Programmed yeast gives researchers new control over cancer immunotherapy

The immune system is complex in itself, and its encounter with cancer cells makes the picture even more challenging. Immune cells try to recognise and kill cancer cells, while cancer cells attempt to evade or deactivate them. This becomes a constant information war, in which cancer cells alter or obscure the signals that immune cells use to recognise them.

Immunotherapy strengthens the body’s own immune system in this battle – and is a rapidly advancing field of research.

For some cancer patients, tumours disappear within months of immunotherapy. The body’s immune system is boosted by the treatment, launches an attack, and the cancer cells are destroyed.

For others, the opposite happens. The treatment works at first – and then loses its grip again, without it being clear exactly why.

This is one of the biggest challenges in immunotherapy. And without understanding when and why a treatment loses its effect, it is difficult both to improve existing therapies and to predict who will become cancer-free – and who risks relapse.

This incomplete understanding is a critical barrier to the development of the next generation of immunotherapies.

When cancer cells change the signal, the immune system loses its grip

Part of the explanation lies in the fact that cancer cells constantly change the small “labels” on their surface that the immune system uses to recognise them. These labels are called antigens and act as the signals immune cells use to decide whether a cell should be attacked or ignored. When the signal becomes weaker or unclear, the immune attack can be weakened – or stop altogether.

“One of the hardest things about immunology is that the immune system is both highly complex and constantly changing,” says senior researcher Michael Krogh Jensen from the Technical University of Denmark.

“If we can test hypotheses in a system where we know exactly what immune cells are responding to,” says postdoctoral fellow Marcus Deichmann, “we can begin to see which treatments make sense – and which do not – in different scenarios.”

That is precisely why a research team led by Marcus Deichmann and Michael Krogh Jensen has developed a laboratory tool that gives researchers control over which signals immune cells encounter when they attack cancer. By programming yeast to present selected human antigens on their surface, the researchers can control both the type and the strength of the signal.

When the target keeps moving, treatment becomes difficult to test

The fundamental problem is that the target itself changes while researchers are trying to measure the effect. The answer grew out of postdoctoral fellow Marcus Deichmann’s genetic engineering work with yeast cells.

“What I had in my hands was a system with almost 100 percent control,” he says. “You can control the behaviour of yeast cells extremely precisely – and then I thought: why not use that for something that is otherwise extremely complex, such as cancer research and immunology?”

“We knew from the beginning that it was somewhat controversial to genetically modify yeast so that it could communicate with human cells when we started in 2021 – but the initial experiments indicated that it was possible,” says Deichmann.

The idea was simple. Instead of testing possible solutions in a system where everything changes at once, the researchers built a yeast-based model in which they could isolate and control one biological variable at a time – namely the density of antigen on the cell surface – and thereby see exactly what that single change does.

One variable at a time provides control

The yeast cells are programmed to mimic cancer cells by presenting human cancer antigens on their surface. The researchers can control how much antigen is present on each individual cell – from none at all to very high levels – allowing them to simulate different cancer scenarios.

Only the signals defined by the researchers themselves are present, and in precisely controlled amounts. In its simplest form, the yeast functions like a volume knob: the more antigen, the stronger the activation signal received by the immune cells.

“We have worked deliberately to identify the control knobs in the cell that allow us, almost with sniper-like precision, to regulate how many molecules sit on the surface of the yeast cells,” says Michael Krogh Jensen.

“A cancer cell is a bit like a Jackson Pollock painting,” says Deichmann. “Everything is happening at once.”

A deliberately simplified system

It was precisely this frustration that sparked a new way of thinking.

“And then it struck me: in other parts of my work, we use yeast cells precisely because they are extremely controllable and can be efficiently adapted for new biotechnological applications,” says Deichmann. “Why do we accept complexity here – when we actually have the genetic tools to tweak one thing at a time? Unlike a cancer cell, we can treat the yeast cell as a blank canvas.”

Seeing the yeast cell as a “blank canvas” also means that the model deliberately strips away much of the cancer cell’s complexity. That is exactly the point: first to understand what changes in antigen alone do to the immune response – before moving on to more complex systems.

In the long term, the platform can be used for faster and cheaper screening of new immunotherapies and to uncover where the boundary between effect and risk actually lies.

“If we start from something that is controllable,” says Michael Krogh Jensen, “we can begin to simulate where the therapeutic window lies.”

The therapeutic window is the range in which immune cells are sufficiently activated to effectively kill cancer cells – but not so strongly that the treatment either triggers severe side effects or pushes the cells into a dysfunctional state. In short: enough pressure to knock down the cancer – without causing dangerous reactions or immune burnout.

When the signal tips: from effective treatment to risk

CAR-T cells are a form of immunotherapy in which a patient’s own T cells are genetically equipped with an artificial recognition receptor – a kind of new “sight” that allows them to find and kill cancer cells carrying specific antigens on their surface. This receptor is known as a chimeric antigen receptor (CAR).

It is a simple and powerful concept – and a therapy with major potential, as seen in the treatment of certain blood and lymph cancers.

In practice, however, the reality is far more complex. There is substantial biological variation in antigen levels – between patients, within the same cancer type, and even within a single patient. In addition, cancer cells are not static targets, but systems in constant change.

When CAR-T cells attack, a process of biological selection occurs. The cancer cells that are easiest to recognise are eliminated first, while those that are hardest to recognise escape and gain the greatest chance of surviving and eventually dominating – simply because they are not detected in time.

“A selective pressure arises in which cancer cells with low antigen levels can escape the therapy,” says Marcus Deichmann. “In addition, our results suggest that some cancer cells can actively adapt their surface profile to avoid the CAR-T cells they encounter.”

When too much signal becomes dangerous

At the other end of the therapeutic window lie severe side effects caused by overactivation of the CAR-T cells.

“It is not ‘the more signal, the better’,” explains Deichmann. “Excessive activation can overstimulate the immune system, triggering serious immune reactions that require intensive treatment – or push the cells into a poorer functional state, where they lose the ability to effectively kill cancer cells, even though they still recognise the antigen.”

This is one of the main reasons why the effects of CAR-T therapy are so difficult to predict.

“That is why it does not make sense to test only a single antigen level,” says Deichmann.

SCASA reveals the limits of immunotherapy

Using the newly developed yeast technology – called the SCASA system (Synthetic Cellular Advanced Signal Adapter) – the researchers were able to measure activation, function and the limits of CAR-T cell responses. In other words, how strongly the cells are activated, how well they function, and when the response tips as the amount of antigen changes.

“We were able to simulate the entire spectrum of situations that CAR-T cells encounter in real cancers that manipulate their antigen levels to escape therapy,” says Marcus Deichmann. “Not just one idealised level – but the entire grey zone.”

The results confirm that CAR-T cell responses do not behave like a simple on/off switch. Instead, the response changes step by step depending on antigen density.

“It is not that immune cells either respond or do not respond,” says Deichmann. “Even small changes in antigen levels can shift CAR-T cells from effective to weak activation.”

Not all CAR-T designs respond the same way

A central observation was that different CAR-T designs used in approved immunotherapies respond differently to the same levels of cancer antigen.

“Two CAR constructs that look similar on paper can behave quite differently,” says Michael Krogh Jensen. “And that difference can be the difference between a treatment that works – and one that either overreacts or loses its effect.”

The system therefore allows researchers to estimate when new CAR-T candidate designs tip from effective responses to either too weak or uncontrolled activation – before they are tested in animals or patients.

Overall, the results confirm that antigen level is not merely a technical detail but a decisive biological factor. It can determine whether CAR-T cells hit the right level – or whether they act too weakly or respond too aggressively.

Why yeast is a more robust model than cancer cells

The researchers arrived at the same biological conclusions using yeast as with cancer cells – but the yeast model proved far more robust.

In activation experiments, cancer cell models lost up to 98 percent of their antigen in a short time or were killed entirely, while the yeast cells remained stable. This made it possible to measure CAR-T responses without the target cell changing its behaviour during the experiment.

“It gives a completely different robustness to the data when you know that the only thing that changes is the parameter you adjusted yourself,” says Marcus Deichmann.

“Yeast is extremely cheap and robust to work with compared with many other platforms,” he adds. “You do not have a small, limited amount of expensive material – you can always grow more yeast, and you can genetically adapt the model for new studies.”

Traditionally, researchers use cancer cell lines, microspheres or nanotechnological particles to study immune responses such as those from CAR-T cells. But living cancer models often contain either too much biological noise or too little experimental control for systematic investigation.

Cheaper, scalable – and without biological noise

Nanotechnological models can effectively reduce biological noise and provide a high degree of control, but they are often very expensive, require specialised equipment and expertise, depend on sensitive purified cancer antigens, and can typically only be produced in microgram quantities – that is, in very small amounts that quickly become a practical limitation.

Yeast, by contrast, can be grown rapidly in large quantities, is inexpensive to work with, and requires only simple equipment and basic growth media.

The researchers estimate that the material costs for experiments involving immune cell activation can be up to 1,000 times lower when using yeast compared with nanotechnological alternatives.

A central question for the success of the yeast-based approach was what would happen when yeast and human cells were mixed – and whether this could be done without triggering unwanted immune reactions against the yeast itself.

“So far, we have not detected that pure yeast without human antigen triggers a disturbing response in isolated human T cells – and that is precisely what gives the method a potentially completely clean starting point,” says Marcus Deichmann.

When yeast cells are brought into contact with human T cells, an immune response occurs only if the researchers themselves have added an antigen.

One of the team’s next goals is to examine in more detail – and confirm or rule out – whether yeast truly does not induce any alternative immune reactions on its own.

“The result is an artificial but fully controlled test system in which immune cells respond only to signals we have deliberately introduced,” says Deichmann.

From trial-and-error to more predictable immunotherapy

“This is a preclinical development tool,” explains Michael Krogh Jensen. “It does not provide a new treatment tomorrow – but it can make the process of identifying better CAR-T designs much faster and more systematic.”

The SCASA system was not developed as a treatment in itself, but as a tool to understand and design future immunotherapies more rationally. The aim is to develop design principles for immune cell therapies and to systematically measure where the window lies between no effect and dangerous overactivation – so that new designs can be guided toward maximum effectiveness.

“When antigen presentation becomes controllable, we move from iterative trial-and-error to systematic and comparable exploration,” says Marcus Deichmann. “With a controllable model, it becomes possible to test many scenarios quickly in the early drug discovery phase of new therapies.”

One of the most immediate impacts lies in the preclinical phase. Today, the development and testing of CAR-T therapies is extremely demanding and costly. The treatment itself is complex, and as an illustrative example, the cost for treating a single patient can often run between half a million and one million US dollars. Extensive preclinical studies are carried out before patients are treated – and this is where yeast cells could dramatically change the equation.

The work is covered by a patent application, and the researchers are now exploring how the platform can be further developed – not as a treatment in itself, but as a tool that can make future immunotherapy more predictable, more controllable and ultimately safer.