Cancer Vaccines: Ready for Prime Time or Another False Dawn?
Moderna and BioNTech recently reported positive early clinical data for mRNA cancer vaccines in combination with PD-1/L1 inhibitors, so should we be getting excited by this field - again? Cancer vaccines have been like the Loch Ness Monster or the Yeti: many reported sightings but when people look rigorously, they find nothing. There has been a myriad of false dawns (and a company that I co-founded and ran in the late 90s through mid-2000s, Onyvax, only added to the litany). In a fit of hubris, I even co-authored a paper entitled “Cancer Vaccines: Will We Ever Learn?”1 in which we set out the countless approaches to cancer vaccines, explained why each, in their own way, had failed, and then set out the compelling reasons why the Onyvax approach would overcome all of the previous flaws. Then the first of our Phase 2 trials read out and, sadly, the rest is not history. So, has the new mRNA approach finally cracked this intractable nut or are we experiencing another false dawn?
Cancer biology has historically outwitted the vaccine approach
The earliest approach to cancer immunotherapy was pioneered by a New York surgeon, William B. Coley, whose journey started around the turn of the 20th century when he noticed that some cancer patients who survived serious bacterial infections showed remission of their tumors. Following a series of empirical clinical experiments (which would definitely not pass any IRB today), he worked up a composition that became known as Coley’s Toxins, a combination of heat-killed Streptococcus pyogenes and Serratia marcescens. Coley reported remarkable success with his toxins; in total, it is estimated that Coley himself injected more than 1000 cancer patients and published over 150 papers related to the topic2. Nevertheless, the approach was considered highly controversial and, in any case, radiotherapy and chemotherapy gained ground as primary therapeutic approaches to cancer alongside surgery, and Coley’s Toxins were relegated to the Museum of Obsolete Medications (along with cigarettes as a treatment for asthma, cocaine for hay fever and leeches for headaches). For many years immune-based therapies for cancer remained firmly off the table, and further, it was accepted dogma that cancers were not immunogenic.
Fast forward to the post World War II years, and the “immune surveillance hypothesis” began to gain ground and evidence accumulated to suggest that mutated cells arise continually, most of the time detected and destroyed by the immune system, and that cancers only arise when this process fails, and a malignant cell picks up enough mutations to evade and/or disable the immune system and develop into a clinically detectable tumor. There then followed a set of experiments and trials employing immunostimulants; BCG was one of the agents tested. Although intravesical instillation of BCG turned out to be effective in non-muscle invasive bladder cancer, it turned out not to work in other clinical settings including intratumoral injection where local remissions were observed but without effects on distal metastases. Later work with IL-2 showed some efficacy in melanoma and renal cancers, but the effects were not spectacular, and the toxicity was severe. This gave rise to the school of thought that non-specific immuno-stimulation was necessary (to counteract the immunosuppressive properties of cancers) but not sufficient – an additional targeting agent was also needed. (Ironically, by the 21st century, immune checkpoint inhibitors starting with CTLA-4 and moving to PD-1/PD-L1 blockers, disproved this perspective at least in those patients showing remarkable responses to those agents.)
Starting in the 1990s, tumor specific antigens (TSA) and tumor associated antigens (TAA) started to be identified, and subunit vaccines became the flavor of the decade. Cutting a long story short, although effective in the preventative setting (anti-HPV to protect against cervical cancer – however, this is an antiviral vaccine, not an anti-cancer vaccine even if often misnamed as such), only one single-target immunotherapy showed sufficient activity in the clinic: Provenge. Brought to market by Dendreon, this single target immunotherapy was created by loading autologous dendritic cells (DC) ex vivo with a fusion protein combining prostatic acid phosphatase, a TAA, and GM-CSF, a non-specific immunomodulator. Clinically only relatively marginal, it proved to be a commercial failure. No other single target vaccine has been approved by FDA. Incidentally, the use of loaded DC instead of more conventional vaccination techniques was designed to induce a super-potent form of specific immune stimulation based on the understanding of the immune response at the time. However, this implicitly assumed that the "defects" of cancer immunization lay in the upstream part of the response pathway (where DC act); the efficacy of today's immune checkpoint inhibitors which act downstream (on the effector T cells) indicates otherwise.
The failure of single target cancer vaccines led to the hypothesis that multivalent approaches would be needed as a vaccine targeted to a single epitope would simply result in tumor clones that lacked expression of the target evading the immune response. Such “escape loss variants” would replace the cancer cells expressing the target and, clinically, the patient would be back at square one. The thinking developed that a multivalent vaccine, targeting several antigens at the same time, would reduce the risk of such escape in the same way that antibiotic triple therapy prevents the emergence of resistant strains of bacteria. The extreme exemplar was the use of whole (inactivated) tumor cells or lysates as a source of a huge range of antigens, including the presumed many that had yet to be identified as such. Sadly, none of these made it to the market despite multi cumulative billions invested in the technology.
It can be argued that in the battle pitched between cancer vaccinologists and the tumors, the latter has always come out on top.
Where are we today?
A recent review article started with "Research on tumor immunotherapy has made tremendous progress in the past decades, with numerous studies entering the clinical evaluation."3 This sentence could have been written (and equally valid) in each of the past four decades. The field has seen cell-based, peptide-based, viral-based, DNA-based, carbohydrate-based, heat shock protein chaperones and other modalities in single target through highly multiplex approaches. We have seen simple prime-boost approaches akin to infectious disease vaccines, regular monthly boosters and even heterologous prime-boost strategies. All have failed.
The state-of-the-art at the time of writing may be the use of mRNA vaccines (currently viewed very favorably in the post pandemic era) targeting multiple cancer neoantigens (individualized TSA) with enhanced potency through concomitant use of checkpoint inhibitors. It's clearly too early to tell whether this will finally nail it, but there is a long history in the field of cancer vaccines showing highly promising phase 2 results only to flop in phase 3. Examples (of several more that could be cited) are:
- CancerVax and GVAX, two cell-based vaccines based on allogeneic cancer cell lines in melanoma and prostate cancer respectively;
- TroVax, a virally vectored vaccine against antigen 5T4 in colorectal cancer;
- Rindopepimut, a peptide that spans the length of EGFRvIII in primary GBM.
On this track record, prospects for today's hot candidates are far from certain. Regardless of individual programs, however, there are a number of pitfalls that previous generations have fallen into and which threaten the viability of some current approaches.
Tumor heterogeneity
There is as much variation in the antigen expression profile between tumors at different sites within a single patient as between samples from different patients: that was the startling message from a study in the early 2000s by Ken Pienta and colleagues at the University of Michigan School of Medicine4. In a single melanoma deposit, parts can be black (expressing the melanin pigment) and others white proving that there is heterogeneity even on an extremely local scale. The consequence of this is that for a single target cancer vaccine to be effective, it must be both a) expressed on 100% of target cells and b) be essential to cellular survival so that escape loss mutations would be lethal. Such a Holy Grail antigen has yet to be identified. Even approaches that target neoantigens, TSA that are specific to a biopsy from the individual patient, may not be targeting TSA that represent all of the tumor in that patient, and this may also be true even for neoantigens high in the "evolutionary tree" of the tumor's mutational profile.
It follows that multivalent approaches will be necessary, and it is also highly probable that multi-modal approaches will also be essential for sustained responses or even cures.
Confusing immune responses for efficacy
A cardinal sin that has been repeated by generations of cancer vaccine researchers is to perform a subgroup analysis and demonstrate that the patients with the best clinical responses were the patients who had made the best immune responses. The flaw is to confuse correlation with causation; it is entirely possible that patients who were generally fitter, and had immune systems in better shape, were going to survive longer regardless of the vaccine than those in weaker conditions. Whenever a failed cancer vaccine trial failed is "rescued" by subgroup analysis, you can be pretty sure that failure really meant failure.
Historic controls
Clinical efficacy in single arm trials is to be regarded with a highly skeptical eye. Historic controls are simply never adequate. Of course, in special cases (e.g., CD19 CAR-T complete responses in >90% of last-line patients) there is little room for doubt, but those instances are extraordinarily rare and yet to be seen in the cancer vaccine space.
Indication selection and trial design
Speaking of CD19 CAR-T, what would have happened if that generation of CAR-T cells had targeted a solid tumor antigen? Based on what we have since learned, the approach would have failed, and this would have destroyed any incentive for further exploration of the technology. This clearly demonstrates the need to select very carefully the indication and the stage of disease for any cancer vaccine strategy. Historically, the field has been bedeviled by the dilemma of whether to target late-stage patients, where trial timeframes are manageable but clinical hurdles are very high due to the high burden of disease and the immunosuppressive effects of advanced cancers, and aiming at earlier stages of disease with more realistic clinical hurdles but impossibly long timeframes. This remains an issue to this day.
Tumor plasticity
This is not so much a pitfall as an inherent property of cancer which undermines all therapeutic strategies. By definition, cancer is characterized by genetic instability, and tumors continue to mutate as the disease progresses. It stands to reason that any therapeutic modality will select for the subpopulation of cancer cells that can defend against or evade the active therapeutic moiety, be that a small molecule, an antibody or even a CAR T-cell. That subpopulation then becomes resistant disease that grows back and recreates the original problem - usually in a more virulent state. There is no reason to suspect that any of today's technologies can overcome this hurdle.
Conclusion
As Bruce Booth put it in the decade before last, realizing the potential of cancer vaccines is “full of complexity”.5 This remains as true now as it was then. There is always a risk of "premature evaluation", that is of attempting clinical translation of new science before the technologies are sufficiently mature, but the counterargument is that it is always too early until it's too late. Science keeps surprising us, from evidence that the time of day that a cancer vaccine is administered can determine its effectiveness6 to the significant impact of the precise structural arrangements and sequencing of antigens in a multivalent vaccine7, and there will be no end to ongoing refinements in active immunotherapy design. We keep watching and hoping that cancer vaccines will finally take off (but we are not holding our breath).
Bibliography
1 Johnson RS, Walker AI & Ward SJ (2009) Cancer vaccines: will we ever learn?, Expert Review of Anticancer Therapy, 9:1, 67-74, DOI: 10.1586/14737140.9.1.67
2 Reviewed in Carlson RD, Flickinger JC & Snook, AE (2020) Toxins 12, 241; doi:10.3390/toxins12040241
3 Liu J et al (2022) Cancer Vaccines as promising immunotherapeutics, J. Hematol and Oncol 15:28 https://doi.org/10.1186/s13045-022-01247-x
4 Shah RB et al (2004) Androgen-independent prostate cancer is a heterogeneous group of diseases: lessons from a rapid autopsy program Cancer Res 15;64(24):9209-16. doi: 10.1158/0008-5472.CAN-04-2442.
5 Goldman B and DeFrancesco L (2009) The cancer vaccine roller coaster Nature Biotechnology volume 27, pages129–139
6 Wang, C., Barnoud, C., Cenerenti, M. et al. Dendritic cells direct circadian anti-tumour immune responses. Nature (2022). https://doi.org/10.1038/s41586-022-05605-0
7 Teplensky, M.H., Evangelopoulos, M., Dittmar, J.W. et al. Multi-antigen spherical nucleic acid cancer vaccines. Nat. Biomed. Eng (2023). https://doi.org/10.1038/s41551-022-01000-2