SEARCH

SEARCH BY CITATION

Keywords:

  • common disease–common variant;
  • genome-wide association study;
  • missing heritability;
  • psoriasis;
  • rare variant

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

During the past 5 years, genome-wide association studies (GWAS), primarily based on single nucleotide polymorphism markers, have identified many loci as potential psoriasis susceptibility regions. These studies appeared to provide strong evidence because the susceptibility genes are involved in the interleukin-23/T-helper 17 axis of psoriasis immunopathogenesis and/or skin barrier functions. However, the “identified” genes only explained a small proportion of psoriasis heritability, although it is known to be comparatively higher than that of other common diseases. GWAS are based on the hypothesis that disease-causing variants are high frequency variants within populations. However, this hypothesis is problematic because deleterious variants such as those predisposing to specific diseases will generally not be maintained by selection pressure throughout human evolution. This issue also affects psoriasis studies. Here, we review the current paradigm shift in human genetic analyses and its implications for detection of psoriasis-causing variants based on linkage analysis and GWAS, except the well-known psoriasis susceptibility locus HLA-C.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

Psoriasis is a complex and multifactorial disease involving environmental factors and has a high prevalence in those with European ancestry. Psoriasis displays a relatively high heritability, and thus, genetic analysis via linkage and/or association studies have been undertaken worldwide to explore its causative genes. Initial genetic studies have demonstrated a strong association between the HLA-Cw6 allele and psoriasis in various races, although the HLA-C locus only explains part of the genetic predisposition to psoriasis. Recently, genome-wide association studies (GWAS) have been used to identify causative genes of psoriasis and other common diseases. Probably, these genes were involved in a predisposition to psoriasis because the susceptibility genes are involved in the interleukin (IL)-23/T-helper (Th)17 axis of psoriasis immunopathogenesis and/or skin barrier functions. However, these observations have never completely explained psoriasis pathogenesis or directly contributed to medical diagnosis and treatment for psoriasis. In this review, we analyze this problem by investigating the implications of current human genetics for disease and genetic analysis of psoriasis by linkage analysis and GWAS, except the well-known psoriasis susceptibility locus HLA-C.

What is gwas?

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

Researchers have recently begun to reassess the outcomes of GWAS in relation to common diseases. A major issue in recent GWAS is whether the implicated variants and/or genes actually predispose individuals to these diseases. However, the conclusions of researchers are inconsistent. The initial hypothesis known as the common disease–common variant (CDCV) hypothesis requires understanding of the recent history of human genetics.

Linkage analysis traces the concordance between genotypes and phenotypes in a disease pedigree and provides a powerful tool for discovering causative genes, especially for Mendelian diseases, even if a researcher does not have a complete dataset of human genome sequences. A causative gene can be found for a Mendelian disease because the concordance between genotypes and phenotypes throughout disease transmission is not an estimation but strong evidence in itself. In fact, causative genes have been discovered using linkage analysis for many Mendelian diseases, such as cystic fibrosis1 and Duchenne’s muscular dystrophy,2 and the total number of diseases caused by known gene mutations is now over 2600.3 These diseases are rarely observed throughout the world and only account for a minority of all human diseases. Researchers have attempted to apply this method to common diseases, which could potentially elevate medical costs and burden the health services in each country, but it has proved difficult to find the causative genes for these diseases. There have been few successes with human diseases. It is easy to explain this problem. Common diseases that do not exhibit obvious Mendelian recessive or dominant inheritance usually have complex traits that are related to environmental factors, which are associated with locus heterogeneity and low penetrance. Moreover, the risk of a locus contributing to the onset of a common disease is generally much lower than with Mendelian diseases. Thus, the statistical detection power for complex diseases using linkage analysis is insufficient.

But was there a better method than linkage analysis for discovering causative genes of complex diseases? During the mid-1990s, the concept of whole-genome association studies was proposed based on affected familial subjects and unaffected individuals in a population. However, researchers had to wait for the completion of human genome sequencing before pursuing such research. Association studies were not new at that time because research of this type was already conducted by using polymorphic markers in human leukocyte antigen (HLA) genes and other genes of interest. GWAS is simply an expanded version of these earlier studies. The purpose of association studies was to find statistical differences in allele frequencies between case and control individuals within the same population. Even if a variant with the strongest disease association in GWAS did not predispose individuals to a disease, it was predicted that a true causative variant would lie in a region that was physically close to the associated variant. This “proxy” variant would exhibit a different allele frequency, if there was linkage disequilibrium (LD) between both loci. However, it was observed that any conclusions obtained from such studies were only estimates rather than evidence. Lander et al.4–6 argued that a disease susceptibility allele would have a high frequency in a common disease, based on results that were mainly obtained from analyses of Alzheimer’s disease and simulations. This is known as the CDCV hypothesis (minor allele frequency >0.01). Lander5 suggested that a comprehensive catalog of variants in whole genomes could be created if hundreds subjects were investigated by sequencing analysis using the human genome sequence. Moreover, Risch and Merikangas7 argued that future genetic analysis of complex diseases was likely to require large-scale examination using association studies and that many genetic effects that were too weak to be identified by linkage would be detected using association studies to statistically estimate their feasibility. The International HapMap Project was established in the early 2000s to create a dataset of common genetic variants found in humans.8 To date, the International HapMap consortium has genotyped 1.6 million common single nucleotide polymorphisms (SNP) in 1184 reference individuals from 11 global populations.9

Outcomes of gwas of common diseases

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

The New York Times published the following in an article in June 2010:10“Ten years after President Bill Clinton announced that the first draft of the human genome was complete, medicine has yet to see any large part of the promised benefits […] Indeed, after 10 years of effort, geneticists are almost back to square one in knowing where to look for the roots of common disease.” Many researchers and other stakeholders had hoped that the use of GWAS for discovery of causative alleles and genes would lead to novel inventions for diagnosis, prevention and treatment of human diseases. But why was this comment published? It is apparent that most of this optimism was misplaced.

Hundreds of SNP were found to be associated with common diseases by GWAS, but the median OR for modest effect sizes was only 1.33.11 Moreover, 88% of these SNP were located in intronic or intergenic regions.11 A large part of the genetic variance (proportion of heritability) in each disease was not explained even when the entire locus discovered by GWAS was assembled.12,13 Like psoriasis, Crohn’s disease is an autoimmune disease that has also been investigated by GWAS, mainly in Caucasian populations where many susceptibility loci were identified. However, an analysis based on a liability threshold model using the 32 loci discovered by GWAS demonstrated that only 13.43% of the total proportion of heritability for Crohn’s disease could be explained.14–18 Similarly, only 13.20% of the proportion was explained by the same analysis of 23 susceptibility loci for systemic lupus erythematosus (SLE).14 GWAS of neuropsychiatric disorders has been even less successful in explaining heritability. Although the heritability of attention deficit hyperactivity disorder (ADHD) was high (76%),19 the variants discovered by GWAS explained less than 1% of its genetic variance.20,21 Some researchers have referred to the undiscovered variance as “missing heritability”12 and have attempted to use alternative methods to find disease-causing variants with a higher effect size. Thus, the validity of the CDCV hypothesis is being increasingly questioned.

A Serious problem with gwas

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

The CDCV hypothesis has a serious problem. Deleterious alleles that contribute to diseases must be rapidly eliminated under evolutionary pressure so as to preserve and expand human species. Therefore, these deleterious alleles will not be maintained at a high frequency in a population. Furthermore, if a common allele directly influences the phenotype, the element surrounding the variant must be essential for viability, indicating that the allele frequency must be maintained by positive selection. Archeological evidence is generally considered to support the initial spread of humans throughout Africa with an East African origin during the first half of the last 100 000 years, followed by their expansion from this point of origin across the world approximately 50 000–60 000 years ago.22 Ancient alleles derived from this African source share most of the alleles of common variants, including those currently used in GWAS.23 Consequently, any deleterious alleles with high frequency must have been supported by evolutionary forces for a long time. This clearly suggests a discrepancy of some sort. However, Lander et al.6,24 explained this discrepancy using simulations where mildly deleterious alleles can emerge to have a moderate frequency, particularly in populations that have undergone a recent human expansion. Klein et al.25 also suggested that the rapid acceleration of human development in the recent evolutionary timeframe has led to numerous environmental changes that increase the risk of complex common diseases. In contrast, Raychaudhuri simulated de novo mutations propagating in a general population. Even if the mutations had no impact on evolutionary fitness and did not influence the survival of individual organisms carrying the mutations, none of the de novo mutations achieved an allele frequency of more than 1% within 200 generations, and over half of the de novo mutations were eliminated from the population after only two generations.26

Exceptional conditions that allow common variants to cause common diseases

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

Common variants with large effect sizes have been detected by association studies, although these cases represent a minority of all studies. These cases have the following patterns:

  • 1
     Late-onset disease: exfoliation glaucoma is the second most common cause of blindness worldwide and is a late-onset disease. GWAS demonstrated that the OR of a common variant associated with this disease was 20.1.27 Alzheimer’s disease, which was used by Lander et al.4–6 as the basis for the CDCV hypothesis, is also a late-onset disease. Clearly, it is reasonable that a common variant can be a causal variant of late-onset diseases because such diseases do not negatively influence reproductive fitness.28
  • 2
     Balancing selection: under balancing selection a gene shows more variation than expected, because an individual with two different versions of it – heterozygotes – have an advantage over those who carry two copies of the same allele.29 In human evolution, a number of alleles that are pathogenic in the homozygous state can confer significant selective advantages in the heterozygous state.30 For example, glucose-6-phosphate dehydrogenase (G6PD) deficiency is the most common enzymopathy in humans and the most common deficiency allele in Africa (G6PD A−), where it has been shown to confer some resistance to malaria.31
  • 3
     Normal traits: common variants may understandably influence normal traits in organisms. GWAS of hair, skin and eye color, which are highly heritable and visible traits in humans, demonstrated that a common variant in OCA2 was strongly associated with eye color (OR = 35.42) and the variant was under the influence of positive selection.32
  • 4
     Pharmacogenomics: the antimicrobial agent flucloxacillin is a common cause of drug-induced liver injury (DILI), which is an important cause of serious liver disease.33 GWAS of cases and a separate drug-exposed control group indicated that HLA-B*5701 had a highly significant association with DILI, where possession of this allele was associated with an 80-fold increase in the risk of developing the disease (OR = 80.6).33 The allele frequency must be maintained because humans have never been exposed chronically to flucloxacillin throughout their evolution and HLA genes are influenced by balancing selection.34
  • 5
     Modifier: β-thalassemia and sickle cell disease are defined as simple Mendelian diseases, but a common variant in BCL11A was discovered by GWAS rather than linkage analysis, and it was found to contribute to both diseases as a modifier.35–37

These examples clearly show how common variants can influence phenotypes and be maintained, but they are exceptional cases in GWAS. Even under neutral evolution, in the absence of natural selection, polymorphisms will eventually vanish, as allele frequencies slowly fluctuate because of genetic drift until one allele becomes fixed.30 Goldstein also suggested that the apparently modest effect of common variation on most human diseases and related traits probably reflects the efficiency of natural selection in prohibiting increases in disease-associated variants in the population.38 Most GWAS results have shown that deleterious alleles are not maintained as common variants during human evolution. Thus, even if a common allele is associated with a disease and is truly the causative variant, the effect sizes are too small to reveal the role of genetic functionality in disease pathogenesis, thereby excluding any applications in diagnosis, prevention and treatment. Moreover, the majority of variants detected by GWAS have no demonstrable biological significance.24 Thus, it is obvious that GWAS based on the CDCV hypothesis cannot be applied to all common diseases.

A Paradigm shift in human genetics

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

At the time the CDCV hypothesis was proposed39 by other researchers, Pritchard suggested that multiple and recent rare variants could both contribute to common diseases. Pritchard referred to studies on Crohn’s disease as examples of rare variants that influence a disease, in which rare frame-shift and missense variants in NOD2 were found to be associated with this disease that displayed high effect sizes.40,41 Goldstein et al.42 argued that much of the genetic control of common diseases is due to rare and generally deleterious variants that have a strong influence on the risk of disease in individual patients. Rare variants are typically more recent and may yet be subjected to negative selection pressure, which means they could include some relatively deleterious mutations.26 Some researchers have pursued rare variants in common diseases and accumulated empirical evidence based on a common disease–rare variant (minor allele frequency <0.01) hypothesis. However, it is difficult to explore rare variants for each individual in a whole genome. However, new sequencing technology and the CDCV hypothesis shed light on research regarding common diseases in human genetics. Next-generation sequencing (NGS) rapidly produces huge amounts of sequence data, allowing human genetic researchers to analyze a personal genome.43 Moreover, the targeted sequencing of all protein-coding regions (“exomes”) using NGS could reduce costs, while increasing the rate of discovery of highly penetrant variants.44 The first report of exome sequencing in a Mendelian disease suggested that this strategy could be extended to diseases with more complex genetics by using larger sample sizes and appropriate weighting of non-synonymous variants based on their predicted functional effect.44 For common neuropsychiatric diseases, several studies using exome sequencing and a few family subjects or unrelated subjects with schizophrenia,45,46 autism,47 epilepsy,48 cognitive disorder49 and mental retardation50 revealed that rare variants of de novo mutations including insertion/deletion polymorphisms with missense and frame-shift mutations showed a large effect size in each disease. Furthermore, rare de novo copy number variations (CNV) with high penetrance were found in schizophrenia,51–56 autism,57,58 epilepsy59–61 and ADHD62 using array comparative genomic hybridization and/or genome-wide SNP genotyping. In addition to these neuropsychiatric diseases and Crohn’s disease, there have only been a few reports of rare variants linked with a predisposition to common diseases. Hypertension is another highly prevalent disease where GWAS have found many associated variants that have small effect sizes.18,63–66 However, multiple rare variants that contribute to blood pressure were discovered using a candidate gene approach by re-sequencing, which demonstrated that many rare alleles influenced renal salt handling during blood pressure variation and that alleles with health benefits were nonetheless under negative selection.67 Rare variants predisposing to pathogenesis of common diseases have been reported for type 1 diabetes,68 low plasma high-density lipoprotein cholesterol levels,69 hypertriglyceridemia70 and severe early-onset obesity.71 These successful studies are not rare cases. The recent and rapid expansion of human populations has resulted in the presence of many rare variants, and rare variants are often more evident when they are more likely to have dramatic functional consequences.26

Based on available results, it is obvious that rare variants are more highly penetrant for most common diseases than common variants. Lander et al.6,24 suggested that mildly deleterious alleles can have a moderate frequency, particularly in populations that have undergone recent expansion. However, there is no available evidence to support this assumption.

Synthetic associations

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

What are the implications of the common variants detected by GWAS in terms of the CDCV hypothesis? A new interpretation of the results of GWAS is suggested. Variants much less common than the associated one may create ‘‘synthetic associations’’ by occurring, stochastically, more often in association with one of the alleles at the common site versus the other allele.72 In other words, it is presumed that rare causative variants with high effect size lie on a chromosome of which chromosomes were identified as a chromosome with low effect size in GWAS. Thus, rare variants can easily lead to genome-wide significant associations attributed to more common variants, given the sample sizes being considered.42,72–74 Moreover, the causal variants may be megabases away from the common variants that provide a signal of the association and can show the several-fold stronger real risk effects than what is credited to a common variant.42 Because rare causal variants arose recently, they often exist on long-range haplotypes spanning multiple blocks of high LD (as observed in control populations), which recombination has not yet had a chance to further fragment. Several lines of evidences that support this hypothesis have already been recognized,42,74 although some researchers have suggested that synthetic associations are unlikely to account for many common disease GWAS signals,75,76 In contrast, Goldstein proposed the synthetic association hypothesis and argued that the proportion of GWAS signals that are synthetic in origin depends on the genetic architecture of human traits, and this architecture remains largely unknown. Hence, it is uncertain how many GWAS signals may be due to synthetic associations.42,73 This is currently a limitation of human genetics. Goldstein observed that “time will tell” in the resolution of this issue.73

Epidemiology of psoriasis

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

Psoriasis is one of the most common human skin diseases. It is characterized by excessive growth and aberrant differentiation of keratinocytes, but it can be completely reversed with appropriate treatment.77 The abnormal production of inflammatory mediators is believed to play an important role in psoriasis pathogenesis. Emerging data from mice and human studies has highlighted a new subset of Th cells that are partially characterized by their production of IL-17, leading to them being known as Th17 cells.78 Considerable overlap is apparent between the molecular pathways involved in psoriasis and those leading to other inflammatory or autoimmune diseases such as Crohn’s disease, SLE, rheumatoid arthritis and Behçet’s disease.79 The clinical manifestations collectively known as psoriasis are a result of complex interactions between different cell types and molecules with multiple environmental triggers such as physical trauma, drugs, infection and stress.80 The familial nature of this disease, which affects almost 2–3% of Caucasians, has long been recognized.81 However, a lower incidence82 has been observed in the Japanese population where most psoriasis cases are sporadic. Psoriasis patients are found in most countries throughout the world, although it is extremely rare or absent in Aborigines, pre-Colombians, Andean Indians, Amerindians in the remote villages of the Amazon–Orinoco forest, Alaskans, Canadians, and Native Americans of the USA.83 The concordance of psoriasis in monozygotic twins is 35–72%, while it is 12–30% in dizygotic twins.84,85 Psoriasis heritability has been estimated at 60–90%, which is among the highest of all multifactorial genetic diseases.86 For example, twin studies in other diseases have indicated that heritability is 30–50% for hypertension,87 50% for Crohn’s disease88 and 40–60% for rheumatoid arthritis.89 Psoriasis in concordant monozygotic twin pairs appears to be similar with respect to age onset, distribution pattern, severity and course, whereas this pattern was not found in concordant dizygotic twin pairs.90 The age of psoriasis onset shows a bimodal distribution with one peak at 20–30 years and another at 50–60 years.91 Moreover, the HLA-Cw*0602 allele is strongly associated with the early-onset type of psoriasis in various races.92–97 In addition to having a lower age of onset, HLA-Cw*0602-positive patients present more severe clinical symptoms.98 Thus, psoriasis has complex multifactorial features, genetic heterogeneity, high heritability, a broad range of onset age and different prevalence in various populations, suggesting that psoriasis is comparatively similar to other common diseases. With the exception of HLA, there is no reason why deleterious variants predisposing to psoriasis should endure any selection pressures. Taking recent human genetics into consideration, it is unlikely that deleterious common variants for psoriasis, other than for HLA, exist in the human genome.

Linkage analysis of psoriasis

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

Although linkage analysis has been used for investigation of causative genes of psoriasis, no causative genes that predisposed to psoriasis with a large effect were revealed apart from HLA-C. Ten loci have been identified as psoriasis susceptibility regions by linkage analysis (Table 1). Fine mapping has been performed in four of these regions. We review the following three loci identified as psoriasis susceptibility regions by previous linkage studies.

Table 1.   Susceptibility locations detected by linkage analysis in psoriasis family
LocationHGNC* approved gene symbolCandidate geneReference
  1. *HUGO Gene Nomenclature Committee.

1q21PSORS4LCE3B and LCE3C110,111
1pPSORS7 128
3q21PSORS5SLC12A8113,114
4qPSORS3 129
4q31-q34PSORS9 130
6p21.33PSORS1HLA-C102,103,109,113,126,128,130,131
16qPSORS8 103,131
17q25PSORS2SLC9A3R1 and RAPTOR99,102–104,131
18p11.23PSORS10 132
19p13PSORS6TYK2?126

17q25

A psoriasis susceptibility region was localized to the distal region of the human chromosome 17q as a result of the first genome-wide linkage analysis using polymorphic microsatellites in eight multi-affected psoriasis families with Caucasian ancestry.99 There was also evidence of genetic heterogeneity, and although none of the related families showed any association with HLA-Cw6, two unrelated families showed weak association levels. Nevertheless, D17S784 showed strong evidence for linkage,99 although two other linkage studies did not confirm this linkage.100,101 On the other hand, two independent studies provided evidence for linkage in the same region of 17q.102,103 To further refine this location, the group that first found this linkage focused on chromosome 17q23–25 (physical distance = 11.5 Mb) surrounding the D17S784 locus and they conducted non-parametric genetic linkage analysis and association studies using microsatellites and SNP.104 As a result, a microsatellite (D17S1301) with linkage and two SNP haplotypes associated with psoriasis were observed in a segment located 512 kb away from D17S784.104 The same group then conducted a family-based association study and found two loci associated with psoriasis where the regions were separated by 6 Mb.105 One locus was an SNP (rs734232) located 84.7 kb distal to D17S1301, which was identified as a putative binding site for the runt-related transcription factor RUNX1 and a causal variant influencing SLC9A3R1 expression, whereas the other peak was in the third intron of RAPTOR located 852 kb distal to D17S784.106,107 The frequency of risk allele in rs734232 was 48% compared with 42% in the controls,105 indicating that the proportion of heritability of this locus would be very low, even if this locus was truly linked with a predisposition to psoriasis. However, other studies have indicated a lack of evidence for a genetic association with the RUNX1-binding site and RAPTOR,107,108 while all recent GWAS have failed to replicate these associations with psoriasis. A consistent result that explained a relationship between SLC9A3R1 expression and the “risk” allele was provided by electrophoretic mobility supershift and luciferase assays.105 It was unlikely that this observation explained a skin abnormality such as psoriasis because the frequency of the risk allele was high in control subjects. However, this locus has been supported in independent families and studies, although there are several cases of negative data. The concepts of genetic linkage in families and a genetic association between cases and controls are entirely different. Linkage analysis is effective for finding a causative gene under allelic heterogeneity and is not effective under locus heterogeneity; however, association study is the opposite. If this linkage is real and multiple rare variants that predispose to psoriasis are found in this region, some discrepancies in these genetic data might be disclosed because multiple rare variants would emerge due to allelic heterogeneity.

1q21

Chromosome 1q21 was mapped as a psoriasis-susceptibility locus by linkage analysis109 and subsequent fine mapping in an association study110 conducted by the same group. An associated haplotype was detected using microsatellites in the association study, which segregated in only one of the 22 psoriasis families where linkage to the 1q21 region was originally demonstrated.109,110 This indicated that the causative allele frequency in this region should be relatively low in populations. In fact, the haplotype was found at an increased frequency among disease chromosomes compared with that in control chromosomes (6.9% vs 1.9%).110 This region was subsequently supported by GWAS using SNP117 and a genome-wide search for CNV,111 but we cannot understand in detail whether these associations might explain the findings of previous linkage analysis and association studies of psoriasis. However, we can at least speculate regarding this issue. The researchers who analyzed CNV concluded that a susceptibility variant of psoriasis was a deletion of the late cornified envelope LCE3B and LCE3C genes.111 Frequencies of the “risk” deletion in European cases and controls were 69.6% and 64.2% (OR = 1.21), respectively.112 The physical distance between the haplotype associated with psoriasis in a previous fine mapping109,110 and LCE3B discovered by GWAS was 583 kb. If this deletion actually predisposes to psoriasis, these results in the linkage analysis and the association studies must be independent. The chromosome harboring the risk deletion allele could not maintain a long-range LD to the linkage region for long if genetic events such as recombination occur normally. This common deletion allele emerged in a very old chromosome; hence, the region surrounding the deletion site must have had frequent opportunities for recombination in every generation. In contrast, the rare haplotype detected by linkage analysis and subsequent fine mapping may have emerged recently and contain a relatively longer haplotype. Thus, if both results are true, the CNV association observed by GWAS may have attributed to the rare causative variants in the same region based on synthetic associations. LCE expression is upregulated in psoriatic and normal skin that is stimulated by tape stripping, while it is not detected in uninvolved psoriatic and normal skin. Thus, the expression of this gene can be induced in the normal epidermis by disruption of the skin barrier.111 A significant correlation was found between normalized LCE3C expression and copy number.111 However, the conclusion that this deletion is a deleterious variant in the epidermis cannot explain why 42% of healthy individuals of European ancestry112 who do not express these genes will not be psoriasis patients. Interactions with the other proteins may explain this discrepancy, but this deletion cannot be a major psoriasis-predisposing event.

3q21

A linkage analysis in Swedish families identified chromosome 3q21 as a psoriasis susceptibility locus.113 Subsequent fine mapping by the same group isolated the locus to a 250-kb interval mainly by using SNP markers and suggested SLC12A8 as a plausible candidate gene.114,115 However, genetic evidence was not strong and it has not been replicated by other studies including recent GWAS.

We suggest that results of previous linkage analysis must also be reconsidered because GWAS cannot detect causative variants with large effects, while linkage analysis is inadequate for common diseases such as psoriasis. Family subjects with psoriasis investigated by some researchers are important and they will be informative for future sequence-based genotyping.

What have gwas of psoriasis told us?

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

In the past 5 years, 10 GWAS of psoriasis based on the CDCV hypothesis have been published, indicating that 24 loci have associations with psoriasis (Table 2).116 These GWAS identified many genes involved with skin barrier functions (LCE3E and LCE3C),15,111,117 IL-23 signaling (IL23A, IL23R and IL12B),117–122 nuclear factor-κB and interferon signaling (NFKBIA, REL, TYK2, IFIH1, IL28RA, TNIP1 and TNFAIP3)118,121,123 and IL-17 cell responses (TRAF3IP2, TYK2 and IL23R),118,120–122 in addition to HLA-C as psoriasis susceptibility genes. In general, these results appeared to indicate the IL-23/Th17 axis in psoriasis immunopathogenesis. However, what does “identify” mean in these studies? Most variants associated with psoriasis are located in intergenic or intronic regions. Thus, GWAS have not identified psoriasis susceptibility genes and merely identified variants that were statistically associated with psoriasis. These genes are involved in IL-23/Th17 pathway and are physically close to variants associated with psoriasis, which cannot prove that they predispose to psoriasis. Furthermore, few studies have completely explained the relationship between genetic data in GWAS and gene expression.

Table 2.   Susceptibility locations detected by genome-wide association studies in psoriasis
LocationOverlap region detected by linkage analysisReported gene(s)Strongest SNP risk alleleContextRisk allele frequency in controlOR95% CIP-valueAncestryReference
  1. *Psoriatic arthritis stratified analysis. CI, confidence interval; NR, not reported; OR, odds ratio, SNP, single nucleotide polymorphisms; UTR, untranslated region.

1p36.11 IL28RArs4649203-AIntergenic0.7301.131.05–1.227.0E-08European121
1p31.3 IL23Rrs11209026-?MissenseNR1.491.27–1.747.0E-07European121
1p31.3 IL23Rrs2201841-GIntron0.3001.13NR3.0E-08European118
1q21.3PSORS4LCE3B, LCE3Crs4112788-CIntergenic0.4401.411.25–1.586.5E–09European111
1q21.3PSORS4LCE3Drs4112788-?IntergenicNR1.291.19–1.403.0E-10European121
1q21.3PSORS4LCE3D, LCE3Ars4085613-AIntergenic0.4301.321.25–1.397.0E-30Chinese117
2p16.1 RELrs702873-GIntergenic0.5601.121.04–1.204.0E-09European121
2p16.1 NRrs842636-GIntergenic0.5601.15NR6.0E-06European123
2q24.2 IFIH1rs17716942-AIntron0.8601.291.17–1.431.0E-13European121
3p24.3 Intergenicrs6809854-GIntergenic0.1901.141.04–1.261.0E-07European121
5q15 ERAP1rs27524-AIntron0.3601.131.05–1.223.0E-11European121
5q33.3 IL12Brs3212227-CUTR0.5170.640.56–0.907.9E-10European119
5q33.3 IL12Brs2546890-AIntergenic0.5601.541.32–1.79(Panel A)1.0E-20European120
5q33.3 IL12Brs12188300-TIntergenic0.0801.701.50–1.937.0E-17European122
5q33.3 IL12Brs3213094-?IntronNR1.391.26–1.535.0E-11European121
5q33.3 IL12Brs2082412-GIntergenic0.8001.44NR2.0E-28European118
5q33.3 IL12Brs3213094-AIntron0.4501.281.23–1.353.0E-26Chinese117
5q33.1 TNIP1rs17728338-AIntergenic0.0541.59NR1.0E-20European118
5q31.1 IL13rs20541-GMissense0.7901.27NR5.0E-15European118
6p21.33PSORS1HLA-Crs1265181-?IntergenicNR22.6NR1.9E-208Chinese117
6p21.33PSORS1HLA-Crs12191877-?IntergenicNR2.792.35–3.33(Panel A)4.0E-32European120
6p21.33PSORS1HLA-Crs13191343-TnearGene-50.1302.372.16–2.612.0E-72European122
6p21.33PSORS1HLA-Crs10484554-?IntergenicNR4.664.23–5.134.0E-214European121
6p21.33PSORS1HLA-Crs12191877-TIntergenic0.1502.64NR1.0E-100European118
6p21.33PSORS1HLA-Crs2395029-CMissense0.0304.103.10–5.302.0E-26European133
6p21.33PSORS1HLA-Crs10484554-TIntergenic0.1502.802.40–3.302.0E-39European133
6p21.33PSORS1HLA-Crs3134792-?IntergenicNRNRNR1.0E-09European134
6q23.3 TNFAIP3rs610604-?IntronNR1.221.13–1.327.0E-07European121
6q23.3 TNFAIP3rs610604-GIntron0.3201.19NR9.0E-12European118
6q21* TRAF3IP2rs33980500-TMissense0.0801.571.38–1.781.0E-16European120
6q21* TRAF3IP2rs33980500-TMissense0.0701.951.69–2.241.0E-20European122
6q21 TRAF3IP2rs240993-AIntron0.2501.251.16–1.345.0E-20European121
9q34.13 TSC1rs1076160-TIntron0.4801.09NR6.0E-06European118
12q13.3 IL23Ars2066808-?IntronNR1.491.28–1.732.0E-07European121
12q13.3 IL23A, STAT2rs2066808-AIntron0.9301.34NR1.0E-09European118
12q13.2 RPS26rs12580100-AIntergenic0.9001.17NR1.0E-06European123
13q14.11 COG6rs7993214-?Intron0.6501.411.22–1.612.0E-06European133
14q13.2 NFKBIArs8016947-CIntergenic0.5701.191.11–1.272.0E-11European121
14q13.2 NFKBIA, PSMA6rs12586317-TIntron0.7501.15NR2.0E-08European123
16p11.2 FBXL19,POL3Srs10782001-GIntron0.3701.16NR9.0E-10European123
17p11.2 NRrs1975974-GIntergenic0.2301.17NR1.0E-07European123
17q11.2 NOS2rs4795067-GIntron0.3501.19NR4.0E-11European123
19p13.2PSORS6TYK2rs12720356-AMissense0.9001.401.23–1.614.0E-11European121
20q13.13 RNF114,SPATA2,SLC9A8,SNAI1rs495337-Gcds-synon0.5701.21NR2.0E-07European123
20q13.13 SPATA2rs495337-?cds-synonNR1.251.12–1.391.0E-08European134
20q13.12 SDC4rs1008953-CIntergenic0.7901.14NR1.0E-07European123

Were there any important findings in GWAS of psoriasis? King et al. argued that most of the signals with small size effects (OR < 1.5) detected in GWAS are false positive, regardless of the P-value associated with them.28,124 In a risk allele which demonstrated an OR of over 1.5 and lay in a coding region of a functional gene (excluding HLA-Cw*0602), the only SNP was rs33980500 in TRAF3IP2 which encoded ACT1, a signaling adaptor involved in the regulation of adaptive immunity (Table 2).122,124 These studies of European populations also indicated that the association with “psoriatic arthritis” was stronger than “psoriasis vulgaris”.120,122 This SNP is a missense variant that causes a mutation from aspartic acid to asparagine in the protein sequence, resulting in a change in charge (a negative electric charge to nonpolar).120 It is interesting that the variant is located in a region that is more than 90% conserved among different species, which further indicates a possible functional consequence of this change.120 In fact, functional assays have clearly shown that a protein derived from a transcript harboring the risk allele could reduce the efficiency of the interaction with tumor necrosis factor receptor-associated factor 6.124 These results are strong evidence, but may explain only a small proportion of heritability. In contrast, it seems unlikely that the other “associated” variants were causative and they predispose to psoriasis. Even if the variants associated with psoriasis were due to synthetic associations from multiple rare variants, we could not evaluate whether synthetic associations are expected in most psoriasis GWAS without additional evidence such as linkage analysis.

Linkage analysis in German families identified chromosome 19p13 as a psoriasis susceptibility locus.125 This region harbored the TYK2 locus discovered in a GWAS with European ancestry.110 However, we could not evaluate whether the concordance between both studies was due to any genetic factor because the linkage study left a vast region (>20 Mb).

Where is the “missing heritability” in GWAS of psoriasis? Chen et al.126 estimated the genetic variance using a liability threshold model for 10 selected loci from GWAS of psoriasis, indicating that these loci could explain only 11.6% of the proportion. The locus with the largest genetic variance, other than the well-known susceptibility locus HLA-C, was IL12B (rs3213094), and this locus could explain 1.27% of the genetic variance in psoriasis.126 Despite spending vast amounts of funds and collecting huge numbers of subjects, recent GWAS of psoriasis have a very large amount of missing heritability, as is found with other GWAS of other common diseases. Most of the 10 published GWAS of psoriasis indicated that functional study, fine mapping and additional association studies are needed to confirm these results, and they leave the identification of causal variants for future studies. Thus, important and informative results have not been reported and they fail to address this fundamental issue of missing heritability. Some researchers indicate that elevating the OR by interactions between variants associated with psoriasis could account for missing heritability. Two studies suggested that the risk allele of rs27524 was found in an intron of ERAP1, and the risk deletion of LCE3D statistically interacted with HLA-C, respectively.111,121 However, it was unlikely that the evidence was strong or that the interactions would be supported by biological and experimental data. In general, interactions are theoretically possible but there are few clear examples of common variants identified in GWAS that interact strongly with each other or the environment to affect a complex trait.42 Thus, even if we conduct more GWAS with larger sample sizes to capture all the remaining unidentified variants influencing psoriasis, numerous variants would be required to explain missing heritability.38

Future of psoriasis genetics

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References

Common diseases are often complex because they are genetically heterogeneous, where many different genetic defects can result in clinically indistinguishable phenotypes.49 Thus, even if the disease in every affected individual emerges from a different specific cause, each will nonetheless share the disruption of related key biological processes.28 Studies of common diseases in populations may not be appropriate for finding disease causative variants with large effect sizes. The genetic heterogeneity of populations and their biology may be more complex than our expectations. We usually prefer to consider a hypothesis that simplifies the results of a study. The estimated total number of human genes in the human genome database was 26 473 on 7 September 2011.127 However, many scientists estimated that the number would be vastly more before human genome sequencing is complete. A large part of the human phenotype cannot be explained using simple hypotheses, such as the “central dogma”. Therefore, identifying disease susceptibility genes one by one based on a conventional method and hypothesis may not lead to a complete understanding of the pathogenesis of each common disease.

None of the available methods or theories of human genetics can comprehensively reveal the genetic factors underlying the pathogenesis of common diseases. However, future studies of psoriasis can only try to find rare variants that predispose to psoriasis and that have large effect sizes, using sequencing-based genotyping by NGS of familial subjects and/or sporadic subjects that are carefully screened based on strict criteria. Thus, we may need to re-evaluate the global epidemiology and the features of psoriasis in detail and construct a novel hypothesis for disclosing the genetic architecture of psoriasis by carefully selecting the optimal methods and without blindly applying new technology.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. What is gwas?
  5. Outcomes of gwas of common diseases
  6. A Serious problem with gwas
  7. Exceptional conditions that allow common variants to cause common diseases
  8. A Paradigm shift in human genetics
  9. Synthetic associations
  10. Epidemiology of psoriasis
  11. Linkage analysis of psoriasis
  12. What have gwas of psoriasis told us?
  13. Future of psoriasis genetics
  14. References