Better reproducing human physiology, identifying the toxicity or inefficacy of a drug candidate earlier and, whenever possible, reducing reliance on animal models: organoids and organs-on-chips are opening up new possibilities for biomedical research. But turning these innovations into robust, standardised methods recognised by regulatory authorities remains a major challenge. In France, an entire ecosystem is now being structured around this ambition.

In preclinical research, one question remains central: how can we predict as accurately as possible what will happen in humans?

Despite their major role in drug development, animal models have limitations when it comes to translating results to humans. During a session dedicated to organoids and organs-on-chips, the speakers highlighted that around 90% of molecules entering clinical development ultimately fail.

This high attrition rate has a direct impact on both the cost and duration of developing new therapies.

Organoids and organs-on-chips do not promise to eliminate animal testing overnight. Instead, they offer another path: reproducing specific human biological functions with sufficient relevance to answer precise scientific questions before moving into clinical trials.

The two technologies are often discussed together, but they are based on different principles.

An organoid is a three-dimensional cellular structure capable of self-organising and reproducing certain structural or functional characteristics of an organ. These models can notably be generated from stem cells or induced pluripotent stem cells.

Today, they can be used to model a wide variety of tissues and diseases and are particularly studied in oncology, where cells derived from tumours can be used to create models that retain some of the characteristics of the patient.

However, one shortcut should be avoided: an organoid is not a “mini-organ”.

It reproduces only part of an organ’s complexity. The absence of complete vascularisation, insufficient cellular maturity, batch-to-batch variability and difficulties in standardisation remain among the limitations of these models.

An organ-on-chip, on the other hand, introduces an additional dimension: microfluidics.

Cells are organised within small compartments or channels in which flows, mechanical stresses and some of the interactions found in the human body can be recreated. Sensors can also be integrated to monitor different parameters.

The technology therefore goes beyond reproducing a biological architecture: it aims to recreate specific organ functions and aspects of their physical environment.

This is perhaps one of the most important changes in perspective.

The purpose of an organ-on-chip is not necessarily to reconstruct a liver, lung or brain in all its complexity.

Instead, the key is to determine which biological function needs to be reproduced in order to answer a specific question.

For the lung, for example, a device may aim to recreate a barrier between pulmonary and endothelial cells, while incorporating mechanical stresses that mimic certain physiological movements.

For the liver, the question may be completely different: is a molecule likely to cause liver toxicity?

For the nervous system, researchers may want to determine whether a substance can cross a biological barrier or induce neurotoxicity.

This concept of “context of use” is fundamental, particularly from a regulatory perspective. Rather than claiming that a model reproduces an entire organ, researchers must demonstrate that it can reliably answer a clearly defined scientific question.

The two technologies can also be combined.

Placing an organoid within a microfluidic environment makes it possible to combine the biological complexity of a three-dimensional model with the capabilities of a chip: controlled flows, mechanical stresses, sensors and interactions with other cell types.

This approach opens up new opportunities for improving the maturation of biological models and studying more complex interactions.

In the longer term, several systems could even be connected to investigate interactions between different tissues or organs.

The ambition is significant: to progressively reproduce certain systemic human mechanisms without claiming to recreate an entire organism on a chip.

These technologies are already being explored across a wide range of fields.

In oncology, models derived from patient tumours can be assembled into biobanks, providing diverse biological models on which future drug candidates can be tested.

The potential extends beyond personalised medicine. These collections can help pharmaceutical companies better identify the types of cancer most likely to respond to a new molecule before it enters clinical trials.

The nervous system is another particularly promising area.

Netri, a Lyon-based company represented during the session, develops devices designed to recreate the innervation of tissues and organoids. Human nerve cells can extend through microchannels towards a compartment containing the tissue being studied.

The aim is notably to investigate neurotoxicity and pain mechanisms in situations where animal models do not always accurately reproduce human physiology.

Other applications include liver toxicity, cardiotoxicity, neurological diseases, inflammatory conditions, infectious diseases and medical devices.

Behind these developments lies a major scientific and economic objective: identifying drug candidates likely to fail as early as possible.

The further a molecule progresses through development before a toxicity or efficacy problem is detected, the more costly that failure becomes.

Organs-on-chips could help introduce human-derived data earlier in the development process, improving the selection of molecules before they reach clinical trials.

This is particularly important in situations where animal physiology differs significantly from human physiology.

During the session, the speakers referred to studies in which liver-on-chip models were able to identify human toxicities that had not been detected by certain animal models.

The objective is therefore not simply to replace one test with another. It is to make preclinical development more predictive of what will actually happen in humans.

For years, the adoption of these technologies faced a paradox.

Pharmaceutical companies could be reluctant to invest heavily in alternative methods as long as they were not recognised by regulatory authorities. Regulators, meanwhile, needed sufficiently robust data before they could accept them.

A significant change began in the United States with the FDA Modernization Act 2.0 in 2022, which opened the door further to the use of alternative methods in drug development.

Since then, momentum has been building.

However, the aim is not to approve every alternative model indiscriminately. For a method to become a recognised reference, its relevance must be demonstrated for a specific context of use, along with its reproducibility, including across different laboratories.

Such validation can take several years and may involve comparisons with known clinical data.

A scientifically promising technology alone is not enough.

For large-scale use by the pharmaceutical industry, it must be reproducible, automatable, traceable and compatible with existing laboratory equipment.

This is one of the approaches pursued by Netri, which develops its devices using standardised multi-well plate formats to facilitate their integration into automated cell culture and imaging systems.

Standardisation also applies to the biological models themselves.

Cell origin, culture protocols, cellular maturity, materials, experimental conditions and data formats can all influence results.

This raises another fundamental question: who owns the data generated from human cells?

Donor consent, associated clinical data, ownership of results and interoperability are becoming increasingly important issues as these technologies move closer to industrial applications.

The development of organoids and organs-on-chips directly depends on access to high-quality human biological resources.

And this goes far beyond logistics.

Building diverse biobanks could make it possible to better represent human variability, including sex, age, genetic characteristics and population diversity.

Such diversity is particularly important when the aim is precisely to overcome some of the limitations associated with highly standardised biological models.

Access to cells is therefore becoming a scientific and industrial challenge, as well as an issue of strategic autonomy. During the session, the speakers highlighted the current reliance on cellular resources purchased abroad for certain applications.

Against this backdrop, France is now seeking to structure its ecosystem.

As part of the Strategic Contract for the Healthcare Industries and Technologies sector, an initiative dedicated to organoids and organs-on-chips is being co-led by BioValley France and France Biotech, with the aim of bringing together academic research, companies, industry and institutional stakeholders.

The mapping exercise presented during the session has already identified more than 130 French stakeholders involved in technological development, biological model creation or their use.

This work has also led to the publication of a white paper and the development of specific scales for assessing project maturity across three dimensions: technological maturity, biological maturity and context-of-use maturity.

The aim is to better identify technologies that are sufficiently advanced to move towards industrial or regulatory applications.

France has recognised academic teams, specialised companies and industrial players already engaged in these emerging approaches.

But the competition is international.

The United States is moving quickly on the regulatory front, while other countries are investing heavily in these technologies and their clinical applications.

The challenge is therefore to transform France’s scientific expertise into validated, standardised and industrially scalable methods.

This will require stronger coordination between stakeholders, secure access to cellular resources, harmonised biobanks, dedicated training, funding for validation projects and regulatory frameworks developed in cooperation with the relevant authorities.

Because the real breakthrough may not come from a single model capable of perfectly reproducing the human body.

It may instead emerge from a combination of models that are sufficiently precise to answer the right biological question at the right stage of development.

Organoids, organs-on-chips, data and artificial intelligence could therefore contribute to making preclinical research more human-relevant, more predictive and, ultimately, more effective.

Article based on the session dedicated to organoids, organs-on-chips and the structuring of the French ecosystem, presented at Forum LABO.