Plasma cytokine profiling in sibling pairs discordant for autism spectrum disorder

Autism spectrum disorders (ASDs) are a heterogeneous group of severe neurodevelopmental disorders characterized by atypical social interactions, impaired communication, and tendency to engage in idiosyncratic, repetitive, or restrictive behaviors, with onset before 3 years of age. ASDs include autistic disorder, Asperger’s syndrome, and pervasive developmental disorder-not otherwise specified (PDD-NOS) [1]. Although significant progress has been made in the identification of genes and copy-number variants associated with syndromic autism (approximately 10% of the total number of cases with ASD) [2], little is currently known about the etiology of idiopathic non-syndromic autism. The clinical heterogeneity of ASD probably reflects the complexity of its genetic underpinnings, involving multiple contributing loci, genetic heterogeneity, epistasis, and gene-environment interactions [3].

In addition to the results from neurobiological research in ASD, highlighting the pathways involved in neural development, synapse plasticity, structural brain abnormalities, cognition, and behavior, converging evidence point to the existence of altered immune function in ASD, which directly affects some or all these neurological processes [4]. Several immune abnormalities have been reported in ASD, including familial autoimmune disorder clustering and ASDs [5], altered gene expression, resulting in dysfunctional natural killer (NK) cells [6], immune transcriptome alterations in the temporal cortex of subjects with autism [7], and the presence of auto-antibodies to the cerebellum in children with autism [8]. Indeed, children with ASD were reported to have higher levels of auto-antibodies (including anti-myelin basic protein [9], anti-myelin-associated glycoprotein [10], anti-ganglioside [11], anti-neuronal [12], and anti-mitochondrial [13] antibodies) compared with healthy children. Despite the extensive research linking immune irregularities to ASDs, there are no salient findings that have significantly advanced the understanding of the pathogenesis of ASD.

Previous studies [14‐25] have reported altered cytokine levels in subjects with autism with inconclusive results, perhaps attributable to different types of study design, but also probably reflecting the wide heterogeneity of ASD. In addition, one study found no difference in cytokine levels between young ASD children and normotypic controls [26]. Moreover, relatively few cytokines have been examined to date, and recent technologies have opened higher throughput means for quantitatively surveying 10s to 100s of unique cytokines per sample on an array. Previous studies have shown that patients with autism have higher cytokine levels in cases of autism compared with controls or subjects with other developmental disorders [14‐25]. In the current study, we evaluated whether these alterations occur within families with ASD by performing a comprehensive plasma-cytokine profiling in 25 sibling pairs discordant for ASD. We also analyzed the correlations between cytokine levels and clinically relevant quantitative traits (Vineland Adaptive Behavior Scale in Autism (VABS) composite score, Social Responsiveness Scale (SRS) total T score, head circumference, and full intelligence quotient (IQ)). In addition, because of the high phenotypic heterogeneity of ASD, we divided the patients into four defined subgroups (those who were non-verbal, those with gastrointestinal (GI) issues, those with regressive autism, and those with a history of allergies), which encompass common and/or recurrent endophenotypes in ASD, and tested the cytokine levels of these groups [27‐30].

Despite the great wealth of data on autism gained to date, and the possible involvement of the immune system in ASD, the results are still inconclusive. This is probably attributable to the large phenotypic and genetic heterogeneity of ASD [37]. In an attempt to identify the complex immune pathogenetic components underlying ASD, we carried out a comprehensive analysis of cytokines and chemokines in discordant sibling pairs. We explored the hypothesis that the use of discordant sibling pairs may reduce heterogeneity and facilitate the identification of immune underpinnings in ASD.

We did not find any significant differences in cytokine levels between children with ASD and their related healthy siblings. This is in line with a previous study by Saresella et al. [38] which showed that the immune profiles of children with autism did not differ from their typically developing siblings. That study indicated the presence of an ‘autism endophenotype’ that expands immune dysfunction to family members who are seemingly unaffected by the core symptoms of autism. Moreover, anti-brain antibodies were found both in children with autism and in their unaffected siblings [39]. Notably, healthy siblings are characterized by subtle neurologic impairment [40‐43], and there is evidence of atypical social and communication development during infancy [44]. Recent studies also described ASD as the quantitative extreme of a neurodevelopmental continuum in the general population [45], with moderate genetic heritability and a substantial shared twin environmental component [46]. Moreover, the high percentage of macrocephaly we found among the healthy siblings (23.8%) compared with that reported for general population (around 3%) [47] concurs with the evidence of an extended familial endophenotype. Thus, the lack of significant differences between sibling pairs discordant for ASD found in our study is in line with the results of previous studies. It is possible that a common immunogenetic background shared by siblings might eventually lead to different clinical outcomes when an environmental stress (for example, prenatal exposure to environmental toxins, viral and bacterial infections, parental microchimerism, etc.) occurs during development.

However, the cytokine/chemokine levels in our subjects did correlate with the quantitative clinical traits, and these were significantly different when the clinical subgroups were analyzed (Figure 4). It is notable that these differences seem to be attributable to a predisposing immunogenetic background, as no other significant differences were noticed between discordant sibling pairs.

IL-1β appears to be the cytokine most involved in the quantitative traits and clinical subgroups of ASD (Figure 4). Involvement of IL-1β in the physiopathology of autism is generally supported by several studies reporting higher levels of this cytokine in the plasma of children with ASD, high-functioning children with ASD, and adults with severe ASD compared with unrelated controls [48]. It has been shown that peripheral blood cells from subjects with ASD produced higher levels of IL-1β both at baseline and after stimulation with Toll-like receptor (TLR)2 or TLR4 [48]. Moreover, plasma levels of IL-1β were found to correlate with regressive onset [20], a finding consistent with the current results. Ashwood et al. [20] also showed that increased cytokine levels occurred predominantly in children who had a regressive form of ASD. In line with this, we also found increased levels of IL-1β, IL-5, and IL-17, together with total Th2 and Th17 cytokine levels, in children with regressive ASD compared with children with ASD who had no regression. All these findings are compatible with the established evidence that levels of immunomodulatory factors are altered in children with regressive ASD compared with children with non-regressive ASD [49‐52]. It is notable that we found increased levels of IL-5 in children who had a regressive form of ASD. Notably, high levels of IL-5 in mid-gestation maternal serum samples were significantly associated with a 50% increased risk of ASD in the offspring, and especially of regressive autism [53]. Further, an increased level of IL-5 has been reported in male subjects with high-functioning ASD/Asperger’s disorder [17, 25]. IL-5 is produced by Th2 cells and mast cells [54]. Its functions are to stimulate B-cell growth and increase immunoglobulin secretion. An IL-5 transgenic mouse model has two predominant features: 1) a marked increase in B-1 cells, resulting in enhanced serum antibody levels, and 2) an expansion of eosinophil numbers in the blood and eosinophil infiltration into various tissues [55]. The presence of maternal auto-antibodies specific for fetal brain proteins has also been shown to be strongly associated with regressive autism [56].

We found the levels of several cytokines (GM-CSF, IL-1α, IL-1β, IL-2, IL-6, IL-16, and MCP-1) to be inversely correlated with the VABS composite score. Increased levels of IL-1β, IL-6, and MCP-1 had previously been shown to be associated with more aberrant behaviors or impairments in cognitive and adaptive function of patients with ASD [20, 48, 57]. The level of MCP-1 was found to be increased in astrocytes in the anterior cingulate gyrus, and also in the cerebellum and in brain tissue homogenates in patients with ASD [58]. Moreover, a 12-fold increase in MCP-1 was also noted in the cerebrospinal fluid of children with ASD children compared with controls [58]. Notably, it has also been shown that MCP-1 expression is induced by IL-1 [59]. Thus, it is conceivable that increased levels of the two pro-inflammatory cytokines IL-1α and IL-1β might trigger the expression of MCP-1. It should be noted that IL-6 has a crucial role in the development and plasticity of the central nervous system, and it was shown to be significantly increased in the cerebellum of subjects with ASD [60]. Furthermore, IL-6 overexpression in granule cells causes impairments in the adhesion and migration of these cells, possibly altering neural-cell adhesion and migration, and causing an imbalance of excitatory and inhibitory circuits. Moreover, IL-6 exerts a striking negative effect through maternal immune activation on fetal brain development during pregnancy [61]. In addition, a significantly higher level of IL-6 was previously found in the plasma of children with ASD compared with that of typically developing children [20]. Overall, it should be noted that all the cytokines that inversely correlated with the VABS composite score in our study exert a pro-inflammatory function. This finding supports the view that increased levels of pro-inflammatory cytokines can contribute to greater impairments in behavior, characteristic of the core features of ASD, particularly deficits in social interaction and communication [4]. IL-6sR, MIP-1β, and MIP-1δ were inversely correlated with the SRS total score, suggesting a positive beneficial role for higher levels of these cytokines in autism-related social behavior. However, these results should be interpreted with caution, because the role of these cytokines in ASD is still unclear.

We also found a correlation of two cytokine/chemokines (BLC/CXCL13 and TIMP-2) with head circumference, which is the first such report, to our knowledge. Head circumference represents one of the most reliable, consistent, and easily detectable endophenotypes in autism research. Macrocephaly has been consistently recorded in approximately 20% (range 14% to 37%) of patients with autism [36]. Excessive neurite outgrowth and reduced terminal pruning during infancy is believed to play a crucial role in the establishment of macrocephaly [36]. It has been shown that TIMP-2 [62] affects neurite outgrowth, making its strong correlation with head circumference in this study highly plausible. The effect of BLC/CXCL13 on head circumference might result from its interaction with the IL-7 receptor pathway [63], which significantly affects neurite outgrowth [64, 65].

We also found associations between plasma levels of several cytokines (GM-CSF, IL-1β, IL-6, IL-7, IL-11, IL-12p70, IL-13, IL-16, IL-17, M-CSF, and TNF-sRII) with full IQ. It is currently unclear how cytokines might affect IQ during childhood in ASD, thus these data should be treated with caution until further validation can be performed.

Language impairment is a common feature of ASD, and some individuals with ASD never acquire language [27]. Studies conducted on the FOXP2 gene, which is related to the speech-language disorder, unveiled the crucial role of the cerebellum and Purkinje cells in the pathogenesis of speech-language disorders [66]. Foxp2 (R552H) knock-in mice showed severe ultrasonic vocalization, motor impairment, and immature Purkinje cells with poor dendrites and fewer synapses [66]. We found increased levels of GM-CSF, M-CSF, and IL-10 in non-verbal children with ASD compared with verbal children with ASD. Purkinje cells express receptors for both GM-CSF [67] and M-CSF [68], thus it is possible that altered levels of these two growth factors influence the maturation of the Purkinje cells, leading to language impairment. Notably, IL-10 is known to inhibit the action of GM-CSF [69], and also to act in synergy with M-CSF [70], thus, its involvement in non-verbal ASD, together with that of GM-CSF and M-CSF, seems highly conceivable.

We found that children with ASD with GI issues displayed significantly higher levels of IL-1β, IL-2, and IL-6. The relatively high frequency and variable spectrum of GI symptoms, reported by many parents of autistic children, could conceivably stem from a complex combination of abnormal gut microbiome, excessive intestinal permeability, local immune dysreactivity, and possibly pleiotropic roles of autism genes in nervous and gut tissue [71]. In this context, it should be notes that increased concentrations of IL-1β and IL-2 have been reported in endoscopic mucosal biopsy specimens from patients with inflammatory bowel disease [72]. It has also been shown that IL-1β and IL-2 production is significantly increased in active ulcerative colitis and is significantly correlated to its activity index [73]. Increased level of IL-6 may result from high levels of IL-1β and IL-2, two cytokines that stimulate IL-6 expression in human monocytes by independent mechanisms [74].

Correlation and association do not imply causation. Even assuming that pathophysiological mechanisms do link cytokines to the quantitative traits and clinical subgroups analyzed in this study, the nature of these mechanisms remains open to interpretation. Nonetheless, our results provide important clues pointing toward possible mechanisms, which will deserve closer scrutiny in future investigations.

Whereas previous studies on plasma/serum cytokine profiling in ASD used bead-based suspension arrays [17, 20, 25, 26, 57], we used a planar array. Planar microarray comprises reagents for individual tests immobilized as an ordered array or grid of discrete reagent areas (spots) on a flat surface (for example, a microscope slide) [75]. The bead-microarray format comprises encoded microbeads (for example, each with a unique fluorescence signature), and each type of bead is coated with a different reagent [75]. Our planar array is based on sandwich ELISA-based technology, which achieves higher sensitivity and specificity, and is the method of choice for low-abundance proteins [76]. Overall, most studies to date have reported good correlation between bead-based and planar cytokine arrays and between multiplex assays and traditional ELISA (comparable relative changes), thus recommending the continued and expanded use of multiplexing systems in high-throughput screening applications [77].

Overall, the present study reports the lack of significant differences in plasma-cytokine levels between children with ASD and in their related non-autistic siblings. Thus, our results support the evidence that the immune profiles of children with autism do not differ from their typically developing siblings [38]. However, the significant association of cytokine levels with the quantitative traits and the clinical subgroups analyzed suggests that altered immune responses may affect core features of ASD [20]. Because ASD may encompass several distinct phenotypes, the potential partition of ASD subgroups based on immunological parameters and/or associations with worsening behavior may have important implications for diagnosis, and for the design and monitoring of therapeutic treatments of ASD [20]. Clearly, further studies are warranted to confirm and extend the associations found in the present study, and should also include analysis of other relevant inflammatory molecules, such as auto-antibodies [9‐13].