Ecosystems in tumor immunity
The buzzword ‘ecosystem’ has popped like a spring dandelion, and it is now used everywhere in biotech. I’m as guilty as anyone of rapid adoption: the term does capture essential elements of modern biomedical science. Complex and interlaced, with key control nodes at work at all levels – scientific, financial, clinical, commercial – and also dynamic, constantly driving adaptation, and, we hope, innovation. Scientifically the ecosystem connections are easily spotted. CRISPR technology appears in cellular therapies including TCRs and CAR-Ts as we simultaneously learn that the mechanisms of immune checkpoint suppression deployed by tumor cells can derail genetically engineered CAR T cells as readily as normal T cells. Further, those genetically engineered CAR T cells and TCRs owe their existence in large measure to our newly developed ability to sequence tumors at the individual level, with great sensitivity, to identify novel targets. The whole enterprise in turn requires ever faster, cheaper, smaller and more reliable equipment (RNA spin columns and PCR cyclers and cloning kits and desktop sequencers and on and on) and software to handle the data. Enterprises like these in turn drive discovery and innovation.
Within the tumor is another ecosystem – the tumor microenvironment or TME. While TME is a fine term it does blur the notion that this microenvironment is in nearly all cases part of a larger environment and not a walled-off terrarium (perhaps primary pancreatic cancer is an exception, within its fibrous fortress). The tumor ecosystem is a more encompassing term, allowing for the ebb and flow of vastly different elements: waves of immune cells attempting attack, dead zones of necrotic tissue being remodeled, tendrils of newly forming blood vessels, a fog of lactate, a drizzle of adenosine, energy, builders, destroyers, progenitors, phagocytes, parasites, predators. When viewed this way we might wonder how any single drug could treat a tumor, since it is not a singular thing that we attack with a drug, but an ever-changing world we are seeking to destroy.
So it’s hard to do.
Our understanding of the tumor as a complex entity was first informed by pathology, then microscopy, then histology and immunohistochemistry, myriad other techniques and of course genetics, the latter leading to the identification of tumor oncogenes, tumor epigenetics, tumor mutations (referred to above) etc, etc. This ecosystem – that of the cell and it’s mutational hardware and software (genome and exome, or genotype and phenotype) we can hardly claim to understand at all, not matter how many arrows we might draw on a figure for a paper or a review. A few recent examples: we think that tumor cells adapt to immune infiltration in part by engaging CTLA4 expressed on T cells, and when that fails they secrete IDO1, or express PD-L1 on their cell surface, or the tumor cells direct tumor associated cells to do the work for them – maybe monocytes, or macrophages, perhaps fibroblasts, perhaps the endothelium, i.e. the ecosystem. As we know from studying patterns of response to PD-1 and PD-L1 therapeutics, it is hardly so simple, as patients who don’t express the therapeutic target will respond to therapy and patients who express the therapeutic target sometimes, in fact often, will not respond. Which just says we don’t know what we don’t know, but we’ll learn, the hard way, in clinical trials.
The abundance of therapeutic targets and our lack of knowledge is best displayed, with some irony, when we try to show what we do know, as in this figure from our recent paper on immune therapeutic targets:
The picture is static, and fails to represent or visualize complexity (spatial, temporal, random, quantum), and we therefore cannot formulate meaningful hypotheses from the representation. Without meaningful hypotheses we just have observations. With observations we can only flail away hopefully, and be happy when we are right 15 or 20% of the time, as is the case with most PD-1 and PD-L1-directed immune therapeutics in most tumor indications, at least as monotherapies. Why focus so on the PD-1 pathway? Because at least for now, it is the singular benchmark immune therapeutic, stunning really in inducing anti-tumor immunity in subsets of cancer patients.
The success of the “PD-1″ franchise has created another ecosystem, clinical and commercial. The key approved drugs, and the 3 or 4 moving quickly toward approval, are held by some of the world’s largest drug companies (BMS, Merck, Astra Zeneca, Sanofi, Pfizer, Roche). Playing in that sandbox has proven very lucrative for some small companies, and very difficult for many others. There is competition for resources, for patients, for assets and ideas. This has created new niches in the commercial ecosystem, as companies try to differentiate from each other and carve out their own turf – Eli Lilly for example has focused on TME targets, distinguishing itself from other oncology pharmaceutical companies in choice of targets, followed closely of course by smaller contenders – Jounce, with a T cell program directed at ICOS but perhaps more buzz around their macrophage targeting programs, and Surface, whose targets are kept subterranean for now. Tesaro and others are betting on anti-PD-1 antibodies paired rationally with antibodies to second targets in bispecific format. Enumeral is focused on building rationale for specific combinations of immune therapeutics in specific indications, perhaps even for the right subset of patients within that indication. And so on.
Lets imagine you are right now pondering an interesting idea, have a small stake, and want to engage this landscape of shifting ecosystems. What might you do?
Lets start with a novel target. You’ve read some papers, woven together some interesting ideas, formulated some useful hypotheses. The protein has been around, maybe there are patents, but not in the immune oncology space, so you think you might have some freedom to operate. Good, best of both worlds. You dig around, find you can buy your target as purified protein, or find a cell line that expresses the target. Now what? Maybe you would hire an Adimab or Morphosys or X-Body to perform an antibody screen. Different companies, varied technologies, but all directed at antibody discovery. My favorite of this group was X-Body, who had an extraordinary platform to screen human antibody sequences and produce antibodies with really stunning activity and diversity. Juno bought them in early 2015, seeking the antibody platform and a TCR screening platform built with the same technology. I hadn’t seen anything quite so powerful until recently, with the introduction of a novel screening technology from Vaccinex. This platform is about as diverse as the X-Body platform (i.e. ~108 Vh sequences and up to 106 Vl sequences; that’s a lot of possible Vh-Vl pairs). What sets them apart is that the entire selection process happens as full length IgG in mammalian cells rather than surrogates like bacteria or yeast. The net result is a reduction in risk associated with manufacturing. They’ve used it to power their own clinical programs and have selection deals with some big names including Five Prime Therapeutics. Remarkably (I think) you can access their platform to screen targets for your own, i.e. external, use. Their website explains the platform further (http://www.vaccinex.com/activmab/) but here is one nice sample of their work on FZD4 (a nice target by the way):
So now via Vaccinex or someone else you’ve acquired a panel of antibodies that you are ready to test for immune modulatory activity in models that are relevant to immune oncology. You can build out a lab (expensive, time-consuming), find a collaborator with a lab, or find a skilled CRO. The immune checkpoint space was until recently devoid of really focused CRO activity, that is, having diverse modelling capability and careful benchmarking. However, Aquila BioMedical in Scotland, UK placed a solid bet on developing these capabilities around a year ago, and that effort is yielding a terrific suite of assays in both mouse and human cell systems, with multiple readouts, solid benchmarking (e.g. to nivolumab) and careful controls. I like this very much, rich in functional data in a way that a binding assay simply can’t reproduce. Aquila BioMedical seeks to become a driving force in this area, and I like their chances very much: see http://www.aquila-bm.com/research-development/immuno-oncology/ for more information on assays like this IFNgamma secretion assay:
Those are clean and robust data.
Now you come to the point of translation to actual use, that is, targeting an indication. How does one proceed? We can probe the TCGA and other databanks for clues, stare at the IHC data online (not recommended), try to cobble together enough samples to do our own analyses (highly recommended but difficult). The goal is to make some educated guesses about two distinct features of the tumor ecosystem: First, is your target expressed on a relevant cell within the ecosystem (tumor, TME, vasculature, draining lymph nodes, etc) in a specific indication or indications, and second, is that ecosystem likely to respond in a clinically meaningful way to manipulation of your target with your antibody?
That second question is a troubling one. What we are really asking is that we deconstruct the ecosystem and look for clues as to how the therapeutic might impact that ecosystem. What are we looking for during deconstruction? Several things, and they are assessed using diverse techniques, adding to the challenge. First, a highly mutated tumor is more likely to respond to immune therapy, and there are several aspects to these phenomena. One is to understand the underlying genomic changes underpinning the oncogenetics of the tumor: what is driving its ability to outcompete the natural surroundings – in our ecosystem analogy perhaps the tumor can be considered starting out life as an invasive species. Genomic sequencing can accurately identify the mutations that support the tumor, but also a potentially vast array of “passenger” mutations that accumulate when tumors turn off the usual mutation repair machinery. Various algorithms exists that can predict which mutated proteins may be immunogenic, that is, capable of stimulating an anti-tumor immune response. Another method designed to determine if an immune response has in fact be stimulated (and has stalled) is to sequence the mRNA expressed in the tumor: exome sequencing. This will reveal, among other things, what the TCR usage is within the tumor, and that in turn will inform you if there is a very narrow anti-tumor response and a broad one, based on the breadth of TCR clonality. That sounds complex, but really isn’t – suffice to say that a broader TCR response in suggestive of immune potential, leashed T cells awaiting clear orders to attack.
More complex is the nature of those orders, and counter-orders. Various methods are being developed to measure the “quality” of the immune response that confronts the tumor. Are key costimulatory molecules present on T cells that would allow stimulation? Are the T cells instead coated with immunosuppressive receptors? Are the tumor cells masked with inhibitory proteins, are they secreting immunosuppressive factors, have they hidden themselves from immune view by downregulating the proteins that T cells “see” (i.e. the MHC complex). What are the cells within the TME doing? Are they monocytes, macrophages, fibroblasts? Where are the T cells? Within the tumor, or shunted off to the side, at the margin between the tumor and normal tissue? Are NK cells present? And on and on it goes. It seems impossible to answer all these diverse questions.
You might try IHC, as mentioned, or targeted PCR for select genes, and Flow Cytometry to look at the distribution of proteins on various cells, or try deep sequencing. All of this is achievable with equipment, labs and people, or by assembling various collaborators, but all in all, quite a challenge. Very recently an interesting company called MedGenome came to my attention, offering a diverse range of services, starting with neo-epitope prioritization and immune response analyses. These offerings, plus some routine IHC, should give most researchers a comprehensive look into tumor ecosystems, informing indication selection, mechanism of action studies and patient profiling. They explain the technology at http://medgenome.com/oncomd/. This is a schematic they sent me showing their neoepitope prioritization scheme that enriches for peptides that trigger anti-tumor immunity, e.g. in a vaccine setting or perhaps in a cellular therapeutic format.
It’s a good start on democratizing a suite of assays typically available only to specialty academic labs and well-funded biotechs and pharma companies.
So now you’ve gotten your antibodies (Vaccinex), performed critical in vitro (and soon, in vivo) assays (Aquila Biomedical), and analyzed the tumor immune ecosystem for indication mapping (Medgenome).
You’ll have spent some money but moved quickly and confidently forward with your preclinical development program. Your seed stake is diminished though, and it’s time to raise real money. Now what? … now you face the financial/clinical/commercial ecosystem.