Interesting #AACR16 Abstracts and Themes

There were a number of AACR abstracts that I found interesting, and many more that just had provocative titles. Many of the symposiums and plenary sessions just had the speakers and titles of their talks, so I am sure there will be much more presented at AACR than what I can glean from the listed abstracts and titles. However, I decided to give my thoughts on a number of them, below. I kind of went down the rabbit hole, as I normally do, and made this blog post longer than I intended, so feel free to pick and choose the parts that seem most interesting. The AACR meeting starts April 16th, and while I unfortunately will not be able to go, I will absolutely be following the presentations and the coverage of the new research.

CARs that act like TCRs
Eureka Therapeutics: ET1402L1-CART, a T cell therapy targeting the intracellular tumor antigen AFP, demonstrates potent antitumor activity in hepatocellular carcinoma models (ADAP)
Eureka is a company that uses phage display to screen for antibodies that can bind to specific peptide-MHC complexes, so they can have the specificity of a T cell receptor (TCR), but in the modality of an antibody. This allows antibodies to "target" intracellular proteins. They are also developing an antibody against WT1 (called ESK1). This antibody was developed with the Scheinberg lab at MSKCC, where they made the original monoclonal antibody (described here), and more recently turned it into a CD3-bispecific antibody here. Not surprisingly, Eureka is also considering using these antibodies to make the scFv portion of a chimeric antigen receptor (CAR), and appears to have done so here, with their lead antibody ET1402, which apparently is able to target AFP. Interestingly, Adaptimmune also has an AFP program entering clinical trials, although they are using the more traditional engineered TCR approach to target it. I don't think it's clear whether it's better to target peptide-MHC complexes with normal TCRs, or with TCR-like antibody CARs (with costimulatory domains), so it will be interesting to see how it plays out. Eureka's pipeline (although very little is disclosed) is listed here.

CAR Design - new safety switches
UPenn: Reversible regulation of chimeric antigen receptor surface expression (NVS, BLCM, CLLS, JUNO, KITE)
Engineered T cells using CARs or TCRs have the potential for serious side effects, both on-target/on-tumor (CRS), and on target/off tumor (HER2 CAR lung tox), in addition to straight off-target (Titin cross-reactive MAGE-A3 TCR). There are numerous safety switch technologies different groups have developed to either remove the engineered T cells or, more recently, control CAR expression/function. The safety switch designs currently being used in the clinic generally rely on inducible removal of the engineered T cells, such as Bellicum's inducible caspase 9 approach, and Juno's truncated EGFR, although the latter has never been used as a safety switch in the clinic. In UPenn's abstract here, they wanted to design a CAR that could be inducibly down-regulated, to avoid irreversibly ablating the engineered T cells, which occurs with the previously mentioned strategies. UPenn's approach is to attach an inducible degradation component to the CAR construct. I believe the one they are using is this LID degradation system from Stanford published here. Basically, similar to Bellicum's approach, they use rapamycin or a rapamycin-related molecule to specifically bind to a protein, and in this case, reveal a degradation motif that targets the CAR to be degraded by the cell's machinery. From the abstract so far they have only tested it in culture, not in animal models, and they were able to achieve up to 80% reduction in the CAR levels by 24 hours, which they state was able to reduce the CAR-T cell's activity back down to near background levels.

There are a number of other approaches that try to tune CAR-T activity with destroying the engineered cells. One is Bellicum's inducible activation GoCAR-T approach, which uses their inducible dimerization construct to activate costimulatory signaling, as opposed to their CaspaCIDe construct, which activates cell death. Wendell Lim's lab previously published a system where they split the intracellular components of the CAR into two pieces that could be reconstituted into the full CAR after addition of a small molecule dimerizer (using a rapamycin-based molecule) to induce CAR activation. Cellectis also published a recent paper showing they could control that they could control the surface expression of CARs using novel extracellular architecture that could be regulated by a small molecule (also using rapamycin-based molecules).

These approaches are further distinguished from those that try to integrate multiple inputs to make smarter CARs that may be able to expand the potential pool of CAR-T surface targets. A recent example is a strategy from Wendell Lim's lab that was published in back-to-back cell papers which used a synthetic Notch receptor to control expression of a chimeric antigen receptor. Michel Sadelain's lab has previously published both systems that activate in response to the simultaneous presence of two antigens, or activate in response to cells that have one antigen but specifically lack the expression of a second antigen

There will also be a symposium at AACR specifically on next generation CAR approaches, which I expect will, at least in part, focus on the costimulatory domains, highlighted in these recent papers: 4-1BB vs CD28 (June), CD28 CAR + 4-1BBL surface expression (Sadelain).

TRK inhibitors face off (LOXO, RXDX)
Loxo and Ignyta are both presenting their early data for their TRK inhibitors in precision medicine clinical plenary session. While neither of the abstracts for their clinical data are available, both have published/presented case reports of clinical efficacy, as well as resistance (Loxo here & here; Ignyta here. Interestingly, there is a new abstract from Ignyta saying that the resistance mutations Loxo identified to Loxo-101 in the above presentation did not affect sensitivity to Ignyta's entrectinib (as well as other gatekeeper mutations in ALK or ROS1). This is a bit surprising, since one of those mutations was a G667 mutation, which was one of the mutated positions in the patient who became resistant to entrectinib in the paper above. It will definitely be interesting to see how the safety, efficacy, and durability of response shake out between Loxo and Ignyta for what appears to be a good target in the small % (although wide variety) of NTRK-rearranged tumors.

New MAGE-A3 TCR from NCI (KITE)
The NCI (in a CRADA with KITE) will be presenting data from its new MAGE-A3 TCR program in solid tumors in one of the immuno-oncology clinical trial plenary sessions. MAGE-A3 has been a difficult target for TCRs, with off-target neurological toxicity & cardiac toxicity in previous TCRs. This new TCR had some data presented previously showing a PR in 3/12 patients, and importantly no off target toxicity. MAGE-A3 is a cancer/testis antigen that are a class of antigens that appear to be good targets for engineered TCRs, with TCRs targeting another cancer/testis antigen, NY-ESO-1, already having some evidence of efficacy in the clinic in synovial sarcoma, melanoma (KITE), and multiple myeloma (ADAP). MAGE-A3 may also be expressed in a higher percentage of tumors than NY-ESO-1, and thus may have a greater addressable population.

Also, most likely just preclinical, but there will be a poster on an NCI/KITE KRAS TCR as well.

Immuno-oncology early trial updates (Genentech, KITE)
In the two plenary sessions for immuno-oncology trials, it appears there will be new clinical data for PD-1 (&CTLA-4) antibodies, but also some data from new targets in the second plenary session.

One presentation will be on an OX40 agonist from Genentech. OX40 is one of numerous additional T cell surface targets investigators are looking to target in combination with PD-1 or CTLA-4 antibodies. In contrast to those inhibitory receptors, where antibodies are used to block signaling, OX40 is a stimulatory receptor, and the therapeutic antibodies have been designed to activate its signaling

The other presentation will be on pegylated-IL10 from ARMO Biosciences. While there is no data in their abstract, I believe investigators have recently presented data from this program at the ITOC3 conference here (search pegylated), where they saw a 25% PR rate in RCC.

Kite will also be presenting updated CD19 CAR data in NHL, which I am sure will be closely watched.

Targeting Neoantigens (BioNTech, ZIOP, KITE)
Neoantigens are potentially personalized medicine to its most extreme for cancer therapy - directly targeting the presence of specific mutations unique to that patient's tumor. Neoantigens are thought to be a major target source for the endogenous immune response - reviewed here. Amplifying this response specifically, with vaccines or t cells (TILs or engineered TCRs), is a hot area of focus, with a sudden swelling of startups focusing on this approach. As far as I can tell, the only neoantigen vaccine presentations that will be at AACR are from BioNTech, which will present a number of neoantigen vaccines programs in Melanoma, GBM (collaboration with Immatics), and TNBC. While neoantigen vaccines are an interesting approach (and potentially easier to manufacture than neoantigen-TCR T cells), vaccines have had difficulty in producing robust efficacy in cancer, whereas neoantigen T cells (in the form of TILs) have had sometimes impressive efficacy (reviewed here). So I am reserving judgement on the potency of neoantigen vaccines until we start to see more clinical proof of concept.

While it appears Steve Rosenberg will only be presenting the MAGE-A3 TCR data, there are also two recent papers from his group showing they could prospectively identify neoantigen TCRs in peripheral blood by using PD-1 as a marker, and in a collaboration with Ziopharm, use a transposase-based system to make neoantigen-TCR T cells. Kite is also working with Rosenberg to develop neoantigen-specific TCRs. The key will be the feasibility of identifying, cloning, and expressing the neoantigen-specific TCRs, and I will be very interested in seeing how this approach is developed by the numerous groups pursuing it.

The best T cell composition for CAR-T products (KITE, JUNO)
There is also debate in what is the best T cell (or composition of T cells) for a CAR-T product. Kite Pharma will be presenting data saying that in NHL, neither the CD4:CD8 composition nor the differentiation state clearly correlated with response, while the final CAR-T product tends to be less differentiated than the initial T cell population. Paul Rennert reviewed this area in a very nice series of blog posts here. Alexey Bersenev also had a great post on this here.
 
Novartis EGFRvIII CAR-T in GBM : Phase I study of T cells redirected to EGFRvIII with a chimeric antigen receptor in patients with EGFRvIII+ glioblastoma (NVS)
While EGFRvIII seems like a great target for a CAR as it is surface protein and the target is exclusively found on the tumor, I do not believe there have been any significant responses reported for CARs targeting it. I will be interested in seeing if there have been any better responses, as well as the expansion, persistence, and trafficking to the brain (if data is available) for the CAR-T cells.

Targeting Ras
Ras has been a difficult, often called undruggable, but much sought after target in cancer due to its role in a large percentage of human tumors. Recently, there appears to be renewed emphasis on approaches to target Ras directly or downstream, and I will be interested in following what appears to be real progress in approaches to drug Ras.
There are two sessions on targeting Ras mutated tumors: Progress in Treating KRas Mutated Tumors session, and From Chemistry to the Clinic: Inhibition of K-Ras
While most of the abstracts have not been published yet, there are a number of approaches that are interesting:
Covalent inhibitors of Ras - one is described in an abstract here.
Trapping mutant KRas G12C in its inactive state - abstract (title only), paper.
Allosteric inhibition of Ras - abstract (title only), paper.

BET-bromodomain Inhibitors in NUT-midline carcinoma
There seem to be a ton of BET-bromodomain inhibitors in development, starting from JQ1, and now there seem to be multiple beginning to generate clinical data, primarily in NUT-midline carcinoma, where tumors appear to be especially dependent due to frequent NUT-BRD4 fusion events.
In a clinical trials minisymposium at AACR there will be data from: GSK525762, a selective bromodomain (BRD) and extra terminal protein (BET) inhibitor: results from part 1 of a phase I/II open-label single-agent study in patients with NUT midline carcinoma (NMC) and other cancers 

Previously there has been clinical data presented on a compound from Tensha: Clinically efficacy of the BET bromodomain inhibitor TEN-010 in an open-label substudy with patients with documented NUT-midline carcinoma (NMC)

There was also just recently a publication on a Merck-licensed compound that also had activity in NUT-midline carcinoma - here in Cancer Discovery.

Dissecting the complex ecosystem of malignant tumors with single cell RNA-Seq
While there was no abstract for this title, this is an approach that seems like it could significantly improve the resolution with which we can look at all the different cell types simultaneously and separately in a tumor. One of my favorite papers from last year used a cytolytic gene expression signature from large genomic data sets to see what tumor properties correlated with what appeared to be an immune response. I would imagine this analysis could be even more powerful in identifying ways in which the tumor and the immune system co-evolve if we had resolution of all the different cell populations instead of looking at a single tumor signature for both the tumor and all the different stromal cells. Update: The paper for the above presentation appears to just have been published in Science, here.
  
Single-cell TIL profiling (JUNO)
Juno does not appear to be presenting much work on CARs at this AACR, but I actually found this non-CAR abstract very interesting. This approach from Juno seems like it could be very powerful for identifying what is going on in individual TILs as at the single cell level, they claim to be able to sequence the TCR, while also getting a snapshot of that T cell's transcriptome and proteome. I could imagine this level of resolution could lead to the identification of additional T cell pathways of activation or repression in different T cell populations in the tumor, such as PD-1+ TILs or neoantigen-reactive TILs.

Identifying novel immunotherapy targets from human tumor data from PD-1 & CTLA-4 antibody trials:
There are so many potential immunotherapy targets for either monotherapy or combination with checkpoint blockade that it can be overwhelming. An approach I like that may help sort out which are the critical targets is starting with data sets generated from human tumors, particularly the differences in responders versus non-responders to checkpoint blockade. In this way, we can try to let what's going on in human tumors tell us what is important to escape immune surveillance. One can see this approach in a number of abstracts, such as:
The tumor immunity continuum as a framework for rational combinations or Immunogenomics and precision cancer medicine among others.

There have also been a number of recent papers published using approaches like this:
Adaptive resistance to therapeutic PD-1 blockade is associated with upregulation of alternative immune checkpoints

Genomic and Transcriptomic Features of Response to Anti-PD-1 Therapy in Metastatic Melanoma

Melanoma-intrinsic β-catenin signalling prevents anti-tumour immunity.

There are many more interesting projects and themes in oncology research that I'm sure will be presented at AACR, but I will somewhat arbitrarily stop here. Any discussion is welcome here or on twitter - @Festivus159.