Although the human DNA sequence is 99.9% identical between individuals, the variations may greatly affect a person’s disease susceptibility. In elucidating the genetic basis of disease, much of what has been investigated in the pre-Human Genome Project era focused on identifying rare genetic variants (mutations) responsible for >1,500 monogenic disorders such as hypertrophic cardiomyopathy, long QT syndrome, sickle cell anemia, cystic fibrosis, or familial hypercholesterolemia, which are highly penetrant (carriers of the mutant gene will likely have the disease) and inherited in Mendelian fashion (hence termed Mendelian diseases). However, most of the genetic diversity in the population is attributable to more widespread DNA sequence variations (polymorphisms), typically single nucleotide base substitutions [single nucleotide polymorphisms (SNPs)], or to a broader category of previously overlooked structural genetic variants which include short sequence repeats (microsatellites), insertion/deletion (I/D) of one or more nucleotides (indels), inversions,
and the recently discovered copy number variants (CNVs, large segments of DNA that vary in number of copies)⁹ all of which may or may not be associated with a specific phenotype (Fig. 6-2). To be classified as a polymorphism, the DNA sequence alternatives (i.e., alleles) must exist with a frequency of at least 1% in the population. About 15 million SNPs are estimated to exist in the human genome, approximately once every 300 base pairs, located in genes as well as in the surrounding regions of the genome. Polymorphisms may directly alter the amino acid sequence and therefore potentially alter protein function or alter regulatory DNA sequences that modulate protein expression. Sets of nearby SNPs on a chromosome are inherited in blocks, referred to as haplotypes. As it will be shown later, haplotype analysis is a useful way of applying genotype information in disease gene discovery. On the other hand, CNVs involve approximately 12% of the human genome, often encompass genes (especially regulating inflammation and brain development), and may influence disease susceptibility through dosage imbalances. The year 2007 was marked by the realization that DNA differs from person to person much more than previously suspected; equipped with faster and cheaper DNA sequencing technologies, researchers have catalogued more than 3 million SNPs as part of the HapMap Project,¹⁰ published the first diploid genome sequence of an individual human,¹¹ launched the 1000 Genomes Project (sequencing the genomes of 1,000 people from around the world), and begun charting CNVs and other structural variants, thus making understanding of human genetic variation the 2007 Science magazine “Breakthrough of the Year.” Below, we review the common strategies used to incorporate genetic analysis into clinical studies.

Most ongoing research on complex disorders focuses on identifying genetic polymorphisms that enhance susceptibility to given conditions. Often the design of such studies is complicated by the presence of multiple risk factors, gene–environment interactions, and a lack of even rough estimates of the number of genes underlying such complex traits. Two broad strategies are being employed to identify complex trait loci. The candidate gene approach is motivated by what is known about the trait biologically and can be characterized as a hypothesis-testing approach, but is intrinsically biased. The second strategy is the genome-wide scan, in which thousands of markers uniformly distributed throughout the genome are used to locate regions that may harbor genes influencing the phenotypic variability. This is a hypothesis-free and unbiased approach, in the sense that no prior assumptions are being made about the biologic processes involved and no weight is given to known genes, thus allowing the detection of previously unknown trait loci. Both the candidate gene and the genome scan approaches can be implemented using one of two fundamental methods of identifying polymorphisms affecting common diseases: Linkage analysis or association studies in human populations.

Linkage Analysis. Used to identify the chromosomal location of gene variants related to a given disease by studying the distribution of disease alleles in affected individuals throughout a pedigree, linkage analysis has successfully mapped hundreds of genes for rare, monogenic disorders. However, common complex diseases are characterized by a multitude of genes with rare and/or common alleles, which create an apparently chaotic pattern of heterogeneity within and between families. The overall effect of this heterogeneity, together with the potentially weak influence of many loci, places a heavy burden on the statistical power needed to detect individual contributing genes and may be the reason why very few genome linkage scans so far have yielded disease loci that meet genome-wide significance criteria.¹² Furthermore, the nature of most complex diseases (especially for perioperative adverse events) precludes the study of extended multigenerational family pedigrees. Nevertheless, a few positive findings have emerged using this approach: A stroke susceptibility locus was mapped on chromosome 5q12,¹³ risk of myocardial infarction was mapped to a single region on chromosome 14,¹⁴ and a recent meta-analysis of several genome-wide scans for pulse pressure variation, an emerging risk factor for perioperative complications, has identified several linkage bins on chromosomes 22 and 10.¹⁵

Genetic Association Studies. Association studies examine the frequency of specific genetic polymorphisms in a population-based sample of unrelated diseased individuals and appropriately matched unaffected controls. The increased statistical power to uncover small clinical effects of multiple genes¹⁶ and the fact that they do not require family-based sample collections are the main advantages of this approach over linkage analysis. Until very recently, most significant results in dissecting common complex diseases were gathered from candidate gene association studies, with genes selected because of a priori hypotheses about their potential etiologic role in disease based on current understanding of the disease pathophysiology.¹⁷ For example, genetic variants
within the renin–angiotensin system,¹⁸ nitric oxide synthase (NOS),¹⁹ and β₂-adrenergic receptors (AR),²⁰ known to modulate vascular tone, were tested and found to be associated with hypertension. Similarly, the possible effects of polymorphisms on genetic predisposition for CAD²¹ or restenosis after angioplasty²² have been extensively investigated; more recently, two large-scale association studies have identified gene variants that might affect susceptibility to myocardial infarction.²³ As it will be presented in more detail later, accumulating evidence from candidate gene association studies also suggests that specific genotypes are associated with a variety of organ-specific perioperative adverse outcomes, including myocardial infarction,²⁴,²⁵ neurocognitive dysfunction,²⁶,²⁷,²⁸ renal compromise,²⁹,³⁰,³¹ vein graft restenosis,³²,³³ postoperative thrombosis,³⁴ vascular reactivity,³⁵ severe sepsis,³⁶,³⁷ transplant rejection,³⁸ and death [for reviews, see (5,8)].

One of the main weaknesses of the candidate gene association approach is that, unless the marker of interest “travels” (i.e., is in linkage disequilibrium) with a functional variant, or the marker allele is the actual functional variant, the power to detect and map complex trait loci will be reduced. Other known limitations of genetic association studies include potential false-positive findings resulting from population stratification (i.e., admixture of different ethnic or genetic backgrounds in the case and control groups) and multiple comparison issues when large numbers of candidate genes are being assessed.³⁹ Replication of findings across different populations or related phenotypes remains the most reliable method of validating a true relationship between genetic polymorphisms and disease,¹⁷ but poor reproducibility in subsequent studies has been one of the main criticisms of the candidate gene association approach.⁴⁰ However, a recent meta-analysis suggested that lack of statistical power may be the main contributor to this inconsistent replication and proposed more stringent statistical criteria to exclude false-positive results and the design of large collaborative association studies.⁴¹

At last, after several decades of frustrating limitations in the ability to find genetic variations responsible for common disease risk, with the completion of the second phase of the International HapMap Project (a high-resolution maps of human genetic variation and haplotypes)¹⁰ and advances in high-throughput genotyping technologies, the year 2007 marked an explosion of adequately powered and successfully replicated genome-wide association studies (GWAS) that identified very significant genetic contributors to risk for common polygenic diseases like CAD,⁴²,⁴³,⁴⁴ myocardial infarction,⁴⁵ diabetes (types I and II),⁴⁶,⁴⁷ atrial fibrillation,⁴⁸ obesity,⁴⁹ asthma, common cancers, rheumatoid arthritis, Crohn’s disease, and others. GWAS make use of the known linkage disequilibrium pattern between SNPs from the human HapMap and the new high-density SNP chip technology to comprehensively interrogate between 65% and 80% of common variation across the genome, with even higher coverage being possible using statistical imputation techniques. The largest, most comprehensive GWAS to date was conducted by the Wellcome Trust Case-Control Consortium, investigating the association between 500,000 SNPs and 7 common diseases in 2,000 cases and 3,000 shared controls, which identified 25 independent association signals at stringent levels of significance (P < 5 × 10⁻⁷).⁴² Interestingly, variants in or near CDKN2A/B (cyclin-dependent kinase inhibitor 2 A/B) conferred increased risk for both type II diabetes (odds ratio (OR) 1.2, P = 7.8 × 10⁻¹⁵) and myocardial infarction (OR 1.64, P = 1.2 × 10⁻²⁰), which may lead to a mechanistic explanation for the link between the two disorders. This finding also highlights the power of GWAS to identify variants outside described genes: While one of the signals occurs in the CDKN2A/B region, the other much stronger association signal occurs >200 kb from these genes, in a gene desert, and thus would not have been picked up by a candidate gene approach. Identifying the mechanism by which this variant may affect CDKN2A/B expression will provide new insights into the regulation of these important genes.

Genomic approaches are anchored in the “central dogma” of molecular biology, the concept of transcription of messenger RNA (mRNA) from a DNA template, followed by translation of RNA into protein (Fig. 6-3). Since transcription is a key regulatory step that may eventually signal many other cascades of events, the study of RNA levels in a cell or organ (i.e., quantifying gene expression) can improve the understanding of a wide variety of biologic systems. Furthermore, while the human genome contains only about 25,000 genes, functional variability at the protein level is far more diverse, resulting from extensive posttranscriptional, translational, and posttranslational modifications. It is believed that there are approximately 200,000 distinct proteins in humans, which are further modified posttranslationally by phosphorylation, glycosylation, SUMOylation, oxidation, and disulfide structures. The “static” view of constitutive DNA sequence variants potentially involved in the pathophysiology of perioperative complications can be complemented by a “dynamic” view that integrates their functionality. There is increasing evidence that variability in gene expression levels underlies complex disease and is determined by regulatory DNA polymorphisms affecting transcription, splicing,
and translation efficiency in a tissue- and stimulus-specific manner.⁵⁰ Thus, analysis of large-scale variability in the pattern of RNA, microRNA (miRNA), and protein expression both at baseline and in response to the multidimensional perioperative stimuli (dynamic genomics) using microarray and proteomic approaches provides a much needed understanding of the overall regulatory networks involved in the pathophysiology of adverse postoperative outcomes. This information is complementary to the assessment of genetic variability at the DNA sequence level using various genotyping techniques as described in previous sections (static genomics). Such dynamic genomic markers can be incorporated in genomic classifiers and used clinically to improve perioperative risk stratification or monitor postoperative recovery.⁵¹ This emergent concept of molecular classification involves the description of informational features in a training dataset using changes in relative RNA, miRNA, protein, or metabolite abundance in the context of genetic predisposition and applying to a test dataset to recognize a defined “fingerprint” characteristic of a particular perioperative phenotype (Table 6-2). For example, Feezor et al. used a combined genomic and proteomic approach to identify expression patterns of 138 genes from peripheral blood leukocytes and the concentrations of 7 circulating plasma proteins that discriminated patients who developed multiple organ dysfunction syndrome (MODS) after thoracoabdominal aortic aneurysm repair from those who did not. More importantly, these patterns of genome-wide gene expression and plasma protein concentration were observed before surgical trauma and visceral ischemia-reperfusion injury, suggesting that patients who developed MODS differed in either their genetic predisposition or their preexisting inflammatory state.⁵²

Alternatively, dynamic genomic markers can be used to improve mechanistic understanding of perioperative stress, and to evaluate and catalogue organ-specific responses to surgical stress and severe systemic stimuli such as cardiopulmonary bypass (CPB)
and endotoxemia, which can be subsequently used to identify and validate novel targets for organ-protective strategies.⁶⁹ Similarly, integrated transcriptomic and proteomic analyses have been used to characterize the peripheral blood molecular response signatures to cardiac surgery with and without CPB, a robust trigger of systemic inflammation.⁷⁰ The study demonstrated that, rather than being the primary source of serum cytokines, peripheral blood leukocytes only assume a “primed” phenotype upon contact with the extracorporeal circuit which facilitates their trapping and subsequent tissue-associated inflammatory response. Interestingly, many inflammatory mediators achieved similar systemic levels following off-pump surgery but with delayed kinetics, offering novel insights into the concepts of contact activation and compartmentalization of inflammatory responses to major surgery.

Several studies have profiled myocardial gene expression in the ischemic heart, demonstrating alterations in the expression of immediate-early genes (c-fos, junB), as well as genes coding for calcium-handling proteins (calsequestrin, phospholamban), extracellular matrix, and cytoskeletal proteins.⁵³ Upregulation of transcripts mechanistically involved in cytoprotection (heat shock proteins), resistance to apoptosis, and cell growth has been found in stunned myocardium.⁷¹ Moreover, cardiac gene expression profiling after CPB and cardioplegic arrest has identified the upregulation of inflammatory and transcription activators, apoptotic genes, and stress genes,⁵⁴ which appear to be age-related.⁵⁵ Microarray technology has also been utilized in the quest for novel cardioprotective genes, with the ultimate goal of designing strategies to activate these genes and prevent myocardial injury. Preconditioning is one of such well-studied models of cardioprotection, which can be induced by various triggers including intermittent ischemia, osmotic or redox stress, heat shock, toxins, and inhaled anesthetics. The main functional categories of genes identified as potentially involved in cardioprotective pathways include a host of transcription factors, heat shock proteins, antioxidant genes (heme-oxygenase, glutathione peroxidase), and growth factors, but different gene programs appear to be activated in ischemic versus anesthetic preconditioning, resulting in two distinct cardioprotective phenotypes.⁵⁶ More recently, a transcriptional response pattern consistent with late preconditioning has been reported in peripheral blood leukocytes following sevoflurane administration in healthy volunteers, characterized by reduced expression of L-selectin as well as downregulation of genes involved in fatty acid oxidation and the PCG1α (peroxisome-activated receptor γ coactivator 1α) pathway,⁵⁹ which mirrors changes observed in the myocardium from patients undergoing off-pump coronary artery bypass grafting (CABG) (OPCAB; Table 6-2).⁵⁸ Deregulation of these novel survival pathways thus appears to generalize across tissues, making them important targets for cardioprotection, but further studies are needed to correlate perioperative gene expression response patterns in end organs such as the myocardium to those in readily available potential surrogate tissues such as peripheral blood leukocytes.

The transcriptome (the complete collection of transcribed elements of the genome) is not fully representative of the proteome (the complete complement of proteins encoded by the genome), since many transcripts are not targeted for translation, as evidenced recently with the concept of gene silencing by RNA interference. Alternative splicing, a wide variety of posttranslational modifications, and protein–protein interactions responsible for biologic function would remain therefore undetected by gene expression profiling (Fig. 6-3). This has led to the emergence of a new field, proteomics, studying the sequence, modification, and function of many proteins in a biologic system at a given time. Rather than focusing on “static” DNA, proteomic studies examine dynamic protein products, with the goal of identifying proteins that undergo changes in abundance, modification, or localization in response to a particular disease state, trauma, stress, or therapeutic intervention [for a review, see (72)]. Thus, proteomics offers a more global and integrated view of biology, complementing other functional genomic approaches. Currently available methods for proteomic analysis include protein extraction, separation by two-dimensional gel electrophoresis or chromatography, followed by identification using mass spectrometry. Although rapidly improving, these methods are currently limited by sensitivity, specificity, and throughput. Several preclinical proteomic studies relevant to perioperative medicine have characterized the temporal changes in brain protein expression in response to various inhaled anesthetics⁷³,⁷⁴ or following cardiac surgery with hypothermic circulatory arrest.⁷⁵ This may focus further studies aimed to identify new anesthetic binding sites and the development of neuroprotective strategies. Furthermore, detailed knowledge of the plasma proteome has profound implications in perioperative transfusion medicine,⁷⁶ in particular related to peptide and protein changes that occur during storage of blood products. The development of protein arrays and real-time proteomic analysis technologies has the potential to allow the use of these versatile and rigorous high-throughput methods for clinical applications and is the object of intense investigation.

As part of the preoperative evaluation, anesthesiologists are involved in assessing the risks of perioperative complications. It is commonly accepted that patients who have underlying cardiovascular disease are at risk for adverse cardiac events after surgery, and several multifactorial risk indices have been developed and validated for patients undergoing noncardiac surgical procedures (such as the Goldman or the Lee Cardiac risk index), as well as cardiac surgery (such as the Hannan or Sergeant scores). However, identifying patients at the highest risk of perioperative infarction remains difficult, and risk scores, while potentially valuable for population studies, are not an ideal tool for directing care in an individual patient.¹³⁰ Genomic approaches have been used in the search for a better assessment of the individual coronary risk profile. Numerous reports from animal models, linkage analysis, family, twin, and population association studies have definitely proven the role of genetic influences in the incidence and progression of CAD, with a heritability of death from CAD as high as 58%; furthermore, hazardous patterns of angiographic CAD (left main and proximal disease), known major risk factors for perioperative cardiac complications, are also highly heritable. Similarly, genetic susceptibility to myocardial infarction (MI) has been established through multiple lines of evidence¹⁴,²³ including a recent well-powered and replicated GWAS.⁴⁵ While these studies do not directly address the heritability of adverse perioperative myocardial events, they do suggest a strong genetic contribution to the risk of adverse cardiovascular outcomes in general.

Despite advances in surgical, cardioprotective, and anesthetic techniques, the incidence of perioperative myocardial infarction (PMI) following cardiac and vascular surgery in several large randomized clinical trials has been reported at 7% to 19%¹³¹,¹³² and is consistently associated with reduced short- and long-term survival. In the setting of cardiac surgery, PMI involves three major converging pathophysiologic processes, including systemic and local inflammation, “vulnerable” blood, and neuroendocrine stress.⁵ In noncardiac surgery, pathophysiology of PMI is not so clearly understood, but a combination of two mechanisms appears predominant: (1) Plaque rupture and coronary thrombosis triggered by perioperative endothelial injury from catecholamine surges and pro-inflammatory and pro-thrombotic states; and (2) prolonged stress-induced ischemia and tachycardia in the setting of compromised perfusion. Extensive genetic variability exists in each of these mechanistic pathways, which may combine to modulate the magnitude of myocardial injury. However, only a paucity of studies exist relating genetic risk factors to adverse perioperative myocardial outcomes, mainly following CABG surgery (Table 6-3).³²,⁸⁵,⁸⁹