Pharmacogenetics is considered as a prime example of how personalized medicine nowadays can be put into practice. However, genotyping to guide pharmacological treatment is relatively uncommon in the routine clinical practice. Several reasons can be found why the application of pharmacogenetics is less than initially anticipated, which include the contradictory results obtained for certain variants and the lack of guidelines for clinical implementation. However, more reproducible results are being generated, and efforts have been made to establish working groups focussing on evidence-based clinical guidelines. For another pharmacogenetic hurdle, the speed by which a pharmacogenetic profile for a certain drug can be obtained in an individual patient, there has been a revolution in molecular genetics through the introduction of next generation sequencing (NGS), making it possible to sequence a large number of genes up to the complete genome in a single reaction. Besides the enthusiasm due to the tremendous increase of our sequencing capacities, several considerations need to be made regarding quality and interpretation of the sequence data as well as ethical aspects of this technology. This paper will focus on the different NGS applications that may be useful for pharmacogenomics in children and the challenges that they bring on.
Pharmacogenetics refers to the influence of DNA variants on drug response, the knowledge of which can facilitate selection of the optimal drug, dose, and treatment duration and avert adverse drug reactions [
The current golden standard for detecting pathogenic variants—single nucleotide variations or small indels—is Sanger sequencing [
The introduction of next generation sequencing (NGS) brought about a technological revolution among genetic screening tools, as it now becomes possible to screen the whole exome—the coding regions of our DNA—and even the complete genome in a single experiment [
Because of this success, these screening techniques are slowly starting to make their way as a diagnostic tool. Certainly for complex diseases for which several genes have been identified—the sequence analysis of which is laborious, time consuming, and expensive—the idea of sequencing all 23.000 genes in the exome in a single reaction is an alluring alternative. Similarly, in a field such as pharmacogenetics, with different variants in different genes influencing the final drug response in an individual patient, such parallel sequencing techniques can provide the promptness which would be required in a clinical setting. This shift to the more extensive screening assays has induced an evolution from pharmacogenetics to pharmacogenomics [
In this paper, we will consider the characteristics of NGS, the different means by which NGS technology can be applied, and set out a concept that we think would be feasible to use NGS-based pharmacogenetics in a present-day clinical pediatric setting.
The human genome is the entirety of an individual’s hereditary information, including both the coding and noncoding regions of DNA and RNA, while the human exome encompasses the coding regions of the genes—the exons—equivalenting ~1%-2% of the total haploid genomic sequence [
Genomic variation in an individual in numbers. All data is given for an individual exome or genome. The number of nonsynonymous variants, which induce a change in amino acid, has been specified as these are more likely to have a functional effect than synonymous variants (where the amino acid remains the same despite the nucleotide change), although functional synonymous variants have been described.
Number of variants | |
---|---|
Similarity between two individual genomes | 99.5% |
Whole genome sequencing variant uptake | 3,5 million SNP variants |
Whole exome sequencing variant uptake | 20.000–100.000 variants |
Coding variants in the genome | 20.000–25.000 variants |
Nonsynonymous coding variants in the genome | 9.000–11.000 variants |
Following the Human Genome Project, which set out to sequence the three billion nucleotides of the human genome, several high throughput technologies were developed. Among these, NGS has known a rapid evolution in a few years time, increasing throughput and reducing costs by continuous improvement of several analysis platforms [
Schematic representation of the principles of massive parallel sequencing using beads. Double-stranded DNA (a) is fragmented into single-stranded DNA (b) which is subsequently coupled, via adaptors, to agarose beads by oligonucleotides complementary to the adaptor (c). These beads are submerged in an emulsion where amplification of the single DNA fragment occurs (d). Subsequently, the bead is placed into a well (e), already equipped with all reagents for sequencing (small beads). Within these wells, parallel sequencing of the different DNA fragments occurs, resulting in the generation of the genetic code of the fragment (f).
General limitations of massive parallel sequencing include error rate, which is higher than Sanger sequencing, warranting confirmation of identified causal variants by conventional sequencing methods [
NGS can be used in a targeted manner or can be applied as a whole exome of whole genome diagnostic tool. Every one of these approaches has its advantages and weaknesses which will be discussed in the following, with respect to pharmacogenomics in children.
In a targeted assay, NGS is used for parallel sequencing of a selection of genes. There are two ways to go about selecting the genes of interest, which determines the molecular technique that will be applied: either a microarray-based target enrichment approach or a targeted analysis of whole exome/genome sequencing (WES and WGS, resp.) can be used. In the first, direct hybridization of the patient’s DNA to an oligonucleotide array, containing probes complementary to the selection of target genes, is performed and then analyzed by NGS, thus generating sequence data only of the genes of interest [
The major advantage of the array-based target enrichment assays is that these selective tests can be optimized to have full coverage of the genes of interest, hence reaching high sensitivity and specificity. Moreover, for each gene it is possible to design the array to just cover the sequence (exon, intron or promotor) in which a particular SNP is present. Further, by generating sequence data only of the genes of your interest, the risk for incidental findings or ethical issues on generated but unanalyzed sequence data—both discussed in the following—is minimized. A good pharmacogenomic example of such an analysis would be coumarine treatment. The drug response to warfarin—a paradigm for drugs with a narrow therapeutic index—is determined by variants in several genes, including VKORC1, GGCX, CYP2C9 and, CYP4F2 [
Vitamin K cycle and warfarin metabolism. Inactive zymogens, among which are the VK-dependent coagulation factors II, VII, IX, and X, are activated by gamma-carboxylation by the gamma-glutamyl carboxylase (GGCX). The cofactor for this carboxylation step is VK, which is transformed into VK epoxide. This epoxide is then reduced by the VK-epoxide reductase of VKORC1 to quinone. Warfarin specifically blocks the initial reduction step, while CYP4F2 catalyzes the formation of hydroxyvitamin K out of quinone.
An important observation is that the functional variants in, for example, the VKORC1 gene reside mostly in noncoding regions such as the introns and promotor of the gene [
Nevertheless, it must be remembered that drugs such as warfarin, where most of the variance in metabolism and clearance can be captured by analyzing a handful of genomic variants, represent only a proportion of drugs with a narrow therapeutic index; for many other drugs, this will not be the case, confronting us with the main limitation of this type of targeted analysis which is the limited flexibility in design. It can be expected that for many drugs, novel genes and variants will be discovered which will have a pharmacogenomic effect in addition to the ones known to date. Though this number may be rather small for a very specific topic such as warfarin biology, the pharmacogenetics of ADHD or asthma treatment—for both of which variants in non-coding regions were described (Table
Identified pharmacogenes in ADHD and asthma.
Gene | Variants | |
---|---|---|
Asthma | ADRB2 | p.Arg16Gly, p.Gln27Glu |
AC9 | p.Ile772Met | |
CRHR1 | c.122-1310C > A (intronic) | |
TBX21 | p.His33Gln | |
LTC4S | g.24030224A > C (promotor) | |
CYSLTR1 | p.Phe309Phe | |
ALOX5 | 5′ UTR | |
GSDML | c.236-1199G > A (intronic) | |
| ||
ADHD | DRD4 | 48 bp allele |
DAT1 | 3′ UTR | |
5-HTT | del-1212-1255 (promotor) | |
SNAP-25 | 3′ UTR | |
COMT | p.Val158Met | |
ADRA2A | g.31585029G > C |
ADRB2: agonists of beta-2 adrenergic receptor; AC9: adenylyl cyclase type 9; CRHR1: corticotropin releasing hormone receptor 1; TBX21: T-box 21; LTC4S: leukotriene C4 synthase; CYSLTR1: cysteinyl leukotriene receptor 1; ALOX5: arachidonate 5-lipoxygenase; GSDML: gasdermin B; DRD4: dopamine receptor 4; DAT1: dopamine transporter 1; 5-HTT: serotonin transporter; SNAP-25: synaptosomal-associated protein; COMT: catechol-O-methyltransferase; ADRA2A: adrenergic alpha2-receptor; UTR: untranslated region.
This implies that with every newly identified gene, the assay needs to be adjusted or updated to obtain the highest yield of useful information. Though novel generation arrays have already become more “user friendly” to expand the number of targeted genes, it still does not come near the ease by which additional genes can be analyzed in a prospective way using WES or WGS. When applying targeted exome analysis, whole exome sequencing is performed resulting in sequencing data of 23.000 genes. Subsequently, only those genes of interest are filtered out to analyse variants. When a novel gene is identified, it is easy to go back to the initial sequence data, access the sequence of the new gene, and analyze variants. However, as mentioned, the major limitation of this approach is the lack of good sequence data of non-coding regions, making this technique only useful to obtain a pharmacogenetic profile for those drugs of which the relevant variants are in coding regions. One such example is the treatment of acute lymphoblastic leukemia (ALL) in children [
Identified pharmacogenes in childhood ALL. All are affecting the coding regions of the respective genes.
Gene | Variants |
---|---|
TPMT | p.Ala80Pro, p.Ala54Tyr, p.Tyr240Cys |
GSTT1 | Large deletion |
GSTM1 | Large deletion |
GSTP1 | p.Ile105Val |
MTHFR | p.Ala222Val, p.Glu429Ala |
GGH | p.Thr151Ile |
TMPT: thiopurine methyltransferase; GST: glutathione-S-transferase family; MTHFR: methylenetetrahydrofolate reductase; GGH: gamma-glutamyl hydrolase.
Since the completion of the Human Genome Project in 2001, sequencing of the complete personal genome has become a technical reality [
The current feasibility of pharmacogenetic implementation of WGS data has been shown in published personal genomes such as the Lupski or the Venter genome [
One of the issues that has risen recently is whether the characterization of the genomic sequence of an individual should become the standard of care. This issue has great relevance to the pediatric population, and several arguments can be conceived why this data should or should not be available as early on as possible, ideally in the neonatal period.
There can be two rationales to gain knowledge of this genomic information: the first would be to improve preventive medicine by identifying causal mutations or risk alleles associated with so-called actionable diseases, that is, diseases for which preventive measurements of screening have been shown useful for improvement of prognosis. After sequencing, the whole genome data set is completely analyzed, with the intrinsic risk to also unveil risk alleles or mutations for nonactionable disorders such as, for example, neurodegenerative diseases. The second rationale would be to have rapid access to the genetic background of an individual if there is an acute disease episode, for diagnostic and pharmacogenetic purposes. This would imply that the uninterpreted data is stored, and—when, for example, the patient develops symptoms of asthma—the sequences of those genes known to be involved in drug response of beta-agonists can be quickly assembled and searched for variants that may guide treatment options.
Though both scenarios can be seen as the ultimate refinement of personalized medicine, there are several practical, ethical, and legal considerations that need to be made before this can be implemented, most of which also apply for WES.
Besides technical issues related to storage and access of the data and data analysis (method and quality of analysis as well as the bioinformatic—hard- and software—capacities), the main practical issue remains the interpretation of the sequence data (Figure
Workflow for pharmacogenomics using WES or WGS. After mapping to the reference sequence and variant calling, variants need to be filtered. Besides the known functional variants and the normal variations, it can be expected that a considerable number of novel variants in genes involved in drug metabolism will be found. For these, additional functional assays will have to be performed to confirm whether they have functional relevance or not.
Several ethical considerations need to be made prior to routine clinical implementation, in adults but particularly also in children. The huge amount of personal medical data produced by NGS, the fact that some will be irrelevant, that some may be relevant for diseases beyond the primary reason for the test and that some may be difficult to interpret and hence unclear, make that all ethical issues raised before on genetic testing now come together in a single test [
A first issue is the consent. Because of its specific nature, certainly WES and WGS require a different kind of consent compared to the routine genetic tests. As mentioned previously, a potential benefit of these screening technologies for the patient may be the early detection of actionable diseases, the symptoms of which can occur in childhood. Besides the obvious advantages for followup and prognosis, it must be taken into account that being confronted unexpectedly with the knowledge that the child may develop one or more diseases can bring about significant psychological burden for the child and the parents. While this will be so for variants associated with high risk to develop disease, -suddenly, an additional medical track needs to be established for this novel health problem-, the psychological effect may be even more pronounced when a variant is discovered with mediocre or low penetrance and hence increased uncertainty about the future of the child. Therefore, it seems imperative that prior to WES or WGS for a given diagnostic question, the patient and/or his parents are informed about other diseases for which the test can reveal information and what the implications can be of each of these. The consent that is given will need to not only stipulate rigorously which diseases are being (indirectly) looked at, but also mention the possibility of variants of unknown significance, of which it remains uncertain what the effect may be. Needless to say that this will require a much more extensive pretest counseling compared to the molecular testing that is routinely used to date as well as posttest (psychological) followup.
To respect the right of the child not to know, it has been a standard policy not to perform presymptomatic tests in children, when the onset of the disease is in adulthood [
All issues mentioned previously underline that a thorough debate addressing the medical, ethical, and psychological aspects of WES and WGS with respect to the child, the parents and the rest of the family is necessary prior to the diagnostic implementation of these techniques in, for example, pharmacogenomics. Such debate has begun to occur within the genetic community, and international consensus and guidelines will need to be drafted regarding late-onset disorders, carriership of recessive diseases, and actionable or non-actionable childhood diseases. However, because of the extensive impact these screening strategies can have, a more global public debate is necessary to inform the public about these novel possibilities and their challenges and to think about what people really want to learn from such a genetic test when sufficiently informed.
A second issue focusses on the interpretation of sequence data. Should patients be informed of variants of uncertain significance? The main argument not to provide such details is that it does not give additional information to the patient at the time of consultation. However, as knowledge increases, more may become known about these variants; this could imply that what was once a variant of unknown significance may turn out to be of immediate relevance to the patients health or that of his family members. Knowledge of these variants, even when their meaning is initially unclear, may be useful in the followup. In this respect, the question whether the physician or diagnostic facility has the duty to recontact patients when the interpretation of their sequence data changes over the years has not been answered. If this were to be the case, a fully automated informatics system would be needed to regularly screen the stored genomic data of every patient and match all variants with the current literature. To our knowledge, such large-scale systems which are flawless are not yet available, making the duty to recontact for these large data sets nearly impossible at this time.
Legal issues that can arise around WES/WGS include the storage and access of genomic sequence data and the question of who can gain access: the individual himself, his treating physician(s), insurance companies, police, and so forth [
The technical revolution in sequencing analysis tools has lead to new perspectives for personalized medicine in general and in pharmacogenetics/genomics specifically. Next generation sequencing and its applications have increased our ability to unravel the genetic code of an individual with significant improvement of the speed of the analysis. On the other hand, the implementation of these assays brings about several considerations regarding sensitivity, data analysis, and interpretation as well as ethical aspects. Of the current NGS technologies, the array-based approach seems to be the most feasible one for pharmacogenomic applications in childhood. Its targeted nature avoids incidental findings while offering sufficient coverage of coding and noncoding regions of genes of interest. With little doubt, the future perspective will be the application of WGS as a diagnostic tool, also in pharmacogenomics. However, many questions need to be addressed before implementing this screening technique in the clinic, including technical challenges, interpretation difficulties, and ethical considerations. Most importantly, the implementation of NGS requires the establishment of genotypes with clinical utility and guidelines on how to use them. Though the topic of research in many areas, more effort will have to go to validating genotypic data and developing clinical algorithms using them.