Broad-Spectrum Cancer Treatments
A proposal for a research program
Introduction and Principles
The conventional wisdom is that we can’t “cure cancer” because every cancer is different.
And it’s true that cancer is genetically and biochemically diverse, and that targeted cancer therapies generally only work on very specific sub-populations.
But some of the oldest, most reliable, and still most commonly used classes of cancer treatment -- radiotherapy and cytotoxic chemotherapy -- are also broad-spectrum. That is, they’re effective on many (though not all) types of cancer, based on selectively targeting properties that many cancers have in common (their rapid cell division).
Today, we rarely search for new treatments that work on many types of cancer.
Partially this is because of the belief that we won’t find any, and partly because cancer treatment is so valuable that individual researchers and biopharma companies are rewarded for any successful cancer research program even if it has narrow applicability. Individual incentives point towards highly targeted therapies and small patient populations.
On the other hand, of course, a single drug that’s effective on, say, half of cancer patients would be much better for the world than a drug that only helps a few thousand people.
Why should we believe broad-spectrum cancer treatments are possible?
Any characteristic that distinguishes a wide range of cancer cells (across different tumor types and mutational signatures) from healthy cells can potentially serve both as a diagnostic test and a therapy, by either labeling cancer cells, or by combining with a cytotoxic agent to selectively destroy them.
The “hallmarks of cancer” provide a framework for thinking about what capabilities cancers have in common, and there are lots of them: proliferation, apoptosis resistance, angiogenesis, metastasis, etc.
We may also think of cancer cells as dysregulated and generically more vulnerable to certain types of generic stresses than healthy cells. Many chemotherapies introduce genotoxic stress (DNA replication damage) or oxidative stress; hyperthermia-based therapies introduce heat stress. Cancers also appear to be selectively vulnerable to mechanical stress1, electrical stress, and nutrient-deprivation stress.
New broad-spectrum cancer therapies are rare but are occasionally discovered.
One recent example is AOH1996, which causes damage to cancer cells during DNA replication. It shows toxicity in over 70 cancer cell lines but has no significant effect on healthy cells (and thus has much higher selectivity, which generally means a better ratio of effectiveness to side effect severity, than older cytotoxic chemotherapies). It also blocks tumor growth in 4 types of mouse tumor models.2
What else is out there like AOH1996? And how could we find such things?
In the rest of this post, I’ll give some evidence for hope, from several directions, that broad-spectrum cancer treatments are possible.
This basically adds up to a research program (or a collection of research programs) that aims to actually cure cancer, as time- and cost-efficiently as possible. I could imagine this being shaped like a new philanthropic fund, a new research center, a new ARPA-H program, or other structures.
New Ways To Attack Cancer Cells
There are a number of intriguing possibilities that have emerged in recent years, potentially allowing us to systematically distinguish cancer from healthy cells, across many or all cancer types and healthy cell types.
Some of these are more tentative or speculative than others; I’m not claiming, of course, that all of them will pan out. But hopefully they’ll suggest some research directions that focus on what cancers have in common, rather than how they differ.
Historically “Undruggable” Cancer-Causing Mutations
The most commonly altered genes, across cancers, are traditionally considered “undruggable.”
Why?
It’s easiest to develop a drug to bind to a protein on the surface of a cell.
But the first mutations that drive a cell towards cancer are genes that code for proteins inside the cell, where it’s harder for drugs to reach. It’s transcription factors — genes that affect the expression of other genes — that remove the genome’s protections against cancer. And these are generally expressed in the nucleus (where the DNA is) or the cytosol (the interior of the cell.)
The biggest oncogene of all is p53. 50-60% of human cancers contain mutations in the p53 tumor suppressor gene; but, because p53 is a nuclear protein, it is difficult to target pharmacologically.3
Likewise, out of 20,066 human primary tumors, 94% have mutations in tumor-suppressor genes and 93% in oncogenes, many of which code for cytosolic or nuclear proteins that are also rarely targeted by cancer therapies.4
Cancer vulnerability screens, using tools like CRISPR or RNAi, give us causal information about which gene knockouts/knockdowns result in cancer cell death. When combined, we find that there are many “pan-dependency” genes, whose knockouts/knockdowns cause cell death in all 403 cancer cell lines, including some well-known oncogenes like Myc and BRD4 and tumor-suppressor genes like CDK1.5
In other words, the mutations that are causally upstream of most cancers — the mutations which unlock the other mutations that make cancer cells proliferate uncontrollably, metastasize, resist apoptosis and immune attack, and generally behave like cancer cells — the mutations that are the most effective intervention points for killing essentially all types of cancer — are mostly untouched by any existing cancer drugs.
But that can change.
With the rise of gene therapy and mRNA vaccines, we have developed a lot of delivery mechanisms to get therapies inside cells, from viral vectors to lipid nanoparticles. The same sorts of nucleic acid payloads that can selectively kill cancer cells in the lab by detecting activated oncogenes or mutant tumor-suppressor genes, could be packaged inside modern intracellular delivery mechanisms and sent to kill cancer cells in patients.
Phospholipid Ethers
Phospholipid ethers, normally found in cell membranes, selectively accumulate in many types of cancer cells relative to healthy cells.
This is because cancer cells are prone to lipid rafts, clumps on the cell membranes which contain these fatty compounds.
This means we can use the distinctive abnormalities of cancer cell membranes to selectively identify — and kill — cancers.
A particular phospholipid ether analog, studied in 2014, showed tumor uptake in 92% of rodent tumors (xenograft, mutant, and naturally occurring wild-type) across more than 20 tissue types, as well as 4 human tumors.6
This selective tumor uptake implies the possibility of both tumor imaging/detection and tumor targeting, using a single method for a wide variety of cancer types.
Endogenous IgG
Immunoglobulin G is a family of antibodies, each specific to a different antigen that the body identifies as potential “invaders”. The striking fact is that if you look at all the IgGs, regardless of target, they cluster around tumors — all tumors, regardless of cancer “type” — much more than around healthy tissues.
Endogenous IgG is more concentrated in malignantly transformed organs than in healthy organs, across five different mouse cancer models, localizing to the same regions recognized as pathologic upon histological analysis.7
Patient-derived tumor cells are frequently coated with IgGs, in the majority (80%) of a range of cancers, while samples from healthy tissues rarely are.8
Once again, a particular molecule (or class of molecules) that’s common across cancers and rare in healthy cells can be used as either a cancer detection or cancer targeting strategy.
One could imagine, for instance, an antibody-drug conjugate that attaches a cytotoxic (cell-killing) drug to an IgG-detecting antibody, so that cells coated with a lot of IgG will be killed at a higher rate.
Negative Surface Charge
Most cancer cells are negatively charged.
This is a simple electrochemical distinguishing characteristic of many types of cancers. The negative cell surface charge can be measured by the accumulation of positively-charged nanoparticles around the cell.
Out of 22 randomly-selected cancer cell lines and 4 types of primary normal cells, all cancers accumulated positively-charged iron nanoparticles (which can be attracted with a magnet), while no healthy cells did.9
While positively charged nanoparticles carry toxicity risk in vivo,10 they can still potentially be developed as highly specific diagnostics for detecting cancer cells in blood or tissue samples.
Membrane Depolarization
Cancer cells, relative to healthy cells, have depolarized cell membrane resting potentials.11
This is a very old and reliably observed effect. As far back as 1938, it was even reported possible to detect tumors with a voltmeter on the surface of the body.12 (Though Michael Levin has reported his lab was not yet able to replicate that result.)13
Metastatic phenotypes can be induced by causing membrane depolarization, while hyperpolarization can cause tumor resistance.
In other words, you can “tune” cells’ disposition to behave like cancer with electricity. Exploring this mechanism might lead to bioelectrical methods of cancer treatment or detection.
Strategies for Pan-Cancer Translational Research
We can imagine a research program to conduct a systematic search for therapies and mechanisms that are effective across many cancers, not just a specific sub-sub-type.
There is already a field of “pan-cancer” research that compiles biobanks and datasets of many types of tumor samples and cancer cell lines, and compares their genomes, transcriptomes, and response to knockout/knockdown/genetic modifications.
However, most of this research is not focused on finding “pan-cancer” treatments; if anything, they consider it a bad sign for a drug target when it’s “pan-essential” (i.e. when knocking it out kills all cancer types) because that might mean it’s essential to survival in healthy cells as well.
A different mindset, however, could use this sort of research in a different paradigm, screening intentionally for “pan-essential” targets, but also testing them in many healthy cell types to see if they’re selective for cancer.
A systematic pan-cancer research program could:
Use diverse biobanks and datasets, across many healthy and cancerous cell, tissue, and tumor types, to continue systematically discovering targets that are “pan-essential” for many cancers but not essential for the survival of healthy cells.
Take advantage of new intracellular delivery techniques and tools like inducible caspases to selectively destroy cells with abnormalities common across cancers (like mutated TP53).
Investigate pan-cancer selective vulnerabilities to physical stresses (electrical, thermal, mechanical, etc) in vitro and in vivo.
Discover agents with strong differential uptake between tumors and healthy tissues, as diagnostic and targeting strategies.
In other words, one could attempt to generalize beyond the existing hints that there are cancer-specific features (like IgG clustering, phospholipid ether clustering, membrane depolarization, etc) and conduct a big search for all the cancer-specific features or mechanisms of selective cancer eradication.
Historical Inspiration: The NCI
Perhaps the ideal, but most ambitious, structure for this program would be a single new research institute with a coherent top-down direction.
In order to keep this project focused, targets and mechanisms would be evaluated for broad-spectrum applicability through systematic screens across diverse cancers, and only then followed up with development of specific therapies that target those mechanisms.
An openness to diverse and unconventional approaches would need to be tempered with an insistence on pursuing only therapeutic strategies with strong, unambiguous, and generalizable effects, using standard benchmarks shared organization-wide.
The inspiration is the mid-20th-century National Cancer Institute, which developed most of the chemotherapies still in use today in a shockingly short timeframe, “under one roof” with a strong research culture seeded by almost the entire WWII-era staff of the Chemical Warfare Service.
Big, systematic searches can succeed. The National Cancer Chemotherapy Program, initiated in 195514, is responsible for the discovery of the chemotherapy agents in most common use today, mostly before the 1980s.
“From 1955 to 1975 up to 40,000 agents per year were selected for screening, largely on an empirical, random basis,” including chemical synthetics, fermentation products, and extracts from 35,000 plant species (including the rose periwinkle Vinca rosea, which produced the successful chemotherapy agent vincristine.)
The NCI’s preclinical research narrowed things down amazingly well; only 150 out of 700,000 potential chemotherapies (about 0.02%) were deemed promising enough to enter the clinic, and of those, 40 showed some anti-cancer efficacy (27%, much higher than the 3.6% of modern clinical-stage cancer drug candidates that receive FDA approval.)15
As 95% of the cost of drug development goes to clinical trials, the cost-effective strategy is to have very high standards for preclinical evidence before conducting trials on humans. The NCI seems to have done this, and gotten correspondingly speedy and successful results.
To some extent the old NCI may have benefited from “low-hanging fruit” discoveries which modern researchers can no longer pick; but researchers today have much more powerful tools and far more knowledge, so we may be able to match their achievement.
Back then, the goal was to literally cure cancer. Perhaps it can be again.
Lien, Sheng-Chieh, et al. "Mechanical regulation of cancer cell apoptosis and autophagy: roles of bone morphogenetic protein receptor, Smad1/5, and p38 MAPK." Biochimica et Biophysica Acta (BBA)-Molecular Cell Research 1833.12 (2013): 3124-3133.
Gu, Long, et al. "Small molecule targeting of transcription-replication conflict for selective chemotherapy." Cell chemical biology 30.10 (2023): 1235-1247.
Baugh, Evan H., et al. "Why are there hotspot mutations in the TP53 gene in human cancers?." Cell Death & Differentiation 25.1 (2018): 154-160.
Sinkala, Musalula. "Mutational landscape of cancer-driver genes across human cancers." Scientific Reports 13.1 (2023): 12742.
Krill-Burger, J. Michael, et al. "Partial gene suppression improves identification of cancer vulnerabilities when CRISPR-Cas9 knockout is pan-lethal." Genome Biology 24.1 (2023): 192.
Weichert, Jamey P., et al. "Alkylphosphocholine analogs for broad-spectrum cancer imaging and therapy." Science translational medicine 6.240 (2014): 240ra75-240ra75.
Rich, Barrie S., et al. "Endogenous antibodies for tumor detection." Scientific Reports 4.1 (2014): 5088
Mazor, Roei D., et al. "Tumor-reactive antibodies evolve from non-binding and autoreactive precursors." Cell 185.7 (2022): 1208-1222.
Chen, Bingdi, et al. "Targeting negative surface charges of cancer cells by multifunctional nanoprobes." Theranostics 6.11 (2016): 1887.
Fröhlich, Eleonore. "The role of surface charge in cellular uptake and cytotoxicity of medical nanoparticles." International journal of nanomedicine (2012): 5577-5591.
Chernet, Brook, and Michael Levin. "Endogenous voltage potentials and the microenvironment: bioelectric signals that reveal, induce and normalize cancer." Journal of clinical & experimental oncology (2013).
Burr, Harold Saxton, Gx M. Smith, and Leonell Clarence Strong. "Bio-electric properties of cancer-resistant and cancer-susceptible mice." The American Journal of Cancer 32.2 (1938): 240-248.
Levin, Michael. Private communication.
Emil Frei, III. The National Cancer Chemotherapy Program.Science217,600-606(1982).
Interesting post! One note -traditionally, "undruggable" didn't really have anything to do with protein localization. For small molecules, getting inside a cell is not typically a major problem -for example, virtually all classical chemotherapeutic agents target intracellular processes, and most antibiotic targets are intracellular, with the major exception of the cell wall. Rather, druggability was primarily a question of whether a potential target has one or more potential small molecule binding pockets (for example, a catalytic site or allosteric regulatory pocket), and whether the binding pocket can be specifically targeted without off-target effects on similar proteins.
For example, KRAS was traditionally considered undruggable because it had no clear binding pockets other than a GTP binding site that binds GTP with extremely high affinity -in the past few years drugs that target a specific oncogenic KRAS mutation have been developed because that mutation produces a cysteine on the protein surface that can be targeted for covalent bonding. Until Gleevec, kinases were considered undruggable because their ATP binding pockets were too similar to each other, and ATP concentrations were too high in the cell. And when I started grad school, protein-protein interactions were thought to be undruggable, as protein-protein binding interfaces are typically large and relatively flat, with no clear binding pockets; since then, a number of PPI inhibitors have been developed, such as the NS5A inhibitors for hepatitis C.
Cell surface localization *does* matter enormously for antibody therapeutics, since antibodies are massive molecules that have no desire to get inside cells. These days so many companies are focused on antibodies that perhaps some people have started to think of cellular localization as the primary determinant of druggability -but it was not always thus.
An old colleague of mine dedicated ~10 years building an argument that echoes these. Directionally, it seems correct that we're not getting proportionate progress at NCI. I'd love to see more high risk bets that pay dividends like radiopharma / adc / checkpoint inhibitors or even bigger new categories and think it's kind of disappointing we have underinvested there
https://www.taylorfrancis.com/chapters/edit/10.4324/9781315674162-13/cancer-spreads-reconceptualizing-disease-katherine-liu-alan-love-michael-travisano