Implications of the Human Genome for Understanding Human Biology and Medicine

The capacity to sequence the entire genomes of free-living organisms, and to analyze such genomes in their entirety, has considerable implications for understanding human biology and medicine.^{1 - 2} Our genomic sequence provides a unique record of who we are and how we evolved as a species.³ The knowledge fostered by understanding the genome might clarify which human characteristics are innate and which are acquired, as well as the interplay between heredity and environment in defining susceptibility to illness. Such an understanding will make it possible to study how our genomic DNA varies among cohorts of patients, and especially the role of such variation in the causation of important illnesses and responses to pharmaceuticals.^{4 - 6} We may also be able to use new approaches to investigating complex aspects of the human condition, such as language, thought, self-awareness, and higher-order consciousness. The study of the genome (genomics) and the associated protein content (proteomics) of free-living organisms will eventually make it possible to localize and understand the function of every human gene, as well as the regulatory elements that control the timing, organ-site specificity, extent of gene expression, protein levels, and posttranslational modifications that define health or illness. For any given physiological process, we will have a new paradigm for addressing its evolution, development, function, and mechanism in causing disease and in affecting the onset and outcome of disease.⁷

One noteworthy finding is the relatively low number of genes in the human genome.^{1 - 2} A gene, in this context, is defined as a locus of cotranscribed exons, which ultimately result in the production of a peptide or protein. There are a number of computational tools used to identify, enumerate, and compare genes within a species and between species. These computational methods integrate gene prediction models with different types of experimental and computational evidence to impart a stringency requirement to this process, and have been applied by ourselves and others.^{8 - 10} The text and subtext of biology prior to the availability of the sequence for the human genome was that the number of genes in an organism would in some fashion reflect its "complexity." There were expectations that the human genome would contain 100 000 genes or more.^{11 - 13}

The genomic sequences of 4 multicellular eukaryote genomes have been published over the past 3 years. The approximate gene count for the fruit fly (Drosophila melanogaster) is 14 000 genes¹⁰ ; for the roundworm (Caenorhabditis elegans), 19 000 genes⁸ ; and for the mustard plant (Arabidopsis thaliana), 26 000 genes.⁹ A comparison of gene numbers among the genomes of different species is in many ways more important than the total number of genes found in a genome of any single species. Those who might be tempted to use the number of genes to explain human complexity might then pause to consider that, by this measure a human being, with approximately 30 000 genes,^{1 - 2} is roughly a fly plus a worm or the equivalent of a plant. This assessment of the number of human genes is based on results from analyses in which there are stringency requirements used in conjunction with the computational algorithms. Thus, absolute count may be less important than comparisons with published genomes. Similarly, those who were expecting very large numbers of newly discovered genes as targets for pharmaceutical interventions might need to reassess their expectations. The number of genes independently reported by both groups^{1 - 2} is far fewer than the expectations based on prior experimental analysis of expressed sequence tags (ESTs) or computational analysis, which estimated that humans would have 70 000 to 120 000 genes.^{12 - 14} The genomic sequence and the gene complement predicted by both groups may be accessed on the Web (at http://public.celera.com/cds/login.cfm [for noncommercial purposes only¹ ]; and at http://www.ensembl.org/ [data generated by the International Human Genome Sequencing Consortium² ]). The sequencing of the human genome suggests that we must look beyond gene number per se (at least protein-coding genes) as we attempt to understand human complexity, future targets for pharmaceutical research, and implications for medical practice (Table 1).^{1 - 2 ,4 - 5 ,7 ,15 - 46}

Surveying the landscape of the human genome leads to several other observations. Only about 1% of the genome is spanned by exons (regions that code for proteins), while just under 25% is contained within introns (regions between exons within genes that are spliced out in the creation of messenger RNA and do not code proteins), and about 75% of the genome is contained in intergenic DNA.^{1 - 2} Thus, genes often exist in nonrandom clusters or gene-rich "oases," separated by what appear to be large "deserts" of several hundreds of thousands of nucleotide codes that do not appear to encode genes. There is no simple explanation for why natural selection has taken this path in the evolution of the human genome, but we believe it is premature to conclude that such "deserts" lack biological or medical importance.

The human genome is filled with blocks or "elements" of repetitive nucleotide codes whose function is still a mystery. It has been known for many years, and amply confirmed with the sequencing of the genome, that human DNA contains large and complex families of such repeat elements.^{1 - 2} These include the long interspersed repetitive elements (LINEs) and short interspersed repetitive elements (SINEs), which include Alu sequences that arose with the evolution of primates, including humans.⁴⁷ These sequences represent a distinct class of retrotransposon-amplified repeat DNA. During primate evolution, these DNA elements could be replicated and transposed to new sites in the genome.^{21 ,47} They comprise approximately 10% of the human genome.^{1 - 2} Their biological function and role in natural selection has remained an enigma.

Yet in surveying the landscape of the human genome, a striking and nonrandom distribution of Alu sequences is evident. They appear to preferentially colocate within gene-rich regions of the genome.^{1 - 2} One inference is that the biological role of these Alu sequences, the effects of nucleotide variations within such elements,²¹ and their ability to mediate recombination events^{17 - 18} will be important in understanding their regulatory effects^{19 - 21} on gene function and disease. Further investigations are required to add to the known examples where Alu sequence variations have been shown to affect biology and clinical conditions.^{17 - 21}

Such elements had previously been characterized as "selfish" DNA (ie, DNA whose existence seems related to replication purposes only),⁴⁸ having no direct impact on medicine or natural selection. The availability of the human genome sequence suggests that this view should be revised since it appears possible that such repeat elements may indeed contribute to the causation of human diseases.

The human genome reveals a remarkable level of duplication.^{1 - 2} Although the biological impact of duplication in generating gene superfamilies is well established, the first comprehensive view of the genome-wide landscape has revealed the widespread impact of 2 distinct mechanisms of duplication.

These 2 forms of duplication are very different: one form mediated at the DNA level (segmental duplication), and another mediated at the RNA level (retrotransposition). Both mechanisms produce paralogs—a term for genes that make their appearance in more than 1 copy in the genome (albeit with possible modifications).

Among humans, the extent of the segmental duplications is 10- to 100-fold greater than that observed in the fly and worm genomes. There are more than 3500 genes in over 1000 genomic blocks ranging in size up to chromosomal lengths, that have shown a duplication with linear preservation of order on another chromosome.¹ This is illustrated by an examination of chromosomes 18 and 20 in Figure 1 and a more global representation of this phenomenon in Figure 2.^{1 ,49}

This process might be analogized to a kind of internal genomic colonization. The clinical relevance of such events is emphasized by our finding of paralogous, disease-causing genes, of validated shared ancestry, on both duplicated segments. A disease-causing gene is defined as a gene in which sequence variants are linked to the causation of a disease. Notable among these are genes involved in hemostasis, complement fixation, transcriptional regulation (such as the homeobox proteins associated with developmental disorders), metabolic disorders, and voltage-gated ion channels associated with cardiovascular conduction abnormalities.¹ In many cases, there is a disease-causing gene with a paralog on the duplicated segment, whose linkage to a disease is not currently recognized.¹ It is possible that an understanding of segmental duplication will provide new insights into the pathogenesis of disease. To be sure, every duplication event will not lead to a paralog that results in the same pathophysiologic consequences. However, it might well be possible to use genomics to demonstrate a unity for disparate diseases. It may also be possible to understand and explain adverse reactions and side effects of drugs through their previously unknown collateral activities against paralogs.

The other remarkable finding is the extent of gene duplication that has resulted from retrovirus-based transposition of gene transcripts.¹ The ancestors of humans encountered retroviruses capable of transcribing RNA to DNA (reverse transcription). Indeed, such viruses are not extinct, as diseases such as acquired immunodeficiency syndrome (AIDS) amply confirm.⁵⁰ The human genome carries the results of many such encounters. Gene duplication by this process in effect creates paralogs that lack introns and often occur in multiple copies scattered randomly throughout the genome. The medical implications of this form of gene duplication are similar to those that apply to segmental duplication. In addition, the degree of identity between the source gene and the retrotransposed gene is often very high, thus leading to the possibility of confounding in DNA- or protein-based diagnostic tests. It is important to note that changes in coding or noncoding regulatory regions in these paralogs, leading to different functions or expression patterns, may be one way of providing an increased functional repertoire in the human genome.

Earlier, we mentioned that the number of protein-coding genes was considered to be low relative to expectations prior to the sequencing of the genome.^{12 - 13 ,51} Does an analysis of the full set of proteins (ie, the proteome) help us resolve the issue of human beings not appearing to carry many more genes than a fruit fly, a roundworm, or a plant? Indeed, we do note that the average human gene makes more proteins, and more complex proteins, than its invertebrate counterparts. A number of such features are worth detailing. These include the evolution of new protein domains (well-defined regions on a protein that show structural and functional conservation), duplications or expansions of domains (domain accretion), as well as greater combinatorial diversity (domain shuffling) in human beings (Figure 3).^{1 - 2 ,52 - 53} In addition, certain genes produce more than 1 type of protein, using alternative transcriptional start sites and RNA splicing. Finally, posttranslational modifications, wherein the translated protein is subjected to a wide range of biochemical modifications, may potentially give rise to a significantly larger set of functional proteins than would be predicted by the gene count. Table 1 provides a summary of the medical relevance of these features.

Extensive protein domain shuffling is observed in the human proteome, and this would serve to increase or alter combinatorial diversity to provide an exponential increase in protein-protein interactions. Moreover, certain special genes show patterns for generating combinatorial diversity at the protein level. For example, immunoglobulins and the T-cell receptors show clonal DNA shuffling or rearrangements to increase the immune repertoire, while the cadherins show exon transsplicing (a form of RNA shuffling that mixes and matches exons to create diversity in the final messenger RNA) to generate increased extracellular interactions.^{54 - 55} All of these factors taken together contribute to a complexity not captured by examining gene number alone.

Many proteins (and protein domains) found in humans evolved early in the animal lineage and hence have orthologs (evolutionary counterparts) in invertebrate genomes. However, several noteworthy vertebrate-specific domains exist, especially within proteins involved in developmental, homeostatic, and nuclear regulation. These proteins have profound implications in understanding human development, malignant transformation, and stem-cell biology. In addition, proteins related to acquired immunity, complement fixation, and hemostasis are either unique or show a considerable expansion in the human genome compared to known invertebrate genomes (Figure 4).^{1 ,52 - 53} Thus, we find several instances where evolution has harnessed "old" domains to provide novel distinct domain architectures in the human when compared to the fly or worm; that is, "new" proteins created using old domains (Figure 3B). Examples include the serine proteases, which occur with a widely diverse set of protein domains in the plasma proteases (coagulation, complement, and fibrinolytic systems), and the recruitment of the immunoglobulin fold into molecules of the acquired immune system, eg, antibodies, major histocompatibility complex, and cell adhesion receptors.^{1 - 2} Also, in concordance with the greatly increased neuronal complexity in the human compared to the fly and worm, there is an increase in the number of members of protein families involved in neural development, structure, and function (Figure 4).^{1 ,56 - 57} These include neuronal growth regulators, as well as classes of voltage-gated ion channels that play a vital role in neuronal network formation and in electrical coupling. Understanding how these components interact to generate the neuronal infrastructure in humans will have an impact on therapeutic modalities to address neuronal injury, as well as provide insights into new ways to diagnose and treat neuropsychiatric disorders. Proteins involved in apoptosis (programmed cell death), a central effector mechanism that regulates cellular physiology, are also greatly expanded in humans.^{1 ,58} The central role for this process in neurodegenerative diseases,⁵⁹ malignancy, and inflammatory conditions⁶⁰ related to extrinsic mediators (eg, pathogens) and intrinsic mediators (eg, cardiovascular disease, inflammatory bowel disease) constitute areas of intense current investigation. Therapeutic interventions that can modulate the apoptotic process will likely have major effects on some of the most devastating clinical illnesses that afflict humankind.⁶¹

However, a focus on the genomic DNA sequence alone will not be sufficient to resolve all the important problems of medicine and biology. The availability of the human genome sequence will substantially enhance the power of proteomics.^{6 ,44} In the near future, once the sequence of any unknown protein is determined in any human fluid or cell culture (eg, by a technology involving mass spectroscopy for separation and identification of proteins), there will now virtually always be a "hit," or "match," between proteins and their genes.⁴⁴ The applicability of this approach to better understand disease processes in humans, as well as to facilitate drug discovery and development,^{43 ,62 - 63} will undoubtedly increase as genomes of additional model organisms (eg, the mouse, rat, and dog) become available. Such approaches will also enhance the capacity for detecting novel microbes and their protein complements, either pathogens or commensals, both of which have profound implications for enhancing microbial diagnosis and developing improved antimicrobial therapeutics.⁶⁴ It will also be possible to link proteins and their posttranslational modifications to the pathophysiology of illnesses.^{43 - 44} Many of these modifications likely affect the activity and disposition of proteins in health and disease. One special form of posttranslational modification involves protein cleavage, which is essential to the activity of certain proteins (of which insulin and other hormones are classic examples), including those involved in the apoptotic process. Ultimately, the number, complexity, and modifications of proteins encoded by human genes all contribute to the complexity of human biology, and underscore that not all answers lie at the level of genomic information per se. Advances in proteomics^{44 ,62 - 63 ,65 - 67} will thus likely enhance the next generation of diagnostics as well as guide therapeutics in ways that were previously impossible or exceedingly difficult (Table 2 and Table 3).^{1 - 2 ,4 - 6 ,22 ,24 - 25 ,31 ,43 - 44 ,62 - 63 ,65 - 84 85 - 137}

The study of the genome supports the fundamental unity of human beings. We all share at least 99.9% of the nucleotide code in our genome.^{1 ,138} And yet it is remarkable that the diversity of human beings at the genetic level is encoded by less than 0.1% variation in our DNA. In any physician's practice, patients are predisposed to different conditions, respond to the environment in variable ways, metabolize pharmaceuticals differently,^{4 ,7} vary regarding dose-response relationships for common drugs, and have a range of susceptibilities to adverse effects of therapeutic agents (even when there is no obvious difference in individual pharmacokinetics or biochemical pharmacology).^{4 ,7 ,24 ,70}

The most common form of DNA variation in the human genome is the single-nucleotide polymorphism (SNP).^{69 - 70} Put simply, an SNP is the substitution of one purine or pyrimidine base for another at a given location in a strand of DNA. Generally, SNPs are biallelic (only 2 choices exist at a given site within a population). The nomenclature defining a mutation (a change in DNA that may affect phenotype⁷ [http://www.nhgri.nih.gov/DIR/VIP/Glossary/pub_glossary.cgi]) can be somewhat arbitrary and relative. By convention, when a substitution is present in more than 1% of a given target population and causes no discernibly abnormal phenotype, it is called a variant or polymorphism.^{7 ,69 - 70} Single-nucleotide polymorphisms can affect gene function, or they can be neutral. Neutrality is sometimes inferred if an SNP does not alter protein coding (ie, a change in an exon that encodes a different amino acid). In practice, this inference can be wrong. It is also worth noting that an SNP may be subtly responsible for an abnormal phenotype, but only in the context of a given environment (or the simultaneous presence of SNPs in other locations), without which an abnormal phenotype is not expressed.^{69 - 70}

We now have a genome-wide survey of several million variants, with precise nucleotide localization, in an ethnogeographically divergent group of individuals.^{1 ,138} In comparing chromosomes from any 2 randomly selected individuals, we know there is an average of 1 variation for every 1250 nucleotides. These variations can occur within exons, with synonymous (no change in amino acid) or nonsynonymous (a change in amino acid) alterations in code, or they can occur outside exons within intronic or intergenic regions of the genome. Less than 1% of all known SNPs encode a direct amino acid change of the ultimate protein product of a gene.¹ Therefore, there are only thousands (not millions) of genetic variations that directly contribute to the structural protein diversity of human beings.¹ We, and others, are currently performing large-scale resequencing and genotyping to define the frequency of these variations in various populations.

While such changes are certainly important to medicine, this finding implies that future medical research will need to also focus on the contributions of polymorphisms in noncoding regions or intergenic regions of the genome, something that was previously difficult or impossible to do. Thus, SNPs in proximity to various regulatory regions,²⁵ some of which exist at a great distance from the regulated gene in either 5′ or 3′ directions, are likely to be important. By the same token, SNPs in introns may have an unexpected role in the causation of human disease.^{26 ,139} Finally, SNPs in genes whose final product is an RNA may also be of unexpected importance.¹⁴⁰

An understanding of the human genome and its DNA variation will allow a rapid expansion of the medical applications of pharmacogenetics.^{4 ,24} There is a number of clear examples where DNA variation, primarily, but not exclusively, in the form of SNPs has implications for clinical research and medical practice (Table 1, Table 2, and Table 3).^{4 ,7 ,24 - 25 ,76 - 105}

These include polymorphisms that influence the clinical course or response to therapy. Thus, angiotensin-II type-1 receptor polymorphisms can have an impact on the severity of congestive heart failure as well as the response to angiotensin-converting enzyme inhibitors^{141 - 142} ; β-adrenergic receptor variants may alter airway hyperreactivity and response to β-agonists administered through metered-dose inhalers¹⁴³ ; and the apolipoprotein E4 allele affects onset of disease and the differential response to anticholinergic agents in patients with Alzheimer disease.¹⁴⁴ Also, the bioavailability of drugs is affected by polymorphisms in genes that code for proteins regulating drug metabolism and disposition (eg, MDR1, a drug efflux pump which regulates digoxin levels,¹⁴⁵ CYP2C19, which regulates omeprazole metabolism,¹⁴⁶ and CYP2C9 or 2C19, which regulate tolbutamide and phenytoin metabolism⁷⁸ ). These have clinical consequences. The availability of genome-wide data on DNA variation is thus likely to expand progress in prevention, diagnosis, and treatment customized to the needs of a specific patient, rather than to a statistical average. In addition, SNPs provide a new tool for familial linkage and population-based association studies to speed the identification of genes as targets for new diagnostics and therapeutics.^{4 ,24 ,69 - 70} In this context, it will soon be possible to integrate information on DNA variations in human populations with an understanding of entire networks of genes. Again, this would have been difficult or impossible prior to the sequencing of the entire genome. Since most common human diseases culminate from long-standing interactions between many genes and environmental factors (including lifestyle), predicting the contributions of genes in complex disorders will remain a challenge for medicine for many years to come.

The modest number of human genes means that we must explore mechanisms that generate the complexities inherent in human development and the sophisticated signaling systems that maintain homeostasis. There is a large number of ways that the functions of individual genes and gene products are regulated. An overview of these mechanisms and their relevance to disease and therapeutic intervention is discussed briefly and enumerated in Table 1.^{1 - 2 ,4 - 5 ,7 ,15 - 46}

The key point is that certain observations at the clinical level provide unique opportunities to understand how the genome functions as an integrated system. Thus, the study of mendelian disorders has led to unique insights regarding the functions of more than 1000 genes.⁴⁹ However, many common disorders, including cancer, asthma, type 2 diabetes mellitus, cardiovascular abnormalities, and neuropsychiatric illness, cannot be generally explained on the basis of variation in a single gene—that is, they are polygenic in origin.⁶⁸ Other illnesses are manifestations of (1) the process of creating triplet repeats^{22 - 23} (eg, Huntington disease, spinocerebellar ataxia, fragile X syndrome); (2) abnormalities of certain epigenetic phenomena, such as gene imprinting^{31 ,35} (eg, Prader-Willi syndrome³³ and Beckman-Wiedemann syndrome³⁴ ); (3) abnormalities of mitochondrial genes¹⁴⁷ (eg, MELAS [myelopathy, encephalopathy, lactic acidosis, and stroke-like episodes] syndrome, Kearns-Sayre syndrome); and (4) somatic mutation or mosaicism^{148 - 149} (eg, McCune-Albright syndrome, paroxysmal nocturnal hemoglobinuria, cancer). In addition, there is growing evidence that conditions such as Prader-Willi syndrome are caused by a variant in genes whose product is an RNA molecule, not a protein per se (Table 1).^{29 ,140} Thus, understanding the physiological roles of noncoding RNA and its modifications may contribute to understanding of the causation of specific diseases.^{37 ,140 ,150 - 151}

While the use of genomic sequence data to identify genetic determinants of disease has already shown significant progress (Table 2),^{22 ,31 ,68 - 72} the availability of the genomic sequence, and the development of high-throughput experimental and computational technologies, heralds a new era in our understanding of disease processes (Table 3). The use of computational homology–based approaches to identify new drug targets and to predict their structures in facilitating rational drug design is revolutionizing the drug development process.^{73 - 74} Efforts to incorporate genomic approaches into various stages of the drug development process for the development of novel and improved therapeutic agents, as well as for optimizing patient stratification and following clinical outcomes in treatment trials, are currently under way (Table 3).^{5 ,24 ,65 ,67 ,112 ,116} The practical benefits in the clinical practice of medicine will become increasingly apparent when there is a complete integration of genomic information with the phenotypes of clinical disease (Table 2 and Table 3).

We are in the midst of a major paradigm shift in biology and medicine¹¹⁹ ; the process of studying genes in isolation has now shifted toward exploring networks of genes involved in cellular processes¹⁵² and disease, identifying molecular "portraits" of disease based on tissue or organ involvement, and ultimately defining the biochemical readouts that are specific to clinical conditions. The era of "personalized" medicine will evolve as a parallel process, in which DNA variations recorded in human populations will be integrated into the above paradigm, to guide a new generation of diagnostic, prognostic, and therapeutic modalities designed to improve patient care (Table 3).¹⁵³

Basic science researchers have the task of deciphering the biological meaning of the 2.9 billion nucleotide codes that comprise the human genome. Physicians may be called on to interpret the scientific implications for their patients. However, physicians may also be called on to address complex historical and societal issues, which are either induced or revived by the sequencing of the human genome, in their everyday practice.

One fundamental issue is the extent to which knowledge of the genomic DNA sequence allows prediction of the essence of who we are, including the determination of risk for illness in various settings. There are some who may view the genome in a deterministic way, believing that the human condition will ultimately be seen entirely as a manifestation of sequence information and computation. We do not subscribe to such a view. Nevertheless, an individual's DNA is, in a sense, the ultimate personal identifier; thus, some patients may fear that the advent of new genomic technologies will affect their livelihoods and standing in the community. It is ironic that approximately 1 week prior to the publications of the human genome, an agency of the US government went to court for the first time to block a private employer from compelling its employees to submit to genetic testing in work-related injury inquiries, threatening dismissal for noncompliance.^{154 - 155} Thus, physicians and other health care providers are likely to have an interest in state and federal legislation protecting patient privacy and prohibiting discrimination on the basis of genetic testing. Patients who have suffered such discrimination in the past or fear it in the future are perhaps unlikely to view the scientific achievement of sequencing the human genome as an entirely positive accomplishment. We believe legislation to protect genetic privacy and prevent discrimination is essential to progress in genomics research.

There is also a complex social and political history related to human genetics. At various times in the past, many societies, including our own, adopted theories of race and genetics as the justification for political oppression against vulnerable groups.^{156 - 157} James D. Watson, the founding director of the Human Genome Project at the National Institutes of Health, provides an important perspective on these issues.¹⁵⁸ It is possible that medicine, even today, is affected by subtle and unrecognized biases. Thus, it has been argued that the medical community may wish to mark the milestone of the recent sequencing of the human genome as a time to discuss how such biases influence medical education, clinical research, and medical practice.¹⁵⁹ In such a discussion, we would offer that an analysis of the genome^{1 - 2} reveals a fundamental unity for all human beings. Our task now is to use the tools of modern genomics to prevent, diagnose, and treat illnesses, and, at the same time, to try to ensure that the benefits of genomics research extend fairly to all members of society.