Improving the Efficacy and Accessibility of CAR Therapies

My lab is called the Therapeutic Innovation and NK cells (THINK) lab. We mostly work on natural killer cells and their application in cancer immunotherapy, although we also do still work on T cells – they’re both very interesting tools, suitable for different jobs.

Within our focus on NK cells, we have several project categories. Our biggest project focuses on optimizing the NK cell “chassis” – for example, removing checkpoints or factors involved in exhaustion, to improve them as therapeutics. To do this, we’re using a form of gene editing called base editing, which allows you to do multiplex edits to a cell without damaging it, by changing one base into another. In comparison, older tools like CRISPR can break the backbone of DNA, causing extreme translocations or inducing death by apoptosis. We can also achieve the same end by programming the cell as it grows, for example with small molecule compounds that push the cells down a certain path or lock them into a certain beneficial state. We hope that with some programming with small molecules or cytokines and some gene editing, we can create an ideal NK cell that can deliver a therapeutic transgene, such as a CAR.

We’ve also spent a lot of time working on NK cryopreservation. Unlike T cells, NK cells don’t cryopreserve very well – but if you can’t cryopreserve them, you don’t have a product. A lot of the trials that have been done so far with CAR NK cells have been single-center trials, where altered NK cells have been infused right away, as a fresh product, which can’t really be commercialized.

Beyond that, I think successful immunotherapy is rarely going to be single agent. I see a lot of benefits for combination immunotherapy: CAR cells and other agents. We’re just wrapping up a clinical trial, the first in the world to use CAR T cells with oncolytic viruses together, and now we’re designing our own oncolytic viruses to use with NK cells. Cancer has a wound healing phenotype that doesn’t usually attract NK cells. Instead, tumors tend to collect macrophages, then reprogram them to make the tumor worse. They also collect T regulatory cells, which can create a barrier to successful immunotherapy. But if you can infect the tumor with an oncolytic virus, you’ve created the perfect context to attract CAR T, or CAR NK cells, maximizing exposure to the target tissue.

CAR T cells are potentially very powerful cells, because they’re all very highly antigen-specific, but they’re almost too good at doing their job. They produce high levels of cytokines, replicate rapidly and can cause a lot of damage, resulting in potentially life-threatening side effects. The two most common are cytokine release syndrome (when too many cytokines are produced, causing an immune cascade), and immune cell-associated neurotoxicity syndrome (ICANS) (excess central nervous system inflammation and disturbance caused by excessive cytokine production). Unfortunately, this means that CAR T-cell therapy can only be given in tertiary medical centers (of which there are only around 30 approved centers in the US), to an admitted patient with available space in an emergency room. This limits the accessibility and availability of CAR T cell treatments, as does the cost of these treatments, and the fact that the T cells must be taken from the patient’s own body.

NK cells provide an alternative in several ways. Since NK cells don’t undergo large clonal expansions and produce a different cytokine profile from T cells, they are less likely to trigger such severe side effects. Additionally, NK cells don’t cause graft-versus-host disease, so they could be made into an allogenic therapeutic. There have been several ongoing projects using CAR NK cells against CD19 that have shown great results, but also show that they can be used as a safe allogenic therapeutic, without causing CRS or ICANS. Therefore they could potentially be given in outpatient centers or community hospitals, at a lower cost, making them far more accessible than CAR T-cell therapies.

In addition, T cells are so antigen-specific (either via their own T-cell receptor or a CAR) that if they don’t see their specific antigen, or the cancer mutates, tumor escape can occur. In comparison, NK cells have evolved to recognize dozens of different stress signals, such as the consequences of DNA damage and metabolic stress, which are almost universal in cancer. This potentially gives CAR-NK cells a very broad activity that could help prevent tumor escape. 

One of the shared challenges (though less relevant to NK cells because of their broad activity) is antigen selection, especially for solid tumors. We were lucky with the hematological malignancies, there were obvious targets in the B cell markers. In solid tumors, we’re targeting tumor-associated antigens that are over-expressed on tumors but are still present in healthy tissues. This leads to untargeted off-tumor toxicity, which can limit the dose of cell therapies.

We also have to deal with antigenic heterogeneity and antigen escape. Whatever your target antigen, expression won’t be uniform and the likelihood is that a percentage of tumor cells won’t be destroyed and can potentially repopulate. You need to engage multiple antigens to solve this problem, but that also potentially means stacking toxicities. We haven’t seen many CAR NK cells for solid tumors enter clinical testing yet, but one antigen that’s received attention recently is claudin18.2 (CLDN18.2) in pancreatic and esophageal cancers. CLDN18.2 is usually expressed in tight junctions, not really visible to the immune system. Tumors depolarize membranes, so they’re no longer neatly stacked, and CLDN18.2 becomes visible – making it a great antigen to target. However, even this isn’t foolproof; high doses of CLDN18.2 CAR T cells can cause gastric bleeds – an on-target toxicity.

We certainly have our work cut out for us to design effective CARs that don’t cause toxicity. We’re trying to make these CARs smarter, but we also need to make sure that once they’re inserted into the target cell, they don’t overload it, or domain swap with each other to cause tonic signaling (low level activation in the absence of target antigen), or cause any secondary malignancies.

There’s a lot that the regulators expect you to do pre-clinically in order to get an investigational drug application approved in the US (or a clinical trial application in the UK and elsewhere). One very important thing is to screen your CAR for off-target toxicity, to make sure it isn’t going to cross react with a second tissue or antigen and cause damage.

The clinical trial design provides a lot of levers to control safety – for example, starting with a low dose of CAR without lymphodepletion, with staggered administration and dose escalation. Lymphodepletion is usually needed to give physical space for your CAR cells to occupy, and to encourage the body to return to homeostatic immune cell levels by producing cytokines that encourage immune cell expansion. Without lymphodepletion, expansion of CAR T or CAR NK cells is very limited, reducing the chance of damaging side effects or toxicities. It’s also important to have treatment for toxicities readily available to mitigate any issues. Overall, patients are very carefully managed during trials.

One potential way of overcoming safety issues like toxicity is to design solutions into the cell therapy. CARs can be designed with on- or off-switches that can be manipulated using small molecule drugs or monoclonal antibodies. Kill-switches are also possible, which – when the controlling small molecule is administered – will wipe out any CAR cells. However, there is then no chance of patient benefit then because the cell therapy is eliminated. It's much better to have safety by design, rather than an emergency off-button.

There are a lot of options for safety management, but some of them are more precise than others. If our goal is to make these therapies accessible in outpatient clinics, we need something that is very smart, very safe and hopefully manages itself, without the need for a lot of expert physicians to control it.  

Like any class of drugs, CAR therapies don’t have a right to exist, they could be swept away by future innovations. We’ll always do what's simple and what works best, and some things are just too complicated. Recently, we've seen a big pullback from gene therapy, despite the curative promises, because the drugs were prohibitively expensive. So, we need to fix the financial toxicity of CAR cells as well as the medical toxicities.

One of the most exciting things happening at the moment is programming the patient’s own cells in vivo, by delivering CARs using viral vectors or mRNA and lipid nanoparticles. There are already multiple trials in progress that are looking very promising for B cell malignancies. For successful use of CAR therapries against solid tumors we need to really work on improving the cell chassis – like we’re doing in our lab – to make the immune cells more robust, more resilient to the immunosuppressive tumor microenvironment. This would be ex vivo cell therapy, but it could still be allogenic and off-the-shelf, which would greatly improve the cost and the safety.

There are still some challenges to overcome, but I really see in vivo CAR therapeutics as the future. Engineered, gene-edited therapies for solid tumors are going to be necessary in order to get the kind of successful, long-term results we need.