Is Cancer a Genetic Disease?

I recently had the opportunity to meet two charming young patients: One, a 32-year-old female with an extremely rare malignancy that arose in her kidney and the other a 33-year-old gentleman with widely metastatic sarcoma.

Both patients had obtained expert opinions from renowned cancer specialists and both had undergone aggressive multi-modality therapies including chemotherapy, radiation and surgery. Although they suffered significant toxicities, both of their diseases had progressed unabated. Each arrived at my laboratory seeking assistance for the selection of effective treatment.

With the profusion of genomic analyses available today at virtually every medical center, it came as no surprise that both patients had undergone genetic profiling. What struck me were the results. The young woman had “no measurable genetic aberrancies” from a panoply of 370 cancer-causing exomes, while the young man’s tumor revealed no somatic mutations and only two germ-line SNV’s (single nucleotide variants) from a 50 gene NextGen sequence, neither of which had any clinical or therapeutic significance.

What are we to make of these findings? By conventional wisdom, cancer is a genetic disease. Yet, neither of these patients carried detectable “driver” mutations. Are we to conclude that the tumors that invaded the cervical vertebra of the young woman, requiring an emergency spinal fusion, or the large mass in the lung of the young man are not “cancers”? It would seem that if we apply contemporary dogma, these patients do not have a cancer at all. But nothing could be further from the truth.

Cancer as a disease is not a genomic phenomenon. It is a phenotypic one. As such, it is extremely likely that these patients’ tumors are successfully exploiting normal genes in abnormal ways. The small interfering RNAs or methylations or acetylation or non-coding DNA’s that conspired to create these monstrous problems are too deeply encrypted to be easily deciphered by our DNA methodologies. These changes are effectively gumming up the works of the cancer cell’s biology without leaving a fingerprint.  

I have long recognized that cellular studies like the EVA-PCD platform provide the answers, through functional profiling, that genetic analyses can only hope to detect. The assay did identify drugs active in these patients’ tumor, which will offer meaningful benefit, despite the utter lack of genetic targets. Once again, we are educated by cellular biology in the absence of genomic insights. This leaves us with a question however – is cancer a genetic disease?

In Cancer Research: An Awakening?

In 2005, as the Iraq War reached a low point with casualties mounting and public support dwindling, Sunni tribesman in the Anbar Province arose to confront the enemy. Joining together as an ad hoc army these fighters turned the tide of the war and achieved victories in the face of what had appeared at the time, to be overwhelming odds.

I am reminded of this by an article in The Wall Street Journal by Peter Huber and Paul Howard of the Manhattan Institute that examined the bureaucracy of drug development. It raised the question: Are new cancer treatments failures or is the process by which they are approved a failure? They describe “exceptional responders” defined as patients who show unexpected benefits from drug treatments. Using molecular profiles, they opine, scientists will unravel the mysteries of these individuals and usher in an era of personalized medicine. Thus, rigid protocols that use drugs based upon tumor type e.g. lung vs. colon fail because they do not incorporate the features that make each patient unique – an awakening.

The example cited is from Memorial Sloan-Kettering where a patient with bladder cancer had an unexpected response to the drug Everolimus (approved for kidney cancer). Subsequent deep sequencing identified a genetic signature associated with sensitivity to this drug. While it is a nice story, I already knew it very well because it had been repeated many times before and would in the past have been dismissed as an “anecdote.” It is precisely because of its rarity that it has been repeated so many times.

The WSJ analysis strikes a familiar chord. For decades, we have decried the failure of rigid clinical trials that underestimate a patient’s unique biology yet cost millions, even billions of dollars, while denying worthy candidates new treatments under stultifying disease-specific designs.

We pioneered phenotypic (functional) analyses (the EVA-PCD platform) to examine whole cell models as we explored drug response profiles, novel combinations and new targets. It is regrettable that these WSJ authors, having raised such important issues, then stumble into the same tantalizing trap of molecular diagnostics, and call for bigger, better, faster genomic analyses.

Cancer patients need to receive treatments that work. They do not particularly care why or how they work, just that they work. These authors seem to perpetuate the myth that we must first understand why a patient responds before we can treat them. Nothing could be further from the truth.

Alexander Fleming knew little about bacterial cell wall physiology when he discovered penicillin in 1928, and William Withering knew nothing about the role of muscle enzymes in congestive heart failure when he discovered digoxin extracts in 1785. Would anyone argue that we should have waited decades, even centuries to apply manifestly effective therapies to patients because we did not have the “genes sequenced?’

We may be witness to an awakening in cancer drug development. It may be that a new understanding of individualized patient response will someday provide better outcomes, but platforms with the proven capacity to connect patients to available treatments should be promoted and applied today.

Toward A 100% Response Rate in Human Cancer

Oncologists confront numerous hurdles as they attempt to apply the new cancer prognostic and predictive tests. Among them are the complexities of gene arrays that introduce practicing physicians to an entirely new lexicon of terms like “splice variant, gene-rearrangement, amplification and SNP.”

Although these phrases may roll of the tongue of the average molecular biologists (mostly PhDs), they are foreign and opaque to the average oncologist (mostly MDs). To address this communication shortfall laboratory service providers provide written addenda (some quite verbose) to clarify and illuminate the material. Some institutions have taken to convening “molecular tumor boards” where physicians most adept at genomics serve as “translators.” Increasingly, organizations like ASCO offer symposia on modern gene science to the rank and file, a sort of Cancer Genomics for Dummies. If we continue down this path, oncologists may soon know more but understand less than any other medical sub-specialists.

However well intended these educational efforts may be, none of them are prepared to address the more fundamental question: How well do genomic profiles actually predict response? This broader issue lays bare our tendency to confuse data with results and big data with big results. To wit, we must remember that our DNA, originally provided to each of us in the form of a single cell (the fertilized ovum) carries all of the genetic information that makes us, us. From the hair follicles on our heads to the acid secreting cells in our stomach, every cell in our body carries exactly the same genetic data neatly scripted onto our nuclear hard-drives.

What makes this all work, however, isn’t the DNA on the hard drive, but instead the software that judiciously extracts exactly what it needs, exactly when it needs it. It’s this next level of complexity that makes us who we are. While it is true that you can’t grow hair or secrete stomach acid without the requisite DNA, simply having that DNA does not mean you will grow hair or make acid. Our growing reliance upon informatics has created a “forest for the trees” scenario, focusing our gaze upon nearby details at the expense of larger trends and insights.

What is desperately needed is a better approximation of the next level of complexity. In biology that moves us from the genotype (informatics) to the phenotype (function). To achieve this, our group now regularly combines genomic, transcriptomic or proteomic information with functional analyses. This enables us to interrogate whether the presence or absence of a gene, transcript or protein will actually confer that behavior or response at the system level.

I firmly believe that the future of cancer therapeutics will combine genomic, transcriptomic and/or proteomic analyses with functional (phenotypic) analyses.

Recent experiences come to mind. A charming patient in her 50s underwent a genomic analysis that identified a PI3K mutation. She sought an opinion. We conducted an EVA-PCD assay on biopsied tissue that confirmed sensitivity to the drugs that target PI3K. Armed with this information, we administered Everolimus at a fraction of the normal dose. The response was prompt and dramatic with resolution of liver function abnormalities, normalization of her performance status and a quick return to normal activities. A related case occurred in a young man with metastatic colorectal cancer. He had received conventional chemotherapies but at approximately two years out, his disease again began to progress.

A biopsy revealed that despite prior exposure to Cetuximab (the antibody against EGFR) there was persistent activity for the small molecule inhibitor, Erlotinib. Consistent with prior work that we had reported years earlier, we combined Cetuximab with Erlotinib, and the patient responded immediately.

Each of these patients reflects the intelligent application of available technologies. Rather than treat individuals based on the presence of a target, we can now treat based on the presence of a response. The identification of targets and confirmation of response has the potential to achieve ever higher levels of clinical benefit. It may ultimately be possible to find effective treatments for every patient if we employ multi-dimensional analyses that incorporate the results of both genomic and phenotypic platforms.

The Changing Landscape in Non-small Cell Lung Cancer (NSCLC)

In October 2012, we published a study of patients with metastatic NSCLC whose treatment was guided by EVA-PCD laboratory analysis. The trial selected drugs from FDA approved, compendium listed chemotherapies and every patient underwent a surgical biopsy under an IRB-approved protocol to provide tissue for analysis.

The EVA-PCD patients achieved an objective response rate of 64.5 percent (2-fold higher than national average, P < 0.0015) and median overall survival of 21.3 months (nearly 2-fold longer than the national average of 12.5 months).

Non-small cell lung cancer

The concept of conducting biopsies in patients with metastatic NSCLC was not only novel in 2004, it was downright heretical. Physicians argued forcefully that surgical procedures should not be undertaken in metastatic disease fearing risks and morbidity. Other physicians were convinced that drug selection could not possibly improve outcomes over those achieved with well-established NCCN guidelines. One oncologist went so far as to demand a formal inquiry. When the hospital was forced to convene an investigation, it was the co-investigators on the IRB approved protocol and the successfully treated patients who ultimately rebuffed this physician’s attempt to stifle our work.

With the publication of our statistically superior results and many of our patients surviving more than 5 years, we felt vindicated but remain a bit battle scarred.

I was amused when one of my study co-authors (RS) recently forwarded a paper authored at the University of California at Davis about surgical biopsies and tumor molecular profiling published by The Journal of Thoracic and Cardiovascular Surgery. This single institution study of twenty-five patients with metastatic NSCLC reported their experience-taking patients with metastatic disease to surgical biopsy for the express purpose of selecting therapy. Sixty four percent were video assisted thoracic (VATS) wedge biopsies, 16 percent pleural biopsies, 8 percent mediastinoscopies, 12 percent supraclavicular biopsies and 8 percent rib/chest wall resections. Tissues were submitted to a commercial laboratory in Los Angeles for genomic profiling.

The authors enthusiastically described their success conducting surgical procedures to procure tissue for laboratory analysis. Gone was the anxiety surrounding the risk of surgical morbidity. Gone were the concerns regarding departure from “standard” treatment. In their place were compelling arguments that recapitulated the very points that we had articulated ten years earlier in our protocol study. While the platforms may differ, the intent, purpose and surgical techniques applied for tissue procurement were exactly the same.

What the Cooke study did not describe was the response rate for patients who received “directed therapy.” Instead they provide the percent of patients with “potentially targetable” findings (76 percent) and the percent that had a “change in strategy” (56 percent) as well as those that qualified for therapeutic trials (40 percent). Though, laudable, changing strategies and qualifying for studies does not equal clinical responsiveness. One need only examine the number of people who are “potential winners” at Black Jack or those who “change their strategies” (by changing tables/dealers for example) or, for that matter, those who qualify for “high roller status” to understand the limited practical utility of these characterizations.

Nonetheless, the publication of this study from UC Davis provides a landmark in personalized NSCLC care. It is no longer possible for oncologists to decry the use of surgical biopsies for the identification of active treatments.

As none of the patients in this study signed informed consents for biopsy, we can only conclude that the most august institutions in the US now view such procedures as appropriate for the greater good of their patients. Thus, we are witness to the establishment of a new paradigm in cancer medicine. Surgical biopsies in the service of better treatment are warranted, supported and recommended. Whatever platform, functional or genomic, patient-directed therapy is the new normal and the landscape of lung cancer management has changed for the better.

Two Women with Metastatic Breast Cancer – Same Age, Same Disease, Two Very Different Functional Profiles

A day in the life of advanced breast cancer. Two different 37-year-old breast cancer patients, both mothers of young children, were seen in consultation on the same day.

The first had been referred by a colleague who was concerned that the patient’s ER positive breast cancer had disseminated to her brain despite aggressive standard chemotherapy. She was to undergo a craniotomy and a portion of fresh tumor would be submitted from the surgery to Rational Therapeutics for EVA-PCD functional profiling.

The second mother had metastatic triple negative breast cancer, which recurred after aggressive standard chemotherapy. She underwent neo-adjuvant treatment (preoperative) but at the time of her surgery, there was no evidence of response to the treatment. By the time we met her, only months into her diagnosis, new areas of metastatic disease were cropping up daily.

The EVA-PCD assay results on these two “similar” patients were entirely different.

The results of the first patient with the ER positive tumor and brain metastases clearly identified treatments directed toward the PI3K pathway, with or without chemotherapy. We are recommending a combination of Everolimus plus chemotherapy.

The second patient had a completely different profile. Indeed, the degree of drug resistance was quite striking. A three-drug combination was among the most active from almost two dozen drugs tested.  The other option appeared to be a new class of drugs called the cyclin dependent kinase (CDK) inhibitors.

On a functional level, we used targeted drugs to probe for sensitivity to inhibitors of these cancer signal pathways. Unlike genomic profiles that tell you whether the gene is present or absent, we can tell whether the gene is driving the tumor. Functional profiling.

One patient is now under my care and the other will begin treatment under the care of a colleague in Orange County, CA. We will await results of these assay-directed therapies and wish these two young patients every success.

Gastric Cancer: A Call for Patient Selection

Gastric cancer is the fourth most common cancer worldwide with more than 930,000 diagnoses and 800,000 deaths attributed to this disease each year. Although relatively uncommon in the U.S., constituting only 2 percent of new cancers, in countries like Korea it makes up 20 percent of all new malignancies.

Among the causes are Helicobacter pylori infection, diets rich in smoked food, a high intake of nitrates and nitrites and cigarette smoking. A rare but aggressive form of the disease is associated with a gene mutation known as CDH1. The high frequency of metastatic disease at the time of initial diagnosis often precludes surgery, leaving systemic chemotherapy as the principal treatment option.

A recent report in The Annals of Oncology (No improvement in median survival for patients with metastatic gastric cancer despite increased use of chemotherapy: Bernards N. et al, Annals of Oncology. November, 2013) describes a retrospective analysis by Dutch investigators who examined the use of chemotherapy in patients with inoperable gastric cancer.

In total, 4,797 cases were examined from 1990 to 2011. Over this time, the proportion of patients presenting with metastatic disease increased from 24 percent in 1990 to 44 percent in 2011. At the same time, palliative chemotherapy use increased from 5 percent to 36 percent. Younger patients and those of higher socioeconomic status had the largest increase in chemotherapy use, while older patients, those with linitis plastica and those with multiple metastases had lower chemotherapy use. Despite the significant increase in the use of chemotherapy, the median survival for patients was unchanged at 15 weeks in 1990 and 17 weeks in 2011 (P = 0.1).

Over this period, early treatment regimens like 5-fluorouracil (5FU) and FAM were largely replaced by combinations like Docetaxel/Cisplatin/5-FU (DCF), Cisplatin/Irinotecan, Epirubicin/Oxaliplatin/Capecitabine (EOX) and Carboplatin/Taxol. While response rates and palliative benefits have continued to improve, this has not translated into improved overall survival. This reflects a dilemma that has confronted medical oncologists for decades.

For many years, clinical trialists have held that one cannot assess the benefit of a treatment by comparing responders to non-responders. That is, time to progression and survival must compare all patients on a given treatment arm to those on the control arm. Their rationale was that “one must treat all patients to obtain the benefit seen in some.”  Put differently you cannot “cherry pick” your winners and losers. It was said that this proscription was needed to avoid selection bias. But as any medical or nonmedical person would recognize, people who respond to treatment do better than those who do not. Lacking the ability to identify responders upfront, these trialists have insisted upon a one-size-fits-all approach to the detriment of clinical therapeutics and drug development.

With the dawn of the molecular era we see chinks in the armor of these trial designs as investigators now question why everyone should receive a treatment if only a small percentage will benefit. In gastric cancer, HER2 over-expression, found in 20-25 percent of patients, is now routinely used to identify patients who will respond to trastuzumab. But what of the other 75-80 percent of patients who do not carry HER2 and for whom there are no widely used determinants of clinical response? Do the results of Bernard article suggest that these patients should not receive therapy?

The Bernard article offers an interesting insight into what may be the future of medical oncology. As cancer therapy is increasingly scrutinized, not only for response or palliation but also for overall survival, patients may soon be denied treatments unless the results of the therapy rise to this, the highest level of evidence, for the entire population of treated patients.

Would it not be preferable to use laboratory analyses, like the EVA-PCD®, to select among treatment candidates before subjecting all patients to the risk and expense of toxic chemotherapy? In this regard, the author’s comments are poignant: “Identification of the subgroup of patients which benefit from palliative chemotherapy is of the utmost importance to avoid unnecessary treatment.” As a laboratory investigator engaged in the field of drug selection science (functional profiling), I couldn’t agree more.

Neuroblastoma Response to Therapy Trumps Age

In April of 2013, we received a tissue sample from investigators in Victoria, Espirito Santo, Brazil. The pediatric oncologist involved requested assistance in the management of a four-year-old child with Stage IV (metastatic) neuroblastoma.

The patient was originally diagnosed in February with abdominal pain and a tumor. The tumor was identified by ultrasound as a large left-sided retroperitoneal mass. The patient was treated with the combination of doxorubicin plus cyclophosphamide. Within a month, it was evident that his “high risk neuroblastoma” would require stronger chemotherapy. Doses were adjusted upward and cisplatin was added. As the patient’s tumor infiltrated his bone marrow, his tolerance of chemotherapy became limited. By early June, after recovering from severe infectious complications, with no evidence of response to treatment, he was taken to surgery.

For background, neuroblastoma is the third most common malignancy of childhood. It arises from sympathetic ganglia (nerve cells) and presents in different forms. It has long been recognized that these tumors can be driven by an up-regulation of the oncogene MYCN. Children above the age of 1.5 years and those with wide dissemination are at highest risk.

We received this patient’s tissue and immediately isolated the malignant populations. As the tumor is only identified in children, it was a somewhat unusual occurrence for our laboratory. In addition, the patient had already received extremely aggressive treatment without benefit. We chose among the drugs that we considered potentially active for study and proceeded with our analysis.

The EVA-PCD assay results were highly instructive. First, those drugs that the patient had already received (platins, alkylating agents) were clearly inactive. Second, the signal transduction inhibitors like imatinib, and everolimus were also inactive. What was striking however, was the extraordinary degree of sensitivity to taxol that placed this patient among the most sensitive patients we have ever tested. The profile for taxol also extended to two taxol-based combinations: taxol plus platinum and taxol plus gemcitabine. However neither combination revealed significant synergy, suggesting that taxol was the principally active agent.

biology.arizona.edu

As I considered our laboratory findings in the context of the contemporary pediatric neuroblastoma literature, several interesting threads emerged. The first, was that investigators in Leiden, Netherlands had described a microtubule associated protein (MAP) encoded double cortin-like kinase gene (DCLK1) in neuroblastoma patients.  The second was the very early but promising work using aurora kinase inhibitors in this disease. It became evident that these observations had their nexus at microtubule function. In keeping with the adult literature this would clearly support classes of drugs that induce G2-M arrest in the cell cycle. I reasoned that the taxanes were highly appropriate for this child based both on our findings and these related molecular correlates.

We contacted the physician in Brazil, and recommended a taxol-based treatment program. It became evident that neither taxol, nor the related carboplatin plus taxol or taxol plus gemcitabine regimens, were in this pediatric oncologist’s lexicon for neuroblastoma. Our report included references to clinical trials in adult tumors where these combinations have been broadly applied. However, it was going to require a certain amount of creative thinking for this well-trained pediatric oncologist to cross walk our “adult” recommendations to this child in need. Fortunately, with the assistance from our collaborators in Sao Paolo, the physician agreed to use our combination in this child who was failing standard treatment.

The results were prompt and dramatic. Within a single cycle of therapy, virtually all symptoms resolved. The child began to eat well and gain weight and despite chemotherapy, the bone marrow function rapidly recovered and the blood counts normalized. With completion of two cycles, a repeat CT scan revealed complete resolution of measurable disease.

I have corresponded with the pediatric oncologist and expressed our delight with the outcome and of her willingness to work with us. This case represents not only a transnational collaboration (the subject of a recent ASCO presentation) but also the successful cross-fertilization between the pediatric and adult oncology specialties. We are deeply gratified on both accounts.

ASCO Update: Personalized Cancer Care – Our Contributions

As part of our ongoing blog postings we like to include recent presentations and publications. On July 9, I described our ASCO presentation exploring crizotinib, “Functional Profiling Leads to Identification of Accurate Genomic Findings.”

To conclude the review of our other presentations from that meeting, here is a brief summary of our work.

The first of the two was our international collaboration in personalized medicine for the treatment of advanced and drug-refractory cancers: “Clinical application of human tumor primary culture analyses.” The study reviewed the results of 67 patients from institutions across Brazil.

Tumor samples were transported by overnight courier to California for drug response profiling. A broad array of tumors were included. The overall success rate provided actionable results in 62 of 67 patients (92 percent). More than 75 percent of the studies provided results for between 8 and 16 drugs and combinations with a median of 12 reported. Several strikingly good responses were observed, including novel combinations identified in the laboratory. This study confirms the feasibility of international collaboration and reflects the globalization of medical care delivery.

The final study published by ASCO was also a collaborative effort with SageMedic of Larkspur, CA, The Ludwig Maximilians University Munich, Germany and the Weisenthal Cancer Group. The study was a meta-analyses that examined the sensitivity and specificity of human tumor primary culture studies and the efficacy of drug therapies selected, based on laboratory findings. In aggregate there were 28 retrospective and 15 prospective trials included.

The overall sensitivity was 0.92 (95 percent C.I. 0.89 – 0.95), and specificity of 0.72 (95 percent C.I. 0.67 – 0.77) with an area under the curve for the ROC of 0.893 (SE = 0.023, p < 0.001). When clinical outcomes were examined, it revealed a two-fold improvement for assay-guided therapy for standard of care (odds ratio 2.04, 95 percent C.I. 1.62 – 2.57, p <  0.001). Finally, the one-year survival rate for assay-guided therapy proved superior (OR 1.44, 95% C.I. 1.06 – 1.95, p= 0.02).

As can be seen from this well conducted meta-analysis, there is a wealth of evidence to support the use of human tumor primary cultures for the selection of chemotherapy.

Functional Profiling Leads to Identification of Accurate Genomic Findings

The 2013 American Society of Clinical Oncology annual meeting, held May 31 – June 1, in Chicago, afforded the opportunity to report three studies.

Crizotinib (Xalkori)
Mechanism of Action

The first, “An examination of crizotinib activity in human tumor primary culture micro-spheroids isolated from patients with advanced non-small cell lung cancer,” reports our experience using the EVA-PCD platform to examine the drug crizotinib. This small molecule originally developed as an inhibitor of the oncogenic pathway MET, was later found to be highly active in a subset of cancer patients who carried a novel gene rearrangement for anaplastic lymphoma kinase (ALK). It was this observation that lead to the drug (sold under the name Xalkori) being approved for the treatment of advanced ALK positive lung cancer. The subsequent observation that this same drug inhibited yet another gene target known as ROS-1 found in a subset of lung cancer patients, has led to its use in this patient population.

Our exploration of crizotinib activity identified a series of patients who received the drug and responded dramatically. This included both ALK positive and ROS-1 positive patients. One patient however, appeared highly sensitive to the drug in our studies, but was found negative for the ALK gene rearrangement by genomic analysis. We repeated our functional analysis only to the find again, the same high degree of crizotinib sensitivity. I felt confident the patient should receive crizotinib, but at the time the drug was not yet commercially available and he didn’t qualify for the protocols, as he was ALK negative.

I scoured the country looking for a way to get the patient treated with crizotinib. From Sloan Kettering to UCLA, no one could help. And then, in collaboration with my abstract co-author Ignatius Ou from UC Irvine, we decided to repeat the ALK analysis. That proved to be a very good idea. For the patient was indeed positive for ALK gene rearrangement by second analysis and subsequently responded beautifully to a treatment for which he would not otherwise qualify. Once again, phenotype trumped genotype. (The complete story of this patient can be found in Chapter 19 of Outliving Cancer.)

A final patient in the series represented a particularly interesting application of functional analysis. The patient, a young woman with an extremely rare pediatric sarcoma, had failed to respond to multiple courses of intensive chemotherapy and her family was desperate. As she approached the end of her third year in high school, it looked unlikely that she would reach her senior year. A portion of her tumor was submitted for analysis. The results confirmed relative resistance to chemotherapeutics, many of which she had already received and failed, but showed exquisite sensitivity to crizotinib. Indeed, our inclusion of crizotinib in the analysis reflected our intense effort to identify any activity for this previously refractory patient.

We reported our findings to the pediatric oncologist and encouraged them to consider an ALK rearrangement analysis, despite this particular pathway not being on anyone’s radar prior to our study. The result – a positive gene rearrangement. This led to a successful petition to the drug company for the use of this agent for an off-label indication. The response was prompt and dramatic, and remains durable to this day, nearly a year later. Again, the phenotypic analysis guided us to the correct genomic finding.

Our other presentations at this year’s meetings will be reported in future blogs.

Gee (G719X) Whiz: Novel Mutations and Response to Targeted Therapies

In a recent online forum a patient described her experience using Tarceva as a therapy for an EGFR mutation negative lung cancer. For those of you familiar with the literature you will know that Lynch and Paez both described the sensitizing mutations that allow patients with certain adenocarcinoma to respond beautifully to the small molecule inhibitors.  The majority of these mutations are found in Exon 19 and Exon 21, within the EGFR domain. Response rates for the EGFR-TKI (gefitinib and erlotinib) clearly favor mutation positive patients. Depending upon the study, mutation positive patients have response rates from 53 – 100 percent, generally around 70 percent, while mutation negative response patients have a response rate of 0 – 25 percent, generally about 10 percent.

So why don’t all the mutation positive patients respond and conversely why do some mutation negative patients respond?

The story outlined in this online forum gives some insight. The individual in question carried a rare, and only recently recognized, Exon 18 mutation known as a G719X. This uncommon form of mutation had previously been unknown and few laboratories knew to test for it. Nonetheless, G719X positive patients respond to erlotinib and related agents. Indeed, there may be reason to believe that the more potent irreversible EGFR/HER2 dual inhibitor HKI-272, may be even more selective for this point mutation.

The excellent and durable response described by this individual, would not have been possible had the patient’s first physician followed the rules. That is, had her physician refused to give erlotinib to an (putatively) EGFR mutation negative patient she might well not be here to tell her story. More to the point, her good response (a clinical observation) led to the next level of investigation, namely the identification of this specific EGFR variant

The lessons from this experience are numerous. The first is that cancer biology is complex and, to paraphrase E.O. Wilson, was not put on earth for us to necessarily figure it out. The second, is that molecular biologists can only seek and identify that which they know about apriori.  To wit, if you don’t know about it (G719X) and you don’t have a test for it, and you don’t know to look for it, then it’s a virtual certainty that you aren’t going to find it.

The premise of our work at Rational Therapeutics is that the observation of a biological signal identifies a candidate for therapy whether we understand or recognize the target. Crizotinib was originally developed as a clinical therapy for patients who carried the CMET mutation. Serendipity led to the recognition that the responding subpopulation was actually carrying a heretofore-unrecognized ALK gene rearrangement. Sorafenib was originally evaluated for the treatment of BRAF mutation positive diseases. Yet it was the drug’s cross-reactivity with the VGEF tyrosine kinases that lead to its broad clinical applications. Each of these phenomena represents accidental successes. Were it not for the clinical observation of response in patients, the investigators conducting these trials would have been unlikely to make the discoveries that today provide such good clinical responses in others.

To put it quite simply, these patients and their disease entities educated the molecular biologists.

When we first identified lung cancer as a target for gefitinib, and began to administer the closely related erlotinib to lung cancer patients, neither Lynch nor Paez had identified the sensitizing EGFR mutations. That had absolutely no impact upon the excellent responses that we observed. It didn’t matter why it worked, but that it worked.  While the EGFR story has now been well-described, might we not use functional analytical platforms (functional profiling) to gain insights into the next, and the next generation of drugs and therapies that target pathways like MEK, ERK, SHH, FGFR, PI3K, etc., etc., etc. . . .