My eclectic tastes frequently carry me far away from where I started. But for this post I’d like to return to my first love, breast cancer, and more particularly to the breast cancer genome, for here something novel and mind-altering is now occurring.
We over-use the term “revolutionary,” but the truth is that we are living in the midst of a scientific revolution in our understanding of human disease. Data is coming at us in such a high rate that even the most diligent amongst us is having trouble keeping up. The last few months have seen an explosion of genomic data across many disease types.
This explosion is a function of changing technology (particularly the rapidly falling cost of genomic analysis) and the creation of broad consortia (most prominently the TCGA and the ICGC) supported by governmental and philanthropic largesse.
Within the past two months several thousands (by my count) of breast cancer genomes have been published, some broad views of the landscape, some more focused. When you recall that the first deep sequencing of breast cancer is less than three years old, this is astonishing.
What lessons are we learning?
First, breast cancer is an amazingly complex disease. Philip Stephens and colleagues, writing in the June 21st issue of Nature (which is stuffed with breast cancer genomics papers) identified several novel driver mutations in genes such as AKT2, ARID1B, CASP8, CDKN1B, MAP3K1, MAP3K13, NCOR1, SMARCD1, and TBX3. Most mutations are passengers, random events not affecting tumor biology, and whether a mutation is a passenger or a driver is not always easy to determine, and sometimes context-dependent.
But what really impresses is the sheer complexity of the mutational landscape: looking at 100 cancers, the authors “found driver mutations in at least 40 cancer genes and 73 different combinations of mutated cancer genes….Most breast cancers differ from all others.” There was substantial variation amongst these 100, with 28 cases having only a single identifiable driver, but some having as many as six. The authors suggest that their method probably underestimates the number of mutated genes in each case. Ouch.
Serena Nik-Zainal and colleagues in the International Cancer Genome Consortium published a fascinating article in the May 25th issue of Cell entitled “The Life History of 21 Breast Cancers.” This is a genuinely fascinating analysis, recommended to anyone interested in cancer biology.
The authors analyze breast cancers as living, evolving, dynamic cell populations. Their modeling suggests that each individual breast cancer has what they call a “most recent common ancestor” (a term derived from evolutionary biology) that occurred early in the molecular history of the cancer, and that most of the cancer’s history “is spent driving subclonal diversification and evolution among the nascent cancer cells.” These subclones fight it out over time, under the evolutionary pressures occurring in a cancer, until one subclone eventually becoming dominant.
A huge number of mutations stack up before this dominant subclone emerges: in one case described by the authors, the dominant subclone (65% of the cancer) had ~15,600 mutations present. The poetic conclusion to this paper states, “Thus we glimpse a model of long-lived, but sparse, lineages of cells passively accumulating mutations until provoked into a major quest for tumor dominance. It is only when this subclone has grown sufficiently populous that the tumor mass becomes clinically detectable.”
These genomic alterations affect therapeutic outcome, to no one’s great surprise. This brings us to a third paper, by Matt Ellis and colleagues, published in the same issue of Nature as the Stephenson report. Here the authors looked at an actual therapeutic scenario (in contrast to most current genomic analyses, which ignore the fact that breast cancer is a treatable disease). They obtained breast cancer tissue from patients treated with preoperative letrozole, performed genomic analyses, then looked for patterns of response and resistance.
The first, and in some ways most interesting, finding is that resistance is a quantitative as well as a qualitative problem: responding tumors had half as many mutations as resistant tumors. The second finding is a daunting one: There were an enormous number of separate mutational events associated with resistance to hormonal therapy.
You look at this paper and think: this isn’t going to be easy. But the authors are quick to point out that while many of the mutations associated with resistance are rare, many are also treatable (in theory, at least) with existing drugs. But we won’t be able to do this the old way, with mass populations treated blindly. We’ll need to test and then treat, or fail to meet the challenge of hormone resistance.
Getting back to the Nik-Zainal paper for a moment, one of the interesting lessons is that successful cancers (if success can be defined by becoming large enough to cause trouble) are inherently antisocial: a dominant clone smashes its way to the top, becoming king of the hill.
But the papers, taken together, suggest that our very human response—and modern science is the most human of activities—to this antisocial phenomenon is an essentially social, collaborative, interactive process. For all three papers are examples of the increasingly common phenomenon of team science.
The life history paper has a total of 49 authors, an army working on behalf of the Breast Working Group of the International Cancer Genome Consortium. The Ellis paper has 59 authors, and the Stephens article has 64. Medical papers increasingly resemble particle physics papers, with their legions of co-investigators, and that is a very good thing for medical science, if not for scientists facing antiquated promotion and tenure committees.
Why a good thing? Jonah Lehrer’s recent Imagine: How Creativity Works (on my “Best Books of the Year” list, and the year only half through: read it!) discusses the research of Ben Jones at the Kellogg Business School. Jones analyzed 19.9 million peer-reviewed papers and 2.1 million patents from the last 50 years. He found that average team size has increased by 20 percent per decade, and that the most-cited papers (the “home-run” papers) are the most likely to be team efforts.
If we ever defeat breast cancer, it will be due as much to our social as our scientific skills. Our ability to mobilize patients, physicians, laboratory scientists, biostatisticians, bioinformaticians, and technologists towards a common goal will be our patients' salvation.