approaches. For example, for SARS-CoV2, researchers tested seven clinically approved drugs on airways models made of human cells, measuring the ability of these drugs to prevent the virus getting inside the cells (Si et al., 2020). When the drugs were tested at concentrations and flow rate equivalent to those found in human blood, they found that only two of them showed great promise in terms of preventing infection and, soon after, a clinical trial was set up to assess the effects of one of these drugs in people with COVID. Additionally, a human stem cell-based system showed that a biological therapeutic could be effective in a rare neuropathy disease (Rumsey et al., 2022) and data from an organ chip demonstrated that a medication used in Type 2 diabetes could prevent chemotherapy-induced kidney toxicity (Cohen et al., 2021). This is hopefully the start of a shift in the regulatory paradigm and is indicative of enhancing flexibility to enable accelerated access to safe and effective drugs for all patients.