In the 1980s and early 1990s, close to 15% of patients treated with the early immunotherapy Interleuken 2 (IL2) saw their cancers go into remission. “It was amazing,” says K Dane Wittrup, Carbon P. Dubbs professor in Chemical Engineering and Biological Engineering at MIT. “What was not great was that every single patient who was treated with it almost died from it.”

Yes, 8% of those whose tumours responded to IL2 left cancer behind for years and even decades afterwards, but, with their blood vessels leaking and their bodies ballooning, the first place anyone went to was intensive care.

Cancers may capitalise on the fact – tricking, suppressing and even recruiting immune cells – but the human immune system limits itself for a good reason. Conditions like sepsis, which can kill in a matter of hours, arise when it doesn’t. Or take the way the body’s defences treat transplanted tissue. Doctors replace organs with the best of intentions, but, to the white blood cells it keeps pumping around the recipient’s body, a donor heart is an alien invasion. “If you take a non-matched organ and just implant it, the immune system can kill off several pounds of tissue remarkably quickly,” notes Wittrup. Without suppression, it’s powerful enough to end the life it exists to protect. Immunotherapies have to tap into that power without making it toxic. In oncology, that means retraining the immune system to recognise, target and destroy cancer cells, while trying to keep the rest of the patient’s body out of the crossfire. It is a spectacularly difficult balance, but over the past decade, we have begun to see just what can be achieved if we get it right. “Cancer adapts, but so can the therapy,” stresses Wittrup. “If you can initiate a response that’s vigorous enough, the immune system will track subsequent variations in the cancer cells.”

Wrecking balls

He describes two general approaches. Firstly, by administering cytokines like IL2, which activate T-cells, doctors can manipulate “nature’s way of stepping on the gas”. The problem with naked IL2 infusions is that they don’t just direct the immune system to accelerate in the direction of cancer cells. Overstimulated T-cells screech into blind corners and pedestrianised zones all over the body, crashing almost indiscriminately into whatever crosses their paths.

To avoid causing off-target damage, then, immunotherapies need to pay attention to the rules of the road. “The immune system is very orchestrated,” explains Wittrup, “it doesn’t turn on everything, everywhere, forever – it turns on certain things in certain places for certain times. And somehow, we just sort of entered into it as a community initially saying, well, let’s just soak the person in [cytokines] for as long as they can handle it. But that means you can't give them much because you just overactivated them, and it’s not how those very same molecules are used by the immune system.”

We’re getting better at mimicking that. Patients receiving CAR T-cell therapies have their own immune cells extracted and tuned to target specific antigens on cancer cells. Unfortunately, at present they can only be used against very specific B cell cancers, as antigen targets on solid tumours and other types of leukaemia also appear on important healthy cells that can’t easily be replaced. Even in approved use cases, any cancer cells without the target antigen escape, increasing the risk of mutation and recurrence. “And with a therapy that’s as potent and effective as a CAR T-cell, there is a risk of toxicity,” warns physician-scientist David Scheinberg, head of Experimental Therapeutics at the Memorial Sloan Kettering Cancer Centre – “severe toxicity”.

Copying one element of the immune system too closely can also subject therapies to the same pitfalls. Like their unaltered peers, expensively engineered CAR T-cells eventually stop killing cells with target antigens and shut down from exhaustion. This is a response to overactivation – another example of how the immune system limits itself. As Scheinberg points out, it’s a useful defence mechanism. “You don’t want your immune system to be overactivated all the time.”

Today’s best implemented immunotherapies take a different approach. In Wittrup’s terms, immune checkpoint inhibitors (ICIs) “take the brakes off” T-cells. They’re still liable to cause crashes – 66– 86% of patients experience immune-related adverse events, which can strike almost anywhere in the body – but over the past decade, anti-PD1 and anti-CTLA-4 antibodies have started to change how scientists think about tackling cancer. About 20-30% of patients with metastatic melanomas treated with ICIs experience a complete response. Therapy can even be discontinued at that point, as the risk of relapse is estimated to be less than 10% over the following five years. Outside of immunotherapy, that’s unprecedented.

“Instead of just shifting the survival curve a little bit at the middle point, but ending up with everybody in the same unfortunate outcome, now the phrase is ‘raising the tail’,” explains Wittrup. “[Doctors and researchers] don’t like to use the word, but these people seem to be cured. So now the problem becomes, ‘How do we make that a larger number?’ As opposed to, ‘I have no idea what might work’, now it’s ‘we know this is working, how do we make it work all the time?’”. Without increasing the toxicity, that is. It’s much harder to justify potentially life-threatening adverse effects when treating less aggressive forms of cancer.

Toxic waste

That’s where Wittrup and Scheinberg come in. Both are trying to build on the success of ICIs with immunotherapeutic techniques specifically designed to minimise toxicity. Wittrup calls it “spatio-temporal programming” – only giving the immune system the green light in specific circumstances. In his case, that means anchoring cytokines in the tumour space. Scheinberg, meanwhile, is converting CAR T-cells into “micropharmacies”, which distribute lethal small-molecule drugs directly to cancer cells.

That analogy is particularly useful for demonstrating how these approaches differ from the systemic way cancer drugs are usually delivered. Scheinberg’s ‘SEAKER’ (Synthetic Enzyme-Armed KillER) cells are directed to tumours by chimeric antigen receptors, but their primary mode of attacking cancers is by unmasking a separately infused small molecule prodrug called AMS, which was discovered by Derek Tan, chair of the Sloan Kettering Institute’s Chemical Biology Programme and Scheinberg’s collaborator for the project. Using specially engineered enzymes to activate highly cytotoxic molecules that the body can’t otherwise produce, these T-cells are able to destroy antigen-negative tumour cells and avoid the immune suppressing effects of the tumour microenvironment. The technology is being developed for human trials by CoImmune, which employs Scheinberg as a scientific advisor.

Importantly, the T-cell-directed targeting system also confines the toxic effects of the small molecule drug, which is too poisonous to administer systemically, to the tumour space. Like a good pharmacy, SEAKER cells ensure that the correct doses of the correct medicines get to the right place at the right time. Scheinberg even jokes about sending in tiny pharmacists to process prescriptions – it’s spatiotemporal programming the way healthcare systems already do it.

“At some point, the physician or nurse would infuse the prodrug,” he explains, “so you can time when you want this to be active. It could be a week later or a month later or multiple times. And you can do it once or you can stop. It gives control over when this is happening, which is different than a [traditional] CAR T-cell, which is always on.”


Close to this percentage of patients in the 1980s and early 1990s, treated with the early immunotherapy Interleuken 2 (IL2), saw their cancers go into remission.



The majority of patients experience immune-related adverse events, which can strike almost anywhere in the body – but over the past decade, anti-PD1 and anti- CTLA-4 antibodies have started to change how scientists think about tackling cancer.


The risk of relapse in the five years after metastatic melanomas therapy with ICIs is successfully discontinued

Nature Reviews

“As an oncologist for 30 years, I've seen all the problems with the therapies. The drugs are extremely toxic. They’ve vastly improved, but we have a long way to go.”

David A Scheinberg

That said, SEAKER T-cells that end up elsewhere in the body could also activate the prodrug, though Scheinberg notes that they haven’t caused notable toxicities in animal models. “The drugs, once made, are rapidly cleared out of the body anyway, so we think that the local accumulation at the tumour is what makes this effective without being toxic,” he explains. This is all still to be tested in human trials, but there’s a chance that the combinatory approach could also lower some of the toxicities associated with the current use of CAR T-cells. SEAKERs maintain their ability to activate prodrugs after their traditional immune functions have been exhausted, so the treatment may not require as many cells.

Cat and mouse

Just as impressively, Wittrup’s team has found that cytokines specifically engineered to stay in one tumour can train T-cells to attack others elsewhere in mouse models. Using protein evolution techniques, they first created IL-2 and IL-12 molecules that stick like Velcro to collagen, the main component of the tumour microenvironment, and (with Darrell Irvine, associate director of MIT’s Koch Institute for Integrative Cancer Research) IL-12s that use the century-old vaccine adjuvant aluminium hydroxide to anchor themselves at their injection sites. In effect, they installed extra brakes on the cytokines before using them to step on the gas in T-cells. The two techniques (which have been licensed to Cullinan Oncology and Wittrup’s own Ankyra Therapeutics, respectively, and are expected to enter clinical trials in the next year) should work with most if not all interleukins, and, at least in mice, they appear to deliver systemic efficacy with localised toxicity.

“There’s rapid communication among all the different immune cells in your body,” Wittrup explains. “The art going forward for us is to make sure that having generated this T-cell response, we support it sufficiently to get therapeutic effects systemically, but not so dramatically that we cross right back over to toxic territory.” The best way to do that? ICIs. “It doesn’t look terribly creative, because it is one of the best-selling drugs in the world, but anti-PD1 really works well to support those T cells when they leave the tumour.”

As well as making the ‘step on the gas’ strategy viable, then, the work of Wittrup and his colleagues has the potential to make ICI therapies effective for more patients, without creating extra toxicities. In tests led by Wittrup’s former student Noor Momin, cytokines that bound to collagen were even found to enhance the effectiveness of CAR T-cells. The immune system is made up of interlocking, synergistic elements, so why shouldn’t immunotherapies be the same?

“It’s a paradigm shift,” says Wittrup, connecting his work to the broader trend towards immunotherapy in oncology. “For so many years, the drug-makers’ job was defined as kill every cancer cell. That’s not it anymore. Now, it’s essentially to vaccinate. And you think about things differently if you’re making a vaccine.” It’s about working on the immune system’s own terms.

That’s what drew Scheinberg to immunotherapy in the first place. “As an oncologist for 30 years, I’ve seen all the problems with the therapies. The drugs are extremely toxic. They’ve vastly improved, but we have a long way to go. And we’re hoping that this might be one further step in that evolution.”