APPLICATION OF MOLECULAR TECHNOLOGIES TO CLINICAL MEDICINE
Along the continuum from health to disease, there are several important points where clinical decision making is now directly influenced and advantaged by molecular technologies. Risk estimates for developing some diseases can be defined during health and possibly even at birth using a variety of DNA analyses. Molecular signatures from technology platforms that measure the expressed genome (RNA, proteins, metabolites) can be used to define physiologic states in response to our environment and predict future clinical outcomes. These approaches also form the basis for a new molecular classification and taxonomy of disease and diagnosis. They can also provide more precise ways to screen for and detect disease at its earliest molecular manifestations, often preclinically. In addition, the selection of certain drugs may now be guided by a patient’s underlying genetic makeup as well as the molecular makeup of the disease. Given that a disease’s evolution from baseline risk often occurs over many years, periodic molecular profiling defines a novel form of health care monitoring that focuses on disease prevention and proactive management rather than the current paradigm of acute intervention and crisis response.
Use of molecular technologies across the continuum from health to disease. Various molecular technologies may be used to complement the traditional approach to evaluating at the time points indicated.
|TIME POINT IN CLINICAL DECISION MAKING||CANCER||CARDIOVASCULAR DISEASE|
Familion five-gene profile
|Screening||HPV genotypes||Cervical||Corus CAD||CAD|
|Diagnosis||Cancer Type ID
|Cancer of unknown primary
Ovarian mass malignancy
|Prognosis||Oncotype DX (21-gene assay)
MammaPrint (70-gene assay)
HER2/neu, ER, PR
|Breast||TnI, BNP, CRP||ACS|
NGS of somatic variation for targeted therapy
|Monitoring||CTCs||Tumor recurrence or progression||AlloMap gene profile||Transplant rejection|
ACS = acute coronary syndromes; BNP = brain natriuretic peptide; CAD = coronary artery disease; CRP = C-reactive protein; CTCs = circulating tumour cells; ER = oestrogen receptor; HPV = human papillomavirus; LQTS = long QT syndrome; NGS = next-generation sequencing; PR = progesterone receptor; TnI = troponin I.
A key question in medicine is to what extent genetic variation influences the likelihood of disease onset, affects the natural history of disease in combination with the environment, or provides clues relevant to the management of disease. In addition, it is not just the human genome that is relevant to an individual’s state of health. The genomes of thousands of microorganisms that constitute our microbiota are also relevant to human phenotypes, and insights from their genomes are providing new approaches for the diagnosis, study, and treatment of disease.
Genome-wide association study (GWAS) debuted in 2005 with the identification of variants in the complement factor H gene as a cause of age-related macular degeneration. GWAS has been a transformative approach to identifying common genetic variation across the entire human genome in an unbiased fashion, offering an unprecedented opportunity to uncover new biologic pathways of disease. GWAS has been carried out on large cohorts of patients and controls across numerous traits and diseases, revealing hundreds of common genetic variants associated with those traits.
Since 2001, the cost of sequencing a human genome dropped from $3 billion to less than $10,000. Next-generation sequencing (NGS) technology can now read approximately 250 billion bases in a week and allows direct measurement not just of common variants but also theoretically of all variations in a genome. It is estimated that the population frequency of germline variants is approximately 1 in every 1000 of the 3.2 billion nucleotide positions, giving rise to approximately 3 million variants in a given human genome. The challenge lies in figuring out the meaning of variants, many of which occur in noncoding regions (introns) of the genome whose function is largely unknown. Until the significance of the noncoding variants is understood, the focus clinically has been on exome sequencing (which examines variation in the coding sequence of exons that are translated into proteins), where mutations have predictable effects on downstream protein structure. Exome sequencing, which represents only about 1.5% of the 3 billion nucleotides that constitute the human genome, is still less expensive to perform than whole-genome sequencing . Individuals typically carry several hundred rare and potentially deleterious coding region variants. The first successful clinical applications of exome sequencing in 2009 revealed the diagnosis of patients with Freeman-Sheldon syndrome, and it is increasingly being explored for clinical applications, including the clinical diagnosis of rare genetic diseases, the selection of cancer treatments based on molecular characterization of the tumour, and the tracking of infectious disease outbreaks in real time.
Despite strong statistical associations linking genetic variants with complex diseases, the low relative risk of the disease alleles (generally <2) limits their use for disease predisposition testing and risk assessment. There are notable exceptions, including genetic variation underlying breast cancer, Lynch syndrome, and celiac disease, where some variants have enabled preventive treatment or screening of family members. Despite these technologic advances in genomics, a simple family history continues to be among the best tools to identify risks for common diseases. In fact, for conditions with high heritability, such as cardiovascular disease, family history is a much stronger predictor of disease than any single or combination of genetic/genomic markers. One model suggests that neither family history nor genetic testing should be used as a standalone but that the real power for disease prediction, risk assessment, and differential diagnosis comes from their combined use.
More than 3500 mendelian disorders have a known molecular basis. However, there are nearly as many suspected mendelian traits for which the molecular basis remains to be identified. The potential for clinical sequencing to find the underlying cause and identify treatment options for these rare, sometimes debilitating diseases has led to the formation of various large national and international rare disease consortia. In some specialized clinical centres and through programs such as the National Human Genome Research Institute Undiagnosed Diseases Program, clinical sequencing is being offered to patients with suspected genetic diseases, the so-called diagnostic dilemmas. Early results from these clinical sequencing programs suggest that the success rate of disease gene identification is about 50%, offering hope for a diagnosis to thousands of individuals with previously undiagnosed or untreated rare disorders.
A natural outcome of identifying genes for rare mendelian disorders is the application of these findings to earlier detection, at birth (newborn screening), in utero (prenatal diagnosis) or before conception (carrier testing). Newborn screening—mandatory, state-supported public health programs meant to protect newborn children by screening them for rare, treatable (and thus preventable) disorders at birth—has been steadily increasing from an average of five conditions in 1995, to a panel of 31 core disorders and 26 secondary disorders currently recommended by the U.S. Department of Health and Human Services. 8 Rapid whole-genome sequencing (about 50 hours from test to result) was recently reported, and because the number of conditions considered for newborn screening will undoubtedly grow, rapid whole-genome sequencing can potentially broaden and foreshorten differential diagnoses, resulting in fewer empirical treatments and faster progression to genetic and prognostic counselling.
Cell-free foetal DNA circulating in maternal blood was isolated, amplified, and sequenced noninvasively through a sample of maternal plasma in 1997, In 2008, NGS technologies were used successfully to identify foetal aneuploidy from cell-free foetal DNA in maternal plasma. Clinical trials of the new method rapidly followed, and by late 2011, noninvasive prenatal testing of trisomy 21 by sequencing of maternal plasma DNA was being offered on a clinical and commercial basis in the United States and China. Noninvasive prenatal testing eliminates the need for invasive procedures, while also greatly expanding the number of genetic variants that have traditionally been detected in utero.
Before conception, carrier screening enables couples to assess their risk for having a child with a recessive mendelian disorder and to use this information to guide their reproductive decisions. There are more than 1000 rare, recessive mendelian disorders for which the underlying genetic mutation is known. Although individually rare, these can have a sizable public health impact considering that each person is estimated to carry on average 2.8 mutations for known severe recessive disorders, and the impact of screening could be substantial in terms of reduced disease morbidity and mortality in the population.
Several genomic markers of efficacy, adverse events, and dosing of therapeutics have been discovered, but their uptake into clinical practice has been variable, despite their clear actionability. In some cases, such as with the HLA-B*5701 genotype for the HIV drug abacavir and HLA-B*1502 for the antiseizure drug carbamazepine, carriers of these genotypes should avoid the drug entirely to eliminate a specific serious adverse event. In other cases, such as thiopurine S -methyltransferase (TPMT) for mercaptopurine or CYP2C9/VKORC1 for warfarin, adjusting the dose of drug based on genotype can help to avoid toxicity and improve efficacy. Actionability is not enough to ensure diffusion of pharmacogenomics testing into clinical practice, as exemplified by the antiplatelet drug clopidogrel, for which despite having a U.S. Food and Drug Administration black box warning for efficacy in individuals carrying the CYP2C19 genetic variant, there is no clear consensus among physicians on its use. In hepatitis C treatment, on the other hand, the IL28B genotype test not only has proved highly predictive of response to pegylated interferon/ribavirin used to treat chronic hepatitis C virus infection but also has seen rapid and widespread adoption in the clinic. Genetic markers that predict reduced therapeutic efficacy may face a high hurdle for established drugs, unless evidence supporting clinical validity and utility of the test is indisputable.
SELECTED MOLECULARLY GUIDED THERAPEUTICS AND INDICATIONS
|Cancer treatment regimens||Oncotype DX 21-gene assay||Breast cancer : 21-gene expression score linked to the likelihood of breast cancer recurrence in women and the magnitude of benefit from certain types of chemotherapy and hormonal therapy|
|Irinotecan||UGT1A1||Colon cancer : variations in theUGT1A1 gene can influence a patient’s ability to break down irinotecan, which can lead to increased blood levels of the drug and a higher risk for side effects.|
|Carbamazepine||HLA-B*1502||Epilepsy and bipolar disorder : serious dermatologic reactions and HLA-B*1502allele. Before initiating therapy, testing for HLA-B*1502 should be performed in patients with ancestry in populations in which HLA-B*1502 may be present.|
|Abacavir||HLA-B*5701||HIV : test determines patients most likely to experience an adverse hypersensitivity reaction.|
|Multiple diseases : these tests are used as an aid in determining treatment choice and individualizing treatment dose for therapeutics that are metabolized primarily by the specific enzyme about which the system provides genotypic information.|
|Warfarin||CYP2C9||Venous thrombosis/stroke : increased bleeding risk for patients carrying either the CYP2C9*2 or CYP2C9*3 alleles|
|Warfarin||VKORC1 (vitamin K epoxide reductase)||Venous thrombosis/stroke : single-nucleotide polymorphisms in theVKORC1 gene (- 1639G/A allele) associated with lower dose requirements|
|Clopidogrel||CYP2C19||Coronary artery disease: increased risk for stent thrombosis and secondary events following percutaneous interventions in patients with theCYP2C19*2 variant|
|Immunosuppressive drugs||AlloMap gene profile||Heart transplantation : blood gene expression score to monitor a patient’s immune response after cardiac transplantation and to guide immunosuppressive therapy|
|Pharmaceutical and surgical prevention options and surveillance||BRCA1, BRAC2||Breast cancer : guides surveillance and preventive treatment based on susceptibility risk for breast and ovarian cancer|
|Pharmaceutical and lifestyle options for disease prevention||Familion 5-gene profile||Heart disease : guides prevention and drug selection for patients with inherited cardiac “channelopathies” such as long QT syndrome that may lead to cardiac rhythm abnormalities|
|TPMT (thiopurineS -methyltransferase)||Acute lymphoblastic leukaemia : patients with inherited little or no TPMT activity are at increased risk for severe toxicity from conventional doses.|
|Maraviroc||CCR5 promoter and coding polymorphisms||HIV: determines whether the patient is likely to respond or not|
Cancer arises as a result of somatic DNA mutations that confer a growth advantage on the cells in which they have occurred, giving rise to tumours. Comparison of the genetic profiles of tumours and the surrounding normal tissue (gene expression profiling) can reveal the acquired DNA variation that drives growth and that may reveal targets for treatment.
The idea of pairing medicines with specific tumour markers in a targeted fashion became a reality in the mid-1980s when detailed molecular studies of breast tumours led to the discovery of human epidermal growth factor receptor-2 (HER-2), a biomarker overexpressed in approximately 30% of breast tumours and associated with adverse outcomes. Subsequently, trastuzumab (Herceptin), a humanized monoclonal antibody targeting HER-2, was developed in 1998 and was shown to have increased efficacy in patients whose tumours tested positive. HER-2 testing of tumour is now part of the standard work-up and management of breast cancer. In the past decade, other examples of cancer therapies with companion diagnostics have emerged. For example, EGFR mutation testing has markedly improved the efficacy of gefitinib and erlotinib, small molecule drugs for the treatment of non−small cell lung cancer that target EGFR. In metastatic colorectal cancer, tumours with mutated KRAS are usually resistant to treatment with cetuximab and panitumumab, leading the American Society of Clinical Oncology and the U.S. Food and Drug Administration (FDA) to recommend withholding the drugs in these patients. NGS now allows a comprehensive assessment of actionable tumour markers that indicate the potential for a specific therapeutic to have efficacy in a given tumour. In 2011, two cancer drugs received accelerated approval by the FDA for use with a companion diagnostic test: (1) crizotinib for the treatment of patients with locally advanced or metastatic non−small cell lung cancer with its companion diagnostic designed to detect the EML4-ALK fusion gene, and (2) vemurafenib for the treatment of patients with metastatic or unresectable melanoma positive for BRAF V600E mutations. The International Cancer Genome Consortium and the Cancer Genome Atlas represent international collaborative efforts to define the spectrum of mutations found in tumours, mapping the genomic landscape of cancer. These efforts will provide a foundation from which to develop additional therapeutic strategies against new targets. However, even when successful, the results may be short-lived as therapeutic resistance evolves. Thus, although NGS is a promising new tool for surveying cancer genomes, it may not be a panacea for cancer genomic medicine.
MOLECULAR MARKER INFORMED CANCER THERAPIES (TARGETED THERAPEUTICS)
|BIOMARKER||DRUG||CANCER TYPE||FDA DRUG LABELING RECOMMENDED OR REQUIRE OESTROGEN RECEPTOR||Tamoxifen||Breast||Yes|
5-FU = 5-flurouracil; ALL = acute lymphoblastic leukaemia; CML = chronic myelogenous leukaemia; FDA, U.S. Food and Drug Administration.
We can now rapidly sequence the genomes of microorganisms—both the commensal bacteria that regularly inhabit our bodies (the human microbiome ) 12 as well as the pathogenic infectious agents that cause acute and sometimes fatal diseases. 13 The Human Microbiome Project recently published a study of the microbial populations inhabiting various human body sites and provided reference sequences for many taxa in health individuals as well as their correlation with host characteristics, including ethnicity, age, and body mass index. There are now emerging associations of human microbiota and diseases such as diabetes, asthma, psoriasis, atherosclerosis, and obesity. Moreover, strategies to modify the gut microbiome are being explored as treatments for inflammatory bowel disease, including the use of faecal transplantation or engraftment of microbiota from a healthy donor into a recipient. The human microbiome will play an important role in molecular medicine because microbial composition can be altered noninvasively through diet or the use of probiotics or antibiotics.
In infectious disease, diagnosis by NGS may supplant the need to first grow microorganisms in culture, previously a major impediment to pathogen identification. For example, in 2003, sequencing of samples from infected patients with the severe acute respiratory syndrome identified the causative agent as a coronavirus. Comparison of sequences of multiple isolates of an organism from a single epidemic gives a picture of the organism’s evolution, allowing one to infer where the outbreak began and how the infection spread. Sequencing has been used to determine the origins of historical outbreaks of cholera, tuberculosis, and the 2009 H1N1 influenza. The clinical application of NGS to infectious disease was highlighted recently when the source of carbapenem-resistant Klebsiella pneumoniae in a hospital outbreak was identified by sequencing isolates of the bacteria—in real time—from infected individuals and examining the genetic differences.
Beyond DNA sequence, measures of gene expression, proteins, metabolites, and epigenetic changes are being used to generate comprehensive profiles of biologic systems in health and disease. Many of the computational challenges of analysing these large, complex data sets are being addressed to yield next-generation biomarkers that are multianalyte, diagnostic, prognostic, and predictive. A growing number of marketed tests now typically measure protein or RNA levels, often with complex algorithms, enabling diagnosis and prognosis. One example is Oncotype DX (Genomic Health Inc., Redwood City, CA), a test that examines expression of 21 genes in tumour tissue to determine the likelihood of disease recurrence in women with early-stage hormone oestrogen receptor−positive breast cancer. The test, which is currently covered by many major insurance companies, analyses expression levels and converts them to a recurrence risk score that has been shown to help guide treatment in patients, reduce overall health care costs, and improve outcomes. Other examples include MammaPrint (Agendia Inc., Irvine, CA), which analyses the expression of 70 genes to determine whether patients are at high or low risk for breast cancer recurrence; OVA1 (Vermillion, Inc., Austin, TX), a five-protein test that gauges whether a woman’s ovarian mass is malignant and requires surgery; AlloMap (XDx Expression Diagnostics, Inc., Brisbane, CA), an 11 blood gene RNA signature for monitoring rejection after cardiac transplantation; and Corus CAD (CardioDx, Inc., Palo Alto, CA), a 23-gene blood RNA signature to screen for obstructive coronary artery disease.
Despite their complexity, in vitro diagnostic multianalyte index assays (IVDMIAs) like these are finding their way to the clinic. The 2007 draft guidance from the FDA suggested that IVDMIAs are used to make critical health care decisions and thus should be regulated by the FDA. Some of the marketed IVDMIAs have demonstrated analytical and clinical validity, but evidence of clinical utility is usually lagging. Moreover, the very nature of IVDMIAs presents challenges to insurers, who grapple not only with limited data on clinical utility but also with how to reimburse such tests that comprise both a laboratory component and an associated algorithm used to score risk, the latter part being integral to realizing the test’s value.
The success of some IVDMIAs is evidence of the power of computational biology but also of the importance of advocacy and financial resources that the commercial developers of these tests must bring. Companies developing IVDMIAs are able to finance key studies aimed at demonstrating clinical validity, navigate regulatory hurdles, advocate for coverage by insurance companies, and disseminate their tests through marketing to health care providers. Their efforts offer valuable lessons on the effective translation of complex molecular tests to medicine.
The large-scale study of proteins, proteomics, allows for both protein identification and differential expression between two physiologic states (such as health and a specific disease). Quantitative proteomics, in which global differences in protein abundances are measured, continues to be a priority area for biomarker discovery and molecular medicine. This area has been dominated by stable isotope approaches, but recent label-free quantitative methods have been developed that rely on the measured intensity of a peptide ion and compares this to its intensity in other samples. Label-free methods have the advantage of higher throughput and fewer sample manipulation steps. Multiple—or selected-reaction monitoring of specific peptides within biofluids allows quantitation of absolute abundance of proteins in clinical samples. Although this technology is relatively immature in its applications to human health and disease, compared with RNA and metabolic profiling, it is anticipated that these methods, combined with the development of mass spectroscopy technology, will advance proteomics to more routine use in disease classification and diagnosis, prognosis, and pharmacogenomics within the next several years.
A metabolic profile is very similar to some of the traditional targeted profiles, such as a lipid profile, although it is more comprehensive. Metabolomics measures changes in the metabolic or chemical milieu that are downstream of genomic and proteomic alterations. It is estimated that humans contain approximately 5000 discrete small molecule metabolites, and the identification of metabolic fingerprints for specific diseases may have particular practical utility for the development of therapies because metabolic changes immediately suggest enzymatic drug targets. Similar to genomics and proteomics, metabolomics may be useful in disease diagnosis, prognosis, and drug development. In particular, metabolomics will likely be a valuable tool in assessing drug toxicity. Targeted mass-spectroscopy-based metabolic profiling has also been increasingly applied to studies of human diseases and conditions. These tools are being applied to diverse areas, such as diabetes, obesity, cardiovascular disease, cancer, and mental disorders.
In order for molecular medicine to be practiced, it must be woven into current systems of health care delivery, with due consideration given not only to the providers of health care but also to the organizations in which they practice as well. Implementation scientists have outlined various aspects that need to be considered in order for molecular medicine to take hold in the clinical setting. Beyond the scientific soundness of the molecular or genomic test, measured by a strong evidentiary base and regard for potential benefits and harms, there is consideration of how the new test will integrate into the clinical workflow. Consideration should be given to aspects such as access to a laboratory certified by the Clinical Laboratory Improvement Amendments of 1988, methods for sample preparation and transport, test ordering, and receipt and delivery of results. Genomic test implementation is complicated by issues of privacy, complex interpretation of results, and the need to involve third parties for counselling in some cases; they may require the development of new systems to accommodate them.
A robust means of integrating genomic and molecular data into electronic health records will be required, with consideration of not only data storage formats and privacy issues but also appropriate decision support tools for prompting their use at the point of care and delivering results in an easily interpretable format. Currently, there are several examples of decision support tools, such as Warfarin Dosing, but they are typically standalone tools and not part of routine clinical workflow. To maximize their effectiveness, such tools should be integrated into electronic health records. Tapping into the collective knowledge and experience of various institutions working in this space would greatly facilitate this effort. Ultimately, a national, standardized technical architecture for integrating clinical decision support into electronic health records will be required. Notable efforts in this space include those of Health Level 7, an organization that provides interoperability standards for the exchange, integration, sharing, and retrieval of electronic health information. Through their Clinical Genomics Workgroup, this organization has developed a standards guide for genetic testing that includes document templates to support integration of genetic testing into electronic health records. 1 Appropriate clinical decision support, provided in the context of the electronic health record, will greatly facilitate the diffusion and uptake of genomic medicine.