*HQHVDQG1RQ0HQGHOLDQ'LVHDVHV 'HDOLQJZLWK&RPSOH[LW\ %HUWUDQG-RUGDQ
Perspectives in Biology and Medicine, Volume 57, Number 1, Winter 2014, pp. 118-131 (Article) 3XEOLVKHGE\-RKQV+RSNLQV8QLYHUVLW\3UHVV DOI: 10.1353/pbm.2014.0002
For additional information about this article http://muse.jhu.edu/journals/pbm/summary/v057/57.1.jordan.html
Access provided by RMIT University Library (16 Aug 2015 09:11 GMT)
Genes and Non-Mendelian Diseases dealing with complexity
Bertrand Jordan
ABSTRACT The first decades of the new medical genetics (1980 to 2000) were
marked by resounding successes, with the identification of the genes responsible (when defective) for muscular dystrophy, cystic fibrosis, or Huntington’s disease, to name just a few of the several thousand Mendelian genetic conditions whose causes are now known. In contrast, the search for genes involved in common disorders such as diabetes, hypertension, schizophrenia, or autism failed miserably in the 1990s, with inconsistent and conflicting results—although the strong genetic component of these diseases (that also involve environmental factors) was (and still is) beyond doubt. From 2000 on, thanks to huge progress in genomic knowledge, technology, and analytical methods, it became possible to reliably identify genes influencing the risk of complex conditions, using the so-called GWAS (Genome-Wide Association Study) approach. Yet many problems remain, such as the vexing question of the “missing heritability,” or the difficulty of translating these scientific results into genetic tests with real clinical validity and utility. Autism is one of the cases in which a strong genetic component has been demonstrated, but where the search for causative genes remains difficult and attempts at developing valid genetic tests have failed, because of the many genes involved and possibly of the heterogeneity of the condition.
Faculté de Médecine de Marseille, UMR 7268 ADES, Aix-Marseille Université/EFS/CNRS, Espace éthique méditerranéen, Hôpital d’adultes La Timone, 264, rue Saint-Pierre, 13385 Marseille Cedex 05, France. E-mail: [email protected].
Perspectives in Biology and Medicine, volume 57, number 1 (winter 2014): 118–131. © 2014 by Johns Hopkins University Press
118
Genes and Non-Mendelian Diseases
A
lmost every human disease has both a genetic and an environmental compo-
nent. Even a classical inherited condition such as hemophilia can be influenced by external factors—in fact, most of the pathogenic effects of the mutation can be avoided by judicious injections of clotting factor, leading to a nearly normal life expectancy. For infectious diseases, often considered as essentially environmental, there are well-documented inherited differences in susceptibility, one of the most striking being the resistance to HIV infection of homozygous carriers of the delta-32 mutation in the CCR5 gene (Dragic et al. 1996). Nevertheless, a number of conditions are “mostly genetic,” others “mostly environmental,” while many have significant components from both sides, as for example diabetes. Another important distinction is that between “monogenic” diseases, where the cause is a mutation in a single gene, the very same in all patients (although the mutations themselves may vary, as well as the severity of the complaint), and “complex” or “multigenic” disorders, in which particular alleles of a number of genes contribute to the genetic component of the disease: type 1 and type 2 diabetes, for example, are multigenic, while hemophilia A, as well as cystic fibrosis or Huntington’s disease, are all monogenic or “Mendelian” conditions. Thus, in the real world, a given disease is usually influenced to some extent by both heredity and the environment, and its genetic component most frequently involves more than one gene. Yet the prevailing image of genetic disorders is that they are due to a single “bad gene” and that the environment matters little:“DNA is destiny,” in other words.This largely stems from the fact that most of the successes of the new medical genetics, in the 1980s and 1990s, have indeed concerned conditions in which the genetic factor was predominant and involved a single gene.This is not surprising, as these are the least difficult cases and were therefore the first to be elucidated. However, these examples have durably shaped our expectations and are responsible for many current misunderstandings. In this article I will briefly recall the principles and results of work on monogenic diseases, then discuss complex disorders, both in general and for the particular case of autism, a particularly difficult and contentious case.
Monogenic Diseases: A Triumph of the New Medical Genetics Monogenic or Mendelian hereditary diseases are those inherited conditions whose transmission in family pedigrees clearly follows Mendelian rules (hence the term), either autosomal dominant (such as Huntigton’s chorea), autosomal recessive (such as cystic fibrosis), or X-linked (such as hemophilia A and B) (Figure 1). In a few cases, the gene involved had been discovered before the recombinant DNA era, using what was the classical approach at the time—in other words, by painstakingly identifying the defective protein and then, also painstakingly, isolating the corresponding gene. At the beginning of the 1980s, it was shown that a new approach—initially called
winter 2014 • volume 57, number 1
119
Bertrand Jordan
Figure 1 Classical Mendelian inheritance schemes. Circles are females; squares males; filled symbols are affected individuals; half-filled ones are carriers. Top left: autosomal dominant, a single copy of the defective gene causes the disease (as in Huntington’s disease); top right: autosomal recessive, the condition is manifest only if two defective genes are present in the same individual (as in cystic fibrosis); bottom: X-linked recessive, the condition is not manifest in females because their second X chromosome carries a functional gene (as in hemophilia A and B).
“reverse genetics,” then, more appropriately, “positional cloning”—could allow identification of “disease genes” without prior knowledge of the protein involved or of the etiology of the condition (Botstein et al. 1980).1 This opened the way for a revolution in human genetics and led to the identification of dozens, then hundreds, of genes responsible (when defective) for as many inherited diseases. The crucial point was the realization that a disease gene could be “mapped” (its approximate location on our chromosomes determined) by following, in suitable families, the association of the disease state with a number of “polymorphic markers” covering our whole genome.2 Once such a genetic association was found, the gene was said to be “mapped”: it was expected to lie within a relatively small region, one or two chromosomal bands, corresponding to a few megabases of DNA. The 1 “Disease genes” are those that cause the disease if they are inactive or otherwise defective, but that are present in active form in normal individuals 2 Polymorphic markers are DNA features that are variable in the population, such as restriction fragment length polymorphisms (RFLPs) or single nucleotide polymorphisms (SNPs).
120
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 2 “Reverse genetics” (later called “positional cloning”) in the 1990s. Top: Mapping or “localization” of the disease gene by following polymorphic markers (represented as restriction fragment length polymorphisms) in affected families, resulting in the designation of a region on the long arm of chromosome 5 (5q12 to 5q14, typically 10 to 20 megabases). Middle: Analysis of the region thus designated and isolation of “candidate genes,” denoted as “transcribed sequences” (TS). Such a region may contain dozens of genes. Bottom: Studies on these genes by sequencing or other techniques until a particular one, transcribed sequence 15 (TS15), is found to be systematically defective in patients, which designates it as the “disease gene.” Consequences (short- or long-term) once this is achieved are indicated at the bottom.
accuracy of this “localization” depended on the number of polymorphic markers available and the size of the families used for the association study. Once this was achieved, it was possible to move to the second step, a detailed molecular analysis of the region designated by the genetic study, to find out which genes lie in the region and to test them one-by-one in patients and controls. When, finally, a gene was found that was systematically defective in patients and functional in controls, this
winter 2014 • volume 57, number 1
121
Bertrand Jordan
was almost certainly the gene involved in the disorder. This process is illustrated in Figure 2, together with some of the tools used at the time for each of the two steps. Originally proposed by Botstein and coworkers in 1980, this procedure quickly led to the mapping of the genes involved in Huntington’s disease (1980), retinoblastoma (1984), and cystic fibrosis (1985). Positional cloning was still quite cumbersome, and the actual identification of the genes often took a number of years after the initial mapping. For example, it took four years to find the genes for cystic fibrosis, and in the case of Huntington’s disease, the gene was mapped in 1980 but only identified in 1993. Nevertheless, positional cloning represented a huge improvement over previous methods. Over the following years, the tools used in both steps became much more powerful. The first genetic maps, based on just a few hundred cumbersome polymorphic markers, were replaced by vastly superior maps incorporating thousands of highperformance markers, notably with the detailed map developed at Genethon (France) in the 1990s by Jean Weissenbach’s team (Dib et al. 1996). Progress in cloning and sequencing technology facilitated the establishment of extensive gene catalogs and the detailed characterization of genes. Accordingly, the rate of progress picked up significantly in the mid-1990s, as shown in Figure 3. Meanwhile the Human Genome Project (HGP), aiming ultimately at the complete sequence of the 3 billion nucleotides of our DNA—a task many believed to be impossible at the beginning—got underway and provided increasingly sophisticated maps and tools for human medical genetics. Thus, by the year 2000, more than a thousand genes involved in human genetic diseases had been conclusively identified, corresponding to almost all Mendelian conditions for which more than a handful of patients exist worldwide. This huge success has not so far resulted in equivalent progress in the treatment of these conditions. Of course, identification of the gene involved in a disease gives access to the sequence of the corresponding protein, and it is a great help in understanding the pathological mechanism. But, for example, although the defective “huntingtin” was identified in 1993, we still do not fully understand how it causes the neuro-degeneration observed in Huntington’s disease. The great expectations of gene therapy for conditions such as hemophilia, cystic fibrosis, or Duchenne muscular dystrophy have not materialized, although the method has scored (limited) success in recent years (Wirth, Parker, and Ylä-Herttuala 2013); at the same time, more conventional, drug-based approaches have not been accelerated as much as was hoped by the identification of the protein targets involved in the corresponding disorders. Understanding, in detail, the biology of a disease takes much time and effort and is a necessary step before the development of novel therapies can be attempted. Molecular genetics has, however, provided efficient diagnostic tools that make it possible to quickly characterize patients and apply the best treatment available; it has also enabled prenatal (and in some cases preimplantation) diagnostics that allow families in which a genetic disease is present to safely foster healthy children, certainly no small achievement. 122
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 3 Progress in gene identification. Number of “disease genes” identified every year from 1981 to 2000. The current total (2013) is over 5,000. Source: http://omim.org/statistics/geneMap.
The public perception of human genetics has been strongly influenced by events in this period, by the resounding successes of medical genetics and also by the media coverage of the ongoing HGP, which gave an oversimplified view (“The gene for disease X”) of the very complex genotype-to-phenotype relationship, as well as an overly optimistic outlook on the short-term medical results. For example, a major French charity campaigned, in 1993, with the motto: “Genes for healing” (Des gènes pour guérir), suggesting strongly that once the genes were found, appropriate treatments would almost immediately be developed. These rosy prospects have not materialized, and this has led to some disappointment with medical research among the general public.
Complex Diseases Are Very Complex Encouraged by the resounding successes scored with Mendelian diseases, many groups attempted to replicate this achievement with conditions that are much more frequent and are known to have a significant hereditary component. Some obvious examples are diabetes, hypertension, and Crohn’s disease, as well as psychiatric syndromes such as schizophrenia or bipolar disorder. While these conditions are clearly influenced by the environment, the personal history, and the lifestyle of individual, they also have a genetic element that is evidenced by the familial clustering of cases and by the high concordance between identical twins compared to fraternal twins.The magnitude of this genetic influence is expressed as heritability, a parameter
winter 2014 • volume 57, number 1
123
Bertrand Jordan
that varies from zero, for a condition exclusively determined by the environment, to one (or 100%), for a purely genetic affliction such as Huntington’s disease. The disorders mentioned above generally have heritabilities of approximately 0.5 (50%), thus the genes involved should in principle be approachable using positional cloning methods. The numerous attempts made in the 1990s, however, gave very disappointing results—so disappointing that a review published in 1996 on the genetics of bipolar disease (called, at that time, manic-depressive syndrome) was entitled “A ManicDepressive History” (Risch and Botstein 1996). A later analysis of research on the genetic bases for schizophrenia was called “The Maddening Hunt for Madness Genes” (Moldin 1997). Indeed, almost each of the many papers published on mapping of disease genes (the first step of positional cloning) pointed at a different chromosomal localization, even—in some cases, when the families used in the genetic study were the same. The technology used at the time was obviously not robust enough to deal with anything more complex than a straightforward Mendelian affection, and the tests applied to evaluate the statistical significance of the findings were woefully inadequate. Indeed, the only solid conclusion that could be drawn from these early studies was that the (undisputed) genetic component of these diseases involved several, probably many, genes, each having a small effect on the probability of developing the condition—and therefore being quite difficult to identify.
The GWAS Approach The situation changed in the early 2000s, thanks to the availability of the complete human genome sequence (published in “final” form in 2003) and to the development of a very dense genetic map based on hundreds of thousands of single nucleotide polymorphisms (SNPs or “snips”). Simultaneously, miniaturized devices (microarrays or “DNA chips”) were developed that could “score” many snips in an individual’s DNA—in other words, determine which allele is present in the person’s DNA for each of the 500,000 snips detected by the array.The cost of these systems soon became low enough to allow studies on hundreds or even thousands of individuals. More rigorous data analysis methods were also implemented, and it was realized that the detection of genes influencing complex diseases would require the recruitment of thousands of patients and controls in order to achieve sufficient statistical power. This led to the concept of Genome-Wide Association Studies (GWAS), in which, typically, the DNAs of each of a thousand patients and as many controls are analyzed for several hundred thousand snips in large projects usually involving a number of laboratories and considerable amounts of money (McCarthy et al. 2008). The rationale of these studies is that if a particular allele of a particular snip is significantly more frequent in patients than in controls, then this indicates that a gene influencing the disease must be located in the (genetic and
124
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 4 Results of a large GWAS study for seven complex diseases. For each graph, the wavy line represents the associations observed between the disease and a given location along the 23 chromosomes. Because of the strong possibility of random associations, only those whose likelihood (“p”) of being due to random noise is smaller than 10–5 are considered to be significant. The vertical scale corresponds to p on an inverted log scale (the higher the point the smaller p, and therefore the more significant the peak observed).
chromosomal) vicinity of this snip. Such an approach scans the whole genome in an unbiased fashion, and the statistical tests performed ensure that the association observed is really significant and not due to random fluctuations in the data. Indeed, results from GWAS studies have been found to be very reliable and are almost always replicated in later analyses involving more snips or more patients. Once a region of the genome is designated in this fashion, the existing human genome sequence gives immediate information on the genes present in the interval, allowing the research to progress towards gene identification. Figure 4 shows the results of a GWAS study performed in 2007 that is useful for discussing the type of information obtained (Wellcome Trust Case Control Consortium 2007). This particular study investigated seven multigenic diseases, winter 2014 • volume 57, number 1
125
Bertrand Jordan
with 2,000 patients for each and 3,000 controls; the statistical analysis is shown as a graph representing the 23 chromosomes and the associations observed. There is a fluctuating background of random associations; the significant ones (when the probability “p” that the signal is due to random noise is smaller than 10–5) appear as peaks above this background. The picture obtained for three conditions—coronary artery disease, rheumatoid arthritis, and type 1 diabetes—is easy to interpret: a large peak indicating association at one major locus, revealing a gene that has a strong influence on the disease, plus some minor peaks that point to genes having smaller, but still significant, effects. For bipolar disorder and hypertension (although the heritability is not in doubt), no obvious peak appears, an indication that the genetic component of these conditions is spread out over many genes, each of which has a very small, almost undetectable effect. This may also indicate that the condition is heterogeneous—that we are mixing up different diseases with different genetic etiologies under the term “bipolar disorder” or “hypertension,” thus making the GWAS approach ineffective. Finally, Crohn’s disease and type 2 diabetes are intermediate cases with no “major” gene but several quite significant associations, an indication that the genetics underlying these conditions involve a relatively small number of genes. GWAS studies have been very widely performed: more than 2,000 of them have been published to date, notwithstanding the fact that these are intensive, expensive experiments. They have, indeed, reliably identified loci, and in many cases genes, that influence the likelihood of developing particular diseases. For example, in type 2 diabetes, where previous attempts had not consistently pinpointed any genes, GWAS identified several that turned out to be more-or-less directly involved in insulin secretion, and that were confirmed in several successive studies (Torres, Cox, and Philipson 2013). Limitations of GWAS
In spite of this progress, the quantitative impacts of the genes identified are disappointingly small and limit the diagnostic value of these findings. The effect for a particular gene is expressed in terms of the “relative risk,” or the risk for an individual carrying the unfavorable allele compared to the standard risk in the population. In most cases, the relative risk for alleles found in GWAS studies is of the order of 1.2 to 1.3, exceptionally reaching 1.5. These small relative risks, while scientifically interesting in order to understand the etiology of the disease, and significant in terms of studies involving a large group of people, have little diagnostic or prognostic value at the level of the individual. As a comparison, the relative risk of lung cancer for a regular smoker is around 10 (10 times the standard risk in the population), a value that does have significant predictive value for the person involved. Moreover, the relative risks found in GWAS studies do not add up to the observed heritability: summing up the many small effects of the loci identified, one can only explain, typically, 10–30% of the known heritability of the condition.This
126
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
is the “missing heritability” issue that has not yet been adequately solved, although part of the answer probably lies in the very rare variants (alleles that are infrequent in the population), which for technical reasons are not detected in GWAS studies (Maher 2008). Coming back, for example, to type 2 diabetes, the genes identified by GWAS collectively account for only 10% of the known heritability. Ongoing analyses using complete genome sequencing of patients and controls may resolve this issue, although they are obviously very expensive and therefore relatively limited in scope. In summary, the studies on genes underlying complex diseases have made very significant progress after difficult beginnings in the 1990s, and current methods do reliably identify loci and genes influencing the likelihood of developing such disorders, leading to progress in the understanding of pathological mechanisms. However, many questions—such as the “missing heritability” issue—remain unresolved, and the prognostic value of these findings is very limited, as we shall see with the example of autism, admittedly a particularly difficult case.
The Elusive “Autism Gene” Autism spectrum disorder (ASD) is a complex syndrome that includes a series of characteristic behavioral problems that impair communication with others and, in many cases, prevent formal education and social integration. It was, in the past, labeled as an “infantile psychosis” and believed to be caused by disturbance of the mother-child relationship (the “refrigerator mother” as voiced by Leo Kanner in 1943 and emphasized by Bruno Bettelheim in the 1970s). However, since the mid-1980s, a strong genetic component has been demonstrated, mostly by twin concordance studies.The concordance, or frequency with which a child is affected if a twin sibling is diagnosed as autistic, ranges from 60–80% for identical twins, while it is only 10–20% for fraternal twins (Ritvo et al. 1985). Since in both cases each member of the twin pair has been exposed to essentially the same environment, this difference between monozygotic twins (almost 100% identical at the gene level) and fraternal twins (no more related than any two members of the kindred) proves the existence of a significant genetic component. The value quoted for the heritability of autism is somewhat variable between different studies, but it lies between 40– 80%. In addition, recent neurological investigations and, in particular, functional imaging of the brain, indicate clear differences between patients and controls and point to defects in synapse function in autistic patients. The prevalence of ASD has increased significantly in the last decades (probably because of wider diagnostic criteria and growing awareness of the issue) and is now of the order of 1%, making this syndrome a significant public health issue. It is, therefore, unsurprising that the search for an “autism gene” was initiated in the 1990s, at the same time as the pursuit of the genetic components for schizophrenia or manic depression. Needless to say, nothing came out of these studies, except for
winter 2014 • volume 57, number 1
127
Bertrand Jordan
the realization that there is no single or even major autism gene. As technology progressed in the 2000s, more reliable methods were used. GWAS studies implicated a few loci, with—as usual—low relative risks (Autism Genome Project Consortium 2007). It was also found that copy number variations (insertions and deletions in the DNA) were more frequent in patients than in controls, and that some of them were (loosely) specific for autism. Finally, de novo mutations, those occurring in germ cells and present in the child but not in the somatic tissues of the parents, were also found to be associated with autism (Abrahams and Geschwind 2008). In the past few years, 226 genes and several hundred specific genetic alterations have been found to be connected with autism (Matuszek and Talabizadeh 2009). Although it may be reassuring that many of these genes and genetic alterations involve genes that are expressed in the brain and expected to be involved in synaptic function, the resulting picture is complex and confusing (Persico and Napolioni 2013). Part of this difficulty is probably due to the fact that, with the widening of diagnostic criteria, different diseases with distinct genetic etiologies may be lumped together under the ASD umbrella. Given this huge complexity, it would appear that the prospects for diagnosis of autism by DNA analysis are very limited. Yet there is strong demand for such a test, particularly from families that already bear the burden of an autistic child and are uneasy about the evolution of a younger sibling. It is true that early diagnosis is highly desirable, as the existing therapies (mostly behavioral) work best if initiated when the infant brain is still in a flexible state, before two years of age. Accordingly, a number of companies in the United States are marketing tests that claim to evaluate the risk of autism in a child by analysis of the child’s DNA, looking at mutations, polymorphisms, or copy number variants. These tests have, however, very limited value, as shown by their sensitivity and specificity values. To put this into perspective, a good test, such as the widely used HIV screening test, has very high sensitivity, of the order of 99%, ensuring that it misses very few really infected patients. It also has good specificity, at 97–98%, so that it classifies only a few unaffected people wrongly, and positive results are then checked by a more specific (and more expensive) method to eliminate these “false positives.”Yet, for one of the autism tests currently on the market, the sensitivity is only 46%, and the specificity 80.5%, according to the manufacturer’s own figures (Integragen 2012). Such tests are essentially worthless in clinical terms, and their imperfect specificity may result in a negative impact, by classifying perfectly normal children as at elevated risk of autism. However, probably influenced by the early successes of medical genetics with Mendelian diseases, parents, and even physicians in many cases, tend to see a DNA assessment as reliable and conclusive, giving black-and-white results. Thus, these tests are in demand because of the parents’ anxieties and the wish of physicians to investigate all possible avenues and to avoid any potential claim of malpractice. What can genetics contribute to the elucidation and therapy of autism? Many of the genes that have been identified do make sense, as they are involved in neural
128
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 5 The continuum of genetic diseases, from (almost) monogenic to (very) multigenic.
function or communication—there are just too many of them. In some cases, several genes are actually involved in the same pathway: identifying such groups can simplify the picture and help us to understand the underlying biological mechanism of the syndrome. Much of the complexity, however, is probably due to the fact that several different genetic etiologies can produce similar symptoms that we collectively label as “autism.” It is to be hoped that further genetic studies will allow dissection of this combination and provide objective criteria to classify patients into distinct groups (that may also have subtle phenotypic differences). Once this is done, the identification of genes important in a particular group will become much more effective. Eventually, this should translate into really specific treatments that address the particular molecular deficiency present in each group of patients.
Conclusion The apparent contradiction between the indisputable fact that autism is (largely) a genetic syndrome and the failure to isolate the gene for autism stems from the lasting impression left by the early success of medical genetics with monogenic, Mendelian diseases. As we have seen, complexity is the rule, in terms of the importance of both genes and environment and their interactions, and also in terms of the genetic component itself, which almost always involves more than one gene. Mendelian conditions represent just one end of a spectrum (Figure 5), but because of their simplicity and of early success in discovering the underlying genes, they remain dominant in our mental model of the genotype-to-phenotype relationship. Our difficulty in understanding complex diseases is particularly strong for psychiatric
winter 2014 • volume 57, number 1
129
Bertrand Jordan
disorders because of the absence of fully objective, quantitative diagnostic criteria, which leads to the strong likelihood of heterogeneity in the patient sample. Nevertheless, I hope to have shown that genetic analysis can be reliably used for complex diseases, as long as we do not apply the reasoning appropriate only for Medelian disorders. I would also like to stress that failure to find “the gene for” a given condition does not negate the importance of genetics in its occurrence, that this influence is indeed mediated by particular alleles of a (possibly large) number of genes, and that the gene itself, even though its definition is fuzzier today than a decade or two ago, remains a fundamental and operational concept in experimental biology, as well as in the study of evolution.
References Abrahams, Brett S., and Daniel H. Geschwind. 2008. “Advances in Autism Genetics: On the Threshold of a New Neurobiology.” Nat Rev Genet 9 (5): 341–55. Autism Genome Project Consortium. 2007. “Mapping Autism Risk Loci Using Genetic Linkage and Chromosomal Rearrangements.” Nat Genet 39 (3): 319–28. Botstein, David, et al. 1980.“Construction of a Genetic Linkage Map in Man Using Restriction Fragment Length Polymorphisms.” Am J Hum Genet 32 (3): 314–31. Dib, Colette, et al. 1996. “A Comprehensive Genetic Map of the Human Genome Based on 5,264 Microsatellites.” Nature 380 (6570): 152–54. Dragic, Tatjana, et al. 1996. “HIV-1 Entry into CD4+ Cells is Mediated by the Chemokine Receptor CC-CKR-5.” Nature 381 (6584): 667–73. Integragen. 2012. http://www.arisktest.com/home.htm. The figures quoted for sensitivity and specificity come from an internal publication that is no longer available from the company website: Carayol, J., et al. 2012. Technical Report: Assessing Genetic Risk for Autism Spectrum Disorder in Siblings of Children with Autism using the ARISk™ Test. Maher, Brendan. 2008. “Personal Genomes: The Case of the Missing Heritability.” Nature 456 (7218): 18–21. Matuszek, Gregory, and Zohreh Talabizadeh. 2009. “Autism Genetic Database: A Comprehensive Database for Autism Susceptibility Gene-CNVs Integrated with Known Noncoding RNAs and Fragile Sites.” BMC Med Genet 10: 102. For current figures, see http://wren.bcf.ku.edu/. McCarthy, Mark I., et al. 2008. “Genome-Wide Association Studies for Complex Traits: Consensus, Uncertainty and Challenges.” Nat Rev Genet 9 (5): 356–69. Moldin, S. O. 1997. “The Maddening Hunt for Madness Genes.” Nat Genet 17 (2): 127–29. Persico, Antonio M., and Valerio Napolioni. 2013. “Autism Genetics.” Behav Brain Res 251: 95–112. Risch, Neil, and David Botstein. 1996. “A Manic Depressive History.” Nat Genet 12 (4): 351–53. Ritvo, E. R., et al. 1985. “Concordance for the Syndrome of Autism in 40 Pairs of Afflicted Twins.” Am J Psychiatry 142 (1): 74–77. Torres, J. M., N. J. Cox, and L. H. Philipson. 2013. “Genome Wide Association Studies for Diabetes: Perspective on Results and Challenges.” Pediatr Diabetes 14 (2): 90-96.
130
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases Wellcome Trust Case Control Consortium. 2007. “Genomewide Association Study of 14,000 Cases of Seven Common Diseases and 3,000 Shared Controls.” Nature 447 (7145): 661–78. Wirth, Thomas, Nigel Parker, and Seppo Ylä-Herttuala. 2013. “History of Gene Therapy.” Gene 525: 162–69.
winter 2014 • volume 57, number 1
131
Perspectives in Biology and Medicine, Volume 57, Number 1, Winter 2014, pp. 118-131 (Article) 3XEOLVKHGE\-RKQV+RSNLQV8QLYHUVLW\3UHVV DOI: 10.1353/pbm.2014.0002
For additional information about this article http://muse.jhu.edu/journals/pbm/summary/v057/57.1.jordan.html
Access provided by RMIT University Library (16 Aug 2015 09:11 GMT)
Genes and Non-Mendelian Diseases dealing with complexity
Bertrand Jordan
ABSTRACT The first decades of the new medical genetics (1980 to 2000) were
marked by resounding successes, with the identification of the genes responsible (when defective) for muscular dystrophy, cystic fibrosis, or Huntington’s disease, to name just a few of the several thousand Mendelian genetic conditions whose causes are now known. In contrast, the search for genes involved in common disorders such as diabetes, hypertension, schizophrenia, or autism failed miserably in the 1990s, with inconsistent and conflicting results—although the strong genetic component of these diseases (that also involve environmental factors) was (and still is) beyond doubt. From 2000 on, thanks to huge progress in genomic knowledge, technology, and analytical methods, it became possible to reliably identify genes influencing the risk of complex conditions, using the so-called GWAS (Genome-Wide Association Study) approach. Yet many problems remain, such as the vexing question of the “missing heritability,” or the difficulty of translating these scientific results into genetic tests with real clinical validity and utility. Autism is one of the cases in which a strong genetic component has been demonstrated, but where the search for causative genes remains difficult and attempts at developing valid genetic tests have failed, because of the many genes involved and possibly of the heterogeneity of the condition.
Faculté de Médecine de Marseille, UMR 7268 ADES, Aix-Marseille Université/EFS/CNRS, Espace éthique méditerranéen, Hôpital d’adultes La Timone, 264, rue Saint-Pierre, 13385 Marseille Cedex 05, France. E-mail: [email protected].
Perspectives in Biology and Medicine, volume 57, number 1 (winter 2014): 118–131. © 2014 by Johns Hopkins University Press
118
Genes and Non-Mendelian Diseases
A
lmost every human disease has both a genetic and an environmental compo-
nent. Even a classical inherited condition such as hemophilia can be influenced by external factors—in fact, most of the pathogenic effects of the mutation can be avoided by judicious injections of clotting factor, leading to a nearly normal life expectancy. For infectious diseases, often considered as essentially environmental, there are well-documented inherited differences in susceptibility, one of the most striking being the resistance to HIV infection of homozygous carriers of the delta-32 mutation in the CCR5 gene (Dragic et al. 1996). Nevertheless, a number of conditions are “mostly genetic,” others “mostly environmental,” while many have significant components from both sides, as for example diabetes. Another important distinction is that between “monogenic” diseases, where the cause is a mutation in a single gene, the very same in all patients (although the mutations themselves may vary, as well as the severity of the complaint), and “complex” or “multigenic” disorders, in which particular alleles of a number of genes contribute to the genetic component of the disease: type 1 and type 2 diabetes, for example, are multigenic, while hemophilia A, as well as cystic fibrosis or Huntington’s disease, are all monogenic or “Mendelian” conditions. Thus, in the real world, a given disease is usually influenced to some extent by both heredity and the environment, and its genetic component most frequently involves more than one gene. Yet the prevailing image of genetic disorders is that they are due to a single “bad gene” and that the environment matters little:“DNA is destiny,” in other words.This largely stems from the fact that most of the successes of the new medical genetics, in the 1980s and 1990s, have indeed concerned conditions in which the genetic factor was predominant and involved a single gene.This is not surprising, as these are the least difficult cases and were therefore the first to be elucidated. However, these examples have durably shaped our expectations and are responsible for many current misunderstandings. In this article I will briefly recall the principles and results of work on monogenic diseases, then discuss complex disorders, both in general and for the particular case of autism, a particularly difficult and contentious case.
Monogenic Diseases: A Triumph of the New Medical Genetics Monogenic or Mendelian hereditary diseases are those inherited conditions whose transmission in family pedigrees clearly follows Mendelian rules (hence the term), either autosomal dominant (such as Huntigton’s chorea), autosomal recessive (such as cystic fibrosis), or X-linked (such as hemophilia A and B) (Figure 1). In a few cases, the gene involved had been discovered before the recombinant DNA era, using what was the classical approach at the time—in other words, by painstakingly identifying the defective protein and then, also painstakingly, isolating the corresponding gene. At the beginning of the 1980s, it was shown that a new approach—initially called
winter 2014 • volume 57, number 1
119
Bertrand Jordan
Figure 1 Classical Mendelian inheritance schemes. Circles are females; squares males; filled symbols are affected individuals; half-filled ones are carriers. Top left: autosomal dominant, a single copy of the defective gene causes the disease (as in Huntington’s disease); top right: autosomal recessive, the condition is manifest only if two defective genes are present in the same individual (as in cystic fibrosis); bottom: X-linked recessive, the condition is not manifest in females because their second X chromosome carries a functional gene (as in hemophilia A and B).
“reverse genetics,” then, more appropriately, “positional cloning”—could allow identification of “disease genes” without prior knowledge of the protein involved or of the etiology of the condition (Botstein et al. 1980).1 This opened the way for a revolution in human genetics and led to the identification of dozens, then hundreds, of genes responsible (when defective) for as many inherited diseases. The crucial point was the realization that a disease gene could be “mapped” (its approximate location on our chromosomes determined) by following, in suitable families, the association of the disease state with a number of “polymorphic markers” covering our whole genome.2 Once such a genetic association was found, the gene was said to be “mapped”: it was expected to lie within a relatively small region, one or two chromosomal bands, corresponding to a few megabases of DNA. The 1 “Disease genes” are those that cause the disease if they are inactive or otherwise defective, but that are present in active form in normal individuals 2 Polymorphic markers are DNA features that are variable in the population, such as restriction fragment length polymorphisms (RFLPs) or single nucleotide polymorphisms (SNPs).
120
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 2 “Reverse genetics” (later called “positional cloning”) in the 1990s. Top: Mapping or “localization” of the disease gene by following polymorphic markers (represented as restriction fragment length polymorphisms) in affected families, resulting in the designation of a region on the long arm of chromosome 5 (5q12 to 5q14, typically 10 to 20 megabases). Middle: Analysis of the region thus designated and isolation of “candidate genes,” denoted as “transcribed sequences” (TS). Such a region may contain dozens of genes. Bottom: Studies on these genes by sequencing or other techniques until a particular one, transcribed sequence 15 (TS15), is found to be systematically defective in patients, which designates it as the “disease gene.” Consequences (short- or long-term) once this is achieved are indicated at the bottom.
accuracy of this “localization” depended on the number of polymorphic markers available and the size of the families used for the association study. Once this was achieved, it was possible to move to the second step, a detailed molecular analysis of the region designated by the genetic study, to find out which genes lie in the region and to test them one-by-one in patients and controls. When, finally, a gene was found that was systematically defective in patients and functional in controls, this
winter 2014 • volume 57, number 1
121
Bertrand Jordan
was almost certainly the gene involved in the disorder. This process is illustrated in Figure 2, together with some of the tools used at the time for each of the two steps. Originally proposed by Botstein and coworkers in 1980, this procedure quickly led to the mapping of the genes involved in Huntington’s disease (1980), retinoblastoma (1984), and cystic fibrosis (1985). Positional cloning was still quite cumbersome, and the actual identification of the genes often took a number of years after the initial mapping. For example, it took four years to find the genes for cystic fibrosis, and in the case of Huntington’s disease, the gene was mapped in 1980 but only identified in 1993. Nevertheless, positional cloning represented a huge improvement over previous methods. Over the following years, the tools used in both steps became much more powerful. The first genetic maps, based on just a few hundred cumbersome polymorphic markers, were replaced by vastly superior maps incorporating thousands of highperformance markers, notably with the detailed map developed at Genethon (France) in the 1990s by Jean Weissenbach’s team (Dib et al. 1996). Progress in cloning and sequencing technology facilitated the establishment of extensive gene catalogs and the detailed characterization of genes. Accordingly, the rate of progress picked up significantly in the mid-1990s, as shown in Figure 3. Meanwhile the Human Genome Project (HGP), aiming ultimately at the complete sequence of the 3 billion nucleotides of our DNA—a task many believed to be impossible at the beginning—got underway and provided increasingly sophisticated maps and tools for human medical genetics. Thus, by the year 2000, more than a thousand genes involved in human genetic diseases had been conclusively identified, corresponding to almost all Mendelian conditions for which more than a handful of patients exist worldwide. This huge success has not so far resulted in equivalent progress in the treatment of these conditions. Of course, identification of the gene involved in a disease gives access to the sequence of the corresponding protein, and it is a great help in understanding the pathological mechanism. But, for example, although the defective “huntingtin” was identified in 1993, we still do not fully understand how it causes the neuro-degeneration observed in Huntington’s disease. The great expectations of gene therapy for conditions such as hemophilia, cystic fibrosis, or Duchenne muscular dystrophy have not materialized, although the method has scored (limited) success in recent years (Wirth, Parker, and Ylä-Herttuala 2013); at the same time, more conventional, drug-based approaches have not been accelerated as much as was hoped by the identification of the protein targets involved in the corresponding disorders. Understanding, in detail, the biology of a disease takes much time and effort and is a necessary step before the development of novel therapies can be attempted. Molecular genetics has, however, provided efficient diagnostic tools that make it possible to quickly characterize patients and apply the best treatment available; it has also enabled prenatal (and in some cases preimplantation) diagnostics that allow families in which a genetic disease is present to safely foster healthy children, certainly no small achievement. 122
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 3 Progress in gene identification. Number of “disease genes” identified every year from 1981 to 2000. The current total (2013) is over 5,000. Source: http://omim.org/statistics/geneMap.
The public perception of human genetics has been strongly influenced by events in this period, by the resounding successes of medical genetics and also by the media coverage of the ongoing HGP, which gave an oversimplified view (“The gene for disease X”) of the very complex genotype-to-phenotype relationship, as well as an overly optimistic outlook on the short-term medical results. For example, a major French charity campaigned, in 1993, with the motto: “Genes for healing” (Des gènes pour guérir), suggesting strongly that once the genes were found, appropriate treatments would almost immediately be developed. These rosy prospects have not materialized, and this has led to some disappointment with medical research among the general public.
Complex Diseases Are Very Complex Encouraged by the resounding successes scored with Mendelian diseases, many groups attempted to replicate this achievement with conditions that are much more frequent and are known to have a significant hereditary component. Some obvious examples are diabetes, hypertension, and Crohn’s disease, as well as psychiatric syndromes such as schizophrenia or bipolar disorder. While these conditions are clearly influenced by the environment, the personal history, and the lifestyle of individual, they also have a genetic element that is evidenced by the familial clustering of cases and by the high concordance between identical twins compared to fraternal twins.The magnitude of this genetic influence is expressed as heritability, a parameter
winter 2014 • volume 57, number 1
123
Bertrand Jordan
that varies from zero, for a condition exclusively determined by the environment, to one (or 100%), for a purely genetic affliction such as Huntington’s disease. The disorders mentioned above generally have heritabilities of approximately 0.5 (50%), thus the genes involved should in principle be approachable using positional cloning methods. The numerous attempts made in the 1990s, however, gave very disappointing results—so disappointing that a review published in 1996 on the genetics of bipolar disease (called, at that time, manic-depressive syndrome) was entitled “A ManicDepressive History” (Risch and Botstein 1996). A later analysis of research on the genetic bases for schizophrenia was called “The Maddening Hunt for Madness Genes” (Moldin 1997). Indeed, almost each of the many papers published on mapping of disease genes (the first step of positional cloning) pointed at a different chromosomal localization, even—in some cases, when the families used in the genetic study were the same. The technology used at the time was obviously not robust enough to deal with anything more complex than a straightforward Mendelian affection, and the tests applied to evaluate the statistical significance of the findings were woefully inadequate. Indeed, the only solid conclusion that could be drawn from these early studies was that the (undisputed) genetic component of these diseases involved several, probably many, genes, each having a small effect on the probability of developing the condition—and therefore being quite difficult to identify.
The GWAS Approach The situation changed in the early 2000s, thanks to the availability of the complete human genome sequence (published in “final” form in 2003) and to the development of a very dense genetic map based on hundreds of thousands of single nucleotide polymorphisms (SNPs or “snips”). Simultaneously, miniaturized devices (microarrays or “DNA chips”) were developed that could “score” many snips in an individual’s DNA—in other words, determine which allele is present in the person’s DNA for each of the 500,000 snips detected by the array.The cost of these systems soon became low enough to allow studies on hundreds or even thousands of individuals. More rigorous data analysis methods were also implemented, and it was realized that the detection of genes influencing complex diseases would require the recruitment of thousands of patients and controls in order to achieve sufficient statistical power. This led to the concept of Genome-Wide Association Studies (GWAS), in which, typically, the DNAs of each of a thousand patients and as many controls are analyzed for several hundred thousand snips in large projects usually involving a number of laboratories and considerable amounts of money (McCarthy et al. 2008). The rationale of these studies is that if a particular allele of a particular snip is significantly more frequent in patients than in controls, then this indicates that a gene influencing the disease must be located in the (genetic and
124
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 4 Results of a large GWAS study for seven complex diseases. For each graph, the wavy line represents the associations observed between the disease and a given location along the 23 chromosomes. Because of the strong possibility of random associations, only those whose likelihood (“p”) of being due to random noise is smaller than 10–5 are considered to be significant. The vertical scale corresponds to p on an inverted log scale (the higher the point the smaller p, and therefore the more significant the peak observed).
chromosomal) vicinity of this snip. Such an approach scans the whole genome in an unbiased fashion, and the statistical tests performed ensure that the association observed is really significant and not due to random fluctuations in the data. Indeed, results from GWAS studies have been found to be very reliable and are almost always replicated in later analyses involving more snips or more patients. Once a region of the genome is designated in this fashion, the existing human genome sequence gives immediate information on the genes present in the interval, allowing the research to progress towards gene identification. Figure 4 shows the results of a GWAS study performed in 2007 that is useful for discussing the type of information obtained (Wellcome Trust Case Control Consortium 2007). This particular study investigated seven multigenic diseases, winter 2014 • volume 57, number 1
125
Bertrand Jordan
with 2,000 patients for each and 3,000 controls; the statistical analysis is shown as a graph representing the 23 chromosomes and the associations observed. There is a fluctuating background of random associations; the significant ones (when the probability “p” that the signal is due to random noise is smaller than 10–5) appear as peaks above this background. The picture obtained for three conditions—coronary artery disease, rheumatoid arthritis, and type 1 diabetes—is easy to interpret: a large peak indicating association at one major locus, revealing a gene that has a strong influence on the disease, plus some minor peaks that point to genes having smaller, but still significant, effects. For bipolar disorder and hypertension (although the heritability is not in doubt), no obvious peak appears, an indication that the genetic component of these conditions is spread out over many genes, each of which has a very small, almost undetectable effect. This may also indicate that the condition is heterogeneous—that we are mixing up different diseases with different genetic etiologies under the term “bipolar disorder” or “hypertension,” thus making the GWAS approach ineffective. Finally, Crohn’s disease and type 2 diabetes are intermediate cases with no “major” gene but several quite significant associations, an indication that the genetics underlying these conditions involve a relatively small number of genes. GWAS studies have been very widely performed: more than 2,000 of them have been published to date, notwithstanding the fact that these are intensive, expensive experiments. They have, indeed, reliably identified loci, and in many cases genes, that influence the likelihood of developing particular diseases. For example, in type 2 diabetes, where previous attempts had not consistently pinpointed any genes, GWAS identified several that turned out to be more-or-less directly involved in insulin secretion, and that were confirmed in several successive studies (Torres, Cox, and Philipson 2013). Limitations of GWAS
In spite of this progress, the quantitative impacts of the genes identified are disappointingly small and limit the diagnostic value of these findings. The effect for a particular gene is expressed in terms of the “relative risk,” or the risk for an individual carrying the unfavorable allele compared to the standard risk in the population. In most cases, the relative risk for alleles found in GWAS studies is of the order of 1.2 to 1.3, exceptionally reaching 1.5. These small relative risks, while scientifically interesting in order to understand the etiology of the disease, and significant in terms of studies involving a large group of people, have little diagnostic or prognostic value at the level of the individual. As a comparison, the relative risk of lung cancer for a regular smoker is around 10 (10 times the standard risk in the population), a value that does have significant predictive value for the person involved. Moreover, the relative risks found in GWAS studies do not add up to the observed heritability: summing up the many small effects of the loci identified, one can only explain, typically, 10–30% of the known heritability of the condition.This
126
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
is the “missing heritability” issue that has not yet been adequately solved, although part of the answer probably lies in the very rare variants (alleles that are infrequent in the population), which for technical reasons are not detected in GWAS studies (Maher 2008). Coming back, for example, to type 2 diabetes, the genes identified by GWAS collectively account for only 10% of the known heritability. Ongoing analyses using complete genome sequencing of patients and controls may resolve this issue, although they are obviously very expensive and therefore relatively limited in scope. In summary, the studies on genes underlying complex diseases have made very significant progress after difficult beginnings in the 1990s, and current methods do reliably identify loci and genes influencing the likelihood of developing such disorders, leading to progress in the understanding of pathological mechanisms. However, many questions—such as the “missing heritability” issue—remain unresolved, and the prognostic value of these findings is very limited, as we shall see with the example of autism, admittedly a particularly difficult case.
The Elusive “Autism Gene” Autism spectrum disorder (ASD) is a complex syndrome that includes a series of characteristic behavioral problems that impair communication with others and, in many cases, prevent formal education and social integration. It was, in the past, labeled as an “infantile psychosis” and believed to be caused by disturbance of the mother-child relationship (the “refrigerator mother” as voiced by Leo Kanner in 1943 and emphasized by Bruno Bettelheim in the 1970s). However, since the mid-1980s, a strong genetic component has been demonstrated, mostly by twin concordance studies.The concordance, or frequency with which a child is affected if a twin sibling is diagnosed as autistic, ranges from 60–80% for identical twins, while it is only 10–20% for fraternal twins (Ritvo et al. 1985). Since in both cases each member of the twin pair has been exposed to essentially the same environment, this difference between monozygotic twins (almost 100% identical at the gene level) and fraternal twins (no more related than any two members of the kindred) proves the existence of a significant genetic component. The value quoted for the heritability of autism is somewhat variable between different studies, but it lies between 40– 80%. In addition, recent neurological investigations and, in particular, functional imaging of the brain, indicate clear differences between patients and controls and point to defects in synapse function in autistic patients. The prevalence of ASD has increased significantly in the last decades (probably because of wider diagnostic criteria and growing awareness of the issue) and is now of the order of 1%, making this syndrome a significant public health issue. It is, therefore, unsurprising that the search for an “autism gene” was initiated in the 1990s, at the same time as the pursuit of the genetic components for schizophrenia or manic depression. Needless to say, nothing came out of these studies, except for
winter 2014 • volume 57, number 1
127
Bertrand Jordan
the realization that there is no single or even major autism gene. As technology progressed in the 2000s, more reliable methods were used. GWAS studies implicated a few loci, with—as usual—low relative risks (Autism Genome Project Consortium 2007). It was also found that copy number variations (insertions and deletions in the DNA) were more frequent in patients than in controls, and that some of them were (loosely) specific for autism. Finally, de novo mutations, those occurring in germ cells and present in the child but not in the somatic tissues of the parents, were also found to be associated with autism (Abrahams and Geschwind 2008). In the past few years, 226 genes and several hundred specific genetic alterations have been found to be connected with autism (Matuszek and Talabizadeh 2009). Although it may be reassuring that many of these genes and genetic alterations involve genes that are expressed in the brain and expected to be involved in synaptic function, the resulting picture is complex and confusing (Persico and Napolioni 2013). Part of this difficulty is probably due to the fact that, with the widening of diagnostic criteria, different diseases with distinct genetic etiologies may be lumped together under the ASD umbrella. Given this huge complexity, it would appear that the prospects for diagnosis of autism by DNA analysis are very limited. Yet there is strong demand for such a test, particularly from families that already bear the burden of an autistic child and are uneasy about the evolution of a younger sibling. It is true that early diagnosis is highly desirable, as the existing therapies (mostly behavioral) work best if initiated when the infant brain is still in a flexible state, before two years of age. Accordingly, a number of companies in the United States are marketing tests that claim to evaluate the risk of autism in a child by analysis of the child’s DNA, looking at mutations, polymorphisms, or copy number variants. These tests have, however, very limited value, as shown by their sensitivity and specificity values. To put this into perspective, a good test, such as the widely used HIV screening test, has very high sensitivity, of the order of 99%, ensuring that it misses very few really infected patients. It also has good specificity, at 97–98%, so that it classifies only a few unaffected people wrongly, and positive results are then checked by a more specific (and more expensive) method to eliminate these “false positives.”Yet, for one of the autism tests currently on the market, the sensitivity is only 46%, and the specificity 80.5%, according to the manufacturer’s own figures (Integragen 2012). Such tests are essentially worthless in clinical terms, and their imperfect specificity may result in a negative impact, by classifying perfectly normal children as at elevated risk of autism. However, probably influenced by the early successes of medical genetics with Mendelian diseases, parents, and even physicians in many cases, tend to see a DNA assessment as reliable and conclusive, giving black-and-white results. Thus, these tests are in demand because of the parents’ anxieties and the wish of physicians to investigate all possible avenues and to avoid any potential claim of malpractice. What can genetics contribute to the elucidation and therapy of autism? Many of the genes that have been identified do make sense, as they are involved in neural
128
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases
Figure 5 The continuum of genetic diseases, from (almost) monogenic to (very) multigenic.
function or communication—there are just too many of them. In some cases, several genes are actually involved in the same pathway: identifying such groups can simplify the picture and help us to understand the underlying biological mechanism of the syndrome. Much of the complexity, however, is probably due to the fact that several different genetic etiologies can produce similar symptoms that we collectively label as “autism.” It is to be hoped that further genetic studies will allow dissection of this combination and provide objective criteria to classify patients into distinct groups (that may also have subtle phenotypic differences). Once this is done, the identification of genes important in a particular group will become much more effective. Eventually, this should translate into really specific treatments that address the particular molecular deficiency present in each group of patients.
Conclusion The apparent contradiction between the indisputable fact that autism is (largely) a genetic syndrome and the failure to isolate the gene for autism stems from the lasting impression left by the early success of medical genetics with monogenic, Mendelian diseases. As we have seen, complexity is the rule, in terms of the importance of both genes and environment and their interactions, and also in terms of the genetic component itself, which almost always involves more than one gene. Mendelian conditions represent just one end of a spectrum (Figure 5), but because of their simplicity and of early success in discovering the underlying genes, they remain dominant in our mental model of the genotype-to-phenotype relationship. Our difficulty in understanding complex diseases is particularly strong for psychiatric
winter 2014 • volume 57, number 1
129
Bertrand Jordan
disorders because of the absence of fully objective, quantitative diagnostic criteria, which leads to the strong likelihood of heterogeneity in the patient sample. Nevertheless, I hope to have shown that genetic analysis can be reliably used for complex diseases, as long as we do not apply the reasoning appropriate only for Medelian disorders. I would also like to stress that failure to find “the gene for” a given condition does not negate the importance of genetics in its occurrence, that this influence is indeed mediated by particular alleles of a (possibly large) number of genes, and that the gene itself, even though its definition is fuzzier today than a decade or two ago, remains a fundamental and operational concept in experimental biology, as well as in the study of evolution.
References Abrahams, Brett S., and Daniel H. Geschwind. 2008. “Advances in Autism Genetics: On the Threshold of a New Neurobiology.” Nat Rev Genet 9 (5): 341–55. Autism Genome Project Consortium. 2007. “Mapping Autism Risk Loci Using Genetic Linkage and Chromosomal Rearrangements.” Nat Genet 39 (3): 319–28. Botstein, David, et al. 1980.“Construction of a Genetic Linkage Map in Man Using Restriction Fragment Length Polymorphisms.” Am J Hum Genet 32 (3): 314–31. Dib, Colette, et al. 1996. “A Comprehensive Genetic Map of the Human Genome Based on 5,264 Microsatellites.” Nature 380 (6570): 152–54. Dragic, Tatjana, et al. 1996. “HIV-1 Entry into CD4+ Cells is Mediated by the Chemokine Receptor CC-CKR-5.” Nature 381 (6584): 667–73. Integragen. 2012. http://www.arisktest.com/home.htm. The figures quoted for sensitivity and specificity come from an internal publication that is no longer available from the company website: Carayol, J., et al. 2012. Technical Report: Assessing Genetic Risk for Autism Spectrum Disorder in Siblings of Children with Autism using the ARISk™ Test. Maher, Brendan. 2008. “Personal Genomes: The Case of the Missing Heritability.” Nature 456 (7218): 18–21. Matuszek, Gregory, and Zohreh Talabizadeh. 2009. “Autism Genetic Database: A Comprehensive Database for Autism Susceptibility Gene-CNVs Integrated with Known Noncoding RNAs and Fragile Sites.” BMC Med Genet 10: 102. For current figures, see http://wren.bcf.ku.edu/. McCarthy, Mark I., et al. 2008. “Genome-Wide Association Studies for Complex Traits: Consensus, Uncertainty and Challenges.” Nat Rev Genet 9 (5): 356–69. Moldin, S. O. 1997. “The Maddening Hunt for Madness Genes.” Nat Genet 17 (2): 127–29. Persico, Antonio M., and Valerio Napolioni. 2013. “Autism Genetics.” Behav Brain Res 251: 95–112. Risch, Neil, and David Botstein. 1996. “A Manic Depressive History.” Nat Genet 12 (4): 351–53. Ritvo, E. R., et al. 1985. “Concordance for the Syndrome of Autism in 40 Pairs of Afflicted Twins.” Am J Psychiatry 142 (1): 74–77. Torres, J. M., N. J. Cox, and L. H. Philipson. 2013. “Genome Wide Association Studies for Diabetes: Perspective on Results and Challenges.” Pediatr Diabetes 14 (2): 90-96.
130
Perspectives in Biology and Medicine
Genes and Non-Mendelian Diseases Wellcome Trust Case Control Consortium. 2007. “Genomewide Association Study of 14,000 Cases of Seven Common Diseases and 3,000 Shared Controls.” Nature 447 (7145): 661–78. Wirth, Thomas, Nigel Parker, and Seppo Ylä-Herttuala. 2013. “History of Gene Therapy.” Gene 525: 162–69.
winter 2014 • volume 57, number 1
131
![[Dealing with rare diseases in medicine and dermatology -- new developments].](/assets/img/hold.png)
