How to get access to CRISPR after the patent battle

by Robert Kruse


The patent battle between the Broad Institute and the University of California was truly one for the ages, with the stakes for control of the CRISPR-Cas9 system of gene editing at stake. The court recently ruled in favor of the Broad Institute (a joint institution between Harvard & MIT) and Feng Zhang for being the inventor of Cas9 editing in eukaryotic cells, having shown sufficient evidence at the time that CRISPR could not only function in bacterial cells where it is normally found, but also in eukaryotic cells whose genomes are housed in the nucleus and whose DNA is wrapped in histones. This left Jennifer Doudna, Emanuelle Charptentier, and the University of California out in the cold, even though they filed their CRISPR patent first. Instead, the court said their applications are still limited to in vitro and bacterial applications. The University of California is appealing the decision and it is likely this court battle will continue on for years.

From the biotech side, at stake was the fortunes of a few companies. Caribou Biosciences and Intellia Therapeutics ($NTLA) had ownership of the University of California IP, while Editas Medicine ($EDIT) owned the Broad IP rights. These companies have also made other deals with companies for other downstream applications and therapeutic targets. The fallout and ramifications for these companies are still being determined. Clearly being in Editas' position would be favorable, but it should be noted that all these companies have continued making innovations and improvements on the CRISPR platform in the years since the original description. The applications that are key to CRISPR are likely tied up now, so even if Editas might want to push forward with a certain application, they must make sure they aren't blocked from an IP perspective in doing it. An example would be an application for CCR5 deletion in T cells for HIV therapy, or in PD-1 editing in T cells for cancer immunotherapy, which in both cases may have been patented by an investigator at a university or a different company. Interestingly, companies like Sangamo Biosciences which specializes in zinc finger nucleases have also filed patents on CRISPR, showing that they may have a similar strategy.

However, this strategy does still not truly allow free access to gene editing, which would be desired by many companies right now. Here, I will briefly outline my advice to invent around the Broad patent and allow one to freely pursue their CRISPR therapeutic applications.

My advice focuses on one simple aspect: The patent estate of the Broad patent is very specific to the Cas9 protein and the sequences of Cas9 proteins from the different organisms described at that time. It will not be easy for any patent lawyers or judges to claim that the Broad patent extends years in the future to a new Cas9-like protein, or even more preferably, a novel RNA-guided cutting enzyme with little sequence homology to Cas9. This is because the Broad patent has to describe in detail how the guide RNA sequences could be engineered, which part of the DNA nuclease enzyme is required, and how to engineer for better efficacy in human cells. The RNA sequences binding to the protein are unique, the length of the recognition domain in RNA is unique, etc. Simply put, there is no way the Broad patent could prevent someone from using their newly discovered Cas9-like system from doing human gene therapy studies.

The advice that is likely already being pursued by companies around the world is to find these new CRISPR systems that would be unique and not under patent protection. What's funny is that Editas Medicine itself would have been in best position to have lost the CRISPR patent battle since it already owns a Cas9 alternative system. Feng Zhang has published an alternative Cpf1 system that is uniquely distinct from CRISPR-Cas9, but can functionally accomplish the same thing with RNA-guided DNA editing. Editas could have easily shifted their entire programs toward Cpf1 and not missed a beat with their clinical programs.

CRISPR systems appear to be operant in bacteria species across all different phylums on the planet. If one could pull a Craig Venter and begin sequencing the oceans and soils, it is likely that you could find unique Cas9-like systems. As opposed to the original discovery of CRISPR, which took two decades of figuring out what these repeat DNA sequences meant, and trying to figure out what the protein partners could be, new DNA algorithms should be more efficient at picking a signature repeat pattern out, and researchers could take these RNA candidates and use them to pull down in affinity assays protein partners. Yes, it would take a significant amount of time working in isolation, as opposed to dozens of Cas9 groups helping to facilitate the understanding of how the original CRISPR works, but that would be the trade off for having a unique platform. The cost of this endeavor might preclude individual researchers from pursuing this path, and make it the domain of a biotech company willing to throw significant funds and resources toward it. From an academic perspective, the focus will likely remain on using Cas9 as the main platform since those researchers are focused on the applications of CRISPR and don't want to mess around with optimizing a brand new system.

In conclusion, the widespread nature of the CRISPR system in bacteria makes it inherently difficult to control from an IP perspective, since nature has created different variations on CRISPR's theme. This differs significantly from the previous discoveries of programmable zinc finger and TALE domains, which generated so much excitement just because they were perceived to be rare in nature. It is likely that new systems will continue to be discovered, meaning gene editing in the future will have open access, from both an academic and a biotech perspective. The merits of the clinical potential of gene editing will rest then on if it is useful in actually ameliorating disease, rather than owning an entire platform.

10 Favorite New Biotechnologies of 2016 (Part II)

This is part two of my Top 10 favorite new technologies published in 2016, part one can be found here. So without further ado, here are the other five, again not listed in any particular order.

6. Reengineering chimeric antigen receptor T cells for targeted therapy of autoimmune disease (link)

Chimeric Antigen Receptors (CARs) expressed in T cells have been one of the hottest areas of biotechnology in the past few years. CAR-T cells have initially been used to recognize and kill all (CD19+) B cells as a treatment for a number of B cell malignancies. However, CARs have the potential to be much more versatile than simply targeting tumor antigens. Their core properties are a recognition domain and a signaling domain, so they could theoretically be applied to many more applications. For instance, some have tried to engineer FVIII reactive CARs into regulatory T cells to suppress inhibitor antibody development during enzyme replacement therapy for hemophilia (link).

In this paper for technology #6, instead of engineering CAR-T cells to kill all B cells, the authors engineer a CAR to specifically kill B cells that produce autoantibodies against the Dsg3 autoantigen in the autoimmune disease, pemphigus vulgaris (PV). To do this, they use the autoreactive B cell's ability to bind the extracellular domain of Dsg3 as a way to specifically kill these cells. The CAR uses the Dsg3 extracellular domain as the extracellular component (instead of CD19 scFv, for instance) and then uses the typical 4-1BB + CD3zeta costimulatory and killing domains as the cytoplasmic components. They call this a chimeric autoantibody receptor (CAAR-T), shown below:

From Ellebrecht et al., 2016

The authors demonstrate that these CAAR-T cells are able to lyse B cells expressing anti-Dsg3 B cell receptors, even in the presence of anti-Dsg3 antibodies. They also tested the cells using in vivo mouse models by injecting anti-Dsg3 secreting hybridomas, and found the CAAR-T cells were able to significantly reduce anti-Dsg3 titers and protect against skin disease. Further, these CAAR-T cells had no apparent toxicity to normal human skin xenografts.

PV is a potentially fatal autoimmune skin disease caused by these autoantibodies. One current treatment approach is to broadly deplete B cells with Rituximab, which causes high remission rates, but often patients relapse. Additionally, this broad depletion of B cells potentially raises the risk of infection. Thus, targeting just the B cells involved in the disease is an attractive approach, and if the CAAR-T cells remain in patients long term, it may provide a more lasting response, or even a cure.

This paper is a nice example of a creative use of chimeric antigen receptors outside of cancer. Their approach shows promise in a severe disease, and also could potentially be applied to other autoantibody-mediated diseases.

7. Overcoming EGFR(T790M) and EGFR(C797S) resistance with mutant-selective allosteric inhibitors (link)

Kinase inhibitors, such as imatinib, crizotinib and erlotinib, can induce significant tumor regressions in a high percentage of patients with tumors driven by the targeted kinase. Unfortunately, these regressions can be transient, and mutations in the target can occur that reduce the activity of the inhibitor. In the case of EGFR activating mutations, such as the L858R mutation or exon 19 deletions, these can confer sensitivity to EGFR inhibitors. However, secondary mutations, such as T790M limit the activity of these inhibitors. Next generation inhibitors that can inhibit EGFR T790M have been developed, with Osimertinib recently being approved. Unfortunately, resistance arises against these inhibitors as well, with the C797S mutation identified as being able to inhibit the activity of T790M inhibitors.

In this paper, scientists at Novartis screened a large library of compounds for the ability to inhibit EGFR (L858R/T790M) but not inhibit wild-type EGFR. They identified a compound that inhibited EGFR in a novel way from all the previous EGFR inhibitors, which typically bind directly in the active site of the enzyme. This allosteric inhibitor (shown below) binds in a separate site, which is potentially advantageous, because the mutations that prevent activity of other EGFR inhibitors may not have any effect on this novel molecule.

From Jia et al., 2016

From Jia et al., 2016

Impressively, they found that this compound did, in fact, have activity against the new C797S mutation that currently makes EGFR insensitive to all T790M inhibitors. This novel inhibitor did not appear to have activity against EGFR with exon 19 deletions, so potential future clinical trials using compounds based on this molecule will most likely focus only on other EGFR activating mutations.

In addition to being able to target the latest resistance mutation in a key target, I am excited by this approach because allosteric inhibitors potentially have non-overlapping resistance mechanisms. Thus, if one could combine treatment with other EGFR inhibitors, this may delay resistance as any individual tumor cell would have to simultaneously be resistant to both inhibitors to survive. Novartis has a similar allosteric inhibitor approach for the BCR-ABL gene fusion in CML, which is usually effectively treated with kinase inhibitors, but resistance mutations can occur. That allosteric inhibitor, ABL001 has shown activity in CML patients who had become resistant to other inhibitors (link).

Additionally, one of my favorite people on twitter, Tom Marsilje (@CurrentIncurSci), is one of the authors of this paper.

8. Identifying neoantigen-specific T cells from patients

Steve Rosenberg's lab at the NCI has been one of the major drivers of research on how a patient's own T cells can recognize tumor-specific mutations and be used as a therapeutic against cancer. Through work using tumor-infiltrating lymphocytes (TILs) and expanding tumor-reactive TILs ex-vivo, his group has achieved impressive results in patients with melanoma. Further, it appears that the primary activity of these TILs are against patient-specific mutations that form neoantigens T cells can recognize (link). However, difficulties remain in generalizing this research - to generate TILs, for direct use or identification of neoantigen-reactive TCRs, one needs access to significant amounts of tumor tissue. Additionally, it is unknown if neoantigen-reactive T cells will be present in tumors with lower mutational burden than melanoma. Two papers from the Rosenberg lab addressed these questions this past year.

Prospective identification of neoantigen-specific lymphocytes in the peripheral blood of melanoma patients (link)

Previous work by the Rosenberg lab had identified the inhibitory receptor, and checkpoint blockade target, PD-1 as being elevated specifically on the neoantigen-reactive T cells in the TIL population. Here, they asked if this biomarker may allow the identification of neoantigen-reactive T cells from peripheral blood, which is much more easily accessible and less invasive than using tumor tissue.

Studying four melanoma patients, they were able to identify neoantigen-reactive T cells from peripheral blood against at least one, and sometimes multiple, neoantigens in three of the four patients. The TCRs identified from these neoantigen-reactive T cells when transduced into autologous T cells were able to redirect cells to recognize the patient's tumor.

To identify which of the PD-1+ T cells are actually reacting against specific neoantigens, you do need to have at least sequencing data from tumor cells to predict potential neoantigens and screen T cells for reactivity against them. While this could limit the benefit of having a non-invasive source of tumor-reactive T cells, the use of less invasive (circulating tumor cells, needle biopsies) or archival tumor samples, still makes the ability to access tumor-specific T cells from peripheral blood an important advance. Identification of these tumor-specific T cells could allow either the expansion of these cells and reinfusion into the patient, or the identification of TCRs that could be used in a (typically) patient-specific engineered TCR therapy. Numerous companies, such as Neon Therapeutics and Kite Pharma (CRADA with NCI/Rosenberg), are interested in identifying neoantigen TCRs for personalized T cell therapies.

Immunogenicity of somatic mutations in human gastrointestinal cancers (link): 

This second paper (Dec 2015, but we'll sneak it into 2016) addresses the question of whether TCR identification approaches from TILs can be generalized to tumors with low mutation burden that may not have a robust endogenous immune response. Here, Rosenberg's team looked at 10 patients with various gastrointestinal (GI) tumors to see if they could grow TILs from their tumors and identify neoantigen-reactive subsets. In 9 of the 10 patients they were able to identify neoantigen-reactive TILs that were often only present as a small percentage of the TIL population (below).

From Tran et al., 2015

Impressively, even a tumor with as few as 25 mutations produced mutation-specific T cells. 4 of the 10 patients received TILs enriched for the neoantigen-reactive subset, and one has had a response lasting >20 months (that patient's TIL treatment was the focus of a previous publication here).

This result shows that the endogenous immune response, even in tumors thought to be less immunogenic than the tumor types that typically respond to immunotherapy, can still recognize the tumor. This opens the door for improved approaches to harness this response in these cancers.

Since the majority of mutations in a tumor are most likely passenger mutations with no selective benefit to the tumor, the majority of neoantigen-reactive T cells are likely to target these mutations and are expected to be patient-specific. However, one patient, 3995, had T cells that reacted against a driver mutation, KRAS G12D. Targeting this mutation, and other driver mutations with T cells is the focus of the next group of papers, below.

9. T Cell Receptors that target driver mutations

Taking off directly from the last paper, the identification of TILs with TCRs that can recognize driver mutations has a number of implications. Targeting driver mutations (as opposed to passengers) are appealing for a number of reasons. Specifically, since they are important for maintaining the tumor phenotype, it would be more difficult for the tumor to lose the mutation compared to passenger mutations. Additionally, as a driver-reactive TCR could be used across the multitude of patients whose tumors contain this mutation and share the same HLA allele the mutant peptide is displayed on. This contrasts again to passenger mutations, where the TCR recognizing the neoantigen needs to be identified and cloned into an expression construct for each patient individually. Below are a number of examples of different recent attempts to use T cells recognizing driver mutations as therapeutics in cancer.

T-Cell Transfer Therapy Targeting Mutant KRAS in Cancer (link)

The paper in the previous section by the Rosenberg group identified KRAS G12D-specific TILs from a colorectal cancer patient. KRAS mutations are common across many cancers (below), however, it is generally considered an undruggable target with small molecules, so identifying novel ways to target these mutations is an exciting development.

From (open access) Ras review by Stephen et al., 2014

The previously mentioned patient, 3995, was infused with the TIL product, but it only contained 0.002% T cells that recognized the KRAS G12D mutation, and that patient did not have a response after TIL therapy. They identified another colorectal cancer patient also with multiple T cell clonotypes that specifically recognized this KRAS mutation. That patient's treatment with these TILs is the subject of this paper. ~75% of the TIL product recognized the KRAS G12D mutation. This patient had 7 metastases, all of which had objective responses, however one began to progress 9 months later, shown below.

From Tran et al., 2016

That lesion (#3) was removed surgically, and interestingly, but frighteningly, was found to have evidence of losing the HLA allele (HLA-C*08:02) the mutant peptide was being displayed on. This again shows the powerful ability of tumors to rapidly evolve in the face of selective pressures, whether they are small molecules or T cells, as I detailed in a previous post.

However, this patient report is equally, if not more, promising, as it raises the possibility of using this TCR in other patients with the G12D mutation and HLA-C*08:02. Additionally, if an effective TCR recognizing a common KRAS mutation can be found, there is reason to believe that we may be able to identify TCRs that can recognize many more driver mutations as well.

The authors state that the HLA-C*08:02 allele is found in 8% of white and 11% of black individuals, and thus that there is a potentially large population (thousands) of patients who could be amenable to this TCR. However, in a letter to the editor, there seems to be some debate if truly that many patients exist who would have both the right HLA allele and the KRAS G12D mutation.

Others at the NCI have tried to identify TCRs against the more common HLA-A*11:01 allele, present in 14% in Caucasians and 23% in Asian-Americans, by immunizing HLA-A*11:01-transgenic mice with KRAS G12D and G12V mutant peptides (link). They were able to identify TCRs that appeared to specifically recognize the mutant peptide, but not wild-type, and had activity in xenografts of cell lines that contained the mutation. Since these TCRs were generated in the mouse, and not from patients, in my mind they are still significantly less validated in their safety and efficacy compared to the TCR identified from the patient. However, further investigation of these TCRs is warranted. Kite Pharma has licensed KRAS targeting TCRs from the NCI, however, I am currently unclear as to which specific TCRs they licensed.

BCR-ABL-Specific T-cell therapy in Ph+ ALL patients on tyrosine-kinase inhibitors (link)

It has previously been shown that BCR-ABL-specific T cells can be found in patients with this fusion mutation (Ph+ CML & ALL), and their presence correlates with clinical response to kinase inhibitors. In this new paper, the authors attempted to treat 3 Ph+ ALL patients with measurable disease with autologous or allogeneic (from donor post-allogeneic transplant) BCR-ABL-specific T cells. They identified and expanded these T cells from peripheral blood by coculturing with dendritic cells pulsed with BCR-ABL peptides.

All three patients had measurable disease and were on a stable course of TKIs prior to infusion. Patients were treated with multiple infusions of T cells, and each patient achieved a molecular or hematologic complete remission post infusion. One patient had their TKI switched from imatinib to nilotinib to ponatinib, during treatment and likely played a role in the response as well. There was also no significant toxicity from the treatment besides grade II skin GVHD in one patient that resolved, and who had had a previous allogeneic transplant.

Although only 3 patients were treated, this report suggests expanding and infusing BCR-ABL-specific T cells from patients or donors can be an effective strategy in Ph+ ALL patients. Alternatively, BCR-ABL-specific TCRs could be used to make engineered TCR products instead of using endogenous T cells. These results further demonstrate the possibilities of targeting shared driver mutations using mutation-specific T cells.

Lastly, there is an intriguing AACR17 abstract title that suggests TCRs against another recurrent mutation have been identified: Identification of a novel and a shared H3.3K27M mutation derived neoantigen epitope and H3.3K27M specific TCR engineered T cell therapy for glioma (link)

10. A brain–spine interface alleviating gait deficits after spinal cord injury in primates (link)

For something completely different, while I have a very limited understanding of how they work, I find brain-body computer interfaces to be a fascinating area of study. Electrical stimulation has previously been shown to be able to activate the spinal cord in paralyzed animals and humans. However, controlling a complicated movement such as walking requires communication and feedback between the brain and spinal circuits. The authors in this study implanted an electrode array to record brain activity in the area of the brain controlling a specific limb in non-human primates. They also implanted an electrical stimulation device that could be wirelessly triggered. The entire experimental set up is shown below:

From Capogrosso et al., 2016

They tested this on animals with normal limb function to learn the pattern of brain activity and nerve firing. They then implemented this system in animals with one limb paralyzed, and used the information from normal walking to decode when the brain was trying to tell the spinal circuits to fire. They could activate the computer interface to induce the stimulation of these neurons and induce the appropriate timing of muscle activity to allow normal walking. A video (viewable without paywall) describes the approach better than I can and shows the impressive results and it's worth a look.

Apparently, all of the components have individually been approved for investigational use in humans, so a potential human trial may not be that far away.


Honorable Mention
There were many other interesting papers from 2016 that I could not fit into this list of 10, so I've listed a whole bunch more below.

MTAP deletion confers enhanced dependency on the PRMT5 arginine methyltransferase in cancer cells (Two papers: link 1, link 2).
- I like the concept of targeting the effects of genetic deletion of genes - since you cannot gain them back. A couple other papers with a similar approach of targeting passenger gene deletions (link 1, link 2).

Pathological α-synuclein transmission initiated by binding lymphocyte-activation gene 3 (LAG3) - (link).
- Interestingly links α-synuclein involved in neurodegenerative disorders to LAG3, a cancer immunotherapy target.

Proteasome inhibition for treatment of leishmaniasis, Chagas disease and sleeping sickness (link).
- A novel target for a number of poorly treated infectious diseases. Also discussed in Derek Lowe's In the Pipeline blog (here).

Programmable editing of a target base in genomic DNA without double-stranded DNA cleavage (Two papers: link 1, link 2).
- One of the drawbacks of current gene editing approaches (CRISPR etc.) is the creation of potentially dangerous double-stranded breaks. These methods allow the modification of bases without cutting DNA, although there are limitations to this approach as well.

Autoimmune manifestations in aged mice arise from early-life immune dysregulation (link).
- Intriguing paper suggesting that (at least in mouse models) early intervention in autoimmune diseases (using an anti-CD40 antibody, here) can have a more profound impact on preventing later autoimmune sequelae. 

FVIII-specific human CAR T-regulatory cells suppress T-and B-cell responses to FVIII (link).
- Putting CARs into regulatory T cells to suppress autoimmune diseases, here preventing inhibitor antibody formation in hemophilia.

Correcting mitochondrial fusion by manipulating mitofusin conformations (link).
- Mitochondrial fusion dysfunction is central to diseases such as Charcot-Marie-Tooth disease 2A, but modulating fusion has not previously been possible. The authors engineered a cell-permeant minipeptide that can correct the cause of CMT2A disease in cells.

Inhibition of ileal bile acid uptake protects against nonalcoholic fatty liver disease in high-fat diet–fed mice (link).
- Bile acids are interesting targets for nonalcoholic fatty liver disease, and this paper used a small molecule that remodeled bile acid composition and had beneficial effects in a mouse model of NAFLD.

A “Trojan horse” bispecific antibody strategy for broad protection against ebolaviruses (link).
- Novel use of bispecific antibody technology to target antibody to cellular compartment (late endosomes) where they can effectively bind and inhibit ebolavirus proteins.

A Genome-wide CRISPR Screen in Toxoplasma Identifies Essential Apicomplexan Genes (link).
- Uses the power of a CRISPR knockout screen to identify novel essential Apicomplexan (group of parasites including those that cause malaria and toxoplasmosis) genes and potential future drug targets.

Factor XIa–specific IgG and a reversal agent to probe factor XI function in thrombosis and hemostasis (link).
- Suggests targeting Factor XI might be a safer anticoagulant target with less bleeding risk (as currently being pursued by Ionis/Bayer).

Potentiating the antitumour response of CD8+ T cells by modulating cholesterol metabolism (link).
- Identifies a novel target for improving T cell activity, potential to target it systemically or directly by modifying cells used in T cell therapies.

Chronic administration of an HDAC inhibitor treats Niemann-Pick type C disease in a mouse model (link).
- NPC is a severe and rare lysosomal storage disease with no current disease modifying treatments, this study suggests HDAC inhibition may be worth further investigation as a potential target.

Expanding antigen-specific regulatory networks to treat autoimmunity (link).
- Nanoparticles coated with autoimmune peptides bound to MHCII reduce autoimmune disease in a number of mouse models.

Allele-specific inhibitors inactivate mutant KRAS G12C by a trapping mechanism (link).
- KRAS mutations are generally considered "undruggable" (read: difficult to drug), but this paper shows that KRAS G12C mutations actually allow for sensitivity to drugs that bind to KRAS in its inactive state, suggesting a potential therapeutic approach for these mutations.

Antibody-drug conjugate targeting CD46 eliminates multiple myeloma cells (link)
- Identifies CD46 as a potentially interesting target in multiple myeloma and shows activity using an antibody-drug conjugate in xenografts.

Theranostic barcoded nanoparticles for personalized cancer medicine (link).
- Uses nanoparticles containing different cancer drugs and DNA barcodes, the authors developed a system to test multiple drugs directly in vivo to determine which drug will be the most effective.

CRISPR/Cas9 β-globin gene targeting in human haematopoietic stem cells (link).
- Were able to correct β-globin mutations in human HSCs using CRISPR, and to enrich for corrected cells by using a reporter in the homologous recombination template.

NHEJ-based CRISPR gene insertion (link).
- Homologous recombination based gene editing is less efficient than non-homologous end joining (NHEJ) based editing, and is thought to only be usable in dividing cells. The authors developed an approach, using microhomology templates and CRISPR to use NHEJ based editing to insert genes at high efficiencies and in non-dividing cells.

IgG-Fc Trimer potent inhibitor of immune complex induced autoimmunity (link).
- Activation of the Fcγ receptor by autoantibody immune complexes can play a role in autoimmune disease pathogenesis. The authors identified a trimer configuration of Fc constructs to bind the receptor without inducing its activation, and potentially serves as a specific way to inhibit this pathogenic interaction present in autoimmune diseases.

Nanobodies that block gating of the P2X7 ion channel ameliorate inflammation (link).
- ATP released by dying cells can contribute to autoimmune and inflammatory diseases. The authors were able to use nanobodies to block ATP binding to ion channel P2X7, and had efficacy in mouse models of inflammation.


Disclosure: I have no positions in any of the companies mentioned.

Jounce IPO Review

Jounce Therapeutics ($JNCE) is debuting on the public markets today after raising $102 million dollars in its initial public offering. The fully diluted market cap based on the price of the shares sold would be approximately $550 million. Jounce is entering the highly active immuno-oncology (I/O) space, focusing on monoclonal antibodies, with the hopes to identify potent combinations to activate the immune system against cancer. Jounce has a high powered team of scientific co-founders and advisors, including Jim Allison, one of the pioneers of immune checkpoint blockade. They were initially seeded by Third Rock Ventures in 2013, and have picked up additional financial backers during their development. I listened to their IPO presentation (retailroadshow) and below are some of my initial thoughts. Their S-1 prospectus can be found here.

Jounce's Broad Vision
Jounce stated a number of guiding principles about how they wanted to distinguish themselves in the crowded, but exciting, space of immuno-oncology.

Biomarkers
One tenet is that they want to make I/O more personalized, and give the right I/O treatment to the right patients. Compared to certain targeted therapy approaches using kinase inhibitors (e.g. TRK inhibitors in TRK-fusion driven cancers), I/O has often taken a less selective approach. This is because, so far, while some biomarkers in use can be useful (see nivolumab's failure in NSCLC using less PD-L1 selection than Merck), they are not great at predicting at the individual patient level who will respond or not respond, as some with no PD-L1 respond and many with high PD-L1 levels do not. Jounce stated that they want to continue to develop better biomarkers for use in guiding the development of their therapies. To accomplish this, a core part of the company is their "translational science platform", which is focused on characterizing (at both the RNA and protein level) human tumors, and their immune components, to identify potentially useful biomarkers and novel targets.

Targeting multiple immune cell types
Most immunotherapy in cancer so far has been focused on the T cell. While Jounce's first candidate (JTX-2011), an agonistic antibody of ICOS, is intended to modulate the T cell response against cancer, they intend to expand their pipeline to target other aspects of the tumor-immune interaction.

 
They are also focused on tumor-associated macrophages (TAMs), cells that are generally thought to be pro-tumorigenic, but could potentially be reprogrammed back into pro-inflammatory, anti-tumorigenic, cells. There has been increased interest in modulating TAMs for I/O, and targets, such as CSF1R, are being explored by a number of companies. They stated they have identified 10 novel targets to potentially modulate macrophages, and disclosed one of these at AACR last year (abstract here). That target was TIM3, which is also a hot target for T cell biology. Interestingly, they identified a novel interaction of TIM3 with an undisclosed protein on TAMs and developed antibodies that could specifically modulate this interaction. They found that these antibodies did not affect T cell activity, but induced a more pro-inflammatory (anti-tumor) phenotype in macrophages.

Based on their pipeline, they are also interested in targeting regulatory T cells as well as B cells. Jounce is also attempting to target so called "Cold Tumors" that appear to lack a strong immune response at the site of the tumor, and is associated with lack of response to current immunotherapies. A number of targets have emerged to try to improve the immune infiltrate in these tumors, such as STING agonists, CXCR4-CXCL12 blockade, and targeting desmoplasia in tumors like pancreatic cancer.

They are also developing their own PD-1 antibody (JTX-4014), which they intend to use only for combinations.

Targeting ICOS
Jounce's lead program is JTX-2011, which is an agonistic antibody against ICOS (Inducible T-cell Costimulator). Unlike targeting T cell-inhibitory proteins, such as CTLA-4 or PD-1, to boost the activity of T cells, Jounce is hoping to activate this costimulatory molecule. For background, Jounce presented some preclinical data on their antibody at AACR (abstract here) and a good review of ICOS signaling, and its role in immunity and cancer is available here. ICOS is a costimulatory molecule, similar to CD28, that is typically found on activated CD4 T cells.

From Medzhitov et al., 2011

Activating costimulatory molecules, such as CD28, had been previously tried with a "superagonistic" antibody TGN1412, which caused severe cytokine release syndrome, as detailed here. However, there were mistakes in the clinical development of that antibody, besides the target, that led to those problems. ICOS, as the name suggests, is only upregulated on already activated CD4 T cells, including both effector and regulatory T cells. Jounce hopes that this more specific expression of ICOS, such as at the tumor site, will limit broad overstimulation of the immune system. The presence on both effector and regulatory populations could confound attempts to target it. Jounce suggested that their antibody both activates effector T cells at the tumor site, as well as depletes regulatory T cells (that also express ICOS). This is presumably through antibody-dependent cell-mediated cytotoxicity. Depletion of regulatory T cells has previously been shown to be a potentially critical mechanism of action for CTLA-4 antibodies. So my assumption is that their ICOS agonist antibody is also an IgG1 antibody, like ipilimumab, capable of causing depletion of the cells it binds to. This has the potential to also cause depletion of effector T cells, but in the case of CTLA-4 seemed to be more selective for depletion of tumor-resident regulatory T cells, potentially due to their higher levels of CTLA-4 expression. How this plays out clinically with ICOS antibodies will be important to follow, and I assume we'll be seeing clinical biomarker data on effector:regulatory T cell ratios among others.

Jounce is testing its antibody both as a monotherapy, but also in combination with PD-1 antibodies. One of the rationales for this combination is that ICOS+ cells are shown to be increased in patients or mice treated with PD-1 or CTLA-4 blockade (here and here). Notably, ICOS+ cells are increased also in peripheral blood post CTLA-4 blockade in patients, which perhaps might be a concern that this will increase both the potency, but also potential toxicity, of ICOS stimulation in these combinations. Combining ICOS stimulation (through expression of ICOS ligand constitutively, in tumor models) with CTLA-4 blockade has been shown to increase the efficacy of blockade. Jounce presented data showing that, as with many immunotherapy combinations in mouse models, checkpoint blockade efficacy was enhanced by their ICOS agonist antibody.

ICOS clinical development
Jounce stated that preclinical data suggested tumors with the highest numbers of ICOS+ cells were the ones most likely to respond to therapy. This provides additional rationale for combinations (e.g. with PD-1) that increase the number of ICOS+ cells. Additionally, Jounce stated that ICOS positivity, by IHC in tumor specimens, will be an important biomarker they use to guide development. There is variability in the amount of ICOS+ cells across tumor types (typically the more "immunogenic" tumor types), as well as variability within a given tumor type. Their initial phase I/II trial testing JTX-2011 as monotherapy, and in combination with anti-PD-1, will be an all-comers trial, whereas their fast-following phase II trial will be focused on typically ICOS high tumor types and using IHC to enroll at least half of patients who have ICOS+ tumors. GSK also has an ICOS agonist antibody in clinical trials (H/T @PDRennert).

Celgene Partnership
Jounce also has a partnership with Celgene to develop a number of their products. Jounce received $225 million upfront, a $36 Million equity investment, and there is $2.3 billion in potential milestones (biobucks). There is a 60:40 US profit share on JTX-2011, with royalties on ex-US sales, and US profit sharing on 3 other targets for Celgene as well as shared profits globally on JTX-4014 (PD-1). This partnership is starting with a four year research term.

Summary and Catalysts
Jounce is backed by some of the leaders in the immuno-oncology field. Their initial target of ICOS does make rational sense based on its known biology. However, efficacy in mouse models is always difficult to project to human efficacy, and this is especially true for immunotherapy. I agree with their biomarker approach in spirit, and hope they will continue to search for predictive biomarkers for both their ICOS antibody as well as future targets. Trying to give the right immunotherapy to the right patients is something that makes a lot of sense, as I've written about previously. Additionally, I appreciate their endeavors to look beyond T cells for attractive targets, such as macrophages or tumor stroma.

Jounce has stated that their JTX-2011 all-comers monotherapy and PD-1 combo trial will readout 1H17, and their phase II in ICOS-enriched tumors may read out in 2H17. I will be looking forward to following their progress.



Disclosure: I have no position in any of the companies mentioned

My Top 10 New Biotechnologies of 2016 (Part I)

There were many interesting new biotechnologies published in 2016, so it was difficult to narrow it down to just 10. I ended up picking ones that either may have exciting potential to be directly translated into the clinic, or may represent approaches that I think will be important directions where biotechnology can go in the future. Here are the first 5 of my top 10 biotechnologies published in 2016 (in no particular order). Part 2 of my top 10 can be found here.

1. Hematopoietic Stem Cell Transplants Without Myeloablation

There were three papers that came out this year on new ways to deplete hematopoietic stem cells (HSCs) to allow transplantation without the toxic myeloablative conditioning regimens that are currently used. One of the problems with conditioning regimens is they severely compromise the immune system for prolonged periods, making patients susceptible to life-threatening infections. These approaches were all tested in mouse models of transplantation and potentially reduce compromising the immune system. These could immediately be applicable to autologous stem cell transplants and gene therapy approaches of ex-vivo modified autologous cells. In fact, in bluebird bio's gene therapy trials, their main toxicities, so far, have been due to the conditioning regimens. How well these approaches would work in conditioning for allogeneic transplants is less certain, as depletion of the host immune system plays a role in allowing the engraftment of the donor HSCs. There may still be benefits (no genotoxicity & organ damage) if these approaches are combined with immune-depleting approaches (as done in the second paper) to allow allogeneic transplant. Below are the three approaches, with a brief description of each.

CD45 Immunotoxin (link)
This study used an antibody-toxin conjugate (CD45-saponin) to deplete hematopoietic cells. They found this molecule was able strongly deplete HSCs in mice, and allowed faster immune reconstitution with only transient depletion of B and T cells compared to irradiation. This approach allowed high levels of chimerism (>60%) after transplant in immunocompetent hosts. Magenta therapeutics has recently been founded to bring this approach, along with a number of other transplant-improving technologies, to the clinic.

CD47 + cKit antibodies for non-chemo conditioning (link)
In this paper from Stanford, continuing with their evaluation of the effects of CD47 antibodies (previously tested in cancer & atherosclerosis), they used a CD47 mAb to potentiate the effects of c-Kit antibodies to deplete HSCs. c-Kit is present on HSCs and their downstream progenitors. They were able to achieve high levels of chimerism (~60% HSCs) after transplant in immunocompetent hosts. Unlike the above paper, they also tested the ability of this approach in a more relevant model of allogeneic transplantation - transplanting between two different mouse backgrounds. For this they combined their approach with CD4/8 antibodies to deplete T cells to allow further suppression of the immune system and allow allogeneic engraftment, and were able to achieve ~20% HSC chimerism. A concern with this approach, which was brought up in the first paper above, was that c-Kit is also present on cardiac progenitors, gastrointestinal cells, neuronal cells, and cells of the reproductive system. Additionally, in the paper above, they found that attempting to potentiate the c-Kit antibody by conjugating it to saponin did not improve its efficacy, although this could have been for a number of reasons.

Depleting dietary valine permits nonmyeloablative mouse hematopoietic stem cell transplantation (link)
Finally, in this most recent paper, the authors found that depletion of the amino acid valine prevented HSC proliferation both in vitro and in vivo. Valine-free feed in mice also caused partial depletion of immune cells. Valine depletion in feed for three weeks allowed transplantation in immunodeficient and immunocompetent mice, but at much poorer levels of chimerism, ~10-30% vs. 60+%, compared to the above two approaches. Additionally, if valine was reintroduced into diets immediately, it would cause "refeeding syndrome" which was lethal in 10 of 27 mice, but was able to be avoided if valine was reintroduced slowly. This approach seems the furthest away from translation to the clinic of the three.

2. Directed Evolution of Bt Toxin Insecticides to Overcome Insect Resistance Mechanisms (link) 

I thought this paper was really interesting because it addressed a significant problem, resistance to the major insecticide, Bt toxins, by using directed evolution, an approach that is both tremendously powerful and versatile. Powerful because it uses rapid evolution to let nature figure out solutions to difficult problems without much a priori knowledge, and versatile because the general theme of directed evolution can be applied in innumerable ways depending on how you set up the experiment. For Bt toxins to work, they need to interact with proteins on the surface of insect midgut cells. Insects have started developing resistance through mutation or loss of expression of these surface proteins. The Liu lab decided to evolve Bt toxins to bind to a different midgut surface protein allowing continued insecticidal activity.

To accomplish this, the Liu lab has developed a directed evolution strategy called PACE (phage-assisted continuous evolution) which they used here to evolve protein-protein interactions. They have previously used similar direction evolution approaches to evolve gene editing proteins to improve specificity or to identify resistance mutations to HCV drugs. Here, they used PACE to evolve the Bt toxin (Cry1Ac) to bind the insect protein TnCAD, a protein structurally related to the toxin's normal target. The general scheme is shown below. They are using phage which need to bind to and enter bacteria to complete their life cycle. To bind bacteria they need to express the pIII protein, but the system is engineered so that the pIII protein is only produced if the phage expresses a Cry1Ac variant (evolving protein) that can bind to TnCAD (target). There is also a mutagenesis plasmid that causes continuous mutation to allow continued evolution in the reactor. The reactor has constant inflow and outflow to allow selective pressures to generate the enrichment of phages containing TnCAD-binding Cry1Ac variants.

From Badran et al., 2016

In the end, they were able to evolve new Bt toxins that could interact with these alternative targets in insect cells. These Bt toxin mutants were found to be much more potent insecticides, especially against Bt toxin-resistant insects, and could lead to the production of new insecticides.

For more detail, Derek Lowe also wrote an excellent commentary on this paper on his blog.

3. A Replication-Defective Human Cytomegalovirus Vaccine for Prevention of Congenital Infection (link)

Human cytomegalovirus (HCMV) is a generally asymptomatic virus that can cause serious complications in immunosuppressed patients and newborns. There currently is no effective vaccine, and there are multiple reasons why vaccine approaches have been difficult to develop. One interesting obstacle is the ability of HCMV proteins to downregulate MHC-I expression and evade CD8+ T cell responses.

Previous live vaccines were attenuated in multiple ways, including mutations to genes in a pentameric protein complex that prevented viral entry into epithelial cells. For this vaccine, the authors restored this ability. Since this modification could strengthen the infectivity of the virus, they attenuated it further by making two essential HCMV proteins unstable in the absence of an artificial ligand (Shld-1). This system uses a degron tag based on FKBP, which binds, and is stabilized by, the rapamycin-analog Shld-1. To produce the vaccine, ligand is present during production, but is not present once given to the patient. They found the more active virus was able to produce a neutralizing antibody response against the pentameric complex, present in natural immunity to HCMV, but lacking with previous vaccines. It was able to produce this neutralizing antibody response (and CD4+ and CD8+ T cell responses) against HCMV in mouse, rabbit, and non-human primate models.

This vaccine was developed by Merck and a clinical trial is ongoing (link). Sometimes the publication of a product by a biotech/pharma company could suggest that they are no longer interested in it. However, for what it's worth, the authors at least state at the end that "Preliminary data from our ongoing clinical evaluation suggest that V160’s safety and immunogenicity profiles in human are consistent with those described in this preclinical study. V160 is therefore a promising vaccine candidate against congenital HCMV transmission."

Recently, there has been an interest in developing HCMV as a therapeutic vaccine vector to treat other chronic infectious diseases, such as HIV and tuberculosis. Vir Biotechnology recently raised over $150 million primarily centered around the development of HCMV vaccine vectors pioneered by Louis Picker & Klaus Frueh's group. One interesting thing to add is that Picker's group found that disruption of the pentameric complex (loss of Rhesus virus ortholog of UL128 & UL130, with mutations in the pentameric complex restored in Merck's vaccine) was necessary to allow the unusual MHC-E restricted immune responses they see in response to their CMV vaccines.

4. Engineering T Cells with Customized Therapeutic Response Programs Using Synthetic Notch Receptors

Cell therapies have exploded in popularity since the impressive early results of chimeric antigen receptor T cells (CAR-T) in ALL. The surprising potency of these cells, and the fact that they can persist in the body, can differentiate these therapies from other antibody-targeted therapeutics. However, one of the truly unique features of cell-based therapies is the ability of a cell to contain much more information than a small molecule or biologic could. In addition to adding a targeting domain and T cell activation domain, one could modify cells to do a whole host of other processes. Tools are now being developed that will hopefully be able to unlock the programmable potential of cell therapies.

Wendell Lim's group has published three (!) Cell papers this year on their new synthetic notch receptors (synNotch), which allow customizable programming of cell responses in adoptive cell therapies.

The first paper describes how synNotch receptors work. Normal Notch receptor signaling is very simple, the Notch receptor binds to its ligand, which stimulates the separation of an intracellular signaling domain that can go on to affect gene expression. Previously, Notch reporters have been developed that contain the same Notch receptor extracellular domain, but with a transcription factor internal domain that can activate GFP to report when Notch signaling is activated. The Lim lab modified this further by coming up with many different potential extracellular domains beyond the Notch receptor to allow this cassette to respond to different inputs, as well as different intracellular domains to allow it to react differently to these inputs. The schematic is shown below.

From Morsut et al., 2016

In this first paper they use synNotch receptors in a variety of ways in additional cell types other than T cells. Additionally, more than one synNotch circuit can be used even within the same cell.

In their second paper, published back-to-back in Cell, they used a synNotch circuit to increase the specificity of CAR-T cells to require the presence of two antigens. Normally CAR-T cells are directed against a single antigen (CD19 etc.) however, not many cancers express a surface antigen that is specific only to the tumor and not essential normal tissue. To increase specificity, a number of groups have developed approaches to require the simultaneous presence of two antigens on the target, which will hopefully single out tumor cells and reduce activation against normal cells possessing only one of the two antigens. The Sadelain lab has also published approaches requiring target cells to possess two antigens simultaneously or specifically one antigen but lacking a second. The Lim lab used their synNotch circuit to direct expression of the CAR only when the T cell binds a first antigen, with the CAR recognizing cells containing a second antigen. The schematic is shown below.

From Roybal et al., 2016

They make T cells that should react only against cells with two (artificial) antigens, and find that the cells do in fact have specificity for cells expressing both. In vivo, when cells expressing one antigen were injected in one side of a mouse and cells expressing both were injected on the other, the engineered cells primarily localized and prevented the growth of the cells expressing both antigens. The ability to maintain specificity to both antigens will be critical for safety, as there is the potential for the CAR to be expressed after interacting with antigen A, and then be able to move around the body and react with a normal cell that only expresses antigen B. However, the ability of a T cell to make decisions from two inputs instead of one could potentially drastically increase the tumor types and targets amenable to CAR-T targeting.

Lastly, the Lim lab more recently published a paper using synNotch circuits to use T cells to deliver different payloads, such as immunomodulatory antibodies, bispecific antibodies, cytokines, or cytotoxic agents localized to target cells expressing a specific antigen. Some of the possibilities are shown below:

From Roybal et al., 2016

This is one of numerous ways to use T cells to cause local changes, which could be critical for enabling higher local concentrations or decreasing systemic concentrations of molecules that could have undesired systemic effects. The synNotch platform is impressive due to its modular nature and ability to be customized, which perhaps will allow it to be applied to multiple therapeutic applications.

One potential concern with all of these approaches is the potential immunogenicity of these additional proteins being produced by the cell. Although how much, if any, this will limit cell persistence/activity is unknown.

This work is being commercialized by Cell Design Labs, as a spinout from the Lim lab.

5. Targeting of Cancer Neoantigens with Donor-derived T Cell Receptor Repertoires (link)

There has been an increasing realization of the importance of the endogenous immune response against neoantigens, immunogenic mutations present only in a patient's cancer. Clinical trials of tumor infiltrating lymphocytes (TILs) and checkpoint inhibitors, like anti-PD-1 and anti-CTLA-4, have demonstrated that the reaction against neoantigens may be critical for the efficacy of these therapies, and boosting a response against neoantigens is an attractive therapeutic approach. Directly targeting neoantigens, with vaccines or T cell receptors (TCRs) is complicated by identifying which ones to target. Steve Rosenberg's group has done a lot of work with TILs and has demonstrated that T cells present in the patient's own tumor can react against specific neoantigens and induce long term tumor regressions (examples here and here). One questions is whether or not the patient's own immune system has produced a robust response against all the potential immunogenic antigens to allow all potential reactive T cells to be detected. This is what Ton Schumacher's group set out to investigate in this paper, where they attempted to identify if T cells existed in healthy donors that could react against patient-specific neoantigens.

The authors used autologous dendritic cells (from donors) loaded with tandem minigenes expressing different predicted neoantigens present in a patient's tumor, and cocultured them with donor T cells to determine if the T cells could react against any of the neoantigens. Across three patients, they could identify T cells reactive against 11 of the 57 predicted neoantigens. Only 1 of the 11 neoantigen-reactive T cells was detectable in TILs from a patient, indicating this is a way to identify a broader repertoire of TCRs reactive to patient-specific mutations. Additionally, this study attempted to expand our understanding of what makes an immunogenic neoantigen. They found that higher stability of the peptide-MHC complex for a neoantigen was statistically significantly associated with being able to identify T cells reactive against it.

This approach provides a way to prospect for neoantigen-reactive TCRs and expand the pool of potential TCRs beyond what is found in a patient, with potentially significant implications for neoantigen-targeting therapies.