Bellicum adding new T cell tools to the CAR trunk

by Robert Kruse

Chimeric antigen receptor (CAR) T cells have met success in the clinic led by Carl June's group at University of Pennsylvania, Michel Sadelein's team at Memorial Sloan-Kettering, Steven Rosenberg at the NCI, and Malcolm Brenner's group at Baylor College of Medicine, among others around the world. Partnerships have been formed between Novartis, Juno, Kite, and Celgene, respectively to translate technologies from academia to industry. While the therapy has significant benefits in achieving complete remission in the some patients, there are still two main draw backs outside of anti-tumor efficacy that could be solved and make for a better therapy:

1. Risk of cytokine storm following immediate infusion of cells into the patient via recognition of tumor cells

2. After clearance of tumor, on-going response against antigen on normal tissues resulting in chronic inflammation, and/or their complete removal in the case of B cells and CD19

Currently, the first problem is solved by supportive care in the ICU, which is not ideal since the toxicity carries significant risks from patient to patient, and indeed some patients have died in trials. In the CD19 trials, the administration of an IL-6 antibody was required to prevent symptoms of systemic shock. For the second problem, steroids have been shown to treat autoimmunity from CAR T cells in limited cases, but this situation could continue to become worse as CAR T cell persistence improves. In the CD19 trials, complete removal of normal B cells is observed. In order to treat this status, intravenous immunoglobulin (IVIG), pooled from healthy human donors, can be administered into patients to supplement their lost B cells producing antibodies. While many investigators think the B cell aplasia isn't a problem since we have a current treatment for the disease, I would argue that giving someone a costly disease, on the magnitude like having a lysosomal storage disorder, isn't the solution either. Obviously, chronic antibody replacement is a better problem to have than death due to cancer, but immunotherapy should still attempt to solve both problems.

Herein, I highlight an intriguing solution for the current CAR T cell problems that is on hand at Bellicum Pharmaceuticals. As a disclaimer, I have no knowledge of Bellicum's inner workings beyond their site, and am not a stock holder of the company. I merely think their technology is interesting and there is room for a potential collaboration between Bellicum and Novartis, Juno, Kite, or other CAR T cell players in biotech.

In general, chimeric antigen receptor signaling is broken down into two parts: 1) zeta chain signaling mimicking TCR activation and an essential first trigger 2) auxiliary costimulation domains (CD28, 41BB, OX40) that deliver cytokine secretion and proliferation signals

The heart of Bellicum's company is its inducible dimerization platform, which they have exploited for many uses. An unpublished use that I came across on their site is splitting about the CAR into two parts, which they name the "GoCAR-T Technology" (see image below). The first has an antibody targeting tumor antigen linked to zeta chain, like normal CARs. The second has no ectodomain, but instead is completely intracellular with the costimulatory domains linked to a drug regulated dimerization domain. The theory is that T cells could recognize and kill tumors through antibody recognition, but their potency, proliferation, and persistence would be regulated by an oral drug administered to the patient.

Taken from Bellicum Pharmaceuticals website

Why does this matter you ask and what is the advantage? Let us return back to the two central problems of the CAR T cell trials, applying Bellicum's technology to it. For example, one could administer CD19 CAR T cells, but regulate with a small molecule drug the degree of their cytokine secretion and proliferation, thereby preventing the cytokine storm scenario seen in many CAR T cell trials. A physician could add more drug gradually to increase efficacy over time. For the second problem, once the tumor is satisfactorily removed, complete withdrawal of drug would lead to T cells that just have zeta chain signaling only. They could still attack host B cells, but the lack of costimulation would greatly reduce their potency and in the experience of previous trials, the CAR T cell numbers would drop over time and the T cells perhaps become somewhat anergic. This would allow a steady state population of B cells to repopulate, or prevent tissue damage in the setting of other CAR targets. This differs greatly from June's data, where a large population of CAR T cells persist, permanently attack and removing any new B cells (one hypothesis is that this off-target killing actually sustains the anti-tumor response and T cell persistence). If a tumor relapse ever did happen, drug could be administered again to revive the remaining CAR T cells to join the fight again at the tumor.

Bellicum appears to be applying the GoCAR-T cell technology first to PSCA and prostate cancer, but there appears to be wide application to other immunotherapy targets. Of course, Bellicum is also pushing forward it's inducible Caspase-9 death switch for CAR T cell therapy, with a program targeting CD19. The death switch is able to avoid side effects, but has the downside of eliminating the cells, as opposed to temporarily attenuating them and later turning the cells back on, which is beneficial at multiple stages of therapy as mentioned above. Given the cost and expense of administering multiple cell products, it seems advantageous to have regulation of the initial cell product that doesn't cause cell death. Regardless, this technology seems ripe for a potential licensing deal with another player in the CAR industry to earn Bellicum more cash. The idea of pharmacological regulation of T cells could be a game changer in the burgeoning field.

Papers of the Week 4/20-4/26

Outcomes Following Gene Therapy in Patients With Severe Wiskott-Aldrich Syndrome

http://jama.jamanetwork.com/article.aspx?articleid=2275447 

This paper extended the findings of other gene therapy studies in Wiskott-Aldrich Syndrome (WAS). A previous paper described 3 patients with milder WAS who were treated with the same vector. WAS is scored on a scale from 1 being the least severe to 5 being the most severe, and the patients in the original study were all scored either 3 or 4. This new study included 7 patients, 6 of whom had a WAS clinical score of 5, and one with a score of 3. The approach is very similar to that of bluebird, both use self-inactivating HIV-based lentiviral vectors to introduce a functional copy of a mutated gene into CD34+ blood cells collected from the patient & then reinfuse them after the patient undergoes myeloablative conditioning. Both this and the previous WAS study using this vector showed good safety, with "highly polyclonal" insertion sites, and no evidence of clonal expansions. I plan on writing up a longer post more specifically on the safety of these types of gene therapy approaches. In terms of efficacy, here, as well as the previous paper, demonstrated the ability of this approach to improve clinical symptoms. All 3 of the milder patients treated in the previous study broadly had resolution of the multitude of symptoms typical of WAS. In the current paper, 6 of 7 patients had significant improvements in their symptoms, with 1 patient unfortunately dying due to a pre-existing viral infection a few months after receiving the gene therapy. The 6 patients who were able to be followed for up for longer periods of time had the majority of their symptoms resolve or significantly improve, with the exception being low platelet levels remaining relatively low. The authors tried to correlate the improvement specifically for platelet levels to the cell dose the patients received, but the correlation was not very impressive for me to make that conclusion. They also showed that the average vector copy number of peripheral blood cells stayed relatively constant from 10 months out to ~45 months for the longest followed patient. My general thinking has been that vector copy number (quality of the cells) is more important for the peak expression of the gene product, whereas the cell dose (quantity) might correlate more with how fast they recover post-transplant & reach peak expression of the gene. For instance, the cell dose will most likely be lower in bluebird's sickle cell trial vs. B-thalassemia trial because the CD34+ cells need to be collected from bone marrow instead of mobilized blood, which generally results in fewer cells. In general, though, this paper further demonstrates safety and efficacy of a lentiviral-transduced HSC gene therapy for a blood disorder, WAS.

Mutant MHC class II epitopes drive therapeutic immune response to cancer

http://www.nature.com/nature/journal/vaop/ncurrent/full/nature14426.html 

Neoantigens are increasingly being realized as the key targets of the endogenous immune response against cancer. A number of others, such as @PDRennert & @MikeNGladstone have written on neoantigens in cancer here, here & here. Neoantigens are formed from the specific mutations in a patient's tumor. Our ability to predict & validate the immunogenicity of the mutated proteins, which can get processed into peptides that get displayed to the immune system on MHC class I & class II proteins, appears to be improving. Schumacher and Schreiber wrote a fantastic review in Science discussing neoantigen prediction and targeting in cancer immunotherapy. Since neoantigens are specific to each patient's tumor, treatment needs to be individualized to each patient. A number of approaches are emerging, such as adoptive transfer of Tumor-Infiltrating Lymphocytes (TILs), reviewed here, where neoantigen-specific T-cells have been shown to be sufficient to mediate regressions. Kite Pharma has also announced their desire to generate patient-specific neoantigen recognizing TCRs, and I wrote some thoughts on the approach here. Finally, groups are also developing vaccine-based approaches to target neoantigens, with early results from a clinical trial showing expansion of the TCR-repertoire against vaccinated neoantigens. That trial used dendritic cells pulsed with predicted neoantigen peptides for binding to MHC class I, to elicit a CD8+ T-cell response. All of this is good background for this paper, which instead used an RNA-based vaccine to induce expression of predicted neoantigen peptides that could bind to either MHC class I or class II for mouse cancer models. Some interesting findings were that they found a higher proportion of the immunogenic peptides able to bind MHC class II, and eliciting a CD4+ immune response that was able to suppress tumor growth. While the majority of the focus for neoantigen targeting has been on those that bind MHC class I, and elicit a cytotoxic T cell (CD8+) response, there is evidence from adoptive transfer of neoantigen-specific CD4+ T cells that CD4+ cells can mediate responses in patients. The final format of their vaccine was to simultaneously target 5 different predicted neoantigens (predicted to bind either MHC I or MHC II) expressed off of a single RNA transcript. This RNA pentatope vaccine significantly suppressed mouse tumor growth, with evidence of an increased anti-tumor immune microenvironment.

RNA Pentatope Vaccine

Interestingly, they found the vaccine induced an increased CD8+ response against a previously characterized MHC I neoantigen that was not directly targeted by the vaccine. This suggests the vaccine induced antigen-spreading, where an initial immune response against the tumor triggers an immune response against additional antigens. When looking to apply this approach to human tumors, the group looked in the TCGA database to see if human tumors would have sufficient MHC I or II neoantigens based on their prediction method, and they found that there were frequently numerous neoantigens that could potentially be targeted across a range of malignancies. The approach is being commercialized by BioNTech, which has started clinical trials around this approach, including a phase I here.

TP53 loss creates therapeutic vulnerability in colorectal cancer

http://www.nature.com/nature/journal/vaop/ncurrent/full/nature14418.html 

This paper took a clever approach to target the "undruggable" loss of p53, which so frequently occurs in tumors. While tumors frequently have genetic deletions of one copy of the TP53 gene, this deletion tends to span greater than just the TP53 gene. The authors looked for other genes nearby that might get caught in the deleted region, looking for ones that might be "housekeeping" genes that are necessary for cell viability. They found POLR2A, a component of RNA Polymerase II to frequently be co-deleted with p53, and speculated that this might cause increased sensitivity to Pol II inhibitors.

When they tested cell lines with hemizygous loss of POLR2A, they found those cells were more susceptible to inhibition of Pol II by knockdown of POLR2A, or by inhibition with α-amanitin, the deadly fungal toxin. α-Amanitin causes liver toxicity, so they tested an EpCAM antibody-α-amanitin conjugate to improve the therapeutic index. They found this ADC had activity in a number of xenograft models, specifically when POLR2A was deleted. Targeting the unintended consequences of the process of tumorigenesis, such as the previously mentioned neoantigens, or in this case deletion of passengers, is an interesting approach since it increases the number of potential targets. These mutations, however, might not have the selective pressure to be maintained in the tumor, potentially allowing for escape. For further reading, here is a very similar approach that targeted the vulnerabilities exposed by homozygous deletions of redundant housekeeping genes.

Monoclonal antibodies against GARP/TGF-β1 complexes inhibit the immunosuppressive activity of human regulatory T cells in vivo 
http://stm.sciencemag.org/content/7/284/284ra56

The successes of immune checkpoint inhibitors, such as anti-CTLA4 or anti-PD1 antibodies have caused a surge in the search for ways to re-activate the endogenous immune response against cancer. A key immunosuppressive cell thought to be important in tumor progression is the regulatory T cell (Treg). Additionally, part of the mechanism of CTLA-4 antibodies may be through antibody Fc mediated killing of CTLA-4 positive regulatory T cells. While a lack of regulatory T cells causes severe autoimmunity, in mouse models, transient depletion of regulatory T cells reduced tumor growth, while not inducing substantial autoimmunity. This was accomplished with mouse genetic engineering, and so this strategy cannot be directly implemented in humans. Targeting Tregs appears somewhat difficult due to lack of a good surface marker to target. This paper sets out instead to try to inhibit their immunosuppressive activity. It's been previously shown that TGF-β1 is an immunosuppressive cytokine that is produced in an inactive state, but can be activated by Tregs. Direct targeting of TGF-β1 might be dangerous due to important roles in other cell types and situations, including being actually tumor suppressive in premalignant cells. To specifically target Treg activation of TGF-β1, the authors attempted to block the interaction of TGF-β1 with GARP, a protein only Tregs use for activation of TGF-β1. They generated 31 monoclonal antibodies against human GARP, and found some that could block the interaction & significantly diminish human Treg immunosuppressive activity in vitro and in vivo, using a model of GVHD. They couldn't test the antibody's ability to reduce tumor growth in any mouse models of cancer because the antibody doesn't cross react with mouse GARP, so they couldn't inhibit mouse Tregs. This, and potentially other, approaches to either depleting Tregs or inhibiting their function will be interesting to follow as potential cancer therapeutics. 


Other interesting papers this week:

PET probe for imaging of immune cells to track immune responses: 

Noninvasive imaging of immune responses

http://www.pnas.org/content/early/2015/04/17/1502609112.abstract

Two papers on microdevices that can simultaneously inject very small doses of different chemotherapeutics into tumors, in a spatially-separated fashion, to predict chemotherapy response:

An implantable microdevice to perform high-throughput in vivo drug sensitivity testing in tumors
http://stm.sciencemag.org/content/7/284/284ra57.short

A technology platform to assess multiple cancer agents simultaneously within a patient’s tumor
http://stm.sciencemag.org/content/7/284/284ra58.abstract

 

by Dan Marks

How low must HBsAg levels go?

by Robert Kruse


I wanted to write some commentary on HBsAg levels and the therapeutic goal for knockdown that must be achieved in order to have clinical responses. This is in response to some speculation from the current ILC 2015 meeting. I wanted to say as a disclaimer that I believe that siRNA and antisense drugs (Tekmira, Isis, Arrowhead, Analym) could very well work against HBV and mediate clinical response; I am just more apprehensive about the certainty of its clinical success. I also do not own any stock in the companies referenced here, and this is not meant to be investment advice, but a scientific discussion.

My fellow HBV commentator at the RNAi Therapeutics Blog suggests that if levels of HBsAg under 500 IU/mL are achieved, that a synergy between interferon and siRNA could be achieved and cure could happen. He suggested that a month treatment of siRNA could get this low level, and then interferon could finish the job.

Let's assume that theory is right. It makes me nervous that even in patients under this magical threshold, they only had 9% cure rate after an entire year of interferon treatment, which most patients can't even tolerate. Does that mean the siRNA combo plus interferon will only have a 9% cure rate? I would hope not, otherwise the therapy would not be very efficacious. Since siRNA knocks down all viral genes, there is the hope that other mechanisms could be at play that would increase this cure rate.  (edit: @RNAianalyst clarified reported data was for all patients with different HBsAg levels, so 9% cure rate was for combined group. In the link below though, HBeAg negative patients with HBsAg levels of ~200 IU/mL and pegasys and adefovir treatment had 17% cure rate (defined as HBsAg loss) at 2 years after 48 weeks of treatment. Definite improvement over non-responders with higher levels of HBsAg, but not still not a home run like HCV blockbuster medications)

I also want to comment on applying a retrospective clinical cohort analysis to prognostications about the potential of this therapy. The study by Dr. Joerg Petersen is likely true and is similar to other clinical reports I've read on pubmed before, that patients with baseline HBsAg that is lower will respond to treatment better. The problem is that if you compare a patient with 5000 IU/mL vs 500 IU/mL, there are likely a host of differences in their baseline status, namely different levels of ongoing immune responses and/or viral genotype.

The lower HBsAg patient possibly has a more vigorous ongoing polyclonal T cell response against HBV that is keeping the levels of HBsAg low - there must be a reason for the difference in the first place after all. It is likely then that adding interferon to this patient can help push that person's immune system over the top. In the example above, the HBeAg negative patients are negative for that antigen because of a more active immune response against HBV, which lent themselves for a 17% cure rate with IFN added versus HBeAg+ patients who had an 11% cure rate (differing levels of HBsAg modified this as well). Indeed, the type of immune response a person has greatly influences the effect of IFN on the course of HBV.

This situation is analogous to how checkpoint inhibitors for certain tumors are very efficacious, likely because the tumors are more immunogenic, so there is a better immune system at the site ready to fight the tumor. The checkpoint inhibitors can easily push the anti-tumor T cells over the top. On the other hand, the patient with the 5000 IU/mL baseline, even if artificially induced to have lower HBsAg levels to 500 IU/mL, still has a poor ongoing immune response that might not be boosted well by interferon. It's certainly possible that the HBsAg drop could help DC presentation and augment T cell polyclonality, but I don't know how long that might take and I'm not sure if anyone in the field does either. It's possible that their viral antigen specific T cells are already exhausted against the given epitopes and are beyond recovery. The siRNA is analogous to holding the target still, interferon could be bullets, but if the patient doesn't have a gun, ie effective adaptive immunity, it all might be a waste.

There is also the black box of how different viral genotypes secrete HBsAg at different levels, and different viral genotypes respond to interferon differently. This has been reported in several papers on pubmed, and I wonder how this might influence the goal biomarker levels of any siRNA treatment. I would guess in Dr. Peterson's previous study that the patients were all controlled for the same genotype (typically these studies are at centers in a certain area, so the geography dictates similar genotypes), but this is one confounder to look at in any future clinical reports.

In summary, I believe siRNA and ASO drugs are definitely exciting therapeutic modalities and potential game changers for HBV treatment, since they offer a new mechanism to fight the virus. I just want to add caution, though, to benchmarking certain data points and prognosticating success. HBV is one virus, but with a heterogeneous course across patients, the natural history of which is still being learned. The key immunological studies being discussed above have been limited since we can only really gain relevant data from chimpanzees and humans. It will be fascinating as an HBV researcher to see if knockdown could truly reactivate the immune system in all patients. Even if it doesn't in all, there could be a treatment cohort where a 100% cure rate is achieved, like the patients with better baseline status (<500IU/mL) discussed above. However this goes, it will at least be interesting from a basic science perspective.

LEM-onade for the CAR ride

by Robert Kruse

Scientists were buzzing last week with a Science paper featuring a previously unknown protein with key functions in the T cell response against infectious disease and cancer. The protein, LEM, regulates metabolism inside T cells. It was found through a genetic screen of for mutations that increased the T cell response to a chronic viral infection, LCMV, in mice. These mice were found to have a point mutation that led to increased stabilization of LEM mRNA and thus higher LEM protein levels. Interestingly, these mice also had improved immunity when inoculated with the B16 mouse melanoma line, resulting in weakened tumor formation.

The team at Imperial College London has filed two patents on the technology and established a startup called ImmunarT, revolving around adoptive immunotherapy. The startup looks to join the legions of Juno, Kite, Novartis, Cellectis, and more toward chasing adoptive T cell therapy. Given the cloudy IP situation around chimeric antigen receptors, it is possible the company can incorporate that technology and combine it with a vector overexpressing LEM inside T cells. Alternatively, this approach could also be applied to TILs or TCR-engineered T-cells. These T cells in theory should be more active than their normal counterparts. A key question will be whether long-term overexpression of LEM will be toxic to T cells, thereby nullifying the effect. Just because LEM is important at a defined step in T cell maturation and the immune response, doesn't mean that LEM would continue being beneficial. The researchers estimated human trials in 3 years, after more testing in mouse models

Beyond the discovery of LEM, this suggests other spheres of T cell optimization for therapy that aren't being targeted today. Currently, strategies to replace cytokines, guard against immunosuppresion, inhibit apoptosis, and help degrade extracellular matrix have been added to T cells to help improve performance. The LEM paper suggests optimizing cell metabolism as another route, merging biochemistry and immunotherapy. It is possible and likely that other metabolic targets exist, which may include proteins much more familiar in the field which could be overexpressed in T cells. A caveat of the LEM discovery is that it might be functionally redundant with existing improvement mechanisms mentioned above. For example, cytokine signaling could easily active expression at LEM's promoter, since cytokines are one of the master drivers of T cell activity. While an exciting target then, this story could be much ado about nothing. More microarray and pathway analysis is needed to define the effects of cytokines and TCR signaling on T cell metabolism, particularly in regards to LEM. If this research finds that LEM is upregulated by existing mechanisms, then the existing big-time CAR biotech companies should rest easy with their IP situation as is.

What's a good biomarker for HBV clinical trials?

by Robert Kruse

The clinical success of HCV cocktail drugs catapulting Gilead's and Abvie's stock the past year has led to increasing investor focus on potential HBV blockbuster drugs. The current therapies, reverse transcriptase inhibitors like entecavir, inhibits viral DNA production in the serum. The multi-log reduction in viral load, however doesn't lead to SVR in most patients. The viral genomes (cccDNA) inside the liver are stable, and viral antigen production is thought to continue suppresses the immune system.

For the investor, scientist, and physician, there are three principal biomarkers that should be noted in evaluating any preclinical or clinical HBV data. They are HBV surface antigen (HBsAg), HBV DNA (or titer), and HBV e antigen (HBeAg). These can all be drawn from the bloodstream of patients and evaluated by ELISA and qPCR. Liver HBV markers would be ideal but require patient biopsy, which is outside of the normal physician treating practices for HBV and is an expensive procedure compared to routine blood draws. Ideally though, they could be used to assess that the viral genome is reduced or inactivated in the liver, which is necessary for cure. Note that I said inactivation, because it has been shown that HBV cccDNA (circular HBV genome) in patients with resolved HBV infection from an acute scenario still possess remnants of HBV DNA that is identifiable by PCR. This suggests that epigenetic inactivation of cccDNA in addition to direct degradation mechanisms are at play during resolution of HBV infection.

HBsAg has received the most attention in clinical trials, particularly for HBV knockdown efforts with oligonucleotides (Isis with antisense and Arrowhead/Tekmira/Alnylam with siRNA) or with polymers (Replicor). The loss of HBsAg has been hypothesized to relieve immune suppression, that would then activate anti-HBV immunity. This hypothesis largely comes from the observation that HBsAg decreases when a person spontaneously clears infection in rare events. The problem is the chicken or the egg in this deduction. Did some type of seroconversion and antibody production lead to HBsAg clearance (as mimicked by these therapy strategies), or did a T cell response against a different HBV antigen or even a non-related liver disease, lead to these cases of resolved HBV.

One issue with current biomarkers is that HBsAg levels vary widely among patients, depending on disease status and serotype. Indeed, it is unknown what is a sufficient level of HBsAg production in order for the virus to suppress the immune system and HBsAg fulfill its other roles. What is known is that HBsAg can be reduced greatly without any impact on DNA levels. HBsAg is the envelope protein for the DNA virus, but is also secreted in the form of empty particles as well, which outnumber DNA containing particles by 1000 to 1. Therefore, a reduction in envelope won't remarkably change virion production. Without a reduction in DNA titer, then the levels of virus in the liver are never going to change and virus can continue to spread.

It seems then, that looking at DNA levels could be a better biomarker, since it is driven by a combination of viral genes and might offer a more accurate readout of the levels and/or suppression of the HBV genome. For example, while HBsAg levels might fluctuate for secretion inefficiencies and/or promoter mutations, the DNA levels are more accurately driven by the core promoter, which is sensitive to interferon suppression by an ongoing immune response. This is key since the majority HBV therapies rely on the activation of an immune response, directly or indirectly. If HBV DNA is suppressed in the serum dropping multiple logs, it means that genome amplification in hepatocytes is decreasing and no new hepatocytes can be infected.

The HBeAg marker could also be useful, being driven by the same core promoter region. However, since HBeAg seroconversion happens in a good fraction of patients, and they continue to have chronic infection, it might not be relevant for people looking for signs of viral cure. That said, the consensus among ID doctors is that HBeAg seroconversion is a good clinical sign, and it is correlated with lower HBV DNA levels, emphasizing the importance of that marker. Various epidemiological studies have shown that higher HBV DNA levels are associated with higher risk of HCC and cirrhosis. While similar reports have high HBsAg levels as a risk factor in certain settings, this is somewhat inconsistent since other forms of HBV has little HBsAg expression and still see signs of HCC and cirrhosis development, notably occult HBV.

Regardless of marker, a definitive cure for HBV should show significant multi-log reduction in all HBV markers, that is sustained off treatment. The first few patients will be examined by liver biopsy for HBV, just like the Berlin patient was biopsied in various sites for the presence of HIV. That said, in these early clinical trial reports, its much more opaque looking at viral readouts to see potential signs of response, particularly when dosing hasn't been optimized. The same is true for any surrogate studies in woodchuck or duck models. The therapeutic play for HBV won't be avoidance of clinical morbidities like HCC, since that takes many years to manifest, but rather a feast or famine for viral cure. I am personally wary of any studies that show drops in HBV levels in chimpanzees, and don't extend it out more months to see what happens, since it is during that time that the immune response could manifest and should be reported. If efficacious, one expectation could be an intermittent drop in HBV DNA caused by therapy, followed by another multi-log dropped precipitated by the immune response. An effective immune response against HBV is actually largely non-cytopathic, making ALT readouts for immune activity potentially confounding. However, HBV DNA levels are very sensitive to immune responses, making those readouts better.

This analysis can't address the validity of the HBsAg reduction hypothesis (ie how much to reduce in order to see the immune system reactivate?), but it does suggest that HBV DNA levels should be monitored more closely as a sign of treatment efficacy for all modalities. Of course, the confounder is that we have an effective treatment in reverse transcriptase inhibitors, meaning many patients in trials might be on entecavir, for example, and have undetectable DNA levels. This valuable marker is lost then. It is suggested here that unless necessary, trials with drugs as a monotherapy could provide a cleaner readout for HBV antiviral potency.

With the advent of new players (see OnCore and Tekmira merger) in the HBV therapeutic space, more attention will need to be assessed on phase I and II clinical data, and we will continue revisiting this topic of relevant HBV biomarkers here at Biotechr going forward.

Papers of the Week 4/13-4/19

Some interesting papers from last week (April 13th-19th):

Studying clonal dynamics in response to cancer therapy using high-complexity barcoding

http://www.nature.com/nm/journal/vaop/ncurrent/full/nm.3841.html

This was a very cool paper from Novartis that was able to investigate the question of whether resistance mutations for kinase inhibitors are pre-existing, or spontaneously evolve during treatment. They were able to do this by "barcoding" the cells by inserting semi-random 30 nucleotide stretches of DNA into tumor cell lines that would essentially mark each cell within the population uniquely. They tested cell lines that can become resistant to an EGFR inhibitor (Erlotinib) in an EGFR-driven cell line or an ABL1 inhibitor (Imatinib) in a BCR-ABL driven cell line. In multiple replicates of experiments selecting for eventual resistance to each inhibitor, they found the same barcodes surviving in the repeats - which suggests that those cells had an innate resistance to the inhibitors, and did not evolve the resistance through mutation during treatment. One of the ways people hope to get much more durable responses with kinase inhibitors is to combine them, much like what has been done with anti-viral medications for diseases like HIV. If the resistance mutations are already present at low levels in the population, it would be much less likely for one cell to have simultaneous mutations that would confer resistance to both (or more) drugs. For this to work, there needs to be as few or no mutations that overlap in conferring resistance to both drugs. The barcoding system allowed the authors to test for overlapping resistance mechanisms, which they tested with new allosteric inhibitors of ABL1 in the same BCR-ABL cell line. They found the barcodes in resistant populations were largely distinct from those in imatinib-resistant cells. This suggests that the resistance mechanisms are largely non-overlapping, and that the combination of inhibitors could be more successful than either alone. Novartis is currently testing an allosteric ABL1 inhibitor in a phase I clinical trial. This approach of barcoding cells could be quite useful in determining resistance mechanisms, and in finding the best drugs for combination.

Further reading:
Paper from Martin Nowak using mathematical modeling to predict resistance to targeted inhibitors & why combinations of inhibitors with mutually exclusive resistance mechanisms are key to durable responses: http://www.ncbi.nlm.nih.gov/pubmed/23805382

Combination therapy being beneficial is also playing out in the clinic with BRAF + MEK inhibitors > BRAF alone in melanoma: http://www.nejm.org/doi/full/10.1056/NEJMoa1406037
The beneficial effect of this combination is probably not as strong due to multiple overlapping mechanisms of resistance: http://www.ncbi.nlm.nih.gov/pubmed/25673644

STING agonist formulated cancer vaccines can cure established tumors resistant to PD-1 blockade

http://stm.sciencemag.org/content/7/283/283ra52.full?utm_campaign=email-stm-toc&utm_src=email

Aduro timed the publication of this paper well, as it was released the same day as their IPO. They recently partnered this program with Novartis for the development of CDN STING agonists.
This paper details the use of synthetic Cyclic Dinucleotides (CDNs), previously shown to play an important role in the immune response to pathogens. This paper used a co-formulation of CDNs with the GM-CSF secreting vaccine Aduro had previously combined with their modified Listeria vaccine in clinical trials. They named this co-formulation STINGVAX, and administration of STINGVAX had activity by itself against a variety of mouse tumor models. They saw increased CD8+ T-cell infiltration of tumors & an increase in PD-L1 expression on the tumors themselves. Not surprisingly, they tested the combination of STINGVAX with PD-1 blockade and found significantly improved activity in models where either treatment alone had only modest effects. STING agonists look like an interesting and distinct method for stimulating an anti-tumor immune response, so interesting that Novartis was willing to pay over $200 million upfront to partner with them on preclinical data alone.

For further reading on the STING pathway, check out some of the work from Thomas Gajewski's lab, which has previously published on STING in cancer:
http://www.ncbi.nlm.nih.gov/pubmed/?term=sting+gajewski

More on STING agonists from Aduro:
http://www.aduro.com/pipeline/cdn/ 
http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3991150/

The protein LEM promotes CD8⁺ T cell immunity through effects on mitochondrial respiration

http://www.sciencemag.org/content/early/2015/04/15/science.aaa7516.abstract

Impressively, this paper not only discovered and characterized a new protein, LEM, important in T-cell biology, but also suggested it may have uses in cancer therapy. They discovered LEM through a mouse forward genetics mutagenesis screen for mutations that increased the T-cell response to a viral infection model, LCMV. LEM upregulation not only improved the T-cell response to viral infection, but also to the immunogenic B16 melanoma mouse model. They went on to further characterize the role of LEM in T-cell biology to find it played a role in cellular metabolism in T-cells. Considering the mutation they found mediated its beneficial effects through increased protein production of T-cell protein LEM, the most direct application would be to overexpress it in adoptive T-cell transfer (TIL, TCR, CAR-T). This is, in fact, what the discoverers plan on doing, and have already formed a company called ImmunarT around this approach.

Robert writes more about it here:
http://www.biotechr.com/2015/04/lem-onade-for-car-ride.html

Further reading - Another screen that identified potential targets for adoptive transfer:
In vivo discovery of immunotherapy targets in the tumour microenvironment

Heparanase promotes tumor infiltration and antitumor activity of CAR-redirected T lymphocytes

http://www.nature.com/nm/journal/vaop/ncurrent/full/nm.3833.html

This paper looked at why longer-term cultured T-cells made into CAR-Ts were less potent, specifically looking at the expression of heparanse, and their invasive ability. They found the cultured CAR-T cells had lower heparanase expression & reduced invasiveness. Since heparanase degrades a key component of the extracellular matrix, they wondered if overexpression of herparanase in the cultured T-cells would restore their invasive ability and improve activity against "stroma-rich" solid tumors. They tested this with a GD2-CAR against a neuroblastoma xenograft model, and found improved activity of the T-cells if they overexpressed herapanase. This paper was interesting, as it focused on the quality of the T-cells, something not as frequently discussed, and perhaps overlooked. Additionally it suggests a way to additionally modify CAR-T cells to improve efficacy in certain tumor types.

Complement Is a Central Mediator of Radiotherapy-Induced Tumor-Specific Immunity and Clinical Response

http://www.cell.com/immunity/abstract/S1074-7613%2815%2900123-5

It has previously been shown that radiotherapy actually can improve the immune system's reaction against a tumor, however, the mechanisms of how this happens are not very clear. This paper surveyed what immune-related transcripts were upregulated post-irradiation and found that the complement pathway of the immune system appeared to be activated. This was surprising, as this pathway had previously been implicated as being immunosuppressive in the tumor microenvironment. In response to radiation, however, they found complement components C3a and C5a to be necessary for the immune response post-radiation therapy. Interestingly, they found dexamethasone, which blocks complement pathway activation and is commonly given to patients receiving radiotherapy, significantly reduced the immune effects & efficacy of radiotherapy. This paper suggests new pathways that may be important for inducing an immune response to tumors.

Further reading: Radiation and dual checkpoint blockade activate non-redundant immune mechanisms in cancer from Andy Minn's lab.
http://www.nature.com/nature/journal/v520/n7547/full/nature14292.html 

by Dan Marks

Initial thoughts on CD44v6 CAR-T (MolMed)

Originally written April 15th, 2015


CD44v6 CAR-T

MolMed today exercised an option on a CD44v6 CAR-T:
http://www.molmed.com/sites/default/files/uploads/press-releases/2638/2638_1428894007.pdf

I’ll primarily be referring to this paper describing the CAR construct:
http://www.bloodjournal.org/content/122/20/3461.long?sso-checked=true

This CAR targets a non-mutated tumor antigen, CD44v6. Non-mutated self antigens have previously been a dangerous class of antigens to target (such as Her2). Mesothelin is another one, and we will see some of the data from this target at the upcoming AACR, which I would expect will discuss trying to mitigate toxicity. As a guess, perhaps the Mesothelin mRNA CAR-Ts last a short enough time that you don’t get intolerable toxicity.

On to CD44v6 – this is an antigen expressed rarely in normal cells, mainly keratinocytes and monocytes, but seems to be expressed in a range of tumor types, such as AML, MM, pancreatic, and head & neck cancer. CD44v6 was found to be expressed in a high percentage of AML & MM samples. Importantly, they corroborated previous work suggesting expression is important for tumor growth, as knockdown of CD44v6 reduced tumor seeding and progression. This suggests that it might be more difficult for the tumor to lose CD44v6 expression as a way of becoming resistant to the CAR.

They made the CAR construct based off of a previously used antibody in clinical trials that is specific for CD44v6, called bivatuzumab. Injection of the CAR-T cells successfully suppressed growth of human AML & MM xenografts, although it looks like the first time point for evaluation was 5 weeks after T-cell injection. However, this doesn’t model potential on-target off-tumor effects of the CAR, which is the big concern for this type, and really any type, of CAR.

They looked at expression of CD44v6 in normal tissue and found it expressed, as expected, primarily in keratinocytes & monocytes, both at seemingly similar levels (maybe keratinocytes slightly less), and both considerably less than expression in AML samples.

First they tested skin (keratinocyte) toxicity, and surprisingly found the CAR-T cells did not react against keratinocytes. I was surprised the CAR didn’t have skin toxicity, although this was not tested in vivo, only in co-culture of CAR-T cells with keratinocytes. The antibody used to make the CD44v6 recognition portion (scFv) of the CAR construct comes from bivatuzumab. Previously, this antibody had been conjugated to radionuclides to deliver radiation to patients' head & neck tumors. Some responses were seen, toxicity was myelosuppression (presumed to be due to the circulating radionuclide) & grade 2 oral mucositis (potentially on-target toxicity). A second trial in head & neck cancer tested bivatuzumab conjugated to mertanzine (an anti-mitotic drug). Preclinical work suggested reversible skin toxicities, presumed to be due to expression of CD44v6 in skin, and reversible ones were seen in the first patients enrolled. There was some evidence of efficacy, some stable disease and minor tumor responses. However, at the highest dose, the first patient developed a fatal skin toxicity – toxic epidermal necrolysis, even though this patient had only a slightly higher dose than the previous patients, and a lower cumulative dose. There wasn’t evidence of deconjugation of the toxin from the antibody, as the free toxin has other toxicities that weren’t observed.

The authors of the CAR-T study propose that for somewhat unknown reasons the CAR-T may have a wider therapeutic index than the antibody conjugates. This would be somewhat surprising to me, as it generally seems that CAR-T cells can recognize very few molecules of the target on cells, and that they have one of the narrowest therapeutic indices of antibody targeted therapeutics. Perhaps playing with the affinity of the scFv could allow for broader therapeutic index, but this is the same antibody used here as in the previous clinical studies. Maybe there is something structural about the scFv on the surface of T-cell versus floating & conjugated to toxin, or perhaps keratinocytes can suppress T-cells, but it is surprising there seems to be little to no reactivity of the CAR-T cells against keratinocytes.

The CAR-T cells did react against monocytes, and when injected in mice reconstituted with a human hematopoietic system, the CD14+ monocytes appeared to be depleted relatively rapidly, in ~10 days. While the accompanying preview by Richard Morgan suggested that short term monocytopenia is tolerable, long term it is not tolerable: http://www.bloodjournal.org/content/122/20/3392.long?sso-checked=true

The authors dealt with this toxicity by introducing suicide gene constructs to be able to deplete the T-cells after a period of time, to allow reversibility of the monocytopenia. This seemed to work well, however I’m not sure at what duration monocytopenia become intolerable, and if the CAR-T cells will have significant enough activity by that point. From what I can tell, based on the xenograft, at 5 weeks there was some clear activity, but it only took 10 days for the CD14+ cells to be depleted, so it might be difficult to separate activity from toxicity. Although for the CD19 CAR-T cells in humans, significant anti-tumor activity appears to occur very rapidly, so CD44v6 CAR anti-tumor activity could occur more rapidly in humans.

Some of the most impressive data was with the suicide gene, iC9 (I believe this is the same as Bellicum’s suicide gene construct) – which was much better than expression of the other suicide construct thymidine kinase. iC9 killed both activated & resting T-cells and did so in only a few hours.

So with all these concerns, I would take a wait & see attitude about the application of this CAR-T in the clinic. It is differentiated, which is nice, considering the paucity of targets, however, I am concerned about the therapeutic index, specifically if skin toxicity shows up, and even the monocyte toxicity might be too difficult to separate from the anti-tumor activity. If everything works like in the paper, it could be a useful CAR in the clinic, however, for me, the above are some reasons to be cautious.

Let me know any feedback or questions/comments you have, happy to continue discussion.

Further reading:
Phase I Bivatuzumab Mertanzine trial: http://www.ncbi.nlm.nih.gov/pubmed/17062682
CD44(v6) as therapeutic target: http://www.ncbi.nlm.nih.gov/pubmed/20303742

Some thoughts on neo-antigen targeted TCRs $KITE

Post originally written March 2nd, 2015


$KITE Some thoughts on today’s announcement on neo-antigen targeted TCRs.

http://ir.kitepharma.com/releasedetail.cfm?releaseid=899124

First, I do not study cancer immunology, but I enjoy following & learning about different therapeutic approaches, and approaches such as adoptive transfer of T-cells are very interesting.

Kite Pharma today announced they are expanding their CRADA with the NCI to work on developing T-cell receptor (TCR) product candidates that are designed to target neo-antigens specific to each patient’s tumor. This is a fundamentally different approach from the current wave of TCR product candidates currently being developed by $JUNO or $KITE.

I would break down the current wave of TCRs targets into 3 different types of aberrant protein expression, which are thus visible to T-cells through recognition of new peptide-MHC complexes. The first type is targeting the overexpression of a protein in the tumor tissue, such as WT-1. The second type is targeting the re-expression of a protein that is normally restricted in expression to the immune-privileged testis – the so-called cancer-testis antigens, such as NY-ESO-1. The third type is targeting the expression of a viral protein from a virus that is associated with tumorigenesis, such as HBV or EBV viral proteins.

While the aberrant expression of these different classes of proteins all provide potential targets for TCRs, there is increasing evidence that the endogenous T-cell response against the tumor is actually against neo-antigens, which are produced by the processing of proteins that are mutated specifically in the tumor. This is supported by the observation that Tumor Infiltrating Lymphocytes (TILs), which have also been used as an adoptive T-cell therapy, react against neo-antigens specific to tumor mutations: http://www.ncbi.nlm.nih.gov/pubmed/23690473, http://www.ncbi.nlm.nih.gov/pubmed/23644516. Additionally, it is believed that the T-cells that are reactivated by checkpoint blockade, such as PD-1 antibodies, also are reacting against neo-antigens: http://www.ncbi.nlm.nih.gov/pubmed/25428507.

From http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3920180/

From the press release from Kite, it appears they will explore making neo-antigen specific TCRs, most likely using something similar to workflow C in the figure posted in the previous tweet. This figure comes from a very nice review of TCR, CAR-T & TIL approaches, which is publicly available & I would definitely suggest a read: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC3920180/
This means that each patient would get a TCR product that is unique to that patient in the sequence and specificity of the TCR transduced into their T-cells. To me, this seems like a potentially daunting manufacturing challenge.

Workflow A has actually already been applied, quite successfully, in a single patient, and this was published in Science last year: http://www.sciencemag.org/content/344/6184/641

As far as I know, no data has ever been published on a patient(s) being treated by workflow C, so we’ll have to see how this approach actually plays out. I would keep an eye out for further publications from Rosenberg’s group, as they might publish on further developing this approach.

So why add the extra steps of identifying the TCR sequence and generating viral constructs to make TCR-modified T-cells as opposed to just using a mixture of TILs or purifying a specific TIL population, as in workflow A?

It’s not entirely clear to me what the advantages or disadvantages would be, but here is a list of few that I could think of, influenced by some mentioned in this other review from Rosenberg & colleagues:
http://www.nature.com/nri/journal/v12/n4/full/nri3191.html

Potential advantages of putting TCRs into T-cells vs. just using TILs:
Unique to TCRs:
1. Can put the TCR into a given T-cell population (T-cells are heterogeneous, and some populations may be better for adoptive t-cell therapy than others).
2. Can increase TCR affinity for the antigen, but this could decrease specificity
3. Potentially don’t need as many tumor-reactive t-cells to identify & clone the TCR & put this into patient T-cells → it is possible this may improve the ability to grow a T-cell product, as a TIL product suitable for use was only able to be generated from ~55% of patients with metastatic melanoma: http://clincancerres.aacrjournals.org/content/17/13/4550.full
4. Potentially already have scalable production method for TCR-modified T-cells that this workflow would feed into after a number of additional early steps.

Can be done with TILs (but requires different/extra steps, like workflow A):
1. With TCRs, you know characteristics of the receptor you put in & have a pure population of T-cells with the defined receptor → for TILs, you can achieve something similar by selecting TILs with a certain reactivity, as was done in Rosenberg group’s previously mentioned Science paper: http://www.sciencemag.org/content/344/6184/641
2. Can engineer other properties into the T-cell, such as secretion of cytokines, provide co-stimulation, prevent apoptosis, prevent immunosuppression (knockout PD-1) etc. These modifications could also be made directly in TILs as well, but again, you would be starting from a different population of cells.

Potential disadvantages to generating patient specific TCRs vs. TILs:
1. Only targeting one antigen at a time, so might be easy for selection of tumor cells that lack that antigen – especially true if the antigen is simply a passenger mutation & not driving tumorigenesis (many neo-antigens are probably just passengers). Passenger mutations would likely cost little for the tumor to lose them.

Potential advantage or disadvantage – manufacturing:
If you look at the figure posted in the previous tweet, showing the workflow of the different approaches to generating neo-antigen reactive T-cells, you can see you are adding a number of potential steps after the common step of identifying tumor-reactive T-cells. You have to generate single-cell T-cell clones, sequence the TCR & generate a new TCR retro/lenti viral construct to insert this into T-cells isolated from the patient (at this point it is the same as any other TCR product preparation). Making this feasible at a large enough scale is potentially the biggest challenge to this therapeutic approach.

In conclusion, I really like the approach of targeting tumor neo-antigens, these can serve as kind of a fingerprint of the tumor, as they are specific to the mutations the tumor has accumulated. Additionally, it is clear that the immune system can recognize neo-antigens. It will be interesting to see what approaches emerge as the best ways to generate an immune response against neo-antigens. I’d be happy to discuss more if anybody else is interested.