**1. Introduction**

#### **1.1 Diagnosis of ASD**

Autism (also known as classic autism or autism disorder) is a common neurodevelopmental disorder. Typically diagnosed before three years old, autistic children usually present with significant language delays, social and communication impairments, as well as abnormal repetitive and restrictive behaviors. Autism spectrum disorders (ASD) however, refers to a boarder definition of autism. Based on the severity of the clinical conditions, ASD is further divided into three subgroups namely autism (the most severe type of ASD), Asperger syndrome and pervasive developmental disorder – not otherwise specified (PDD-NOS; also called atypical autism) [1-3].

Of note, current diagnosis criteria of these disorders are based on behavior tests, no single biomarker has been clinically accepted, which mainly due to the difficulties for studying cellular and molecular etiology of ASD. First, subjects among different researches lack of comparability because of the diagnostic heterogeneity [4]. Second, the prevalence of ASD is relatively low therefore sample sizes are usually too small for statistical analysis. Third, comparing with other diseases, the young ages of the autistic subjects make biological study difficult. Forth, valid control groups require age-, gender-, IQ- and socioeconomic statusmatched developmentally normal subjects, which most studies failed to satisfy with [5].

#### **1.2 Epidemiology**

ASD is reported to occur in all racial, ethnic and socioeconomic groups, and are about four times more likely to occur in boys than in girls probably due to the extremes of typical male neuroanatomy of autism [6, 7]. Studies in Asia, Europe and North America have identified individuals with ASD with an approximate prevalence of 6/1,000 to over 10/1,000 [8]. Chronologically, the prevalence of ASD increased from 0.8/1,000 in 1983 to 4.6/1,000 in 1999 in Western Australia, while this ratio increased from 6.6/1,000 in 2000 to 9/1,000 in 2006 in United States [9-11]. This increase is probably because of changes and broadening of the diagnostic criteria and due to heightened awareness, but may also reflect, in part, a true increase due to environmental factors acting upon a genetically vulnerable background [12, 13].

#### **2. Immune disorders and autism**

The relationship between immune disorders and ASD has been proposed for decades. Based on the epidemiological data, higher rate of autoimmune conditions, such as rheumatoid

Autoimmune Disorder and Autism 519

and IL-12p40 in the ASD group compared with typically developing controls [32]. All these findings suggest that inflammatory responses may be related to disturbances in behavior. And the characterization of immunological parameters in ASD has important implications for diagnosis, therefore should be considered when designing therapeutic strategies to treat

Using human fetal and adult brains as antigenic substrates, maternal serum antibodies transferred through placenta are detected by four independent research groups, suggesting an association between the transfer of IgG autoantibodies during early neurodevelopment

Singh et al provided more confirmative evidence by studying regional distribution of antibodies to rat caudate nucleus, cerebral cortex, cerebellum, brain stem and hippocampus of 30 normal and 68 autistic children. Autistic children, but not normal children, had antibodies to caudate nucleus (49% positive sera), cerebral cortex (18% positive sera) and cerebellum (9% positive sera). Brain stem and hippocampus were negative. Since a significant number of autistic children had antibodies to caudate nucleus, the author proposed that an autoimmune reaction to this brain region may cause neurological impairments in autistic children [38]. Agreed with this result, Trajkovski et al measured plasma concentration of IgA, IgM, IgG classes, and IgG1, IgG2, IgG3, and IgG4 subclasses in children with autism. Plasma concentrations of IgM and IgG in autistic children were significantly higher in comparison with their healthy brothers or sisters. Children with autism had significantly higher plasma concentrations of IgG4 compared to their siblings. Increased plasma concentration of IgG1 was found in autistic males as compared with their healthy brothers. Plasma concentrations of IgG and IgG1 in autistic females were increased in comparison with IgG and IgG1 in their healthy sisters [39]. More recently, Enstrom et al report significantly increased levels of the IgG4 subclass in children with autism compared with typically developing control children and compared with developmental delayed

However, No consensus has been reached regarding the immunoglobin levels in autistic subjects. Morris and colleagues failed to find any useful biomarker in a small group of subjects, posing question to the current theory [41]. Stern et al found in their study most of the autistic children had normal immune function, suggesting that routine immunologic

The report regarding the relationship between autism and gastrointestinal disorders was seen as early as 1971, when Goodwin et al described 6 of 15 randomly selected autistic children with symptoms of malabsorption [43]. Later Horvath et al investigated 412 autistic children, of which 84.1% had at least one of the eight abnormal gastrointestinal symptoms, comparing with 31.2% of the healthy siblings [44]. However, disagreements exit. Kuddo group and Molloy group failed to find any association between chronic gastrointestinal symptoms and autism based on the literature search or their own sample [45, 46]. Fernell et al tested two independent biomarkers of inflammatory reactions (faecal calprotectin and rectal nitric oxide) in 24 autistic children, but didn't find clear link between active intestinal

investigation is unlikely to be of benefit in most autistic children [42].

and the risk of developing of autism in some children [33-37].

ASD.

**2.3 Immunoglobulin** 

controls [40].

**2.4 Gastrointestinal disorders** 

inflammation and autism [47].

arthritis, autoimmune thyroid disease, asthma, ulcerative colitis, exits in parents of autistic children [14-17]. Another line of evidence supporting immune dysfunction at least partly responsible for ASD comes from large population studies, which suggest maternal immune dysfunctions may be related to a later diagnosis of ASD in the offspring [18]. Furthermore, cumulative evidences support the theory that ASD is caused by a loss of self-tolerance to one or more neural antigens during early childhood. Using western blot for the presence of IgG antibodies against protein extracts from human brain or sera, multiple brain-specific autoantibodies are detected [19, 20]. Other groups measured the plasma concentration of immunoglobulins and/or cytokines, autistic subjects exclusively exhibited abnormal immunoglobin and/or cytokine profiles [21-24]. It's not known yet whether immune activation plays an initiating or ongoing role in the pathology of ASD. But investigations of dynamic adaptive cellular immune function suggested dysfunctional immune activation, which may be linked to disturbances in behavior and developmental functioning [25].

### **2.1 Autoimmune diseases**

Autoimmune diseases are the most common type of immune disorders. And its relationship with autism has been widely studied. Very early study reported an increased number of autoimmune disorders in some families with autism, suggesting immune dysfunction plays a role in autism pathogenesis [26]. Consistent with this result, Sweeten et al investigated the frequency of autoimmune disorders in families that have probands with pervasive developmental disorders and autism, compared with control groups. Autoimmunity was increased significantly in families with pervasive developmental disorders compared with those of healthy and autoimmune control subjects [27]. More persuasive evidence comes from a multicenter study of 308 children with Autism Spectrum Disorder. Regression was significantly associated with a family history of autoimmune disorders. But the only specific autoimmune disorder found to be associated with regression was autoimmune thyroid disease [28].

#### **2.2 Cytokines and chemokines**

Cytokines and chemokines are thought to mediate the pathogenesis of autism, although the exact mechanism remains unclear. Jyonouchi group determined innate and adaptive immune responses in children with developmental regression and autism spectrum disorders, developmentally normal siblings, and controls. Their results indicated excessive innate immune responses in a number of ASD children that may be most evident in TNFalpha production [29]. Similarly, Molloy et al reported children with ASD had increased activation of both Th2 and Th1 arms of the adaptive immune response, with a Th2 predominance, and without the compensatory increase in the regulatory cytokine IL-10 [30]. But Li et al showed that proinflammatory cytokines (TNF-alpha, IL-6 and GM-CSF), Th1 cytokine (IFN-gamma) and chemokine (IL-8) were significantly increased in the brains of ASD patients compared with the controls, but not the Th2 cytokines (IL-4, IL-5 and IL-10). The Th1/Th2 ratio was also significantly increased in ASD patients. Based on these results, the author concluded that ASD patients displayed an increased innate and adaptive immune response through the Th1 pathway, suggesting that localized brain inflammation and autoimmune disorder may be involved in the pathogenesis of ASD [31]. Most recently, Ashwood group used larger number of participants than previous studies and found that significant increases in plasma levels of a number of cytokines, including IL-1beta, IL-6, IL-8

arthritis, autoimmune thyroid disease, asthma, ulcerative colitis, exits in parents of autistic children [14-17]. Another line of evidence supporting immune dysfunction at least partly responsible for ASD comes from large population studies, which suggest maternal immune dysfunctions may be related to a later diagnosis of ASD in the offspring [18]. Furthermore, cumulative evidences support the theory that ASD is caused by a loss of self-tolerance to one or more neural antigens during early childhood. Using western blot for the presence of IgG antibodies against protein extracts from human brain or sera, multiple brain-specific autoantibodies are detected [19, 20]. Other groups measured the plasma concentration of immunoglobulins and/or cytokines, autistic subjects exclusively exhibited abnormal immunoglobin and/or cytokine profiles [21-24]. It's not known yet whether immune activation plays an initiating or ongoing role in the pathology of ASD. But investigations of dynamic adaptive cellular immune function suggested dysfunctional immune activation, which may be linked to disturbances in behavior and developmental functioning [25].

Autoimmune diseases are the most common type of immune disorders. And its relationship with autism has been widely studied. Very early study reported an increased number of autoimmune disorders in some families with autism, suggesting immune dysfunction plays a role in autism pathogenesis [26]. Consistent with this result, Sweeten et al investigated the frequency of autoimmune disorders in families that have probands with pervasive developmental disorders and autism, compared with control groups. Autoimmunity was increased significantly in families with pervasive developmental disorders compared with those of healthy and autoimmune control subjects [27]. More persuasive evidence comes from a multicenter study of 308 children with Autism Spectrum Disorder. Regression was significantly associated with a family history of autoimmune disorders. But the only specific autoimmune disorder found to be associated with regression was autoimmune thyroid

Cytokines and chemokines are thought to mediate the pathogenesis of autism, although the exact mechanism remains unclear. Jyonouchi group determined innate and adaptive immune responses in children with developmental regression and autism spectrum disorders, developmentally normal siblings, and controls. Their results indicated excessive innate immune responses in a number of ASD children that may be most evident in TNFalpha production [29]. Similarly, Molloy et al reported children with ASD had increased activation of both Th2 and Th1 arms of the adaptive immune response, with a Th2 predominance, and without the compensatory increase in the regulatory cytokine IL-10 [30]. But Li et al showed that proinflammatory cytokines (TNF-alpha, IL-6 and GM-CSF), Th1 cytokine (IFN-gamma) and chemokine (IL-8) were significantly increased in the brains of ASD patients compared with the controls, but not the Th2 cytokines (IL-4, IL-5 and IL-10). The Th1/Th2 ratio was also significantly increased in ASD patients. Based on these results, the author concluded that ASD patients displayed an increased innate and adaptive immune response through the Th1 pathway, suggesting that localized brain inflammation and autoimmune disorder may be involved in the pathogenesis of ASD [31]. Most recently, Ashwood group used larger number of participants than previous studies and found that significant increases in plasma levels of a number of cytokines, including IL-1beta, IL-6, IL-8

**2.1 Autoimmune diseases** 

**2.2 Cytokines and chemokines** 

disease [28].

and IL-12p40 in the ASD group compared with typically developing controls [32]. All these findings suggest that inflammatory responses may be related to disturbances in behavior. And the characterization of immunological parameters in ASD has important implications for diagnosis, therefore should be considered when designing therapeutic strategies to treat ASD.

#### **2.3 Immunoglobulin**

Using human fetal and adult brains as antigenic substrates, maternal serum antibodies transferred through placenta are detected by four independent research groups, suggesting an association between the transfer of IgG autoantibodies during early neurodevelopment and the risk of developing of autism in some children [33-37].

Singh et al provided more confirmative evidence by studying regional distribution of antibodies to rat caudate nucleus, cerebral cortex, cerebellum, brain stem and hippocampus of 30 normal and 68 autistic children. Autistic children, but not normal children, had antibodies to caudate nucleus (49% positive sera), cerebral cortex (18% positive sera) and cerebellum (9% positive sera). Brain stem and hippocampus were negative. Since a significant number of autistic children had antibodies to caudate nucleus, the author proposed that an autoimmune reaction to this brain region may cause neurological impairments in autistic children [38]. Agreed with this result, Trajkovski et al measured plasma concentration of IgA, IgM, IgG classes, and IgG1, IgG2, IgG3, and IgG4 subclasses in children with autism. Plasma concentrations of IgM and IgG in autistic children were significantly higher in comparison with their healthy brothers or sisters. Children with autism had significantly higher plasma concentrations of IgG4 compared to their siblings. Increased plasma concentration of IgG1 was found in autistic males as compared with their healthy brothers. Plasma concentrations of IgG and IgG1 in autistic females were increased in comparison with IgG and IgG1 in their healthy sisters [39]. More recently, Enstrom et al report significantly increased levels of the IgG4 subclass in children with autism compared with typically developing control children and compared with developmental delayed controls [40].

However, No consensus has been reached regarding the immunoglobin levels in autistic subjects. Morris and colleagues failed to find any useful biomarker in a small group of subjects, posing question to the current theory [41]. Stern et al found in their study most of the autistic children had normal immune function, suggesting that routine immunologic investigation is unlikely to be of benefit in most autistic children [42].

#### **2.4 Gastrointestinal disorders**

The report regarding the relationship between autism and gastrointestinal disorders was seen as early as 1971, when Goodwin et al described 6 of 15 randomly selected autistic children with symptoms of malabsorption [43]. Later Horvath et al investigated 412 autistic children, of which 84.1% had at least one of the eight abnormal gastrointestinal symptoms, comparing with 31.2% of the healthy siblings [44]. However, disagreements exit. Kuddo group and Molloy group failed to find any association between chronic gastrointestinal symptoms and autism based on the literature search or their own sample [45, 46]. Fernell et al tested two independent biomarkers of inflammatory reactions (faecal calprotectin and rectal nitric oxide) in 24 autistic children, but didn't find clear link between active intestinal inflammation and autism [47].

Autoimmune Disorder and Autism 521

However, unlike monogenic Mendelian disorders, the genetic and clinical heterogeneity of ASD poses a difficult challenge to precisely define the underlying genetics. This complexity has been blamed for the lack of replicability of the many reported chromosomal susceptibility regions. Therefore, multiple parallel approaches are needed for the exploration of the potential loci underlying the etiology of ASD. In general, there are a number of methods available for genetic studies of ASD, with each having different advantages as well as limitations. The most widely used methods include cytogenetic analysis, linkage and association studies, copy number variation and DNA micro-array

A cytogenetic study is the most "classic" of genetic methods. Based on the assumption that ASD is a result of unique rare mutations that present sporadically or "de novo" in the population and are not usually inherited, cytogenetics helps to determine the contribution of chromosomal abnormalities in childhood diseases. Cytogenetics has transitioned from light microscopy to molecular cytogenetics to DNA-based microarray detections of structural variations [60]. Copy number variation (CNV) analysis is a newer molecular cytogenetic approach, aiming to detect the insertion or deletion of DNA fragments typically larger than 50 kb [61]. However, extreme caution must be paid when interpreting CNV analysis since it is very dependent on the specific methods employed, which may partly account for the low

Differing from cytogenetics, linkage studies trace genetic loci that are transmitted with autism in the families of affected individuals. Parametric and non-parametric linkage studies are two typical designs. While parametric analysis requires a model for the disease (i.e. frequency of disease alleles and penetrance for each genotype), and therefore is typically employed for single gene disorders and Mendelian forms of complex disorders, "model-free" non-parametric linkage analysis evaluates whether segregation at specific locations is "not-random". Given the uncertainty of the mode of inheritance in ASD, nonparametric linkage is more widely used, providing suggestive evidence of linkage on almost all of the chromosomes [63]. However, linkage studies are unable to identify mutations in critical genes in highly heterogeneous disorders involving many different genes and

Genetic association studies, including case-control and family-based studies, examine differences in allele or genotype frequencies between two groups [63]. Typically, several microsatellite markers or SNPs are chosen based on linkage studies or biological evidence. The seemingly countless potential candidates make it hard to determine the causative relations between genes and ASD [61]. In addition, although association studies are suitable to identify common susceptibility alleles present in large numbers of patients compared to

Rapid advances in micro-array technologies have substantially improved our ability to detect submicroscopic chromosomal abnormalities. These tools have allowed for highoutput and high-resolution detection of rare and de novo changes in a genome-wide manner. Moreover, newly developed, commercially available whole-exome arrays are increasingly being employed to detect de novo mutations in complex disorders. Based on the fact that the protein coding regions of genes (i.e. exons) habor 85% of the mutations of disease-related traits, whole-exome sequencing offers the possibility to identify diseasecausing sequence variations in small kindreds for phenotypically complicated, genetically heterogeneous diseases when traditional linkage studies are impossible [65-69]. As such,

controls, they usually fail to identify rare, causal mutations [63, 64].

**3.1 Genetic methodology** 

replicability among studies [62].

chromosomal loci [64].

analysis.

Morphological and histological studies provided consistent results with the clinical manifestations. Ileocolonoscopic examinations in 60 children with autism and other developmental disorders revealed that 8% (4/51) affected children but none in controls presented with active ileitis. Chronic colitis was identified in 88% (53/60) affected children compared with 4.5% (1/22) controls [48]. Similarly, another group conducted upper gastrointestinal endoscopy in 36 autistic subjects. 69.4% (25/36) of whom presented with grade I or II reflux esophagitis, 41.7% (15/36) with chronic gastritis, and 66.7% with chronic duodenitis [49].

In addition, biochemical researches reported evidences of abnormal intestinal cytokine profiles. Ashwood et al found enhanced pro-inflammatory cytokine production present in 21 ASD children compared with 65 controls [50]. Furthermore, they investigated the peripheral blood and mucosal CD3+ lymphocyte cytokine profiles in 18 autistic children with gastrointestinal symptoms. In both peripheral blood and mucosa, CD3+ TNFalpha+ and CD3+ IFNgamma+ were increased, while CD3+ IL-10+ were markedly lower in ASD children. And mucosal CD3+ IL-4+ cells were increased in ASD compared with NIC [51]. Similarly, Jyonouchi et al provided evidence that intrinsic defects of innate immune responses in ASD children with gastrointestinal symptoms, suggesting a possible link between GI and behavioral symptoms mediated by innate immune abnormalities [52]. However, DeFelice et al assessed levels of proinflammatory cytokines, interleukin (IL)-6, IL-8, and IL-1beta, produced by intestinal biopsies of children with pervasive developmental disorders but failed to find significant difference between autistic and control groups [53].

How do the gastrointestinal disorders affect brain functions? Currently available pathophysiological studies provided partial explanations. D'Eufemia et al investigated the occurrence of gut mucosal damage using the intestinal permeability test in 21 autistic children without known intestinal disorders. They found increased intestinal permeability in 43% (9/21) autistic patients, but in none of the 40 controls, which suggested an altered intestinal permeability could represent a possible mechanism for the increased passage through the gut mucosa of peptides derived from foods with subsequent behavioral abnormalities [54].

### **3. Genetics of autism**

Similar to several other complex diseases, autism was not widely considered to have a strong genetic component until the 1980s. But increasing numbers of epidemiological and genetic studies are deepening our understanding of the genetic contribution autism. First, it is estimated that about 10% of children with ASD have an identifiable co-occuring genetic, neurologic or metabolic disorder, such as the fragile X syndrome and tuberous sclerosis [55]. Second, the relative risk of a newborn child to have autism, if he or she has an affected sibling, increases at least 25 folds comparing with general population [56]. Third, independent twin studies have suggested identical twins have a 60-90% chance to be concordantly diagnosed with autism, and this risk decreases sharply to the sibling risk of 0- 24% in non-identical twins [57, 58]. However, based on a large scale study of 503 ASD twins in California, Liu *et al* suggest the heritability has been largely overestimated [59]. They found the concordance rate for monozygotic male twins was 57% and for females 67%, while for same sex dizygotic twins the rate was 33%. Fourth, cumulative reports have confirmed mutations or structural variations of a number of specific genes significantly increase the risk of ASD [56].

#### **3.1 Genetic methodology**

520 Autoimmune Disorders – Current Concepts and Advances from Bedside to Mechanistic Insights

Morphological and histological studies provided consistent results with the clinical manifestations. Ileocolonoscopic examinations in 60 children with autism and other developmental disorders revealed that 8% (4/51) affected children but none in controls presented with active ileitis. Chronic colitis was identified in 88% (53/60) affected children compared with 4.5% (1/22) controls [48]. Similarly, another group conducted upper gastrointestinal endoscopy in 36 autistic subjects. 69.4% (25/36) of whom presented with grade I or II reflux esophagitis, 41.7% (15/36) with chronic gastritis, and 66.7% with chronic

In addition, biochemical researches reported evidences of abnormal intestinal cytokine profiles. Ashwood et al found enhanced pro-inflammatory cytokine production present in 21 ASD children compared with 65 controls [50]. Furthermore, they investigated the peripheral blood and mucosal CD3+ lymphocyte cytokine profiles in 18 autistic children with gastrointestinal symptoms. In both peripheral blood and mucosa, CD3+ TNFalpha+ and CD3+ IFNgamma+ were increased, while CD3+ IL-10+ were markedly lower in ASD children. And mucosal CD3+ IL-4+ cells were increased in ASD compared with NIC [51]. Similarly, Jyonouchi et al provided evidence that intrinsic defects of innate immune responses in ASD children with gastrointestinal symptoms, suggesting a possible link between GI and behavioral symptoms mediated by innate immune abnormalities [52]. However, DeFelice et al assessed levels of proinflammatory cytokines, interleukin (IL)-6, IL-8, and IL-1beta, produced by intestinal biopsies of children with pervasive developmental disorders but failed to find significant difference between autistic and control groups [53]. How do the gastrointestinal disorders affect brain functions? Currently available pathophysiological studies provided partial explanations. D'Eufemia et al investigated the occurrence of gut mucosal damage using the intestinal permeability test in 21 autistic children without known intestinal disorders. They found increased intestinal permeability in 43% (9/21) autistic patients, but in none of the 40 controls, which suggested an altered intestinal permeability could represent a possible mechanism for the increased passage through the gut mucosa of peptides derived from foods with subsequent behavioral

Similar to several other complex diseases, autism was not widely considered to have a strong genetic component until the 1980s. But increasing numbers of epidemiological and genetic studies are deepening our understanding of the genetic contribution autism. First, it is estimated that about 10% of children with ASD have an identifiable co-occuring genetic, neurologic or metabolic disorder, such as the fragile X syndrome and tuberous sclerosis [55]. Second, the relative risk of a newborn child to have autism, if he or she has an affected sibling, increases at least 25 folds comparing with general population [56]. Third, independent twin studies have suggested identical twins have a 60-90% chance to be concordantly diagnosed with autism, and this risk decreases sharply to the sibling risk of 0- 24% in non-identical twins [57, 58]. However, based on a large scale study of 503 ASD twins in California, Liu *et al* suggest the heritability has been largely overestimated [59]. They found the concordance rate for monozygotic male twins was 57% and for females 67%, while for same sex dizygotic twins the rate was 33%. Fourth, cumulative reports have confirmed mutations or structural variations of a number of specific genes significantly

duodenitis [49].

abnormalities [54].

**3. Genetics of autism** 

increase the risk of ASD [56].

However, unlike monogenic Mendelian disorders, the genetic and clinical heterogeneity of ASD poses a difficult challenge to precisely define the underlying genetics. This complexity has been blamed for the lack of replicability of the many reported chromosomal susceptibility regions. Therefore, multiple parallel approaches are needed for the exploration of the potential loci underlying the etiology of ASD. In general, there are a number of methods available for genetic studies of ASD, with each having different advantages as well as limitations. The most widely used methods include cytogenetic analysis, linkage and association studies, copy number variation and DNA micro-array analysis.

A cytogenetic study is the most "classic" of genetic methods. Based on the assumption that ASD is a result of unique rare mutations that present sporadically or "de novo" in the population and are not usually inherited, cytogenetics helps to determine the contribution of chromosomal abnormalities in childhood diseases. Cytogenetics has transitioned from light microscopy to molecular cytogenetics to DNA-based microarray detections of structural variations [60]. Copy number variation (CNV) analysis is a newer molecular cytogenetic approach, aiming to detect the insertion or deletion of DNA fragments typically larger than 50 kb [61]. However, extreme caution must be paid when interpreting CNV analysis since it is very dependent on the specific methods employed, which may partly account for the low replicability among studies [62].

Differing from cytogenetics, linkage studies trace genetic loci that are transmitted with autism in the families of affected individuals. Parametric and non-parametric linkage studies are two typical designs. While parametric analysis requires a model for the disease (i.e. frequency of disease alleles and penetrance for each genotype), and therefore is typically employed for single gene disorders and Mendelian forms of complex disorders, "model-free" non-parametric linkage analysis evaluates whether segregation at specific locations is "not-random". Given the uncertainty of the mode of inheritance in ASD, nonparametric linkage is more widely used, providing suggestive evidence of linkage on almost all of the chromosomes [63]. However, linkage studies are unable to identify mutations in critical genes in highly heterogeneous disorders involving many different genes and chromosomal loci [64].

Genetic association studies, including case-control and family-based studies, examine differences in allele or genotype frequencies between two groups [63]. Typically, several microsatellite markers or SNPs are chosen based on linkage studies or biological evidence. The seemingly countless potential candidates make it hard to determine the causative relations between genes and ASD [61]. In addition, although association studies are suitable to identify common susceptibility alleles present in large numbers of patients compared to controls, they usually fail to identify rare, causal mutations [63, 64].

Rapid advances in micro-array technologies have substantially improved our ability to detect submicroscopic chromosomal abnormalities. These tools have allowed for highoutput and high-resolution detection of rare and de novo changes in a genome-wide manner. Moreover, newly developed, commercially available whole-exome arrays are increasingly being employed to detect de novo mutations in complex disorders. Based on the fact that the protein coding regions of genes (i.e. exons) habor 85% of the mutations of disease-related traits, whole-exome sequencing offers the possibility to identify diseasecausing sequence variations in small kindreds for phenotypically complicated, genetically heterogeneous diseases when traditional linkage studies are impossible [65-69]. As such,

Autoimmune Disorder and Autism 523

familial CNVs may include candidate genes relevant to ASD where they are rare in the

1 1p34.2 Regulating Synaptic Membrane Exocytosis 3(*RIMS3)* [90] 2 2q [71, 72]

 2q31 *SLC25A12* [92] 2q24-2q33 *SLC25A12, CMYA3* [75] 2q24-2q33 *SLC25A12, STK39, ITGA4* [77] 2q34 Neuropilin-2 (*NRP2*) [74] 3 3q25-3q27 HTR3C [48] 5 5q31 Paired-like homeodomain transcription factor 1(*PITX1*) [78] 5p14.1 [79] 5p15 *SEMA5A* [80] 6 6q Abelson's Helper Integration 1 (*AHI1*) [88] 6q27 [80] 7 7q22.1-7q31 [93]

Neuronal cell adhesion molecule (*NRCAM*)

4-Aminobutyrate Aminotransferase (*ABAT),* 

16p11.2 [81, 82,

Array comparative genomic hybridization (aCGH) is the most widely used method for detection of CNVs. A seminal early report used aCGH, with a mean resolution of one probe every 35 kb, to study a sample of 264 ASD families. After validation by higher-resolution microarray scans, G-banded karyotype, FISH, and microsatellite genotyping, 17 *de novo*

17 17q11.2 [99] 19 19p13 [99] 20 20q13 [80] 22 22q13 *SHANK3* [89] X Xp22.11 *PTCHD1* [91]

Glutamate receptor, ionotropic, NMDA 2A *(GRIN2A)* 

7q32 NADH-ubiquinone oxidoreductase 1 alpha subcomplex

7q31-7q33 wingless-type MMTV integration site family member 2

11 11p12-p13 [76] 12 12q14 [87] 15 15q11-q13 Angelman syndrome gene (*UBE3A*) [85] 15q11-q13 [83] 15q13 Amyloid precursor protein-binding protein A2 (*APBA2* ) [84]

CREB-binding protein *(CREBBP),* 

2q31-2q33 *GAD1,STK17B,ABI2,CTLA4,CD28,NEUROD1,* 

7q31 Laminin Beta-1 (*LAMB1*),

Table 1. Loci identified by genome wide linkage analysis

**Loci Candidate genes Ref.** 

*PDE1A,HOXD1, DLX2* [73]

<sup>5</sup> *(NDUFA5)* [48]

(*WNT2*) [97]

[94, 95][96]

[98]

90]

normal population.

16 16p11-13

**Chromosome** 

studies in this realm have been increasing in the past several years and there will surely benefit the etiological diagnosis and genetic counseling of ASD in the near future [70].

#### **3.2 Potential loci in autism**

#### **3.2.1 Genome wide linkage analysis**

Although there is accumulating evidence supporting a genetic component to ASD, the specific genes involved have yet to be totally clarified. Genome-wide screening of autistic subjects and their first-degree relatives offers an attractive means to search for susceptibility genes. However there has been a disappointing lack of replication of many of the reported susceptibility regions. The reason for this could be due to the epistasis of many interacting genes. But it may also be due to the genetic and clinical heterogeneity present in ASD [71]. The noted effects of heterogeneity of the samples on the corresponding results, have led to attempts to decrease sample heterogeneity by various ways which include narrowing inclusion criteria and studies of specific, autism-related endophenotypes.

A substantial body of evidence has resulted from genome-wide screening for the susceptibility genes of ASD (table 1). Significant replicability has been found for several chromosomal loci including 2q, 5, 7q, 15q and 16p. Two studies provided suggestive evidence for linkage to chromosome 2q using a two-stage genome screen [71, 72], while association tests for specific candidate genes in the chromosome 2q31-q33 region led to negative results [73]. Additional support for the presence of susceptibility loci on chromosome 2q is given by overlapping positive linkage findings in four other independent genomic scans [74-77].

There are three reports about gene variants on chromosome 5. Philippi found strong association with autism for allelic variants of "paired-like homeodomain transcription factor 1" (*PITX1*), a key regulator of hormones within the pituitary-hypothalamic axis [78]. Two other groups used genome-wide linkage and association mapping studies to analyze chromosome 5 gene variations finding that SNPs located at 5p14.1 and 5q15 respectively were significantly associated with autism [79, 80].

Chromosome 16 linkage results have been fairly consistent in showing a peak at 16p11-13, which strongly suggested a gene in this region may contribute to the risk of ASD [81, 82]. 15q11-q13 is another frequently identified locus by linkage studies. Several genes located in this region have been intensively studied and some have provided very promising results [83-86]. But in all of these linkage reports there is a certain lack of reproducibility, and therefore they require further validation based on using a combination of several methods.

Besides these "hot spots", there are other reports regarding associations of other loci with ASD [80, 87-90], including some evidence of linkage to the X chromosome [91]. However, there is little overlap of these potential loci involving potential candidate genes, suggesting that the genetic background of ASD is full of complexity.

#### **3.2.2 Copy number variation (CNV)**

Rapid advances in genomic DNA microarray technologies have substantially improved our ability to detect submicroscopic chromosomal abnormalities. Novel rare variants have been detected in association with ASD and these can be either *de novo* or inherited. *De novo* or noninherited CNVs are found in 7%–10% of ASD samples from simplex families (having only one child affected, the majority), in 2%–3% from multiplex families, and in ∼1% in non-ASD controls. Further, about 10% of ASD subjects with *de novo* CNVs carry two or more CNVs [100-102]. Inherited CNVs reportedly are found in up to 50% of ASD subjects for whom one of the presumably normal parents also has the duplication/deletion. These

studies in this realm have been increasing in the past several years and there will surely benefit the etiological diagnosis and genetic counseling of ASD in the near future [70].

Although there is accumulating evidence supporting a genetic component to ASD, the specific genes involved have yet to be totally clarified. Genome-wide screening of autistic subjects and their first-degree relatives offers an attractive means to search for susceptibility genes. However there has been a disappointing lack of replication of many of the reported susceptibility regions. The reason for this could be due to the epistasis of many interacting genes. But it may also be due to the genetic and clinical heterogeneity present in ASD [71]. The noted effects of heterogeneity of the samples on the corresponding results, have led to attempts to decrease sample heterogeneity by various ways which include narrowing

A substantial body of evidence has resulted from genome-wide screening for the susceptibility genes of ASD (table 1). Significant replicability has been found for several chromosomal loci including 2q, 5, 7q, 15q and 16p. Two studies provided suggestive evidence for linkage to chromosome 2q using a two-stage genome screen [71, 72], while association tests for specific candidate genes in the chromosome 2q31-q33 region led to negative results [73]. Additional support for the presence of susceptibility loci on chromosome 2q is given by overlapping positive linkage findings in four other independent

There are three reports about gene variants on chromosome 5. Philippi found strong association with autism for allelic variants of "paired-like homeodomain transcription factor 1" (*PITX1*), a key regulator of hormones within the pituitary-hypothalamic axis [78]. Two other groups used genome-wide linkage and association mapping studies to analyze chromosome 5 gene variations finding that SNPs located at 5p14.1 and 5q15 respectively

Chromosome 16 linkage results have been fairly consistent in showing a peak at 16p11-13, which strongly suggested a gene in this region may contribute to the risk of ASD [81, 82]. 15q11-q13 is another frequently identified locus by linkage studies. Several genes located in this region have been intensively studied and some have provided very promising results [83-86]. But in all of these linkage reports there is a certain lack of reproducibility, and therefore they require further validation based on using a combination of several methods. Besides these "hot spots", there are other reports regarding associations of other loci with ASD [80, 87-90], including some evidence of linkage to the X chromosome [91]. However, there is little overlap of these potential loci involving potential candidate genes, suggesting

Rapid advances in genomic DNA microarray technologies have substantially improved our ability to detect submicroscopic chromosomal abnormalities. Novel rare variants have been detected in association with ASD and these can be either *de novo* or inherited. *De novo* or noninherited CNVs are found in 7%–10% of ASD samples from simplex families (having only one child affected, the majority), in 2%–3% from multiplex families, and in ∼1% in non-ASD controls. Further, about 10% of ASD subjects with *de novo* CNVs carry two or more CNVs [100-102]. Inherited CNVs reportedly are found in up to 50% of ASD subjects for whom one of the presumably normal parents also has the duplication/deletion. These

inclusion criteria and studies of specific, autism-related endophenotypes.

**3.2 Potential loci in autism** 

genomic scans [74-77].

were significantly associated with autism [79, 80].

that the genetic background of ASD is full of complexity.

**3.2.2 Copy number variation (CNV)** 

**3.2.1 Genome wide linkage analysis** 


familial CNVs may include candidate genes relevant to ASD where they are rare in the normal population.

Table 1. Loci identified by genome wide linkage analysis

Array comparative genomic hybridization (aCGH) is the most widely used method for detection of CNVs. A seminal early report used aCGH, with a mean resolution of one probe every 35 kb, to study a sample of 264 ASD families. After validation by higher-resolution microarray scans, G-banded karyotype, FISH, and microsatellite genotyping, 17 *de novo*

Autoimmune Disorder and Autism 525

**results**

120]

[158-

174]

179]

196]

**Negative/Unconfirmed** 

**results** 

127] [128-140]

154] [155-157]

163] [164-169]

183] [184-186]

198] [199, 200]

189] [190]

**Genes Loci Positive** 

*RELN* 7q22 [111-

*SLC6A4* 17q11.1-17q12 [121-

*GABR* 15q11-15q13 [141-

*OXTR* 3p24-3p25 [170-

*MET* 7q31.2 [175-

*SLC25A12* 2q31 [180-

*GluR6* 6q21 [187-

*CNTNAP2* 7q35 [191-

*GLO1* 6p21.3-6p21.2 [197,

repeat polymorphisms with liability to autism [116, 117].

reported to have a significant association with ASD [111].

*TPH2* 12q21.1 [201] [197, 202]

Additionally, case-control and family-based studies provided further evidence supporting the association of *RELN* and ASD. Persico identified a *RELN*–related polymorphic GGC repeat located immediately 5' of the ATG initiator codon in Italian and American subjects [120]. Using the similar methods and 126 multiplex ASD families, Zhang *et al* examined the polymorphic CGG-repeat of *RELN* [118]. Family-based association tests showed that larger *RELN* alleles (≥11 repeats) were transmitted more often than expected to autistic children. Independant studies regarding the CGG-repeat of *RELN* have also supported its contribution to the genetic risk of autism [112, 113, 115]. Others have also reported significant differences in the transmission of the reelin alleles of exon 22 and intron 59 SNPs to autistic subjects [114]. However, results have not been uniformly positive. Krebs *et al* performed a transmission disequilibrium test (TDT) analysis of the CGG-repeat polymorphism in 167 Caucasian families and found no evidence of linkage or association [119]. Similarly, another two groups failed to find a significant association of *RELN* CGG

The association between *RELN* and ASD were also found in other ethnic groups besides Caucasian populations. Recently, a significant genetic association between the *RELN* SNP2 (located in intron 59) and ASD was reported in a Chinese Han population, and the combination of *RELN* SNP1/SNP2/SNP3/SNP4, all in strong linkage disequilibrium, were

(*NLGN4Y*)

Table 2. Selected candidate genes

*i. Reelin gene (RELN)* 

3q26(*NLGN1*), 17p13 (*NLGN2*), Xq13 (*NLGN3*), Xp22.3 (*NLGN4*), Yq11.2

*NLGN*

CNVs were confirmed [102]. A Korean group recently reported deletion CNVs at 8p23.1 and 17p11.2 using whole-genome aCGH [103]. Using aCGH with a mean 19 kb resolution, 51 autism-specific CNV were identified in 397 unrelated ASD subjects [100]. Similarly, Qiao and colleagues performed aCGH on 100 autistic subjects and identified 9 CNVs, three of which were unique to their cohort [104]. A Spanish group recently reported the identification of 13 CNVs containing 24 different genes in their sample of 96 ASD subjects [105].

Single-nucleotide polymorphism (SNP) array analysis, primarily developed to determine linkage, now is also employed to determine genomic CNVs [106]. Marshall performed a genome-wide assessment via SNP array analysis. They genotyped proximately 500,000 SNPs for each sample and detected 13 loci with recurrent or overlapping CNVs in a sample of 427 ASD cases [101]. Using SNP markers, another group identified 6 CNVs within a 2.2 megabase (Mb) intergenic Chr 2 region between cadherin 10 (CDH10) and cadherin 9 (CDH9) in a combined sample set of 1,984 ASD probands of European ancestry [107]. In addition, SNP array analysis offers some special advantages in the exploration of potentially relevant gene networks. Two recent reports have provided strong evidence for the involvement of certain genes in important gene networks including neuronal cell-adhesion, ubiquitin degradation and GTPase/Ras signaling [108, 109].

Currently available aCGH methods for identifying CNV typically assay the genome in the 40-kb to several Mb range. Methodological improvements that employ oligonucleotides are providing a high potential resolution down to approximately the 5-kb resolution level for aCGH with genome-wide detection of CNVs [106]. Thus, SNP or oligonucleotide aCGH analysis can detect a CNV as small as a few kilobases. Therefore, it is clear that the higherdensity oligonucleotide or SNP arrays offer the higher resolution for analysis of CNVs in the future.
