If you have an ACS member number, please enter it here so we can link this account to your membership. (optional)

ACS values your privacy. By submitting your information, you are gaining access to C&EN and subscribing to our weekly newsletter. We use the information you provide to make your reading experience better, and we will never sell your data to third party members.



Controlling CAR-T: How scientists plan to make the engineered T cell therapy safer, and work for more cancers

CAR T-cell therapy works wonders for some cancer patients. For others, it is a death sentence. To make the revolutionary therapy work for more people, scientists must devise better ways to control it

by Ryan Cross
May 7, 2018 | A version of this story appeared in Volume 96, Issue 19
A cartoon of CAR t cells.

Credit: Yang H. Ku/C&EN/Shutterstock


Last August, in a milestone decades in the making, Kymriah became the first cell therapy to be approved by the U.S. Food & Drug Administration. Marketed by the Swiss giant Novartis, Kymriah is made by genetically engineering an individual’s own immune cells to attack a form of leukemia.

In brief

CAR T-cell therapy, a treatment that reprograms a person’s immune cells to attack cancer,works wonders for some people with rare blood cancers. But the highly personalized medicine also comes with risks: It can be extremely toxic and, in certain cases, even lethal. If there’s to be hope of using CAR T cells on more common cancers, scientists will need to control the living therapy. Read on to learn how biotech companies and academic researchers are installing control systems to help reduce CAR-Ts’ toxicity and drive the engineered cells into more kinds of cancer.

The approval was based on a groundbreaking clinical study that showed the drug could wipe out the blood cancer in people who had already gone through many unsuccessful rounds of traditional therapy. Just three months after receiving Kymriah, 83% of these people were in complete remission; doctors couldn’t detect any cancer remaining in their blood.

“This was like landing on the moon for this field,” says Usman Azam, former head of Novartis’s cell and gene therapies unit. “It was hard to believe at first, but you can’t argue when you see these children and adults alive and well.”

Kymriah was the first success story for a new kind of treatment called chimeric antigen receptor T-cell therapy, or CAR-T therapy. And its breathtaking results set off a land grab: The same month that Kymriah was approved, Gilead Sciences agreed to pay $11.9 billion for Kite Pharma, which subsequently received FDA approval for its own CAR-T therapy, Yescarta. Later, Celgene paid $9 billion to acquire Juno Therapeutics, a longtime darling of the CAR-T field that doesn’t have an FDA-approved therapy. Meanwhile, venture capitalists have sunk hundreds of millions of dollars into biotech firms with next-generation CAR-T technology.

Unlike conventional drugs that are manufactured in bulk, Kymriah and Yescarta must be made new for every patient. The process starts by removing a person’s T cells, immune system assassins that scour the body in search of pathogens. Scientists then insert a gene for a synthetic protein called a chimeric antigen receptor, or CAR, into the T cells. That added receptor allows T cells, once they are reinfused into the patient, to spot and destroy tumors.

“As recently as a couple of years ago, people would have said CAR-T was like science fiction—or at a minimum didn’t have a commercial appeal,” says Brad Loncar, a biotech investor specializing in immunotherapy. “It couldn’t have materialized more quickly.”

But CAR-T therapy has a ways to go before it lives up to the hype. Kymriah and Yescarta treat only people with rare blood cancers. And personalizing medicine comes at a price: A one-time injection of Kymriah costs $475,000. Add in the related hospital care, and the overall cost of CAR-T therapy can approach $1 million. Health insurers are still grappling with how to pay for that. Although it’s early in Kymriah’s adoption, the therapy brought in only $12 million in sales last quarter.

Investors, doctors, and patients are tempering their expectations for other reasons, too. In the process of killing cancer cells, CAR-Ts can also cause systemic inflammation and brain swelling, side effects implicated in the deaths of more than a dozen people. The expense and risk have made CAR-Ts the treatment of last resort.

While researchers fight to rein CAR-Ts’ toxicity in rare blood cancers, they are also attempting to unleash the therapy on more common solid tumors, such as breast and lung cancers. Doing so will require engineering even more potent cells and possibly treading a fine line between killing cancer and killing the patient.

That’s why a growing number of biotech companies and researchers are devising control systems to make new CAR-T therapies simultaneously safer and more powerful. Installing emergency off switches, adding drug-controlled dials to tune T-cell activity, and building more selective sensors are all in development. Call it CAR-T 2.0.

“We have to figure out as a field how to move these cell therapies from last line to first line in blood cancers—and eventually to all sorts of cancer,” says Michael Gilman, CEO of Cambridge, Mass.-based Obsidian Therapeutics. “But in order for that to happen, these therapies have to be tamed. They have to behave like pharmaceuticals where doses can be controlled and sensitively managed by everyday physicians.”

A chart showing five steps of making and using a CAR-T therapy.
Vein to vein

Manufacturing a CAR-T therapy is an elaborate, personalized process that usually takes two to three weeks from extraction to treatment.

Sources: Novartis, C&EN

The kill switch

Whereas traditional drugs are ephemeral—they begin to break down as soon as they enter the bloodstream—CAR-Ts are a living therapy that multiplies exponentially once the cells spot their cancer target in the blood. As CAR T cells attack tumors, they also release molecules called cytokines, which promote inflammation and recruit even more immune cells to the attack. When this happens too quickly, it can cause a life-threatening immune flare-up called a cytokine storm.

“It’s a runaway response,” says Travis Young, director of protein sciences at the California Institute for Biomedical Research (Calibr). “There’s no way to control if that patient will have a 100-, a 1,000-, or a 10,000-fold expansion of their CAR T cells.”

Doctors currently dampen cytokine release by giving patients corticosteroids and a rheumatoid arthritis drug called tocilizumab. In the event of extreme cytokine storms, however, having an emergency kill switch to eliminate all CAR T cells could be a lifesaver. In fact, clinical trials of multiple therapies are under way that feature such switches.

One of the most advanced comes from Houston-based Bellicum Pharmaceuticals. Bellicum’s safety switch uses two engineered proteins located inside the CAR T cell that dimerize when exposed to a small-molecule drug called rimiducid. Rimiducid activates a protein called caspase-9, which kick-starts the process of CAR T-cell suicide.

Bellicum hasn’t tested its suicide switch in a CAR-T clinical trial yet, but the switch was installed in T cells used in a hematopoietic stem cell transplant trial for blood disorders. FDA put that trial on hold in January after three patients experienced encephalopathy, or brain damage, potentially related to the transplant. A patient whose doctor decided not to use the rimiducid safety switch died, according to Bellicum CEO Rick Fair. The other two, whose doctors opted to use the switch, are alive. FDA lifted its hold on the study in April.

San Diego-based Poseida Therapeutics is also using a switch that is activated by rimiducid. “We like it because it’s fast,” says CEO Eric Ostertag, potentially eliminating some or all CAR T cells within minutes to hours.

An alternative to the small-molecule safety switch is one that uses an antibody. The French company Cellectis tags its CAR T cells with a protein that the cancer antibody rituximab—marketed as Rituxan—can bind. A rituximab injection should cause a patient’s normal immune cells to target and kill the CAR T cells. “The problem with the antibody-mediated switches is that they take days to really work,” Ostertag says.

Last year one man died in a Cellectis trial after developing a cytokine storm. Cellectis CEO André Choulika says there was no time to administer rituximab to that man, but it has been deployed to remove CAR T cells from other people not experiencing severe cytokine release. “We are working on new versions that can be activated or deactivated with small molecules,” he adds.

Poseida, which recently began Phase I trials testing its CAR-T therapy in multiple myeloma, hasn’t had to administer its safety switch yet, Ostertag says.

It’s too early to know if any of these switches can reverse deadly episodes of CAR-T toxicity, and not everyone is convinced the switches are necessary. “I don’t think it is essential for the current generation of CAR-Ts,” says Kristen Hege, vice president of translational development for hematology and oncology at Celgene. “The whole medical community is getting more comfortable in managing these toxicities. But if you can keep all the efficacy and dial down the toxicity, that would be an improvement.”

T cell tuners

Kill switches are a reactionary form of safety, but Bellicum has a second CAR-T design that could help prevent severe toxicity from happening in the first place. It’s a version of CAR T cells with a default state set to off, engineered to turn on only when the CAR binds a cancer antigen and an activating small molecule drug simultaneously. These so-called GoCAR-Ts should allow doctors to increase T-cell activity and proliferation by raising the dose of that drug, which is rimiducid.


Bellicum rewired its GoCAR-Ts to be activated by rimiducid, rather than killed by it, as in the suicide-switch-containing cells. This design could allow doctors to scale back CAR-T activity by reducing or halting administration of rimiducid. But the main goal for GoCAR-Ts is to drive the therapy into solid tumors.

“Most CAR-T studies fail because nothing happens,” says Alan Musso, Bellicum’s chief financial officer. While CAR T cells can easily spot cancer in the blood—sometimes so quickly that they cause cytokine storms—the cells often burn out before they can find and launch an attack on solid tumors buried deeper in the body. Bellicum hopes to change that by using rimiducid to repeatedly pump the gas on GoCAR T cells as they search for solid tumors. Its most advanced GoCAR-T, BPX-601, is in a Phase I trial involving pancreatic cancer.

Designing CAR T cells that can be turned on only by adding a small-molecule drug is an increasingly popular strategy for biotech companies. But their tuning dials differ.

One of those companies is Obsidian, which was founded in 2015 and is based on the discoveries of Stanford University researcher Thomas Wandless. A decade prior, Wandless was studying mutant versions of proteins that didn’t properly fold and were subsequently destroyed by the proteasome, the cell’s protein shredder. Wandless found that small-molecule drugs could coax the proteins into the right shape, thus stabilizing them.

A diagram showing Bellicum's kill switch.
Installing a kill switch

Bellicum Pharmaceuticals installs an emergency off switch in its CAR-T therapies by adding the gene for an engineered protein made of caspase-9 and a binding domain for the small-molecule drug rimiducid. When T cells are exposed to rimiducid, a pair of caspase-9 proteins dimerize, kick-starting the process of cell death.

Sources: Bellicum Pharmaceuticals, C&EN

As stories of CAR-Ts’ success began emerging, Wandless wondered if these misfolded regions of proteins, called destabilizing domains, could be used to better control the therapies. “I just thought it was the perfect test case,” Wandless says.

In theory, engineering the CAR to feature a destabilizing domain would allow doctors to use a small molecule to adjust the activity of the therapy. Without the small molecule, CARs, which are constantly being manufactured by CAR T cells, would be sent to the proteasome and wouldn’t make it to the surface of the cell. Add a stabilizing small molecule, and the CAR would stick around, allowing the cells to enter attack mode.

Obsidian, which in December revealed it had raised $49.5 million in funding, is taking human proteins and mutating them to have destabilizing domains that can be returned to a stable form by binding common small-molecule drugs. One program is using the domains to control CAR expression. Obsidian’s other programs could help turbocharge CAR T cells to better target solid tumors.

A diagram illustrating the synNotch system.
Improving CAR-Ts’ self-driving sensors

When Cell Design Labs' CAR T cell is in the right environment, a receptor called synNotch is activated, prompting CARs to be produced on the surface of T cells. That allows the CAR T cells to attack cancer cells while leaving healthy cells that display the same markers alone.

Sources: C&EN, adapted from Science 2018, DOI: 10.1126/science.aar6711

Solid tumors release a slew of molecules to suppress the immune system and shield themselves from T-cell attack. Interleukin 12 (IL-12) is a potent cytokine known to counteract the tumor’s immunosuppressive environment.

It’s too potent to be taken as a drug, but letting CAR T cells produce their own IL-12 could be a safer way to overcome a tumor’s suppressive barriers. Obsidian is using destabilizing domains to carefully produce and release IL-12 in CAR T cells, as well as another protein called IL-15, which keeps T cells proliferating longer.

Researchers are intrigued by CAR-Ts’ ability to both target cancer itself and enlist the help of self-produced drugs. “CAR T cells really offer so much more than a small molecule or antibody,” says Laurence Cooper, CEO of Boston-based Ziopharm Oncology, one of several other companies developing CAR T cells that produce either IL-15 or IL-12. “These cells really have the ability to adapt to the patient’s needs, to essentially be an autonomous, little robot in your body,” he adds.

Self-driving cells

Some academic researchers are taking an even more complex view of what these cellular robots can do in the body—and are finding themselves being courted by investors as a result. One of those scientists is Wendell Lim, a synthetic biologist at the University of California, San Francisco.

“The way I think about CAR-Ts is a lot like real cars,” Lim says. “We still want the ability to control the gas and brakes on these cells but also have much more sophisticated autonomous control.”

In 2015, Lim published a study demonstrating an advanced on switch, now known as Throttle, in which a small-molecule dimerizer binds to the CAR protein itself to activate the T cell. That year, Lim formed a company called Cell Design Labs based on Throttle and another control system he would publish later. To improve CAR-Ts’ autopilot abilities, Lim devised a flexible sensing system called synNotch. It identifies a signal outside the T cell that activates an engineered pathway inside the T cell to alter its behavior (Cell 2016, DOI: 10.1016/j.cell.2016.09.011).

Peter Emtage, former vice president of synthetic immunology at the genetic engineering firm Intrexon, was watching Lim’s work. When Cell Design Labs launched, “I decided to jump on board for what I considered to be CAR-T 2.0,” Emtage says.

The two approved CAR-T therapies, Kymriah and Yescarta, target an antigen called CD19, found on B-cell cancers. Unfortunately, the CD19 protein is also located on healthy B cells, immune cells that produce protective antibodies. When CAR-Ts work, they kill all CD19-bearing cells, and thus all B cells. Because people can receive regular infusions of the antibodies that their missing B cells would normally make, doctors consider this collateral damage acceptable. But few other targets can be depleted wholesale without widespread harm.

Emtage, who became Cell Design Labs’ chief scientific officer, thought that combining synNotch and CAR in a single CAR T cell could provide a way to more selectively target solid tumors. For example, Emtage explains that scientists would like to make CAR T cells that target a protein called mesothelin, which sits on the surface of many tumors, including some ovarian cancers. But mesothelin is also found on healthy tissues like the lungs.

One way around this issue is to make sure the mesothelin-targeting CAR is produced only when CAR-Ts are in the ovaries. By first crafting synNotch to bind a protein found only in the ovaries and then linking that binding to the transcription of mesothelin-targeting CARs inside the cells, scientists can make CAR-Ts attack mesothelin-flagged cancer cells in the ovary while sparing healthy cells. The example is hypothetical, but Emtage says the concept is not far fetched.

In December 2017, less than two years after Cell Design Labs launched—and without a single program in the clinic—Gilead announced it would acquire the start-up in a deal worth up to $567 million. “It speaks to the unique time in this field,” Lim says.

Ideas for how to control CAR T cells are quickly evolving, making the marketed therapies seem almost rudimentary in comparison. For example, Peter Yingxiao Wang, a bioengineer at the University of California, San Diego, has engineered CAR T cells that remain turned off until a mechanically sensitive ion-channel protein is activated by ultrasound (Proc. Natl. Acad. Sci. USA 2018, DOI: 10.1073/pnas.1714900115). He envisions one day equipping patients with a small ultrasound transducer above their tumors and granting doctors wireless control to turn on CAR-Ts only when they’re near the tumors.

Remote-controlled CAR-Ts probably won’t be tested in humans soon, but the concept is a testament to the gears turning in researchers’ minds as they consider how to expand the use of cell therapy into more cancers and other diseases.

“I think it is a natural progression to go from small molecules to biologics to living cells as therapies,” Lim says. “They all have capabilities and complexities, but for things like cancer, the intelligence and sophistication of a living cell is a huge boon.”

Universal control

As researchers continue to push CAR-Ts into more kinds of cancer, they are creating and testing new CAR proteins that bind different cancer antigens. But Calibr’s Young thinks there’s an easier approach. He has developed a universal CAR that, rather than binding the tumor antigen directly, binds a cancer-targeting antibody that links the CAR T cell to the tumor.

Young says his antibody switches, much like small-molecule-controlled CAR-Ts, would allow doctors to regain “traditional, pharmacokinetic control” of CAR T cells because increasing the dose of the antibody increases CAR-T activity.

Antibody-based switches have their critics: Some say they won’t work fast enough or that they will linger in the body too long. But those developing the technology say the appeal isn’t just safety; it’s in the potential to design broadly acting CAR-T therapies. By making new antibodies that bind the same CAR on one end and different cancer antigens on the other end, researchers hope such a therapy could treat many different cancers. In this universal CAR T cell, “the CAR is the hardware, and the antibody switch is the software,” Young says.

That’s an exciting prospect because, unlike cell-therapy engineering, antibody engineering is a staple of the biotech industry. In fact, a Cambridge, Mass.-based company called Unum Therapeutics is already testing a similar concept in multiple clinical trials, in which one kind of engineered T cell can bind different antibodies targeting different blood cancers. The firm’s first foray into solid tumors will use the FDA-approved antibody trastuzumab to guide T cells to the cancer antigen HER2, which is found in breast and gastric cancers.

Unum’s universal receptors have another advantage too. “We are exploring cocktails of antibodies that simultaneously hit multiple different antigens,” Unum CEO Chuck Wilson explains. That could be helpful for tackling solid tumors marked by more than one antigen.

At the end of April, Wilson Wong, a former postdoc in Lim’s lab, published a paper that took this concept even further. Wong now runs his own research group at Boston University, where he created a control system that uses a pair of universal receptors dubbed zipCARs, each of which binds a different antibody-based switch. The switches each target a different cancer antigen, and the T cell is activated only when both zipCARs are binding both switches (Cell 2018, DOI: 10.1016/j.cell.2018.03.038).

“I am going to be honest. It is the most complicated therapy that you can engineer at this point,” Wong says. “So it is definitely not the first thing people will try.” Regardless, he is pushing ahead to add even more controls to the system. Wong is a cofounder of Senti Biosciences, a cell-therapy company that launched this year.

Young and Wong share a bold vision for CAR-T therapy’s future. A growing area of investment is so-called allogeneic therapies, where the CAR T cells are made ahead of time from a donor’s cells rather than individually for each patient. “I see the CAR-T control field and the allogeneic cell-therapy field as two groups that need to come together in the future to create truly universal, off-the-shelf T-cell therapies that you can provide to any patient,” Young says.

A diagram of universal CAR receptors directed to tumors by antibodies.
A CAR-T for many cancers

Currently, CARs must be redesigned to target new cancer antigens, the markers on a cancer cell’s surface. Calibr's universal CAR binds an antibody linker instead, which in turn bridges the CAR T cell to the cancer antigen. The same CAR T cell can target different tumors by using different antibodies.

Source: C&EN

Such a therapy, which would be manufactured in bulk, could be directed to any cancer with the appropriate antibody switch.

Some researchers are careful to point out that overly complex systems will be difficult to turn into therapeutics. “The more bells and whistles you put into the system, the more ways there are for the system to fail,” says Yvonne Chen, a chemical engineer developing new CAR receptors at the University of California, Los Angeles.

Widespread adoption of CAR-T therapy still has many challenges: streamlining manufacturing, reducing the risk of deadly side effects, and expanding the currently small patient population, to name a few.

Yet just a few years ago, scientists “had real uncertainty about whether this could become a real therapeutic because of the logistical challenges,” says Glenn Dranoff, head of immuno-oncology at the Novartis Institutes for BioMedical Research. Scientists don’t know if CAR-Ts will ever be successful at treating more common cancers, but if the pace of innovation is any indication, hopes are high.

“Our grand hypothesis is that eventually, in 10 or 20 years, cell therapy will be the third pillar of medicine, alongside small molecules and proteins,” Obsidian’s Gilman says. “We’ve just seen the barest glimpse of what cell therapies are ultimately going to do for people.”


UPDATE: This story was edited on May 11, 2018, to state that Ziopharm Oncology is one of several companies developing CAR T cells that produce either IL-15 or IL-12.


This article has been sent to the following recipient:

Chemistry matters. Join us to get the news you need.