In this interview with Bruce Levine Ph.D. (University of Pennsylvania, PA, USA) we find out about his talk on constantly enhancing and improving engineered T cells, which was presented recently at the American Association for Cancer Research (AACR) Annual Meeting (March 29–April 3, 2019, Atlanta, GA, USA). In addition, we also asked about his thoughts on two CAR-T studies that caught the news headlines at AACR 2019.

Could you give us an overview of your talk at AACR on enhancing and improving engineered T cells?

I started by describing CAR-T cells as a different type of drug that is targeted and programmed. Importantly, raw material derived from patients exhibits variability. So how do we learn how to set critical quality attributes and how do you think about dosing and kinetics for designing a drug where the dose of cells that is infused does not necessarily, or frequently does not, correlate with the clinical response? What we now know does correlate with the clinical response is the post-infusion maximum concentration of CAR-T cells achieved and the area under the curve for the first 28 days. Therefore, we have characterized cells by looking at the phenotype of the source cells from the patient, and the final product. We can get a phenotypic flavor for what correlates with clinical response, but there is substantial overlap when you look at things like PD-1 expression and central memory phenotype versus terminal effect or phenotype.

How do we learn how to set critical quality attributes… where the dose of cells that is infused does not necessarily, or frequently does not, correlate with the clinical response?

Again, there is a high degree of overlap and it’s likely that a combination of factors gives us a better indication of manufacturing feasibility and the potency of the cellular product to be infused. Ultimately though it’s how those cells behave in the patient. But it brings up an interesting question, which is – for a drug that is currently expensive do you measure some combination and factors for potential patients before they are collected? So if the overall response rate in lymphoma for Kymriah® for example, is 50%, if you conduct testing and that comes back with a predictor that with this combination of phenotype, the patient looks like an 80% responder or this patient looks like a 20% responder, so does that then inform whom you might treat or what insurance companies might reimburse? That hasn’t come into practice yet, but it’s an interesting thought exercise with what we know about immunology in cancer biology. We have the tools to do this. It’s just how it would be implemented.

We are also working to improve the product in chronic lymphoid leukemia. There is a fairly low overall response rate, but we have conducted a clinical trial combining Imbruvica® (ibrutinib) in patients even failing ibrutinib and found that that vastly improves our overall response rate and that data presented by Saar Gill of Penn Medicine at the 2018 American Society of Hematology (ASH) Annual Meeting & Exposition (December 1–4, 2018, CA, USA). We have other things that we’ve been looking at in manufacturing such as shortening the time of manufacturing, for example,

Other areas of variability include looking at the viability. To address the viability, we had a poster at ASH 2018 looking at the quintiles of product viability and whether that related to clinical response and in fact it did not, and it’s similar to looking at the thinking about the cell dose. What’s important is what happens in the patient. The conclusions on the viability study, which we’ve now expanded and will be presented at the International Society for Cell and Gene Therapy Annual Meeting (May 29–June 1 2019, Melbourne, Australia), is that within the viability of ranges that occurred in these trials between 70% and 98%, no dose response relationship and no clinical response relationship was observed.

Another issue is how to rescue CAR-T non-responders- we have combined CAR-T cells with checkpoint antibody. We have a clinical trial in which we treated a number of patients and shown that we can rescue non-responders (NCT02650999). Kite/Gilead has a trial where they administered CAR-T cells simultaneously with checkpoint antibodies (NCT02926833).

Looking back to where we first infused cancer patients with CAR T-cells in 2010 to now global approvals in US, Europe, Switzerland, Canada, Japan and Australia, it’s really, really remarkable.

As an example of where things are going for some challenging targets, we have the setting of acute myelogenous leukemia (AML) and CD33. Unlike CD19, CD33 is present on the hematopoietic stem cells, so you don’t want to target that, as it could result in aplastic anemia. Saar Gill from the University of Pennsylvania has shown that CD33 can be edited out of the hematopoietic stem cell and the blood cells that mature from those added stem cells are resistant to CAR 33 killing and CD33 is not necessary for hematopoietic differentiation or function or myeloid cell function.

Find out more here.

The idea is create protected hematopoietic stem cells using CRISPR. First, we will collect hematopoietic stem from a donor, then using CRISPR targeting CD33, create CAR 33 knock-out cells from that donor, give the recipient (the patient) the CD33 CRISPRed-out stem cells and then give the CAR-CD33 T cells to target the AML.

To reflect on where we are, the week before I gave a talk at Kyoto University in Japan and met with clinicians that had been treating CAR T pediatric patients in their hospital. Two days later we heard about the approval in Japan by the Ministry of Health, Labor and Welfare, and more recently the payment was approved in Australia. Looking back to where we first infused cancer patients with CAR-T cells in 2010 to now global approvals in US, Europe, Switzerland, Canada, Japan and Australia, it’s really, really remarkable. This is not a one off. There are now two products approved in multiple indications and global approvals, with more on the way!

At AACR some interesting research emerged in the field of CAR-T, the first being that HER2 targeted CAR-T cell therapy showed promise in advanced sarcoma. Could you tell us your thoughts on this study?

The key takeaways from my point of view are that these are encouraging results, though they need to be validated in more patients. Secondly, the combination of the chemo regimens and the CAR T-cells are key, the relative contribution of each to the induction of clinical responses is uncertain. Thirdly, there is an interesting finding suggestive of “epitope spreading”, that is the initial kill released antigens that were recognized by B cells:  analysis of the patient’s blood at different time points showed antibody responses against several intracellular proteins involved in the cell cycle, cell growth, cell signaling, and in tumor processes, such as invasion and metastasis.

HER2 could be a challenge in that it is present at high levels on the tumor, but also at lower levels on other tissues. The researchers did combine with chemotherapy, so it’s also important to know what previous chemotherapy these patients had received and whether this was a novel combination that they received prior to the CAR T-cells.

The combination of the chemo regimens and the CAR-T cells are key, the relative contribution of each to the induction of clinical responses is uncertain.

It looks like the patients received multiple infusions. What we worry about with HER2 is on target, but off tumor toxicity. Now there are approaches to deal with this. One, we have worked on is investigating affinity tuning of the antibody portion of the CAR to HER2, so you can tune it so that it sees the bright expression and it doesn’t see the dim expression. Yangbing Zhao (University of Pennsylvania Perelman School of Medicine, PA, USA) did a very elegant series of experiments in a humanized mouse model to show that yes you can tune the CAR to differentially respond to high expression HER2 tumor. Other approaches are to target to particular tissues or with combinations of targets so that is more specific to the tumor, and again not on target, but off tumor.

Another key CAR-T story from the meeting highlighted mesothelin targeted CAR-T cell therapy as a possible treatment for solid tumors. Can you tell us your thoughts on this?

The mesothelin study at AACR is of interest for the clinical responses and also for directly injecting into the pleural cavity. The investigators did incorporate a suicide switch, but it doesn’t look like it was activated. The group at Sloan Kettering has published on local delivery in a mouse model and it is encouraging to see this has moved to a human clinical trial with nice activity. We have been testing mesothelin CAR’s since 2011 and have injected directly into tumor and intraperitoneally, though mostly intravenously. In our first paper, we did see epitope spreading (see figure 4), as appears to have now occurred in the Baylor trial testing HER2 CAR-T cells as well.

We’ve treated more than two dozen patients with different permutations of mesothelin CARs and different diseases, not only mesothelin but also pancreatic and a couple of patients with ovarian cancer. Again, this study presented at AACR looks encouraging. It validates what we have previously seen, although we have not combined with anti-PD1. These reports together confirm that there is clinical activity with CARs in solid tumors.