This blog was written by Michael Gladstone, Principal at Atlas Venture, as part of the “From the Trenches” feature of LifeSciVC.
“A big asset of VC’s is pattern recognition. It is also often our biggest liability.” An apt observation tweeted by veteran venture capitalist Bob More.
Having just returned from an immunotherapy-fueled AACR meeting, it seems many investors may be at risk of pursuing this particular pattern with a bit too much enthusiasm and a bit too little true novelty.
To preface this, I’ll point out that I love the immuno-oncology field, and I thoroughly believe immunotherapy has incredible promise in the treatment of cancer. It’s the ultimate “personalized” therapy, giving the patient’s own immune system a chance to fight that patient’s unique tumor. This has been covered in much greater depth elsewhere (including in Bruce’s blog), so I’ll stop there.
With that said, here’s a snarky summary of what I think was implied by many AACR presentations and posters:
- Nearly every molecule in anyone’s pipeline is immunotherapy and/or synergizes with immunotherapy
- There’s other stuff going on in cancer research, like epigenetics and targeted therapies, and these things might be interesting too (mainly if you combine them with immunotherapy)
Everything from kinase inhibitors to chemo to vitamin D was posited as immunotherapy. And I actually wouldn’t dispute that most of these things really do impact the immune system in one way or another. But do they do it in a way, or to an extent, that is likely to be clinically meaningful in a definable therapeutic setting?
Evaluating this clinically in almost any single case will be time-consuming and expensive, so evaluating any significant fraction of the proposed options is unlikely to be concluded in our lifetimes (or at least in the lifetime of most current biotech CEOs).
Many of these I/O approaches will fail (besides, of course, Atlas’s investments in CoStim, Surface Oncology, Unum Therapeutics, and F-star). So as an early stage biotech investor, I have to ask myself what in this area has more paradigm shift/quantum leap potential.
Let’s start by revisiting today’s immunotherapies and then consider what developing work is suggesting we consider for future approaches and combos.
How do today’s clinically validated checkpoint therapies work?
Our clinically validated antibody-based immunotherapies work by amplifying the endogenous immune response to tumor antigens. Some combinations of these agents appear to further amplify this response, offering greater efficacy in some settings. But, predictably, they also amplify immune-related toxicity, due to their broad de-repression of the immune system and its sometimes tenuous relationship with self-tissues in the gut, lungs, liver, and other organs.
It remains to be seen how therapies targeting novel T cell coinhibitory/costimulatory targets may compare to those in the clinic today, but it will take lot of time and money before we know. My guess is that while we may see some improvements in the efficacy/toxicity balance in some settings, it will probably remain difficult to uncouple anti-tumor efficacy and treatment-induced autoimmunity to the extent that we’d all like.
So, how do we better uncouple immunotherapy’s efficacy and its toxicity?
My main takeaway this week was that there is increasingly provocative evidence that the adaptive anti-tumor immune response is driven primarily by recognition of somatic neoantigens (and in many cases, a surprisingly narrow range of them), which vary on a patient-by-patient basis as a sort of (evolving) fingerprint of a given tumor. These responses have several advantages over unmutated tumor-associated antigens (TAAs), including true tumor specificity, increased TCR avidity due to lack of negative selection, and improved effector function due to a lack of peripheral tolerance/anergy induction.
This deserves a more thorough discussion than this space (or my own expertise) allows, but in the next few paragraphs I’ll highlight just a few recent pieces of elegant work (much of it presented and discussed at AACR) highlighting the centrality and occasional narrowness of neoantigen-based immune responses.
We’ve known for some time that tumor types with higher typical mutation loads tend to be among the more immunotherapy-responsive (Fig. 1 here was shown probably hundreds of time at AACR). More recently, several papers (e.g., here and here) have demonstrated that even at the patient-by-patient level, increased mutational burdens are correlated with responsiveness to checkpoint inhibitor therapies. Further, we know that an inflamed tumor microenvironment (which promotes presentation of these antigens and immune cell infiltration to respond to them) is also associated with greater responsiveness.
Additional work has delved further into the mechanics of these responses. For instance, a large group led by Robert Schreiber (here) used a genomic/bioinformatics approach to identify neoantigens responsible for the anti-tumor efficacy of checkpoint blockade in a mouse cancer model. Strikingly, a peptide vaccination based on just a single neoantigen recapitulated the checkpoint inhibitor’s efficacy. This shows (at least in mice) that a large degree of anti-tumor efficacy could be mediated by a mono/oligoclonal T cell response, which could be amplified in a highly specific way, reducing (if not obviating) the need for therapies that modulate the whole immune system (and the toxicity that can accompany them). But do we have similar evidence in humans?
We might. Steve Rosenberg’s group (here) at the NCI used a somewhat similar approach to identify a single neoantigen that was recognized by and powerfully activated tumor-infiltrating lymphocytes (TILs) from a human patient with metastatic cholangiocarcinoma. The investigators administered to the patient an expanded population of TILs highly enriched for T cells recognizing this neoantigen, again mediating a powerful and durable clinical response in a nasty metastatic solid tumor.
These specific papers (and associated presentations at AACR) support additional data directionally supportive of this phenomenon. For example, as presented by MD Anderson’s Patrick Hwu, the growing number of patients with deep, durable responses to bulk TIL preparations have experienced very little treatment-associated toxicity. This is quite distinct from the occasionally drastic skin, eye, and ear toxicity seen in some patients treated with TCR-transduced T cells targeting unmutated TAAs expressed in some healthy tissues. Once again, this suggests that anti-tumor efficacy can be derived from truly tumor-specific neoantigens, and immune responses outside of these have uncertain benefit, despite their frequently considerable toxicity.
To me, all of this implies that our current approaches for “removing the break” and “stepping on the gas” on the immune system may still be somewhat heavy-handed. We may indeed find therapeutic windows for some of these drugs/combos for some patients, but in general, we still seem to be activating large portions of the immune system with dubious benefit and occasionally quite meaningful harm. We’re doing this all to derive efficacy that may actually be driven by quite a small fraction of that activated immune system.
Here I’ll literally borrow a page from Hermann Hesse’s terrible book Siddhartha, when the eponymous character in a fleeting moment of clarity proclaims:
“I will no longer mutilate and destroy myself in order to find a secret behind the ruins.”
If Siddhartha were an immuno-oncologist today, he might find himself at MD Anderson, NCI, Memorial Sloan Kettering, or the Netherlands Cancer Institute. What might he be working on?
Perhaps as this early but fascinating work continues to play out, he’d focus on approaches to identify, amplify, or elicit neoantigens. For example:
- Identify the neoantigens to elicit specific immune responses: the Schumacher/ Schreiber approaches are examples of ways to identify a given patient’s neoantigens to design personalized antigen-specific vaccines, TILs, or engineered T cells, decreasing our dependence on broad immune activation
- Generate therapies to improve uptake, processing, and presentation of antigens, to improve the endogenous education and activation of the immune system. Surface Oncology has several programs focused on just this, as do others. Local radiation and intratumoral therapies (TLR agonists, STING agonists, oncolytic viruses, etc.) create, in essence, an in situ vaccination. Aduro’s STING agonists (recently partnered with Novartis) have shown striking preclinical efficacy by further inflaming the tumor microenvironment and promoting antigen recognition and subsequent adaptive immune activation. These technologies and their clinical implementation continue to improve, and they are likely to play an increasingly important role, especially in treatment of poorly-immunogenic tumors.
- Generate more neoantigens: perhaps a bit more out there, but there may be better ways to modulate the processes involved in DNA damage (and its repair) to induce additional mutations, providing the immune system more opportunities to recognize truly tumor-specific mutations to respond to. For example, we already have therapies that inhibit double-strand break repair – in the right setting, might these promote neoantigen generation and give the immune system more to work with?
Just a few ideas here to stir the pot. Entrepreneurs with additional ideas – please reach out to us at Atlas, we’re always happy to discuss new approaches!