Tumors-on-Chips: Debugging the Cancer Research Model

In 2015, an estimated 192 million animals were used for scientific purposes¹. If this does not strike you as an alarming figure, consider that it’s three times the population of the United Kingdom, and more than three times the number of humans who died of any causes in that same year.

Photo by Ricky Kharawala on Unsplash

Many of these animals were bred and disposed of in our attempts to cure what has come to be known as “The Emperor of all Maladies” — a disease which strips our planet of around 9.6 million lives every year — cancer².

The relentless drive towards human medical progress is understandably an extremely grey area when it comes to animal ethics. Most of us, though we might question ourselves in doing so, would justify this sacrifice if it means curing human diseases and saving human lives. Animal use in research evokes ethical concern, but when it comes to cancer, are you in your right mind to question whether the ends justify the means?

I want to convince you that this question deserves much more than the tentative, tiptoeing inspection we have granted it thus far. I think it deserves our most intensive scrutiny — not only from an ethical and moral standpoint but from a scientific one.

Experiments in animals are intended to yield meaningful biological insights that will benefit human medicine. If this intention is not being fulfilled — and there is overwhelming evidence to suggest this is the case³ — can we righteously excuse ourselves to continue down this potholed path?

Mouse-shaped Medical Blank Spaces

Most of the drugs we use to treat cancer today wouldn’t be here without animal models — that’s true — but who’s to say what we might have gained in their place? In a world where animal research was never contrived or considered acceptable in the first place, and we were forced to pool all our efforts into human-based models instead, might we perhaps have been more prosperous in the war against cancer?

Some of the most effective cancer drugs we have today would probably never have progressed to clinical use if the current strict regulatory requirements based on animal testing were in practice at the time of their development.

Consider tamoxifen, an effective endocrine therapy used to treat certain breast cancers. Years after its approval, it was discovered that tamoxifen causes liver tumors in rats. If this had been recognized at the time of preclinical testing, there would have been no hesitation to discard tamoxifen from the pipeline.

Gleevec, the gold standard drug used to treat chronic myelogenous leukemia (CML), is another fortunate survivor of the mistranslation filter. Having shown serious adverse effects, including liver toxicity, in more than four tested species during preclinical testing, Gleevec narrowly avoided discontinuation based on human cell assays which found no evidence of liver toxicity. Courageously, scientists, considering themselves more human than not, decided to pedal forwards. Clinical trials proceeded, confirming Gleevec’s safety in humans, and so a blockbuster was born.

But there is no question about it — there are blank spaces in the way we treat cancer today — blank mouse-shaped spaces strewn throughout medicine that might have held other blockbusters like these. It is unsettling to wonder how many effective cancer treatments have already been discovered and promptly discarded because of non-human warning signals.

Any scientifically-competent person knows to interpret animal findings with a pinch of salt, but perhaps we should be pinching it more generously, or just burying our mousy mistakes in the salt mine.

The Mistranslation Bottleneck

In drug discovery, the average rate of successful translation from animal models to clinical trials in humans is disturbingly poor. When a medicine appears safe and effective in animals, it will enter the first stage of clinical trials, which simply tests whether humans without the disease can feasibly be given the drug without experiencing harmful side-effects. In oncology, drugs that emerge from Phase I clinical trials unscathed have about a 5% chance of reaching marketing approval for use in human cancer patients⁴.

Of course this figure does not even consider that a number of these medicines will later be withdrawn from the market⁵, and most of them will have marginal clinical benefit — usually extending patients’ life-expectancy by less than a year⁶.

Photo by Haley Lawrence on Unsplash

Proof of safety and efficacy in animal studies clearly tells us very little about that drug’s potential to treat a human disease. It turns out, translating mouse science to human science is like reading Shakespeare to a four-year-old and expecting him to reply with his own tasteful sonnets.

A Case of Scientific Paralysis

There are a lot of things we scientists are exceptionally good at. When it comes to animal research, we’ve been mulling over and rethinking it for quite some time — considering, discussing, analyzing, recognizing, improvising, criticizing, plotting and planning, reviewing and concluding. Overwhelmingly, we’ve concluded that animal models are not effective, and yet overwhelmingly, we have continued down that same tortuous path.

That’s another thing scientists are great at — continuing — continuing through wind, rain and snow — because everyone knows cancer won’t be cured on a Sunday in front of the television. But this very perseverance that builds great research might also be doing it a disservice. Isn’t it time we stop “recognizing” and “acknowledging” and “taking into account” everything that’s broken about this model, and actually wake ourselves out of paralysis and do something to fix it?

Plastering Bandages on Broken Limbs

Recognizing the frailty of 2D in vitro models in recapitulating human disease, scientists today rely trustingly on the animal model for “proof-of-concept” — “Proof” being a word we toss around extremely loosely. Animal models are thought to represent a stepping stone in the translation of in vitro findings into clinical discoveries, but more often than not they mislead us down false paths that are riddled with clinical failures.

But rather than eradicating or at least dramatically reducing animal experimentation, the more accepted response is to try to improve upon the animal model by modifying it genetically to a more “humanized” version. We will have to eventually accept that no matter how modified, neither mice nor monkeys will ever accurately reflect the complex processes steering human cancer growth and progression, or human responses to drugs.

We have killed this model so many million times — you would think it would be dead by now. We do after all consider ourselves the most intelligent species. As Einstein mused, “The measure of intelligence is the ability to change”. Plastering bandages on broken limbs will not be enough. This isn’t a model that we’re going to fix by pruning the leaves of our dying tree. We need to examine the roots — tear the tree up from the ground and replant it in a new kind of science.

How do you get the cancer in the animal?

Most non-human animals don’t naturally develop cancer, so in order to study the disease and the medicines we make to treat it, we first have to create cancer in the animal (if you find this thought deeply disturbing, I am with you).

Animal models of cancer come in two flavors: ones in which cancers develop “spontaneously” or “naturally” (meaning, counter-intuitively, a genetically engineered mutant strain which has been induced to develop cancer, by carefully tailoring its genetic makeup), and ones in which cancerous tissue is transplanted into the animal, taken either from a human cancer tissue or from the former “spontaneous” model. In this model, cancer cells might be injected ectopically (usually under the skin), orthotopically (in the organ of origin, to model “natural tumor growth”), or systemically (into the bloodstream, to model metastatic spread).

This second model is particularly problematic in replicating tumor progression. Tumor tissue placed into a foreign microenvironment does not act as though it were at home — it doesn’t speak the language there, it doesn’t understand the culture, or have the faintest idea how to dress to blend in, how likely it is to get pickpocketed, what the street signs mean, or when it’s appropriate to smile at strangers — so even though it has come from a human tumor, it’s not going to act like one — but it’s not going to act like a normal mouse tumor either.

The animals used in these experiments also need to undergo immune suppression so that their immune systems will not react by attacking the foreign transplanted tumor. The blood vessel architecture that develops around a transplanted tumor also does not resemble that which would naturally grow. These discrepancies can dramatically impact how the cancer evolves in its host.

The genetically engineered “spontaneous” model is considered superior, but of course has its own flaws. Although these tumors are structurally more realistic, and patrolled by functional immune systems, they reside in a rodent — an animal which diverged from our human branch of the evolutionary tree about 96 million years ago⁷. Over those years we have changed quite a bit, as mouse and man have traveled down very different paths — we are not victims to the same diseases; we are not responsive to the same medicines; our genes — though on the face rather similar — are like harp strings strung in different order by different hands.

“If you have cancer, and you are a mouse, we can take good care of you.”

The renowned cancer researcher Judah Folkmen is quoted as having said: “If you have cancer and you are a mouse, we can take good care of you”. Its true — scientists have cured cancer in mouse a million times — they’ve cured it in human-transplanted mouse tumors, and they’ve cured it in “spontaneous” mouse tumors, and every time they have to stop themselves from getting a bit excited about that — because where is our cure?

Substances which are potently carcinogenic in humans are often not harmful to mice, and vice versa. Our genes encode proteins and enzymes which differ in small but meaningful ways. We share some biological pathways but not others. We differ markedly in size, metabolic rate, microbiome and immune function⁸. All of these factors influence cancer.

It should not come as a surprise then that while we have cured cancer a million times in mice over the past decades, the human disease outcome is as bleak today as ever. Yet we keep digging there, blindly, in that same grave; digging it up to cure it again. It’s time, I think, we let this broken model of ours die, because no matter how we modify him, and “humanize” him, and plaster bandages on his wounds, our little mouse will never be a man.

New mouse models are being engineered to address some of these challenges — mice expressing “humanized” immune systems, and “humanized” bone tissues are some examples. Large research efforts are being placed on advancing mouse tumor models — “advancing” meaning making them more like humans. But nature will always have its way in the end.

Non-human models will always be bad models, perhaps until we cross some critical threshold of “humanization”, and at that point probably decide that it’s unethical to synthesize cancer in an organism that is so much like us. Or perhaps before we reach that point, someone (louder than me) will point out that the whole endeavor was morally obscene from the outset.

The Tumor as a Multicellular Organ: Why 2D in vitro Models Fail

Ordinary in vitro cell models — typically comprising a 2D layer of cancer cells in a dish of nutrient-containing medium — are useful for studying isolated mechanisms in a single cell type. However cancer is not a cell-autonomous process: That is to say that the complex collection of processes driving the tumor to grow and progress and spread to distant body sites are controlled not only by genes within the cancer cells themselves, but by an army of surrounding non-cancerous cells and extracellular elements constituting the so-called tumor microenvironment.

We have come to view the tumor — not as an inert or self-driven mass — but rather as an interactive multicellular organ which communicates extravagantly with its microenvironment. In this light, the 3D tumor chip model — which we will soon meet — offers unparalleled advantages over traditional cell cultures because it is designed to reflect interactions between multiple cell types as they would occur in the body.

The behavior of a tumor and its response to a medicine can be dramatically transformed by its microenvironment, which is in turn continuously remodeled by tumor cells throughout disease progression. For example, tumors are patrolled by bands of immune cells which can fight tumors like they do infections, and cancer cells develop clever mechanisms to mask themselves from immune attack.

The structure and content of the microenvironment also determines how readily cancer cells can invade into the surrounding tissue, which is the first step leading to metastasis — spreading of the tumor to distant body sites. Tumor cells also direct the growth of new blood vessels in their midst, establishing a flood of oxygen and nutrients which they require to survive. The blood vessel composition of a tumor also dictates the effective passage of drugs from circulation into the tumor site.

These processes and others clearly demonstrate just how limited an understanding of cancer behavior can be drawn from traditional 2D cell cultures devoid of a co-evolving immune system and surrounding vasculature.

Tumors-on-Chips: A New Kind of in vitro

An organ-on-a-chip is not some kind of electronic computer chip like it sounds. It is a so-called “microfluidic” device — a slice of glass or plastic molded with tiny hollow microchannels which are lined with living human cells. These chips have emerged as a promising alternative to animal models for studying disease processes and developing medicines.

Photo by National Cancer Institute on Unsplash

The device to an outsider resembles a clear square of plastic about the size of a USB stick — but on the inside it is intricately seeded with organ-specific human cells, and interfaced with an artificial blood system. Organ chips have been designed which replicate the complex architecture and function of various human organs, reflecting multiple cell types in a single device, and mimicking physiological processes that occur in the human body.

Early examples included lung chips that reflect breathing motion⁹, intestine chips that mimic digestive movements¹⁰ and brain chips that convey the selectively permeable nature of the blood-brain-barrier¹¹. Now scientists have created tumor chips, revealing possibilities for in vitro studies of cancer which were hitherto unimaginable.

Modelling cancer in vitro has many advantages — but not by the reductionist methods we are used to. The in vitro approach allows for precisely controlled conditions and a standard of reproducibility that can never be expected from live models. However, traditional in vitro models are limited in that they fail to reflect the heterogeneous nature of a growing tumor in vivo, and its complex interrelations with non-cancer cells and other elements of its microenvironment.

Tumor chips have been created which reproduce the tumor vascular interface embedded in three dimensions within a physiologically relevant matrix. This is allowing scientists to investigate intricate, multidimensional processes like invasion and metastasis — whereby tumor cells break through the surrounding blood vessel linings and move into circulation in order to travel to distant body sites¹².

The chips also try to reflect the complicated cross-talk between tumor cells and other cells in their environment, such as immune cells and muscle cells that wrap around surrounding blood vessels, and to study how molecules and proteins released into the microenvironment influence tumor behavior and drug responses¹³. These kinds of interactions are inextricably linked to every stage of tumor growth and progression.

The chips have also been used to track the transport of medicines into the tumor niche and to monitor their uptake by tumor cells, which is challenging in live models. The transparent chip offers a window into the disease or organ so that real-time responses can be captured in high resolution by much cheaper and more feasible means than imaging in live animals¹⁴.

Tumor chips are also amenable to high-throughput studies making them highly desirable for preclinical evaluation of drugs¹⁵.

A further compelling feature of the microfluidic system is the incorporation of real tumor tissues obtained from patient biopsies, which could herald great advancements in personalized cancer medicine. This technique would allow drug responses to be analyzed in an in vitro system that is designed specifically to reflect the heterogeneity of each individual patents’ cancer.

Of course it would be most ideal to study these intricacies at the level of the whole organism, but genetic and physiological differences simply make it impossible to draw reliable conclusions from such studies in animals.

Uniting tissue engineering and microfluidic technology, tumor-on-chip systems offer tantalizing possibilities of unraveling human tumor biology and drug responses, while minimizing the time, cost, and ethical concerns of animal research.

All Models are Wrong — Some Models are Better than Others

The complex heterogeneity and evolutionary nature of cancer is never going to be perfectly emulated in a tumor chip system, and probably not in any other system that is not the human body. Recreating the tumor microenvironment and inter-organ interactions in a three-dimensional in vitro system remains immensely challenging. In the past decade we have only glimpsed the tremendous potential of tumor-on-chip systems — the bulk of the iceberg is waiting, and it glistens in the sunlight.

From broken models grows broken science. From broken science grows broken medicine. Decades spent studying biological systems that are evolutionarily, genetically and physiologically different from ourselves has distorted our understanding of human health and disease. This model has squandered resources, obstructed successful drug discovery and ultimately compromised human health. You might be an outstanding scientist with outstanding ideas, but too many outstanding ideas have been wasted on a broken model.

A Cure for Medicine

Limitations of traditional 2D in vitro system have necessitated a heavy reliance on animal models in the war against cancer, because animal models, however genetically different they may be from ourselves, have one undeniable advantage — each one represents the sum of all networks and systems which constitute a whole living organism, and no in vitro replica will ever be as true to nature. The inherent problem with non-human animal models though, is just that — they are not human.

So what is more true to human nature — an entire living organism of a non-human species that happens to work very differently to our own, or a human-like organ made from human cells, that happens to live outside of a body?

Organ chip research has grown exponentially over the past decade, but is still dwarfed by animal experimentation, and so are the funds it receives. But if animal bodies are the wrong place to look for answers — and clearly they are — the costs they now swallow would reap richer fruits if we redirected them towards human-based cancer models.

As this model advances in the coming decade, and as our understanding of this elaborate cancer crosstalk continues to evolve, we might expect that this next generation of in vitro modelling could offer transformative insights into cancer biology and therapy.

Knowing that alternatives exist that are worth our investment, there are two steps to fixing a broken model: First, people with eyes will have to raise their voices. Then, people with voices will have to open their eyes. This is my attempt at eye opening.


  1. https://www.crueltyfreeinternational.org/why-we-do-it/facts-and-figures-animal-testing
  2. https://www.who.int/news-room/fact-sheets/detail/cancer
  3. Akhtar A. The flaws and human harms of animal experimentation. Cambridge Quarterly of Healthcare Ethics. 2015 Oct;24(4):407–19.
  4. https://www.bio.org/sites/default/files/legacy/bioorg/docs/Clinical%20Development%20Success%20Rates%202006-2015%20-%20BIO,%20Biomedtracker,%20Amplion%202016.pdf
  5. Onakpoya IJ, Heneghan CJ, Aronson JK. Post-marketing withdrawal of 462 medicinal products because of adverse drug reactions: a systematic review of the world literature. BMC medicine. 2016 Dec;14(1):10.
  6. Tibau A, Molto C, Ocana A, Templeton AJ, Del Carpio LP, Del Paggio JC, Barnadas A, Booth CM, Amir E. Magnitude of clinical benefit of cancer drugs approved by the US Food and Drug Administration. JNCI: Journal of the National Cancer Institute. 2018 May 1;110(5):486–92.
  7. Nei M, Xu P, Glazko G. Estimation of divergence times from multiprotein sequences for a few mammalian species and several distantly related organisms. Proceedings of the National Academy of Sciences. 2001 Feb 27;98(5):2497–502. https://www.pnas.org/content/98/5/2497
  8. Perlman RL. Mouse models of human diseaseAn evolutionary perspective. Evolution, medicine, and public health. 2016 Jan 1;2016(1):170–6.
  9. Huh D. A human breathing lung-on-a-chip. Annals of the American Thoracic Society. 2015 Mar;12(Supplement 1):S42–4.
  10. Bein A, Shin W, Jalili-Firoozinezhad S, Park MH, Sontheimer-Phelps A, Tovaglieri A, Chalkiadaki A, Kim HJ, Ingber DE. Microfluidic organ-on-a-chip models of human intestine. Cellular and molecular gastroenterology and hepatology. 2018 Jan 1;5(4):659–68.
  11. Wevers NR, Kasi DG, Gray T, Wilschut KJ, Smith B, Van Vught R, Shimizu F, Sano Y, Kanda T, Marsh G, Trietsch SJ. A perfused human blood–brain barrier on-a-chip for high-throughput assessment of barrier function and antibody transport. Fluids and Barriers of the CNS. 2018 Dec 1;15(1):23.
  12. Zervantonakis IK, Hughes-Alford SK, Charest JL, Condeelis JS, Gertler FB, Kamm RD. Three-dimensional microfluidic model for tumor cell intravasation and endothelial barrier function. Proceedings of the National Academy of Sciences. 2012 Aug 21;109(34):13515–20.
  13. Moore N, Doty D, Zielstorff M, Kariv I, Moy LY, Gimbel A, Chevillet JR, Lowry N, Santos J, Mott V, Kratchman L. A multiplexed microfluidic system for evaluation of dynamics of immune–tumor interactions. Lab on a Chip. 2018;18(13):1844–58.
  14. Young AT, Rivera KR, Erb PD, Daniele MA. Monitoring of microphysiological systems: integrating sensors and real-time data analysis toward autonomous decision-making. ACS sensors. 2019 Apr 9;4(6):1454–64.
  15. Du G, Fang Q, den Toonder JM. Microfluidics for cell-based high throughput screening platforms — A review. Analytica chimica acta. 2016 Jan 15;903:36–50.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s