Familial aggregation

Familial aggregation is often assessed by the recurrence risk ratio. The pattern of recurrence risk ratios across different types of relatives can provide valuable information about the origin of a binary trait and can add to the statistical power of linkage studies (14). The recurrence risk ratio (λ_R in relatives of type R relatives is the prevalence of the disease in relatives of type R of affected cases (P_R) divided by the prevalence in the general population (P). If the relatives are siblings, λ_S and P_S would be used. Families are often recruited because they have affected members. This outcome-based sampling is often more informative and increases power. Furthermore, it has obvious benefits for a study aimed at estimating λ_R, the prevalence of disease in a particular subgroup of relatives. However, because the familial determinants of the trait of interest are usually unobserved in a study of familial aggregation, this sampling method can lead to severe ascertainment bias (13).

For many complex diseases, the average λ_R in first-degree relatives is around 2. It tends to be greater the younger the age at onset in the affected individual, to fall as the familial relationship becomes more distant, and to increase as the number of affected relatives of the at-risk individual rises. Although a λ_R of 2 might appear modest, it does suggest that uncovering all sources of familial aggregation might well be worthwhile. A moderate λ_R generally implies the presence of underlying familial risk factors (genetic or non-genetic) that are, at least, an order of magnitude stronger than λ_R itself. Because a simple assessment of familial aggregation takes no account of the underlying biology, one should not assume that evidence of familial aggregation implies genetic effects. For many complex diseases, the non-genetic risk factors identified to date have a modest effect and are weakly correlated in relatives. They, therefore, seem to explain a little familial aggregation (13).

ADs show a tendency towards familial aggregation but the incidence in close relatives of affected individuals is usually low compared to the much higher figures that would be expected if these conditions were Mendelian-like (9). Recurrent associations of ADs in family members of patients have been reported and comprise autoimmune thyroid disease (AITD), systemic sclerosis (SSc), rheumatoid arthritis (RA), and systemic lupus erythematosus (SLE) as the most common ADs among relatives (15-22). Multiple sclerosis (MS), primary biliary cirrhosis (PBC), celiac disease (CD), and antiphospolipid syndrome (APS) have also been registered although less frequently (16,20,22-24). Shared immunopathological mechanisms among Sjögren’s syndrome (SS), SLE, SSc, and AITD as well as a possible familial aggregation of these diseases has been also reported (25). In fact, SS may coexist with other ADs such as RA (26), SLE (27), AITD (28), and SSc (29). In addition, AITD has been the most common disease encountered among first-degree relatives as has also been reported in familial studies of MS (30), vitiligo (VIT) (31), juvenile RA (32), SLE (33), SS (22,28), and type 1 diabetes mellitus (T1D) (34,35) patients. T1D, SS, and SLE share similar susceptibility gene polymorphisms including HLA and non-HLA variants (36-39), which may account for aggregation.

Many disorders have demonstrated a familial aggregation that does not conform to any recognized pattern of Mendelian inheritance. These conditions show a definitive familial tendency but the incidence in close relatives of affected individuals is usually around the lower instead of the much higher figures that would be seen if these conditions were caused by mutations on single genes (40). The impact of genetic predisposition on susceptibility to AD was first identified by the analysis of disease concordance rates in monozygotic twins (ranging from concordance rate ranges from about 10% to 85%) (41,42) (Table 2). The decrease in the concordance rate of siblings compared to the rate for monozygotic twins supports the presence of multiple genes contributing to the autoimmunity phenotype onset (43). In fact, ADs are not inherited in a classical Mendelian pattern, but instead have a complex, yet to be defined mode of inheritance (44-46).

Twin studies are an invaluable source for researchers attempting to distinguish whether genetic or environmental factors (or varying degrees of both) contribute to the development of disease and set the basis for composite disease heritability estimates (43,47). Concordance rates are based on a comparison of disease status between monozygotic (MZ, i.e., developing from a single fertilized egg and, therefore, sharing all of their alleles, MZ), and dizygotic (i.e., developing from two fertilized eggs and, therefore, sharing on average 50% of their polymorphic alleles, the same level of genetic similarity as found in non-twin siblings, DZ) twins. The presumption is that as MZ twins share 100% of their genomic sequence, a phenotype concordance significantly higher than that for DZ twins would be suggestive of predominantly genetic influences while low concordance rates would indicate stronger environmental factors. Concordance rates for most conditions generally support both genetic and environmental influences with varying degrees of each. A combination of environmental and genetic influences has formed the basis of a complex multifactorial model of disease, where a genetically predisposed individual encounters several environmental exposures over a lifetime which culminates in disease development after stochastic exposures (Figure 2). This model has been proposed for several multifactorial diseases, including AD, and the limited applicability of the most robust genomic associations from genome-wide association studies (GWAS) has supported this notion.

Figure 2

A. Phenotype progression. B. Familial and individual liability. The exposures of an individual to environmental risk factors throughout his lifetime will define the vector of disease progression toward an autoimmune disease state depending on his population (more...)

Moreover, twin studies in ADs highlight the complexity and obstacles that may be faced from one disease to the next such as the variability of MZ and DZ concordance rates between reports of the same disease (Table 2). Several ADs, e.g., CD, have a strong genetic component with high concordance rates for MZ twins (43,47). Likewise, for example, SSc has been shown to have a less prominent genetic component with, therefore, more room for environmental factors. T1D, which is perhaps the most studied, discloses a concordance rate for MZ twins ranging from 23 to 64.3% (43). However, due to great variability in concordance rates, the interpretation of the results of these reports needs to be done cautiously. Sex differences have also been observed in twin studies on T1D. The proband and pairwise concordance rate for MZ male twins was 42.4% and 27% respectively, and 7.8% and 4.1% for DZ male twins respectively (48). Female MZ twins had a concordance rate of 43.5% probandwise and 27.8% pairwise. DZ females had a 4.4% and 2.3% probandwise and pairwise concordance respectively (48). A higher female concordance and greater time interval of discordance was noted in a study by Redondo et al. (49). A North American study found that the risk for male DZ twins was as high as that for the MZ male twins and significantly higher than DZ female twins (50).

Age and autoimmunity

Consideration of age as a factor responsible for the onset of ADs at midlife (age, 40–60) has been proposed. Two specific circumstances should be considered. Firstly, detection bias, given that some ADs depict slow sign and symptom progression, making the age of onset not perceptible. Secondly, alteration of the biological functions, which may alter the development of the disease (i.e., decreased apoptosis and increased clonal activation of T cells, or decreased ability to respond to antigenic or mitogenic stimulation, etc.) (51). Nevertheless, age remains an important topic in autoimmunity not only because of the biological implications of aging on the immune system but also because of the setback it constitutes for epidemiological studies which has the common origin of ADs as its goal. The problem with age in epidemiologic studies lies in the fact that many ADs have different ages of onset. For children for example, the most common diseases are T1D, CD, and VIT (52). For young adults, MS, myasthenia gravis (MG), VIT, and SLE are the most frequent (53). Mid-age patients are more likely to have SS (54), SSc (55), and RA (56). Finally, the elderly are more prone to have SS (54), AITD (57), and MG (58). This mosaic of ages constitutes one of the biggest problems in aggregation and co-occurrence studies and can be summarized in two types of setbacks (59). The first is the reduced probability of finding aggregation of ADs for affected patients when age differences are considerable. For example, there is the case of a young child affected with T1D whose parents are young and restricted to a small group of ADs. The opposite scene would be the elderly patient whose parents are already deceased and whose children are too young to present many of these kinds of pathologies. The second type of setback is the one arising when doing co-occurrence studies. The limitation arises when two diseases are so far apart with respect to their time of diagnosis that a necessary and rigorous follow-up will be required in order to find co-occurrence in one patient (59).

Gender, inheritance, and autoimmunity

The vast majority of patients with ADs are female. The reason for this prevalence is poorly understood (Table 2). The more frequent the AD and the later it appears, the more women are affected. Many ideas that are mainly based on hormonal and genetic factors that influence the autoimmune systems of females and males differently have been proposed to explain this predominance. These hypotheses have gained credence mostly because many of these diseases appear or fluctuate when there are hormonal changes such as in late adolescence and pregnancy (60). The proportion of females with AD varies depending on the disease: from 18:1 in AITD to 1:1 in PSO to 1:2 in ankylosing spondylitis (AS) (Table 2). Even for specific populations, reports have described differences in gender and clinical presentation pertaining to sex and RA (61). ADs that are more prevalent in men are characterized by acute inflammation, appearance of autoantibodies, and a proinflammatory Th1 immune response whereas ADs that are more prevalent in women have known antibody-mediated pathologies. Moreover, ADs that are more prevalent in women and that clinically appear in women over age 50 are associated with chronic Th2-mediated pathology (62).

Sex hormones are involved in the susceptibility factors for AD through modulation of Th1/Th2 response. Impact of hormonal changes on the disease course in females is documented in pregnancy: severity of MS and RA has been reported to decrease during pregnancy whereas severity in SLE is either exacerbated or unaffected during pregnancy. This may be explained by high levels of hormones during pregnancy which enhance Th2 response and suppress RA and MS which are driven by Th1 response. In contrast, SLE is Th2 driven and may not be suppressed by the hormones (63).

Theoretically, X-chromosome inactivation and the resultant tissue chimerism might explain the female predisposition to systemic autoimmunity (64,65). In females, half their somatic cells express antigens derived from the paternal X and half from the maternal X. The Burnet-Jerne theory of somatic generation of antibody diversity and forbidden clone elimination states that lymphocytes under maturation in the thymus are killed or suppressed if they present high or no affinity towards a histocompatibility antigen. If this were to hold for self-antigens as well, females would escape expressing one of her parental X chromosomes, which would still be able to react to self. Then lymphocytes happening to pass the selection in the thymus would meet only cells expressing one of the parental X chromosomes and it would be easier to predispose for a dysregulation of self-tolerance in females than in males. This is known as the Kast conjecture (66). Even though specific responses to immunization do not appear to account for the high sex ratio seen on ADs (67), there is still a chance that chimerism among immunological cells could represent a starting point for perpetuating or acquiring an unbalanced self-tolerance. Ultimately, both X monosomy as a resource to produce chromosome instability and haploinsufficiency for X-linked genes have been put forward as possibly playing critical roles in the predominance of ADs in females (68). None of the suggestions presented have a completely proven experimental background, and they are still part of a sex-connotation discussion.

A preferential inheritance of the autoimmunity trait from mothers has been observed in patients with primary SS (19,22), SLE (33), and T1D (34), thus indicating a preferential transmission of susceptibility genes from mothers to their offspring. Maternal transmission of autoimmunity could be influenced by the high preponderance of ADs in females compared to the general population given their greater intrinsic susceptibility to develop these diseases that can potentially arise from sex related physiological factors (69). Nonetheless, other reports have postulated a preferential transmission of the autoimmune trait through the father, specifically T1D (70) and MS (24). Given women’s inherent excess of susceptibility to developing an AD, men would require an augmented risk to overcome the resistance towards autoimmunity relative to women. Thus, men would need a greater content of susceptibility variants that would trigger their phenotype and would also guarantee more frequent transmission of the autoimmunity trait to their offspring. This has been reported previously for MS by Kantarci et al. (24) under the Carter effect, which is defined by the observed higher incidence of the trait in relatives when the index case is the least commonly affected sex (71). Consequently, the incidence of the trait would be expected to be the highest in daughters of affected men and lowest in sons of affected women, a trend that has not been convincingly confirmed (24). Although the Carter effect for pyloric stenosis has been explained by a multifactorial threshold model (72), further research on familial autoimmunity must be done given the possibility of an ascertainment bias of the studied families through the affected probands that would generate a higher proportion of autoimmunity in the affected fathers than in mothers (70).

Segregation of autoimmunity

The purpose of segregation analysis is to use familial phenotype data to resolve major loci predisposing to a trait and estimate the most likely parameters of the mode of inheritance. Although computationally demanding, it is now possible to set up models that include more than one mode of inheritance providing the family structures have sufficient information. Classical segregation analysis has no requirement for observed genotypes. It can be viewed as a special case of investigation of familial aggregation which often focuses on the pattern of aggregation within individual families rather than averaging across the population. The results of a segregation analysis can be very sensitive to inappropriate adjustment for ascertainment (73).

Segregation analysis for ADs alone and together with other ADs has been implemented. A single major gene has been hypothesized to confer susceptibility for autoimmunity indicating that a postulated autoimmune gene is expressed as an autosomal dominant trait with penetrance of approximately 92% in females and 49% in males (44). Furthermore, others have demonstrated the presence of a dominant major gene and strong environmental effects as the most parsimonious model of segregation for VIT (45). However, when analyzing RA together with other ADs, a mixed model fit the data significantly better than the major gene or polygenic models. (46) The above-presented models are initial approaches towards unraveling the dynamics of the polygenic component among families presenting the autoimmunity trait. Still, analyses combining the presence of autoimmune family history that take autoimmunity as a trait remain to be carried outcome.

Genetics of autoimmunity

As multifactorial etiologies, ADs develop from the cumulative effect of diverse events on the immune system (59). It is now clear that they do not begin at the time of the clinical appearance but rather many years before (Figure 2). This implies the possibility of predicting autoimmunity (74). A common origin for diverse ADs is sustained by three levels of evidence (9). The first comes from clinical observations indicating the possible shift from one disease to another and the fact that more than one AD may coexist in a single patient (i.e., polyautoimmunity) (8,75-77), or in the same family (i.e., familial autoimmunity) (Figure 3) (78). The second refers to known pathophysiological mechanisms shared between ADs, and the third corresponds to the evidence implying common genetic factors (77). The importance of this concept focuses on the probability of having multiple ADs simultaneously in one patient, which goes beyond epidemiological inferences. Therefore, family history of ADs should be considered when doing genetic analysis as this new approach incorporates all accepted pathologies for which evidence suggests an autoimmune origin.

Figure 3

General models proposed to explain the genetic component of complex disease. Square: Male; Circle: Female; filled circle and square depicts affected. Adapted from Wright et al. (106).

The human genome is distributed among 46 chromosomes, 22 homologous pairs of autosomes, and one pair of sex chromosomes. The complete set is the diploid complement, while one set is the haploid (i.e., gametic) genome. One chromosome in each of the 22 homologous pairs is derived from the mother and one from the father. The two homologues will have the same sequence of genes in the same positions, but they will usually exhibit sequence variations at several loci and can therefore be distinguished. The human haploid genome is about 3·3 billion bp. Some 3% of the genome consists of coding sequences, and there are 30,000–40,000 protein-coding genes. Any two genomes typically differ by 0.1%, but the DNA sequence may vary between the two versions of the same chromosome several ways (79).

Many different types of DNA sequence variants exist, and they can be classified different ways (e.g., by the physical nature of the sequence variation, by the effect on protein formation, and by the associated susceptibility to a disease). The two most important structural classes are microsatellites and single nucleotide polymorphisms (SNPs). Microsatellites are highly variable, and most people are heterozygous at any given locus. Coding regions tend to not contain microsatellite sequences. SNPs, in contrast, represent variations in a single nucleotide that occurs with high frequency in the human genome (13).

The modern unit of genetic variation is the SNP, typically used as markers of a genomic region with the large majority of them having a minimal impact on biological systems. SNPs in protein-coding regions can be non-synonymous (i.e., missense) or synonymous (i.e., sense), depending on whether they do or do not modify the amino acid sequence in the gene product. Non-synonymous SNPs can also be called coding SNPs. Intronic and intergenic SNPs lie in the non-coding regions. A non-synonymous SNP in a coding sequence is generally more likely than other classes of SNP to affect the function or availability of a protein. The true distribution of disease-associated variants between non-coding and coding sequences is unknown; however, they are by far the most abundant form of genetic variation in the human genome (80).

Common variants (frequency >1%) in the population account for most of the genetic variation observed in the genome (79,81). Rare variants, each found in <1% of the population, make up the remainder of genetic polymorphism but are far more numerous than common variants in total counts given the large size of the human population. In all, 10–15 million common variants and billions of rare sequence variants constitute the human genetic diversity which makes the identification of those variants relevant to particular ADs challenging (82).

Models for complex diseases

Currently, three general models have been proposed to explain the genetic component of complex disease (83): the common disease/common variant (CD/CV) model, the rare alleles of major effect (RAME, Multilocus/multiallele hypothesis) model, and the infinitesimal model (Figure 4). These are neither mutually exclusive nor sufficiently precisely defined to allow them to be distinguished in any particular case, but they provide an essential conceptual background. However, it is also debatable whether they are collectively sufficient. The evidence for CD/CV is underwhelming, RAME is difficult to reconcile as a general explanation for familial risk distributions, and infinitesimal liability does not lend itself to reductionist genetic dissection.

Figure 4

Familial autoimmunity. The vertical axis corresponds to the proband’s disease and each disease individually. On the horizontal axis, diseases present in first degree relatives are shown. Each color belongs to the proband’s disease. The (more...)

The evidence now suggests that genetic susceptibility to common disease is probably due to hundreds or even thousands of alleles that may or may not be rare, that are at least as often ancestral as derived, and that can vary in frequency among human populations. Evidence for epistatic interaction between risk alleles and their role in ADs has started to gain more focus (84). Finally, some diseases yield to GWAS studies more readily than others, which suggests different levels of genetic buffering (85). Below is a brief description of the models that have been proposed to explain the genetic contribution to disease.

1. Common disease/common variant (CD/CV) model (Figure 4) According to the CD/CV model, common disease susceptibility is caused by common polymorphisms, each of which makes a moderate contribution but generally explains no more than five percent of disease susceptibility. By analogy with QTL mapping, it is assumed that up to 20 or so common variants (each with a minor-allele frequency greater than 0.05) are associated with any particular disease, and that affected individuals carry an excess of the risk alleles (see chapters 17 and 18). Note that it is not clear whether the SNPs detected in GWAS studies typically capture common variants or a set of rare major-effect variants that are enriched on the susceptibility haplotype (83).
2. The rare alleles of major effect (RAME) model: This model postulates that some common diseases are actually highly heterogeneous with respect to their etiology (Figure 4). That is, rare variants with frequencies of less than 0.01 can promote disease. Hundreds of such polymorphisms, with homozygous individuals for each polymorphism occurring at a frequency of just 1 in 10,000, could add up to an appreciable incidence in the general population. Double heterozygous interactions could further increase the susceptible population, possibly to common disease status (83). Despite this expectancy, the characterization of causal rare variants has not, so far, been able to explain the observed effects as being the result of common tag variants mapping close to the discovered rare variants. This gives further support to the CD/CV model attributed to common variants with weak effects as disease defining factors (86).
3. The infinitesimal model: The infinitesimal model is gaining in popularity as GWAS reveal that most genetic variation for complex diseases must be due to variants that have relative risks that are less than 1.2 and explain a fraction of a percent of the liability. As disease is often regarded as a discrete state, it is often assumed that a threshold of liability separates cases from controls. However, the need for a threshold assumption has been questioned, at least in relation to autoimmune disorders. If disease liability follows the same distribution of effects as the best-characterized continuous trait, height, then hundreds if not thousands of variants covering a wide range of frequencies are likely to contribute to common diseases (83). It should also be noted that some subphenotypes have stronger genetic factors than the phenotype (i.e., the disease) itself. The infinitesimal model does not immediately explain how susceptibility becomes enriched in certain pedigrees, or why the prevalence of disease has increased so markedly in recent decades.

Genetic correlations

After obtaining evidence of a likely genetic component in the cause of a complex disease (without genotyping genes), the next step is to locate and identify any causative genes (Figure 1). However, for most complex diseases, there are so many candidates and so many genes whose usual effects are completely unknown that candidate gene work is often preceded or accompanied by an attempt to locate regions of the genome that are etiologically relevant. Regardless of assumptions about the genetic model of a trait, or the technology used to assess genetic variation, no genetic study will have meaningful results without a thoughtful approach to characterize the phenotype of interest (87). When embarking on a genetic study, the initial focus should be on identifying precisely what quantity or trait genetic variation influences. The commonly used strategies to find variants that influence disease are linkage and association studies. They are described below.

Genetic linkage Analysis. Major genes for monogenic conditions have been located by linkage analysis (3), but there have been far fewer successes with complex diseases. Genetic linkage analysis is perhaps the best example of a common investigative approach that derives almost entirely from a consideration of the underlying genetics since it relies entirely on the tendency for shorter haplotypes to be passed on to the next generation intact, without recombination events at meiosis. If a marker can be identified that is passed down through a family such that it consistently accompanies the disease of interest, this suggests a gene with a functional effect that is located close to that marker (88).

It is noteworthy that the modified use of traditional linkage approaches remains a useful tool for the study of polygenic diseases, especially if a major locus that contributes to the phenotype is known. It has been observed that, in some cases genetic loci overlap or co-localize between related disorders. Becker et al., based on previous AD linkage studies, first reported eighteen common non-major histocompatibility complex (MHC) loci clusters in 1998 and also hypothesized a shared and common genetic basis for the autoimmune trait (36). Tomer et. al., working with both Hashimoto’s thyroiditis and Grave’s disease in the same analyses reported linkage for AITD (89). Other studies which considered the presence of linkage for specific diseases have found shared autoimmunity loci (9,90). Limitations of genome-wide scans when applied to complex ADs, involve heterogeneity in disease phenotypes, population and ethnic differences, and lack of statistical and analytical models (9).

Association analysis. This approach can be seen as a traditional epidemiology approach applied to genotypes or alleles across a population. Thus, many of the analytical approaches used in epidemiology and biostatistics can be applied directly to the evaluation of associations. Among these are univariate methods and regression analysis. Furthermore, analysis can be extended to deal with data that have a complex correlation structure including: family data, longitudinal data, data naturally subject to geographical or temporal clustering, and/or data collected under a multistage sampling plan and applied to phenotypes in various classes, including binary traits, continuous normally distributed traits, and time to event (survival time) (13). A test of association can be informative even when based on genetic variants that are not functional. It can also be useful to detect linkage disequilibrium (LD) (BOX 1) between disease and a non-functional marker. An association analysis based on a putative functional genetic variant can be called direct, and one based on linkage disequilibrium with a marker indirect. Indirect association analysis allows finer mapping than conventional linkage analysis.

Despite the enormous diversity of genetic variants associated with susceptibility/protection from individual to individual with the same AD, a high degree of similarity is observed between ADs in the pool of genetic variants associated with disease (see chapters 17 and 18). These genetic associations can be dissected into three biologically distinct classes.

1. HLA associations: for most ADs, the strongest genetic associations are with the human leukocyte antigen (HLA) locus. Variations in the HLA region were the first polymorphisms investigated in association studies and turned out to play a major role in most ADs even though precise understanding of the effects is still under investigation owing to the highly polymorphic nature, exceptional LD, and high gene density of this region. Although some linked HLA haplotypes are shared between ADs, most HLA associations seem to be specific for a disease (82). For a complete description of HLA association the reader is directed to chapter 17.
2. Common autoimmune risk variants: Many of the strong loci that have been associated with one AD are also found to be involved in multiple other ADs (See chapter 18). Typically, the allele of these shared variants that causes risk in one AD also causes risk in other ADs, indicating that the alleles participate in a shared immunological process common to the development of multiple, clinically distinct ADs. A degree of complexity exits in which precise mechanisms may differ between subgroups of ADs (92,93).
3. Disease-Specific risk variants: A smaller number of associations are observed that are specific to a single AD with no measured impact on other ADs. This suggests that they might drive a target organ-specific pathway toward disease.

Genome-wide associations studies (GWAS)

Debate has recently erupted in the field of genetics between the CD/CV and RAME variant hypotheses. GWAS are designed to study common variants typically present at an allele frequency of more than 5% and have been exceedingly successful in doing this for ADs (see chapter 18). GWAS exploit LD between SNPs and make it possible to assay a manageable number of variants while still capturing the majority of variation in a given population’s genome. GWAS have been very successful in the identification of novel loci and pathways contributing to AD (See chapter 18). However, the effect sizes reported are usually rather modest, with odds ratios (OR) typically between 1.1 and 1.8. Note that most of the genes identified to date affect more than one autoimmune condition. One interesting observation has been that many genetic loci appear to harbor variants that are associated with multiple, sometimes seemingly distinct traits, and such associations are termed cross-phenotype associations (94). Cross-phenotype association differs from pleiotropy. A cross-phenotype association occurs when a genetic locus is associated with more than one trait in a study, regardless of the underlying cause for the observed association. Pleiotropy occurs when a genetic locus truly affects more than one trait and is one possible underlying cause for an observed cross-phenotype association [for complete review see Solovieff et al. (93)].

Genetic studies have led to important conclusions on the genetic architecture of ADs. A plausible disease trait scenario –multiple environmental variants interacting with several genes to conferred susceptibility on individuals in a population– has been projected (83). According to this theory, individuals would express the disease trait if they were located on the wrong side of the normal distribution. This concept has been widely accepted, but additivity must be assumed for all genetic variants accompanied by an equal effect on the trait. Autoimmunity might involve a genetic distribution that is not as straightforward as a normal distribution but rather an unknown one in which many loci would not add to but complement the individual’s risk of developing the autoimmune phenotype (Figure 3b).

For many and perhaps most traits, the interaction with other genes and environmental factors might be genetically programmed or may be purely stochastic. Most common diseases probably contain major subsets that fall into this sort of causation. Even when the phenotypical manifestation is usually considered to be monogenic, it could be the result of gene-environment interaction (95). This does not mean impossibility in the modeling of a complex trait but a warrant towards generalization.

Mathematical models which assume a continuous distribution of liability to disease in the general population have been employed to explain observed “non-Mendelian” patterns of familial occurrence (96,97). Polygenic determination refers to the mathematical model in which a number of genes with small additive effects provide an underlying genetic predisposition to disease development. Likewise, the term multifactorial describes models in which environmental factors interact with genetic predisposition. This multifactorial model was adapted to account for discontinuous traits by the addition of a threshold, the point of risk distribution beyond which individuals are affected (Figure 3). Moreover, a number of observations in human population and experimental animals have shown consistency with the model of multifactorial inheritance (24,95). The relative merits of this hypothetical model (polygenic multifactorial-threshold and major single gene with incomplete penetrance) have been debated vigorously (95).

Families with multiple affected relatives appear to share common risk alleles with sporadic patients but may have a higher genetic load. A consequence of the polygenic model for complex diseases is that patients are inevitably highly heterogeneous in terms of the particular set of risk alleles they carry. It has been suggested that this may translate into different genetically determined disease mechanisms in subgroups of patients or a common disease mechanism that is complemented by additional pathways that are more or less predominant in different subgroups (82).