SAN DIEGO—In just two decades, precision medicine has gone from futuristic concept to realistic toolbox for clinical physicians. At the 2017 ACR Clinical Research Conference on Nov. 3, the Precision Medicine in Rheumatic Diseases: Hopes and Challenges lecture featured rheumatologists and experts on genetics, genomics, pharmacogenetics and big data who spoke about the latest research in this field and how it may lead to truly personalized rheumatology.
Precision medicine is an umbrella term that may mean different things to different people, said Soumya Raychaudhuri, MD, PhD, associate professor of medicine at Harvard Medical School in Boston, and professor of genetics at the University of Manchester in England.
“What is precision medicine? If you go to the NIH [National Institutes of Health] website and look at the Precision Medicine Initiative, it says, ‘Precision medicine is an emerging approach for disease treatment and prevention that takes into account individual variability in genes, environment and lifestyle for each person.’ But it’s also seen as a medical model customized and tailored to the individual patient,” said Dr. Raychaudhuri, who co-chaired the conference with Lindsey Criswell, MD, MPH, professor of medicine at the University of California, San Francisco.
“My take on it is that precision medicine is very much about technology and data, and integrating [those] data into our model of how we approach patient care. So precision medicine exploits the intersection of clinical information, which we use all the time, and high-resolution individual patient data, to elicit effective treatment decisions,” he said.
The Human Genome Project, a massive effort to map the entire human genome funded by the NIH and completed in April 2003, was precision medicine’s launching pad, leading to such technologies as human genome sequencing, transcriptional profiling and immune cell profiling using mass cytometry, which allows researchers to examine up to 40 biomarkers per cell, Dr. Raychaudhuri said.1 Researchers can now develop a transcriptome for a single cell and use it to sequence thousands of cells, and can also use new epigenetic technologies to “query the genome to determine the areas that are active at any time, allowing us to quickly identify which promoters and enhancers in the genome are active in this particular patient.”
Practical examples of precision medicine in clinical practice already exist, such as:
- Using pharmacogenetics to tailor drug therapy, such as testing for TMPT enzyme activity or a TMPT genotype before starting azathioprine therapy;
- using genetic sequencing technology to define rare diseases; and
- sero-immunotyping a patient to gather data to inform treatment decisions.
With technology becoming more affordable, faster and accessible, “there are exponential expectations” for precision medicine, Dr. Raychaudhuri said.
How can precision medicine help rheumatologists diagnose and treat monogenic, autoinflammatory diseases? One example: When young children from two different states were suffering recurrent strokes, researchers used new gene sequencing techniques to discover recessive, inherited loss-of-function mutations in their CECR1 gene associated with an inflammatory phenotype, including early-onset strokes and vasculitis.2 Eventually, this data led the rheumatologists to treat the children with biologics that halted their strokes, said Daniel Kastner, MD, PhD, Intramural Research Project Director at the National Human Genome Research Institute (NHGRI) of the NIH.
His own interest in genomic medicine dates back 20 years, when he worked on the development of high-resolution genetic mapping of MEFV, the gene mutated in familial Mediterranean fever (FMF) in families of North African and Middle Eastern descent, he said.3
“The idea that you could use tools from the beginning of the Human Genome Project to discover the genes underlying what were, at that time, unexplained conditions, sounded like a really cool thing to do,” said Dr. Kastner. With only a few hundred genetic markers then known, early genomic research required the collection of large amounts of human DNA from donated placentas, he said. “We had to find out what chromosome the gene was on, and we hoped we would eventually find a gene teaching us something about inflammation. That’s exactly what happened. We did find a novel gene, and this protein product nobody knew anything about. There was a long process to figure out what this gene does and how it affects human disease.”
He and his colleagues in the International FMF Consortium discovered that pyrin, a domain on the MEFV gene, allows molecules in the innate immune system to associate with each other, which led to the discovery of a new signaling pathway. Rho-GTPase, which lives inside the cell membranes of certain leukocytes, normally leads to the activation of PKN-1, a kinase that in turn phosphorylates pyrin and prevents the activation of IL-1, an inflammatory cytokine. Bacterial toxins can inactivate Rho-GTPase and inhibit PKN-1, which then inhibits the phosphorylation of pyrin and leads to the formation of this inflammasome and IL-1 production in these patients, he said.
“So this knowledge has helped us understand this disease better, and has helped us treat patients,” said Dr. Kastner. Some FMF patients don’t respond to colchicine, the usual therapy, because the drug affects Rho-GPTase activity. “For those patients, we now know we can use IL-1 inhibitors. It’s made a huge difference in the lives of those people.”
Starting with their discoveries related to MEFV mutations and FMF, Dr. Kastner and other NIH researchers looked at mutations in other genes involved in the same pathway, and they found that several inflammatory diseases, including Muckle-Wells syndrome, familial cold autoinflammatory syndrome and neonatal-onset multisystem inflammatory disease (NOMID), were also related. Although children with NOMID often died before they reached adulthood, genomic research allowed rheumatologists to develop IL-1 inhibitors to treat it, and “now these kids are graduating from high school,” said Dr. Kastner.
Through genomic research, rheumatologists are discovering new gene mutations that lead to more accurate diagnosis and effective treatment of many diseases. When patients thought to have lupus didn’t respond to standard therapy, for instance, NHGRI researchers used exome sequencing to discover a gene mutation in TNFAIP3 that caused haploinsufficiency of the protein A20. This mutation upregulated the NFκB signaling pathway, activated the NLRP3 inflammasome and triggered overproduction of cytokines. The patients didn’t have lupus, but a disorder similar to Behçet’s disease, suggesting biologics might provide effective therapy, he said.4
Precision Medicine in Lupus
How can rheumatologists use precision medicine to understand and manage a complex autoimmune disease such as systemic lupus erythematosus (SLE)? Although we know a great deal about its genetic etiology, “we aren’t as far along in lupus as we are in other complex human diseases like cancer,” said Judith A. James, MD, PhD, Vice President of Clinical Affairs at the Oklahoma Medical Research Foundation in Oklahoma City. “Lupus has had a lot of failed clinical trials. Different patients may have the same diagnosis and meet the classification criteria, but patients may have pathogenic heterogeneity we don’t yet fully understand.”
In lupus, precision medicine may require targeted therapy based on trials of patients who share a molecular pathway, she said. “Lupus is a complicated process. Interrogating just an individual’s DNA or cells or soluble mediators may not [suffice], and we may have to start thinking about putting together different pieces so we can assemble a more cohesive story and identify which patient should be in which group or treated with a specific therapy.”
In the past, lupus patients have been grouped by organ involvement, demographics or certain biomarkers, but personalized medicine research may need to encompass genetics, genomics, immunophenotyping and soluble mediators. More than 100 genetic associations in lupus have been mapped, including a major new study that used Immunochip genotype data from 27,574 individuals of different ethnic backgrounds and revealed novel associations and definition of genetic load in this disease, she said.5 This genetic association study included mostly patients of European descent, so “we need more information on genetic risk factors in African-American, Hispanic and patients from other racial/ethnic backgrounds.” With more comprehensive genome-wide association study (GWAS) data, rheumatologists will learn more about the genetic architecture and genetic load of lupus.
“This information will eventually come into our clinics so we can counsel individuals about their genetic risk and environmental factors, ideally putting this information together to help inform our patients and individuals at risk for autoimmune rheumatic diseases,” she said.
Personalized immunomonitoring, or studying gene expression profiles to identify molecular pathways linked to disease activity, was used in a recent pediatric SLE study.6 Researchers looked at 486 gene transcripts in 158 pediatric lupus patients and 48 healthy controls, and found that children with elevated disease activity had specific interferon and plasmablast signatures.
Because lupus is highly heterogeneous, “each individual patient seems to have different modules that may be important to their disease and their disease activity,” said Dr. James. Transcriptional immunomonitoring is an intriguing tool for lupus research, but “the challenge is that some parts of these modules are driven by genetics, which can lead to constant changes in the modules for subsets of patients. Concurrent infections can also change expression in some modules, so additional investigation, especially in adult patients, can help the practicing clinician know if we can stop the patient’s therapy, or if a patient needs more aggressive therapy.” Additional modules may also help identify patients at higher flare risk before they come into the office, she said.
Current research on soluble mediators, such as complements, autoantibodies, cytokines, chemokines and shed receptors, may help complete the puzzle. In a 2017 study of African-American lupus patients, researchers found soluble mediators that suggested which patients were about to flare.7 This research could help develop an algorithm or “soluble mediator score” to help select patients for future clinical trials, or point to which patients need therapy modifications, she said.
Precision Medicine in JIA
Rheumatology currently views pediatric and adult arthritis as two separate disease families, but they should be viewed as a continuum, said Peter A. Nigrovic, MD, director at the Center for Adults with Pediatric Rheumatic Illness at Brigham and Women’s Hospital, and associate professor of medicine at Harvard Medical School, both in Boston, Mass. Further, using GWAS data to identify similar forms of arthritis across age groups may lead to better understanding of disease mechanisms, he said.
“Adult and pediatric arthritis have historically been described using totally distinct nomenclatures, and this has had consequences,” said Dr. Nigrovic. “Treatment algorithms begin with naming the disease for which they are intended to be used. Insurers approve medications for some diseases and not others. Researchers include only patients with a particular disease in their studies. Any attempt at precision medicine therefore requires that we get the nomenclature right. But it’s not easy to decide what goes with what.”
Disease classification should anchor on the “primacy of pathogenesis,” encompassing genetics and biological features as well as clinical phenotype, he said. As researchers dig more deeply into patient subgroups, both similarities and differences emerge. “Seropositive and seronegative RA are distinguished not only by the presence of certain autoantibodies, but also by the presence of immune complexes and complement fixation in joints, by the abundance of specific T cells in synovial tissues, and by environmental and genetic risk factors. Spondyloarthritis describes another family, with a distinct gender ratio and pattern of affected joints, the prevalence of enthesitis, and the prominent genetic association with HLA-B27.”
Such clear biological differences are much less evident between categories of juvenile idiopathic arthritis (JIA) as currently defined. For example, oligoarticular and seronegative polyarticular disease are more similar than different, while so-called “enthesitis-related arthritis” excludes another enthesitis-related disease, psoriatic JIA.
Dr. Nigrovic highlighted a recent study that showed HLA associations for subtypes of JIA often overlap those in adult arthritis. Hinks, Bowes and colleagues studied 5,000 JIA patients and 14,000 healthy controls, finding that HLA associations for most subgroups of JIA corresponded to those for clinically related forms of arthritis in adults, strongly suggesting pathophysiologic continuity.8 For example, HLA associations for seropositive JIA are identical down to specific amino acids in the HLA binding pocket with those in seropositive rheumatoid arthritis (RA). “So there is in fact substantial overlap between juvenile and adult arthritis,” he said.
“This doesn’t mean children should be lumped with adults for all purposes,” he noted. Children will exhibit some differences, such as in drug clearance, toxicity and the effects of disease and treatment on growth. “But it’s important to keep your thinking clear on this topic. Don’t make the mistake of considering these forms of arthritis different simply because we sometimes need to treat children differently. It’s important to keep questions of etiology apart from those regarding the practicalities of treatment.”
As an analogy, Dr. Nigrovic noted that physicians don’t classify pneumonias as juvenile and adult, but rather by pathogenic agent, recognizing at the same time that the organisms that cause pneumonia can vary widely with age and risk factors.
Juvenile arthritis is different from adult disease in some ways, he said. In particular, arthritis in children is most common between the ages of one and four. These young children often develop disease in only a few joints and are prone to chronic anterior uveitis, a phenotype essentially never seen in adult-onset arthritis. This form of early-onset arthritis may be a distinct disease. However, because it still shares HLA associations with adult seronegative arthritis, “it’s also possible it’s the same disease but arising in a different milieu,” said Dr. Nigrovic. “We can’t really choose definitively between these two hypotheses at this point.”
GWAS data used to identify regions of DNA linked to human disease risk are revealing more about mechanisms involved in juvenile rheumatic diseases even beyond the HLA region, he said.
“These data are important for understanding mechanisms. Each hit corresponds to a pathogenic pathway implicated by human genetics in disease biology.” Importantly, most GWAS hits don’t reflect coding regions, but rather alter segments of DNA that regulate gene expression via binding of regulatory proteins such as transcription factors.
At his laboratory, Dr. Nigrovic and his colleagues recently developed a new method called SNP-seq (i.e., single nucleotide polymorphism-DNA sequencing) that, with support from the Rheumatology Research Foundation, was employed to screen 27 non-HLA loci in JIA to identify such regulatory variants, finding more than 100 candidate variants.9 They also developed a technique called Flanking Restriction Enhanced Pulldown (FREP) and used it to define two new associations between specific regulatory proteins and two JIA-associated SNPs that modulate expression of the T cell regulatory gene STAT4.10
“Arthritis transcends the pediatric–adult boundary. We need to unshackle ourselves from the current nomenclature and to unlock the potential of GWAS to reveal new mechanisms involved in this continuum,” he said.
Age of onset nevertheless remains informative, he said. “In most diseases, early onset means genetic loading, suggesting that pediatric-onset disease may be an especially rich area to seek genetically driven disease.” In a new review in Arthritis & Rheumatology, Dr. Nigrovic and his colleagues also explore the continuity of pediatric and adult arthritis from a genetics perspective.11
Genomics Clues Beyond Rheumatology
Rheumatologists can turn to precision medicine research advances in other diseases and outside sources to find methods and concepts that lead to personalized, effective treatment of rheumatic diseases. Cancer researchers pioneered efforts to find genes that contribute to disease risk.
“With few exceptions, it’s never a single penetrative allele or gene. We’re looking for many genes that may contribute to cancer,” said Josh Stuart, PhD, professor in the Biomedical Engineering Department and Baskin Engineering Endowed Chair at the University of California, Santa Cruz. “How we can use an integrative, pathway-level approach to not only classify phenotypes, but select treatments in a patient-specific way?”
Genomics research in a group of patients with the same type of tumor may not show many commonalities, so researchers drill down to the pathway level to find those links. An advanced prostate cancer patient may have various gene mutations, for example. When researchers group him with others who have similar disease, “we look for alterations in the genome, then look for alterations in the transcriptome. Think of the transcriptome as everything that’s different from a matched, normal sample,” said Dr. Stuart. “We’re hoping the genes that interlink the genomic variations to these more transcriptional phenotypic variations are somehow critical to the signaling of the tumor, and if we reverse them, it will somehow have an effect.” The patient’s pathway profile may suggest specific treatment options.
“First, find the subtype of the cancer. Then, use the transcriptome as a readout. Look at the situation the tumor is in. Are there altered pathways? Look at the tissues of the organ, or the cell of origin. Are the genes altered? Has some acid in the body gone awry?” he said. His laboratory has developed a tool called PARADIGM to analyze all this data, using it to analyze genomic signatures shared across 12 different tumor types.12
The study suggested molecular data could help cluster patients more effectively, and hopefully point to effective therapy. The researchers also analyzed genomic and transcriptomic data to find key pathways in patients with castration-resistant prostate cancer to help develop more effective therapy protocols.13
Researchers now use techniques such as whole-genome sequencing models to identify genetic variants in patients for more accurate diagnosis. Some patients may have rare, monogenic diseases that aren’t well understood, so researchers such as Shamil R. Sunyaev, PhD, professor and Distinguished Chair of Computational Genomics at Harvard Medical School, look for genetic clues both in their extended family tree and in other, unrelated families. One focus of their study is Mendelian disease, caused by a single, inherited locus.
“There are three types of diseases that we look at: undiagnosed genetic disease, known genetic disease and known diseases where all the known genes have been ruled out,” said Dr. Sunyaev. His research team at Brigham Genomic Medicine Program accepts candidate cases to analyze, and has solved 17 so far.
The researchers also analyzed genomic & transcriptomic data to find key pathways in patients with castration-resistant prostate cancer.
“In many cases, we have only one affected individual patient’s DNA and the parental DNA. If we believe this is a monogenic situation, we ask, ‘Can we resolve this case?’” he said. “We make two assumptions. We assume the penetrance is complete, and we assume we must limit the analysis to protein coding mutations. If penetrance is complete, and the mode of inheritance is dominant, and the parents are unaffected, we assume the mutation is de novo.”
In dominant gene scenarios, one or two parents and the child have the variant, but there may be a cousin affected as well, so “this brings the number of genetic pathways down. Does this variant likely have an effect on gene function?” he said. Next steps include identifying other patients or families who may have the variant using typical disease characteristics or genotyping, developing an animal or cell model, or conducting genotype-informed clinical tests. A significant fraction of patients lacks a coding variant that explains the phenotype in rare diseases, and genetic drift may not explain rare events in large populations, he said.
“Why are some patients unique and nobody seems to see their phenotypes elsewhere? Frequently, the phenotype is not loss of function. Sometimes there’s an environmental trigger,” he said. Although some patients with autoimmune disease can have an extreme presentation that makes their disease seem rare or unique, they may have family members with milder forms of the same disease. GWAS data analysis can tell us more about the many genetic variants that can affect presentation in autoimmune diseases.
Genomics, Big Data & Therapy
GWAS data can also help refine pharmacologic therapy by clarifying how distinct patient subsets, based on their genetic or genomic profiles, might metabolize certain drugs differently, said Minoli Perera, PharmD, PhD, associate professor of pharmacology at Northwestern University Feinberg School of Medicine in Chicago, Ill. Pharmacogenetics (i.e., the genome’s role in drug response) is an increasingly important area of precision medicine, she said. Her research focuses on anti-thrombotic medications in African-American patients.
GWAS data on warfarin have shown three genes are associated with dose requirement, which is clinically important given the tight control and therapeutic monitoring required for this drug. But these large genomic studies were performed in mostly white or Asian patient populations, she said.
“African-Americans taking warfarin, on average, require a higher dose than other patients. Known variations explained less of the variability of drug response in these patients,” she said. African-American patients on warfarin also suffer much higher rates of bleeding than other ancestry groups, “so this suggests other population-specific genetic variants may explain the dose requirement in these patients.”
African-Americans typically have a greater genetic variation than other American populations due to their admixture of African and European ancestry, she said. “So using genotypes found in whites to guide therapy led to poor results in African-Americans. There was harm done to them if you used genotyping to determine their warfarin dose.”
Dr. Perera and her colleagues conducted the first GWAS in African-Americans to look at venous thromboembolism (VTE) risk, and they discovered a new biomarker for increased risk of the condition, thrombomodulin (THBD).14 Decreased genetic expression of THBD affects coagulation in African-American patients. They’re now studying how patients with African ancestry might metabolize drugs differently, which has widespread implications across all areas of medicine.
Challenges include developing practical ways to deliver genotyping data to clinicians, as well as access to testing and cost. “What tests will be covered by insurers? In the long run, will genotyping save money over traditional care?” she said. “We’re past the age of genotype-guided therapy. We have genomes, we have transcriptomes and we will soon have microbiomes.”
Big data, already a fixture of big business, is also now in use in large-scale disease research. California-based 23andMe conducts huge genotyping analyses to identify disease risks and potentially preventive interventions. 23andMe has nearly 3 million customers in its database now, according to Robert Gentleman, PhD, vice president of computational biology at the company. The company sells its genetic testing services directly to consumers, who send saliva samples through the mail for analysis.
23andMe conducts genotyping (but not sequencing) and has generated a panel of 23 million SNPs so far. These data reveal variability in the human genome that may drive development of new therapies for rheumatic diseases and others.
Customers join online to receive a saliva-based DNA kit in the mail. 23andMe analyzes the sample and sends the customer detailed data about their health risks. Customers may opt out anytime, and all the data are aggregated and de-identified for privacy.
23andMe says it has collected more than 800 million data points, and it divides its findings into disease-specific cohorts that may prove useful for rheumatology research, such as one cohort of nearly 50,000 psoriasis patients. Self-reported data aren’t always reliably accurate, he said.
Researchers use next-generation sequencing to develop more effective tests. A new blood test measures IgG4/IgG mRNA ratio in patients, which may show active disease in patients with granulomatosis with polyangiitis (GPA). Researchers in the Netherlands developed this IgG4 qPCR test to identify a key protein that correlates to high disease activity in GPA patients. GPA can lead to irreversible organ damage and has a 75% five-year survival rate, said Niek de Vries, MD, PhD, a researcher at Amsterdam Rheumatology and Immunology Center in the Netherlands.
A lack of sensitive, specific disease activity markers for GPA may lead to delayed treatment, undertreatment or overtreatment, said Dr. de Vries. Evidence points to a clear role for B cells, and results from this new blood test may help differentiate patients with active disease from those in remission.
Susan Bernstein is a freelance journalist based in Atlanta.
- Ermann J, Rao DA, Teslovich NC, et al. Immune cell profiling to guide therapeutic decisions in rheumatic diseases. Nat Rev Rheumatol. 2015 Sep;11(9):541–551.
- Zhou Q, Yang D, Ombrello AK, et al. Early-onset stroke and vasculopathy associated with mutations in ADA2. N Engl J Med. 2014 Mar 6;370(10):911–920.
- The International FMF Consortium. Ancient missense mutations in a new member of the RoRet gene family are likely to cause familial Mediterranean fever. Cell. 1997 Aug 22;90(4):797–807.
- Zhou Q, Wang H, Chae J, et al. A dominantly-inherited Behçet-like disorder caused by haploinsufficiency of the TNFAIP3/A20 protein. Pediatr Rheumatol Online J. 2015;13(Suppl 1): O71.
- Langefeld CD, Ainsworth HC, Cunninghame Graham DS, et al. Transancestral mapping and genetic load in systemic lupus erythematosus. Nat Commun. 2017 Jul 17;8:16021.
- Banchereau R, Hong S, Canterel B, et al. Personalized immunomonitoring uncovers molecular networks that stratify lupus patients. Cell. 2016 Apr 21;165(3):551–565.
- Munroe ME, Vista ES, Merrill JT, et al. Pathways of impending disease flare in African-American systemic lupus erythematosus patients. J Autoimmun. 2017 Mar;78:70–78.
- Hinks A, Bowes J, Cobb J, et al. Fine-mapping the MHC locus in juvenile idiopathic arthritis (JIA) reveals genetic heterogeneity corresponding to distinct adult inflammatory arthritic diseases. Ann Rheum Dis. 2017 Apr;76(4):765–772.
- Li G, Martínez-Bonet M, Wu D, et al. High throughput identification of non-coding functional SNPs via Type IIS enzyme restriction. Nature Genet. 2018 [in press].
- Li G, Cunin P, Wu D, et al. The rheumatoid arthritis risk variant CCR6DNP regulates CCR6 via PARP-1. PLoS Genet. 2016 Sep 14;12(9):e1006292.
- Nigrovic PA, Raychaudhuri S, Thompson SD. Review: Genetics and the classification of arthritis in adults and children. Arthritis Rheumatol. 2018 Jan;70(1):7–17.
- Hoadley KA, Yau C, Wolf DM, et al. Multiplatform analysis of 12 cancer types reveals molecular classification within and across tissues of origin. Cell. 2014 Aug 14;158(4):929–944.
- Drake JM, Paull EO, Graham NA, et al. Phosphoproteome integration reveals patient-specific networks in prostate cancer. Cell. 2016 Aug 11;166(4):1041–1054.
- Hernandez W, Gamazon ER, Smithberger E, et al. Novel genetic predictors of venous thromboembolism risk in African-Americans. Blood. 2016 Apr 14;127(15):1923–1929.