MASSIVELY PARALLEL SEQUENCING
Massively parallel sequencing (MPS) is an alias of the probably more popular term next-generation sequencing (NGS). In this article the term next-generation sequencing will be avoided for the simple reason that “next” is a relative term in matter of time. If Sanger’s and Maxam and Gilbert’s ground breaking inventions of DNA sequencing in 1974 are counted as the first generation, then the automation of Sanger’s method could be considered the next or second generation. Further, the next step which in 2005 lead to the development of machines which were able to sequence millions of fragments of DNA simultaneously would certainly have to be called the next-next-generation or third generation[1]. And the next generation of technological improvement is on its way: the 4th generation of sequencing methodology will utilize entire strands of DNA without the need of fragmentation, and will become cheaper, and more precise, and simpler to handle bioinformatically. In order to avoid counting generations of technological advancement, the term MPS, or MPS for short, seems more applicable and will be used herein.
A variety of technical approaches to MPS exist, all of them have been reviewed in detail (for an excellent overview please see, e.g.[2]. Briefly, the general principle relies on (1) the fragmentation of DNA/RNA, optionally followed by fragment size selection; (2) amplification of the fragments; and (3) sequencing of the fragments. Currently, the length of those sequences can be up to 800 bp, depending on the vendor. As the fragmentation step generates random breakpoints in the DNA backbone, so will the sequenced fragments in Step 3 be at a random position of the DNA or RNA. This is where the individual small pieces of sequence information will have to be bioinformatically stitched together, “assembled”, being a challenge to which there are plenty of approaches with slightly different quality, depending on the analysis pipeline of choice.
The applications of MPS are overwhelming and offer never seen before opportunities to study genomes, exomes, transcriptomes, and chromosomal rearrangements and secondary modifications like methylation of DNA and alkylation of RNA. Through its unbiased template-free approach, it is now also possible to sequence DNA and RNA of novel species in de novo assembly analyses and thus accelerate discovery of, e.g., ontological relationships[3], and even discover novel RNA species[4]. Input amounts in the low ng-range for some MPS applications make it possible to study biological samples in a detail which could not have been envisioned before[5]. MPS has found its way to the analysis of single eukaryotic cells or even cell-free DNA in blood samples, e.g. for non-invasive prenatal diagnosis[6].
PERSONALIZED TREATMENT
Following transplantation, drug treatment must be carefully adjusted to prevent rejection. Drug metabolism is influenced by a large variety of factors such as age, gender, disease, dose, drug-drug interaction, and metabolic competence. Differences in the genotype (polymorphisms) can be linked with altered drug metabolism in transplant patients. Tacrolimus for example is primarily metabolized via the CYP450 enzymes. Non-expressors of CYP3A5 metabolize the drug slower than others, hence requiring lower doses than normal expressors[7]. Similarly, in conjunction with age as variant, polymorphisms in the transporter ABCB1 can determine the bioavailability of cyclosporine and mycophenolate mofetil[8]. Another example which illustrates the importance of studying genetic polymorphisms to optimize personal treatment is the occurrence of hypertension after transplantation. In genome-wide association studies polymorphisms have been identified in a number of genes affecting hypertension (e.g.[9]). For overviews of the field please see the recent reviews of D’Alessandro et al[10] and of Kurzawski et al[11]. As these examples and other studies, which cannot be discussed here for space limitation, show, the individual landscapes of polymorphisms in patients need to be assessed to optimize treatment efficacy. Sequencing of genes with standard methods is time-consuming and can deliver ambiguous. MPS technology can be used to study exomes of patients through targeted sequencing of candidate genes and determine polymorphisms which may affect treatment. However, MPS does not always deliver unambiguous results either due to sequence coverage differences and DNA sequence specifics such as guanine-cytosine (GC) content or homopolymers which cannot always be resolved by current MPS technologies alone. At times, one may need to verify the results by alternative technologies to obtain further sequence information.
HUMAN LEUKOCYTE ANTIGEN MATCHING
Alleles of the human leukocyte antigen (HLA) genes are commonly used for organ and bone marrow matching prior to transplantation. Humans vary widely in the composition of antigens arising from alleles of those six HLA genes (A, B, C, DR, DQ, DP). Detection of foreign HLA antigens by the host can lead to strong antibody mediated reactions, thus they can be considered important mediators of immune response. During the graft selection process, it is therefore essential to detect donor-host HLA mismatches, a process commonly performed by Sanger sequencing of the HLA locus. While Sanger sequencing certainly has its merits, technical limitations such as relatively high sequence inaccuracy resulting in sequence ambiguity due to highly polymorphic DNA regions, and limited sequence coverage in a single experiment (only a small number of exons is sequenced systematically, and some important polymorphisms may be located outside the sequenced regions) may make another round of experimental verification necessary in many cases. With ever decreasing costs, MPS has the potential to deliver high-quality sequence data which cover a large proportion of the entire HLA locus[12,13].
IMMUNE SYSTEM
MPS can be applied to many aspects of biological research in the transplantation arena. Exon arrays and RNA sequencing was applied to address the question whether alternative splicing takes place during immune response post-transplant. The group of Grigoryev et al[14] purified human CD2(+) T or CD19(+) B cells, activated them to model early post-transplant immune events and continued to sample from those cell pools over time. Indeed they were able to show that these two cell populations not only regulate gene expression following in vitro stimuli, but also regulate exon usage to generate alternative panels of transcripts which may contribute to the biological pattern of immune response. MPS now permits devising experiments which aim at studying the methylation status of DNA of T cells and B cells before, during and after immune response, e.g., graft rejection. Methylation of promoter regions plays an important role in gene regulation[15]. Changes of the methylation status of genes during immune response during and after treatment hence may give clues about how the expression of genes is regulated, for example in combination with DNA-protein motif discovery. For excellent overviews of MPS methods for the investigation of epigenetic modifications of DNA and RNA, please see[16,17].
METAGENOMICS
16S rRNA pyrosequencing is a variety of MPS used to selectively sequence the highly variable 16S rRNA regions of bacterial genomes, thus providing qualitative and quantitative genus and species information of bacteria present in a sample[18]. The group of Diaz et al[19] used 16S rRNA pyrosequencing to study the bacteriome of the human oral cavity after transplantation. They demonstrated a shift in the composition of the microbiome of the oral cavity during immunosuppression following transplantation[19]. The authors speculate that immunosuppression may create an environment in the oral cavity which could be more permissive for opportunistic pathogens.
A number of groups have focused on characterizing the microbiome of alveolar fluid in relation to lung transplantation. For instance, Borewizc et al[20] have applied 16S rRNA pyrosequencing to study the human lung microbiome after lung transplantation. The authors compiled sequencing data from 12 bronchoalveolar lavage fluid samples from four patients over three time points, two additional samples from healthy, non-transplanted individuals served as controls. Interestingly they found that the microbial diversity increased after transplantation, and that the dominating phyla after transplantation were different from those in healthy lungs. The authors suggest to follow those results under the aspect of the bronchiolitis obliterans syndrome, which is a marker of chronic lung transplant rejection[21].
DIAGNOSIS
In 2012, Wen et al[22] demonstrated that the number of circulating endothelial cells (CECs) increased in whole blood of renal transplant patients undergoing acute rejection, acute tubular necrosis, and chronic allograft nephropathy, when compared to control samples. CEC count decreased after immunosuppressive therapy. The authors attributed the increased CEC count to injury of vessel endothelium in conjunction with endarteritis, and conclude that monitoring CEC numbers can be used as minimally invasive tool to diagnose or prognose poor short-term outcome of renal allografts. Technically, it is not farfetched to design scenarios in which MPS technologies could be applied to monitor the number of CECs in whole blood samples. Whole genome sequencing would not be necessary; an exon-capture set specific for exons of endothelial genes would suffice for qualitative measurement of CECs. On the quantitative side read numbers would have to be normalized against a set of stably expressed genes, identification of which can be challenging, as seen in the microarray arena.
Certainly, similar to other conventional approaches such as microarrays, MPS can be used to develop biomarkers of rejection or tolerance. Despite striving to identify the best matching grafts for hosts, the best matches are not always tolerated. The reason for tolerance, or lack thereof, may be found not within coding region of the HLA, but possibly in surrounding (introns, promoters) or even distant genomic areas. With ever decreasing costs of MPS it will soon be possible to sequence not only exons or exomes in a larger scale than possible or affordable today, but entire genomes. As is the case in other research disciplines it will be necessary to gather genomic sequence information from a sufficient number of individuals to draw significant conclusions. This is the case for mutation analysis [e.g., single nucleotide polymorphisms (SNP)], as well as the analysis of secondary modifications such as methylation when certain biological conditions are compared. Research will see a steady growth of available sequence information which will contribute to discovery and qualification of biomarkers and elucidation of biological processes for the benefit of patients.
BIOINFORMATICS CHALLENGES
There are now many MPS-approaches to sequencing DNA, which will continue to reduce speed and cost of sequencing. When in 2000, still in the pre-MPS era, the drafts of a human genome sequence were published, one would not have thought that already 13 years on, the cost for this undertaking would come down from around $3B to around $5-10K, and the sequencing and analysis time would shrink from 10 years “for a rough working draft” to around 3-4 wk on the average for a complete version. However, decreased sequence raw data generation time and costs mean huge challenges for IT in terms of data storage and transfer of the huge raw data files which can be in the TB range per run, and for bioinformatics data analysis capacities, including quality control, alignment, assembly, annotation, and statistical analysis. No longer is the data generation process the experimental bottleneck, but the analytical side of things. In fact, as Sboner et al[23] phrase it, there is an “unpredictable amount of extra ‘human’ time” which is required for the identification of the best analysis pipelines, software installation, etc. Like in the early days of microarrays experts argue about the approaches to data processing. This leads to an amount of approaches which can be even overwhelming for bioinformaticians themselves (if they would admit it): What is the most precise, fastest, aligner, assembler, normalization method, algorithm to identify SNPs, statistics for differentially expressed genes, differentially methylated sites, etc.? Some methods are listed in[2]. Evaluating which analysis pipeline suits best to which problem and to which IT environment is challenging and time consuming. The final step, the interpretation of the results, is yet another “unknown” time factor which can rarely be done automatically, but requires human intervention. In the end one needs to understand that sequencing cannot in every case provide an immediate answer to all scientific questions. Just like in all other comparative experiments which we have become familiar with over the years, the first step in experiments involving MPS is sampling. Sampling means that individuals are selected which represent the entire group of individuals we are interested in, a process which can be attempted in a variety of statistical approaches of experimental study design, such as randomization, blocking, and randomization[24]. Many sequencing applications do not omit the need for biological replicates, a cost-factor which needs to be considered in the planning phase. Certainly this is true for transcriptome analysis, differential methylation analysis, but also for genome-wide association studies (GWAS). The latter will benefit dramatically from the increased precision and availability of whole genome information in the near future, contributing to the growing number of lead mutations in diseases (for an overview of GWAS studies, http://www.genome.gov/gwastudies). MPS will allow the discovery of rare variants where commonly used SNP arrays will have to fail. Certainly, there are settings where one sample will suffice. These are occasions in which individual information about a genome is investigated, e.g., in cancer-genomics or in rare diseases. This brings our discussion to the aspect of personalized medicine and MPS.
DATA MANAGEMENT AND PRIVACY
Decreasing costs and increasing availability of resources will make MPS a tool for medical research and clinical care. However, routine genome sequencing for patient care brings along important socio-ethical and legal ramifications which are heavily discussed. Crucial concerns arise around patient information to obtain informed consent, data protection and patient privacy protection, data ownership, third-party use, use of incidental findings, and how such (incidental) findings are disclosed to the patient, to name a few[25-27]. On the other hand, sequencing data can be used for a whole range of scientific and clinical applications, becoming accessible via databases across nations. Sequence data can be used e.g. for trait analysis, phylogenetic testing, and expression analysis, bringing along a wide range of possible findings which is difficult to estimate at the time of sampling. Hence, to obtain informed consent from a patient the extent of consent has to be fairly thorough, which may cause frustration and possibly unwillingness to consent, additionally posing risks of study bias due to social background. McGuire et al[28] proposed a tiered consent process with three levels, from intended release of data information on multiple gene loci, to single gene loci, to releasing no data. Sample donors would have to be educated about the risks and benefits of the foreseen use of their data. Data access would have to be restricted according to the intended use at the beginning of the study. Reconsideration of study purposes may enforce re-consenting.
If genomic information is released though, is it possible to fully protect the privacy of sequencing data? Already in the pre-MPS era of 2004, Malin and his team showed that it was possible to link genomic data to named individuals in publicly available records by leveraging unique features in patient-location visit patterns[29]. With the growth of genome sequence databases it should be possible to identify individuals based on their DNA sequence (e.g. SNP pattern), provided a template is present. In 2004, Lin et al[30] published it was possible to de-identify a person by interrogating just 75 SNPs, not many when taking into consideration that SNP databases of human genomes contain hundreds of thousands per genome. Not only the patient’s but also the relatives’ privacy is affected, but may be affected. This has large implications not only on research, but even more importantly on health care systems and national databases. The goal of the Health Insurance Portability and Accountability Act of 1996 is to protect genomic data as personal health information (http://www.hhs.gov/ocr/privacy).
The extent of result disclosure poses another issue. How much does a patient need to learn about the results, especially incidental findings, which were not part of the original study. What are results and who is interpreting them? As Sharp pointed out in a detailed discussion[31], the amount of data and potential findings with all their false positives and negatives, is equally overwhelming for the practitioner as it will be for the study participant. Many mutations may be harmless, and a result-interpretation may again be interpreted as a result by a study participant[31].
These are only a few critical concerns that have to be addressed urgently. The scientific community needs to ensure that the legal and ethical framework which makes social discrimination based on genetic information impossible is appropriate for the developing technology. International databases and cloud computing impose the necessity of international legislation which puts the patient rights first. By ensuring privacy protection, study participation has a chance to be beneficial for the individual, not a potential risk for social exclusion.
OUTLOOK
Over the next years prices per sequenced nucleotide will continue to fall, sequencing machines will become smaller, cheaper and easier to use, eventually making genomic sequencing a standard tool in research and clinics. Despite growing databases, MPS data interpretation will remain a challenge. The legal and ethical frameworks for using MPS data need to be defined on an international level, granting respect to sample-providing individuals as well as the research goals of scientists and clinicians. International consortia need to address the possibility that the current speed of genome research may outrun the pace of legal regulation, and impose adjustments.
P- Reviewers: Kelesidis T, Schuurman HJ S- Editor: Cui XM L- Editor: A E- Editor: Yan JL
