Case 75, an infant, presented to the Clinical Genomics Center at UCLA with a history of multifocal complex partial epilepsy and regression of developmental milestones. The child, as well as the two parents, underwent clinical exome sequencing (so-called trio-CES). You can imagine what it must have been like to be the parents of such a child, their desire to understand what caused such misery, and the hope that something beneficial might emerge from genomic analysis. Sequencing revealed that the child carried a previously undescribed missense variant in KCTN1, which encodes the KCa4.1 protein, a member of the calcium-activated potassium channel protein family.
When I was in training, and for many years thereafter, the description of such a case, and the identification of its etiology, warranted a paper in the New England Journal of Medicine or TheLancet. When the Human Genome Project opened its doors there were only some 25 well-delineated inherited diseases of metabolism. Now, according to the authors of a stunning recent JAMA paper on clinical whole-exome sequencing (2014;312:1880-1887), the Online Mendelian Inheritance in Man website lists some 4,000 disease-gene relationships. What was once impossible has become commonplace, even trivial.
At the UCLA center, 814 patients underwent whole exome sequencing, and 26 percent had a molecular diagnosis. Similar results were seen in 2,000 patients studied at Baylor, where 25.2 percent had a diagnostic mutation detected (2014;312:1870-1879). The Houston authors state that 58 percent of the diagnostic mutations were previously unreported. New diseases are raining down on us faster than any of us can possibly comprehend.
Dazzle
The genomic revolution continues to dazzle. I cannot imagine any more exciting time to be alive. Can there be any time in history when we have learned so much about human biology in so short a time? This seems the equivalent of that period early in the 20th century when physics exploded, with atomic theory, special and general relativity, and quantum mechanics all unrolled over a decade or two, the world utterly changed.
Some of it is cancer biology, and some just (just!) general biology. Starting with the human genome project, and continuing with The Cancer Genome Atlas project and its many relatives, we've seen a technological juggernaut roll through every aspect of human biology-indeed, every aspect of biology. We've gone both deeper and wider, as the two JAMA papers suggest.
Leaving aside the "cool" factor of such work, this new look at human biology has profound implications for physicians and their patients.
When whole exome sequencing becomes cheap and ubiquitous, basically something you get on a newborn like Tay-Sachs testing, what do you do with the results? Do you tell the child's parents, their pediatrician (or geriatrician-some of these things may take a lifetime to emerge), or their insurance company?
And when you go looking, you frequently find something else, something unexpected. In the 2,000 patients tested at Baylor, 92 patients (4.6%) had a medically actionable incidental finding. Looking over the list reported in the paper, I see old friends such as BRCA 1 and 2, PALB2, RAD51D, and RET, among many others. Remember, they were not being tested for this: just incidental findings with the promise of future misery, or perhaps medical salvation.
What to Do with Results of Panel Testing?
One of the current debates going on in guideline committees involves what to do with the results of panel testing, and whether even to order panel testing. If I order a panel test because I suspect my patient has a BRCA mutation, and I find a gene predicting an increased risk of colorectal cancer, or (further afield) hypertrophic cardiomyopathy, what is my ethical and practical responsibility as a physician? What do I owe my patient, or my patient's children, or society (health economics rearing its ugly head almost immediately)?
My personal bias, for what it's worth, is that such broad panel testing is inevitable technologic imperialism, like we see with just about every potentially useful diagnostic technology. There is part of me that wants to shout, just like when viewing some not very bright teenagers in a slasher movie, "Don't look behind that door. There are monsters there." But we always open that door. Always.
And how do the tests affect a patient sitting in the room with the doctor? How does knowledge of a KCTN1 missense variant help a baby with partial complex epilepsy? And if it does not, do we even want to know? We all know Francis Bacon's dictum that knowledge is power, but what happens when knowledge leaves the doctor powerless in the face of a previously undescribed disease?
O brave new world, that has such testing in it.
Most oncologists aren't cancer geneticists, at least not yet, though we may be forced into the role sometime soon. But the broader, deeper aspects of the genomic revolution are clearly affecting how we think about the cancer patients we see. When, a couple of years ago, the first decently sized genomic evaluations of human cancer began to come out, we were all impressed with the large number of mutations, and in particular the ubiquity of rare driver mutations seen across common human cancers.
These studies, it now seems obvious, seriously underestimated the problem for many cancers. The original TCGA work, for instance, tended to focus only on mutations occurring in more than five percent of a particular cancer's cells. This meant, as was recognized at the time, that we were missing many low frequency mutations buried within a cancer's genome. Now we are beginning to see what deep, dark waters there are in the genome's abyss.
Wang and colleagues (Nature, 2014; 512[7513]:155-60) performed single-cell whole genome sequencing on tumors from two breast cancer patients, one ER-positive and one triple-negative. In the first cancer genome studies, one chose an area with high tumor cellularity, ground it up, and took what was essentially a family portrait. If the family had eight adult brunettes and a runty blond baby sitting behind them, you only saw the brunettes. With single cell whole exome sequencing, if you sequence enough cells (the Nature paper sequenced about 50 per tumor), the rare family members pop out. Single cell sequencing allows a collection of individual portraits to complement the family group picture.
The first thing one discovers is the incredible variability of cancer cells. The authors state "No two single tumor cells are genetically identical," which I find somewhat scary. The triple-negative breast cancer they examined didn't even pretend to be a single cancer, having three distinct subtypes buried within the cancer and a myriad of private mutations.
For the past several years the emerging genomics have depressed me even as they have fascinated me. Hypervariation is obviously a bad thing if one is throwing kinase inhibitors at a cancer: the whack-a-mole problem of compensatory resistance mechanisms dooms monotherapy approaches to inevitable failure.
Immune Checkpoints
But the silver lining of genomic hypervariability is now beginning to be seen in the immune checkpoint field. Genomic hypervariability is associated with neoantigen diversity. A recent evaluation of melanoma patients treated with the checkpoint inhibitor ipilimumab demonstrates a link between genomic instability and tumor response: the more mutations per megabase, the better the response to Ipi.
Mutations per megabase is not a great way to select therapy (the same New England Journal of Medicine article presents a signature that is a better predictor of response), but I suspect the concept is sound. The current issue of Nature, as I write this column, has five manuscripts devoted to cancer immunotherapy. Among them are some fascinating pieces of information: the emerging evidence that checkpoint inhibition immunotherapy will be useful in bladder cancer (a disease sorely in need of new therapies), and the importance of PD-L1 expression on tumor infiltrating lymphocytes as a marker of response.
But what really caught my eye was evidence, in two of the papers, that checkpoint inhibition would be particularly successful where there are specific tumor neoantigens. Much of the genome literature in recent years has distinguished "driver" and "passenger" mutations. One can imagine a driver mutation saying to a passenger "You're just along for the ride. I'm the one that matters. I drive the cancer." And this statement is no doubt true when one speaks of kinase inhibition for a cancer. It's been the basis for drug discovery in the last decade: find the mutant growth factor, find a molecule to block it, treat.
But in the world of immunotherapy, the drivers are just chauffeurs; the important guys sit in back, like fat cat bankers in a black stretch limousine. The "passenger" mutations signal the immune system, unleashed by anti-PD1 antibody, to attack the cancer cell. So finding genomic hypervariabilty in a patient's cancer may lead, not to despair, but to a PD1 inhibitor.
Getting back to my earlier concern that knowledge is not always power, it is perhaps best to add a qualifier: "yet." Identifying the genomic disorder underlying a baby's seizures may not be actionable today, but if you believe that progress results from the progressive accumulation of facts, and our ability to weave those facts together into testable hypotheses and new therapeutic approaches, then some day that KCTN1 missense variant test result may be accompanied by a doctor telling the parents "but we've got a treatment for that" and a prescription.
So what if there are 4,000 gene-disease relationships in the online databases? Indeed, so what if it is 10,000 next year? So what if the cancer has myriad mutations? We have lots of assistant professors, and all the time in the world, to solve these problems. That's what we do.
More from George Sledge on His OT Blog!
Read and comment at http://bit.ly/OT-Sledge