We have shown that the expression of IL-4 in maternal serum is altered significantly between ASD affected and matched control groups at 20 weeks’ gestation in a small, but carefully characterised cohort of mothers and children where the child has a diagnosis of ASD by age 10 years.
Previous evidence indicates that aberrations of the immune system may play a role in ASD, (24, 31, 42). Some propose that alterations in cytokine expression could facilitate the classification of ASD subtypes (19, 24, 43) as well as work as biomarkers of response to treatment. In the diagnosis and management of ASD, earlier is better, and identification of reliable biomarkers during pregnancy may allow for targeted behavioural interventions from early infancy. This could also aid the development of targeted pharmacological strategies which have already shown promise in animal models (14), and analogues of which are currently in use in routine medicine practice (44, 45).
IL-4
That we have demonstrated alterations in IL-4 is an interesting finding. In the small number of studies that have examined mid-gestational serum of mothers to ASD affected children, IL-4 is the only cytokine to demonstrate altered expression across all studies (23, 24, 46). Interestingly, while previous authors found levels of IL-4 to be elevated in the ASD affected group versus controls, we have found the opposite. Physiologically, IL-4 is a pleiotropic, generally anti-inflammatory cytokine that functions to suppress the pro-inflammatory milieu. Produced by activated T-cells, NK cells, and mast cell, IL-4 aids the conversion of naïve T helper cells into Th2 cells as well as potentiating the Th2 response (47, 48). IL-4 also has a role in the developmental and maintenance of key regulatory T-cells (Tregs) through STAT6 signalling pathways (49). Tregs are important mediators of inflammation during pregnancy and at the feto-maternal interface (50). We find IL-4 itself at the feto-maternal interface throughout pregnancy (51), indeed in normal pregnancy; levels of IL-4 persist and increase as the pregnancy progresses (52). Low circulating levels of IL-4 during pregnancy have been linked with spontaneous abortions, pre-eclampsia, intra-uterine growth restriction and pre-term delivery (53-55). Failure of the usual pregnancy homeostasis (elevated IL-4 levels) may lead to a more pro-inflammatory pregnancy environment with subsequent effects on maternal health, obstetrics outcomes, and child health and development.
Animal-based studies:
Although there are very few human studies that have examined the molecular links between MIA and ASD, many animal-based studies have addressed the question of MIA and the association of elaboration of cytokines and parallel behavioural changes in offspring. MIA has been replicated in a variety of small animal models: mouse, rat and simian phenotypes of ASD have been created through intrauterine inflammatory exposure (56). These models provide valuable insights into the effects inflammation can have on social and communicative behaviour in progeny (56, 57). Remedial steps have been possible with improvements in and resolution of some ASD traits following blockade of specific inflammatory pathways (IL-6 and IL-17A) (14). This work suggested that these two cytokines in particular are significantly involved in the neuronal dysfunction brought about through MIA (14, 57-59).
In MIA-mouse models of ASD, a commonly used synthetic analogue of double stranded RNA which mimics the effects of viral infection, Polyinosinic:polycytidylic acid or Poly(I:C) (60), is shown to increase IL-17A levels in maternal blood and the postnatal brain of offspring (61). Importantly, there is also an increase in placental mRNA levels of the cytokine, suggesting upregulation of IL-17A activity at the feto-maternal interface. Determining the role alterations of IL-17A have to play in ASD pathogenicity has become a key question over the past few years. In 2016, Choi et al demonstrated persuasively that simulated MIA in murine models leads to elevation in maternal IL-6, leading to downstream activation of maternal Th17 cells. Maternal Th17 cells produce IL-17A that is hypothesised to cross to the foetus via the placenta leading to increased expression of IL-17AR in the foetal brain, contributing to cortical malformations and behavioural abnormalities (14, 62). Conversely, inhibition of IL-17A signalling via IL-17A specific antibodies prevented ASD phenotypes in offspring (14). In support of the synergy between IL-6 and IL-17A, Gene knockout of IL-6 in Poly(I:C) treated dams results in failure to alter IL-17A levels in offspring, which suggests IL-6 acts upstream of IL-17A (63).
Human studies:
Quite a number of human based studies have examined immune/cytokine aberrations in individuals affected by ASD themselves. In Table 6, for simplicity, we have categorised the cytokines measured in our analysis based on their overall function (64-66). We also highlight their role in ASD according to the literature:
Table 6: Cytokines included in our analysis and their roles and relevance to ASD
Cytokine
|
Category
|
Altered in blood/CSF of ASD individual
|
Altered in gestational blood
|
Altered in amniotic fluid
|
Cytokine characteristics
Relevance to ASD
|
TNFα
|
Pro-inflammatory
|
(67-70)
|
(24)
|
(71)
|
Apoptosis of infected cells. Elevated in the CSF and blood of ASD affected individuals (67-69).
|
IL-1β
|
Pro-inflammatory
|
(67, 68, 72, 73)
|
(24)
|
|
A potent pro-inflammatory cytokine involved in both acute and chronic inflammation. Correlated with ASD symptom severity (43).
|
IL-6
|
Pro-inflammatory
|
(67, 68, 70, 72-75)
|
(24)
|
|
Induces production of acute phase proteins and stimulates B-cell antibody production (76). Pleiotropic (affects hematologic, hepatic, endocrine and metabolic function). Thought to impact synapse formation and neuronal migration (77). Potentially mediates IL-17 linked ASD risk in pregnancy (14, 57)
|
IFNγ
|
Pro-inflammatory
|
(31, 67, 74)
|
(23, 24)
|
|
Interfaces between innate and adaptive immune response. Secreted by NK cells, and promotes NK killing. Activates macrophages, which produce IL-12 and -23, stimulating Th1 and Th17 cell respectively. Inhibits Th2 cells. Versatile, with a role in defence against intracellular pathogens, tumour surveillance, autoimmunity, allergy and the protection of the amniotic space during pregnancy (78).
|
IL-17
|
Pro-inflammatory,
Chemotactic
|
(19, 21, 67, 70, 75, 79)
|
(24)
|
|
Derived from Th17 cells, a subset of CD4 cells. Potentiates the innate PMN response throughout inflammation. Postulated to trigger alterations in the blood brain barrier and lead to cortical dysplasia (57).
|
IL-4
|
Pro-/Anti-inflammatory,
Allergy
|
(73)
|
(23, 24, 46)
|
(71)
|
A Th2 derived cytokine, often linked with asthma and allergic type inflammation (80). Dual role: pro/anti-inflammatory properties. Crucially important in mitigating inflammation during pregnancy (primarily through suppression of Th1 T-cells and associated cytokines (IL-2 and IFNγ).
|
GM-CSF
|
Growth factor
|
(81)
|
(24)
|
|
A colony-stimulating factor. Produced by stromal cells, it targets bone marrow, and precursor cells, mediating haematopoiesis.
|
IL-8
|
Chemotactic
|
(22, 72, 74)
|
(24)
|
|
Produced by fibroblasts, neutrophils and macrophages. Chemo-attractant for phagocytes at site of inflammation.
|
Note: In Table 6, reference numbers for supportive scientific literature in parentheses (all are human based studies).
While the cytokine profiles of ASD affected individuals have been well characterised, very few studies have investigated the relationship between mid-gestation cytokine levels and ASD risk in offspring. To our knowledge, only three human studies have examined maternal serum (23, 24, 46), and one more has examined amniotic fluid cytokine profiles in mothers of ASD affected children (71). Below, we have summarised the findings from these studies, which effectively provide all of our current understanding of gestational cytokine profiles in the setting of ASD.
Previous literature on gestational samples analysis in ASD:
Working from the same laboratory and using similar methods, Goines et al (2011) and more recently, Jones et al (2017) both demonstrated elevated mid-gestational cytokine levels between groups of ASD affected children versus controls or children without ASD. Goines et al demonstrated elevated levels of mid-gestation (15 – 19 weeks’ gestation) IFNγ, IL-4 and IL-5 with an associated 50% increased ASD risk. While Jones et al showed elevated levels of mid-gestation GM-CSF, IL-6, IFNγ and IL-1α in the ASD affected group versus children with developmental delay, but not ASD. The authors do not mention the age of the samples used in either study, but the samples used were sourced from the same birth cohort in Orange County, California between 2000 and 2003. In both studies, the samples were initially stored at room temperature and later at - 20°C freezer conditions before long-term storage at - 80°C. This initial handling may have contributed to some cytokine degradation. In the Goines study, ASD cases were matched with neuro-typical controls based solely on child characteristics (sex, birth month and year), something which the authors acknowledge in their limitations. Neither study had access to comprehensive maternal health information during the pregnancy (including intrapartum infections). Nor did they have a record of relevant maternal medical history, all, information important to the interpretation of their findings.
Irwin et al (2018) demonstrated alterations in IL-4, MCP-1 and IL-10 levels in 28-week gestation serum of mothers who birthed ASD affected children (46). Specifically, IL-4 (usually anti-inflammatory or involved in allergic type inflammation (80)) was increased and associated with higher ASD symptomology (as measured by the Social Communication Questionnaire (SCQ)) in offspring. Higher concentrations of IL-10 (anti-inflammatory) were associated with fewer ASD symptoms in offspring (measured by the Social Responsiveness Scale (SRS)), and finally, elevated MCP-1 was associated with fewer ASD symptoms (as measured by the SCQ). The samples used in this analysis were reported to be at least 5 years old. No controls were used in this analysis, instead a large cohort of ASD affected individuals were enrolled, and the 28-week gestation cytokine concentrations were correlated with ASD symptomology at 7 years of age. This is novel in two senses, none has previously assessed the cytokine profile in the third trimester, and none has correlated cytokine findings with severity of ASD symptomology in this way. As with previous authors, they had no access to relevant maternal pre-conceptual medical history or gestational infections data.
Finally, Abdallah et al (2013) examined amniotic fluid samples and found elevated levels of IL-4, IL-10, TNFα, and TNFβ. In a preliminary study (2012), they also identified elevations in MMP-9 in ASD cases relative to controls (82). Advanced sample age is again an issue with the oldest samples in this analysis being 29 years old, the youngest 10 years old. The samples were stored at - 20°C according to local guidance (83). Both the storage conditions and the samples ages are likely to have contributed to significant cytokine degradation (40).
Limitations:
The samples used in our study fall outside the ideal sample age for accurate analysis of cytokines (40). To our mind, this is the single most important limitation confronting studies of this nature. Unfortunately, the shelf life of archived samples is finite, and even samples in long-term ultra-low temperature storage (−80°C) suffer from degradation of cytokines and chemokines over time (40, 41). Retrospective sample analysis, would present an excellent opportunity to study cytokine aberrations in ASD, if the time to ASD diagnosis was shorter. One UK study found that the average delay between concerns first being noted by parents and the child receiving a diagnosis of an ASD was 4.6 years (SD 4.4 years) (84). ASD services continue to be under-resourced (85) and diagnoses are chronically delayed (86). Under current conditions, our experience of retrospective analysis of archival samples suggests that this style of study design is not well suited to addressing this question. Even large-scale population based studies would suffer from the same issues of sample fidelity over longer periods.
To ensure future study designs are capable of accurate mid-gestation cytokine analysis, they should be prospective, and concentrate on early ASD case identification or screening. Early identification should be paramount, the diagnostic stability of ASD is reliably fixed from as early as 14 months old (87) so screening and identification within the first 2 - 3 years of life is possible. Cytokines should be analysed contemporaneously, acute phase reactants such as IL-1β and IL-6 have demonstrated greater than 50% degradation within 3 years even in −80°C freezer conditions (40). IL-4 is stable only for 3 years, while IL-17A, IFNγ, and TNFα, all suffer more than 50% degradation within 4 years at ultra-low temperature storage (40). Basic handling of samples and initial processing requires optimisation to ensure the risk of sample degradation is minimised: (i) Store samples at ultra-low temperatures, (ii) initial processing should be rapid (<1 hour from venepuncture to freezer storage) and (iii) freeze-thaws cycles should be minimised. With robust methods of early screening in place, early confirmatory diagnosis within the first 2/3 years, and analysis of gestational samples within 3 years, it should be feasible to increase the yield and validity of such studies, and greatly reduce cytokine loss through prolonged storage. While this approach would allow for study of children presenting with the earliest signs of ASD, or targeted high-risk groups (ASD affected siblings). It would likely miss those presenting later, including those who are a high-functioning phenotype or of female sex.
Finally, our small sample size is a major limitation, and results should be interpreted with caution. Analysis of IL-4 levels in the groups yielded results on only 16 individuals (6 cases and 10 controls). Attrition of the viable samples was due to a combination of the low absolute concentrations of IL-4 in the samples (likely exacerbated by advanced sample age), concentrations at or below the sensitivity (LLOD) of the MSD multiplex format and high CV values. It is difficult to make inferences about results in samples this size, and larger scale group analysis is warranted.
Strengths:
Although our study has suffered from some of the same limitations as previous studies, our study is strengthened by the quality of our cohort. Each child had a concrete specialist service ASD diagnosis, confirmed by the clinical paediatric fellow. Each child was well characterised clinically and matching was strictly observed. Matching was not only based on child characteristics (Sex, Gestational age, Birthweight), but also on an important maternal characteristic, BMI at 15 weeks’ gestation. This enhanced the validity of our results. In addition to detailed child characteristics, we have also included important information regarding the past medical histories, medication or anti-inflammatory use, and pre-existing inflammatory conditions of the mothers included in the study. We present crucial information about infection rates in the first 20 weeks of pregnancy, all of which presents a major confounder to accurate analysis if this information is absent. Our methods were robust, and we identified two key issues of multiplex assay sensitivity and advanced sample age, and remedied the former through utilisation of ultrasensitive single analyte plates.
Conclusion:
In conclusion, in a carefully characterised maternal-child cohort study we did not replicate the findings of similar mid-gestational studies, but did find some evidence of mid-gestational cytokine aberrations (downregulated IL-4) in the mothers of children with ASD. Reduced levels of IL-4 are linked to a pro-inflammatory state during pregnancy and negative obstetric and foetal outcomes. All studies to date have had similar and significant limitations. Future studies should focus on minimising the time between sample acquisition and analysis, use of best practice for initial sample handling, and early identification and characterisation of cases and their mothers. Future analysis should be serial and include investigation of samples taken from early in pregnancy. The first trimester, and particularly 8 - 12 weeks’ gestation is a crucial period for organogenesis and differentiation, and analysis from this period will help complete the picture of gestational cytokine fluctuations and their effect on neurodevelopment.