The Scientific Drunk and the Lamppost: Massive Sequencing Efforts in Cancer Discovery and Treatment

See allHide authors and affiliations

Science Signaling  02 Apr 2013:
Vol. 6, Issue 269, pp. pe13
DOI: 10.1126/scisignal.2003684


The massive resources devoted to genome sequencing of human tumors have produced important data sets for the cancer biology community. Paradoxically, however, these studies have revealed very little new biology. Despite this, additional resources in the United States are slated to continue such work and to expand similar efforts in genome sequencing to mouse tumors. It may be that scientists are “addicted” to the large amounts of data that can be relatively easily obtained, even though these data seem unlikely, on their own, to unveil new cancer treatment options or result in the ultimate goal of a cancer cure. Rather than using more tumor genetic sequences, a better strategy for identifying new treatment options may be to develop methods for analyzing the signaling networks that underlie cancer development, progression, and therapeutic resistance at both a personal and systems-wide level.

From the geneticists who promised over a decade ago that gene expression analysis was all that was needed to cure cancer, and later said that sequencing the human genome would cure cancer, comes the latest claim: Blind sequencing of massive numbers of human tumors will cure cancer. How could sequencing tumors lead to new cancer cures? The idea would clearly be tenable if (i) there were many new cancer-causing genes yet to be discovered; (ii) one could distinguish critical mutations responsible for causing or maintaining a particular cancer from those that are simply “bystander” mutations, leading to the development of new drugs against the mutant protein; or (iii) unique combinations of tumor-specific mutations could be identified and drive particular therapeutic decisions.

So far, the results have been pretty disappointing. Various studies on common human tumors, many under the auspices of The Cancer Genome Atlas (TCGA), have demonstrated that essentially all, or nearly all, of the mutated genes and key pathways that are altered in cancer were already known. A TCGA study of glioblastoma, for example, showed that eight genes were significantly amplified or mutated, including genes encoding the Ras guanosine triphosphatase–activated protein neurofibromatosis 1 (NF1), epidermal growth factor receptor (EGFR), phosphatidylinositol 3-kinase (PI3K), and the cell cycle regulator retinoblastoma (Rb) (1). All eight were previously known to be important in cancer. Additionally, this study concluded that mutations in the gene encoding the transcriptional regulator and tumor suppressor p53 were particularly common, hardly a surprising finding given that p53 was discovered as a major tumor suppressor well over 20 years ago (2) and has been the subject of innumerable cancer-related studies. Nearly the same gene and pathway information emerged after a massive genomic sequencing analysis of ovarian cancer (3), but here two of the nine genes that were statistically significantly mutated above background were not even expressed in the tumors! A very similar set of pathway genes emerged after an extensive genomic and sequence analysis of ~200 colorectal tumors (4), although genes in the Wnt and transforming growth factor–β (TGF-β) pathways were found to be frequently mutated or methylated in this tumor type. These results largely confirmed an important role for these pathways in colorectal cancer, something that was already widely appreciated (5) before this extensive and expensive study. Finally, another TCGA study, focusing on breast cancer and involving over 800 patients and nearly 350 authors (6), has largely confirmed what was already known: The four basic categories of breast tumors, which had been previously defined on the basis of a combination of clinical and pathological features and mRNA expression analysis, were essentially correct. The study did produce some new findings; for example, not all Her2-overexpressing tumors had similar mRNA expression signatures, and upstream kinases involved in the activation of the mitogen-activated protein kinases JNK, ERK, and p38 may be specifically inactivated in luminal tumors. But, paradoxically, the two “critical” findings implicated as translationally important “new” discoveries—that some hormone-responsive tumors may be particularly dependent on the PI3K pathway and that basal breast tumors and serous ovarian tumors shared many similarities—were already known to the medical and scientific community (7). Clinical trials using inhibitors of PI3K-driven pathways were already under way in subsets of patients with estrogen receptor–positive breast tumors that were resistant to anti-estrogen therapies, before the TCGA report (7). Mutations or epigenetic inactivation of BRCA1 in basal breast cancer (8), together with the strong association of BRCA1 mutations in ovarian cancer, had already led many clinicians to suspect that the two diseases shared a common molecular pathology (9).

Although not providing very many novel insights into the fundamental biology of cancer, TCGA data, in their current form, are nonetheless extraordinarily valuable because they provide information about real human cancers that confirms conclusions drawn from studies of cultured cancer cells, xenograft cancer models, and cancer pathology specimens (Fig. 1). The data show that tumors contain lots of mutations, confirming that cancer is indeed a mutating disease (10). Furthermore, TCGA data give an unbiased glimpse at the types of mutations that occur—such as insertions, amplifications, and specific point mutations, providing important insights into tumor evolution (11, 12)—and the specific combinations of mutations that occur in certain tumor types, although the relative importance or expression of the mutant proteins is not yet clear. In addition, TCGA and other related tumor sequence data have clearly revealed an unexpectedly large genetic heterogeneity within cells of the same tumor, further complicating any attempt to translate bulk genetic data into therapeutic treatment options in a simple, straightforward manner (13). But perhaps the most important message from all the data as a whole is that alterations in signaling pathways are at the very heart of cancer.

Despite the U.S. National Institutes of Health (NIH) spending over a quarter of a billion dollars (and all of the R01 grants that are consequently not funded to pay for this) and the massive data collection efforts, so far we have learned little regarding cancer treatment that we did not already know. Now, NIH plans to spend millions of dollars to massively sequence huge numbers of mouse tumors! The paucity of novel insights arising from sequencing human tumors casts doubt on the soundness of sequencing mouse tumors for advancing therapeutic options for treating human cancer. Why, then, does the scientific community seem to be addicted to sequencing, despite the relative lack of novel insight it provides?

I believe the answer is quite simple: We biomedical scientists are addicted to data, like alcoholics are addicted to cheap booze. As in the old joke about the drunk looking under the lamppost for his lost wallet, biomedical scientists tend to look under the sequencing lamppost where the “light is brightest”—that is, where the most data can be obtained as quickly as possible. Like data junkies, we continue to look to genome sequencing when the really clinically useful information may lie someplace else.

So, where exactly should we go next? One idea might be for sequencing studies to focus on cancer in children, since the genetics are likely to be simpler (assuming children have not had the opportunity to lead such mutation-inducing lifestyles as their parents), or to focus on particularly rare human cancers, such as Waldenstron’s macroglobulinemia, in which 90% of patients have mutations in a single gene (Myd88) (14), or hairy cell leukemia, in which 100% of patients have the exact same mutation in the gene encoding B-Raf (15). This highly focused approach to uncommon cancers may be highly productive.

An alternative idea, however, proposed by some cancer geneticists, is to massively expand the scale of these sequencing efforts to look for smaller and smaller numbers of mutations within commonly occurring tumors that are present in fewer and fewer people. These tumors are likely to contain hundreds or thousands of mutations, and sifting the relevant ones from the bystanders seems an impossible task. Instead, I recommend that research efforts should instead focus on the signaling proteins and networks that result in carcinogenesis, metastasis, and chemoresistance. It is the signaling proteins, not the genes per se, that are responsible for the phenotypes of tumors and for the emergence of therapeutic resistance.

An unbiased look at the emerging data in cancer biology supports such a signaling-focused approach. First, consider three recent game-changing discoveries in human cancer; (i) the identification of mutations in isocitrate dehydrogenase 1 (IDH1) and the altered metabolism in tumors; (ii) the identification of B-Raf mutations as drivers of melanoma; and (iii) the identification of EML4-ALK fusions (echinoderm microtubule–associated protein-like 4 fused to the anaplastic lymphoma kinase) as drivers of lung cancer. None of these pioneering discoveries came from TCGA. The IDH1 mutations were identified by sequencing exons of protein coding genes in glioblastoma (16) similar to the TCGA approach, while the other two came directly from studies highly focused on genes encoding proteins in specific signaling pathways relevant to melanoma (17) or from cloning transforming genes and investigating abnormal phosphorylation patterns in phosphoproteomic studies of lung cancers (18, 19).

Second, essentially all of the current and emerging targeted molecular therapies, such as EGFR-related inhibitors and antibodies, Abl inhibitors, B-Raf inhibitors, ALK inhibitors, and VEGFR inhibitors, are directed at signaling molecules, not at collections of mutant genes. These therapeutic successes may have come even faster, and the drugs may be more effectively used in the future, if cancer research focuses on network-wide signaling analysis in human tumors (20), particularly when coupled with insights that the TCGA sequencing data now provide (Fig. 1). Currently, signaling measurements are hard, not particularly suited for high-throughput methods, and not yet optimized for use in clinical samples. Why not invest in developing and using technologies for these signaling-directed studies? This effort requires some new out-of-the box thinking, not just mass spectrometry data dumps. Mass spectrometry is one important tool that should be further developed, but it is clearly not the only one. If 20% of the current NIH expenditures being devoted to tumor sequencing were redirected toward understand signaling in tumor cells and developing applications and tools to facilitate these kinds of studies in vivo, the results could be amazing!

Fig. 1

Combining genomic, proteomic, and signaling data is key to successfully treating cancer. Tools for genomic and mRNA analysis are well developed. Investment is needed in methods and tools for analyzing the status of tumor signaling networks and the state of the tumor proteome.



View Abstract

Navigate This Article