Brain Transcriptome of Autism Spectrum Disorders and Tourette Syndrome Denes Common Targetable Inammatory Pathways Involving Cytokines, Complement, and Kinase Signalling

Background Neurodevelopmental disorders (NDDs), including autism-spectrum disorders (ASD) and Tourette syndrome (TS) are common brain conditions which often co-exist, however current management focuses on symptom mitigation, with no approved treatments targeting disease mechanisms. There is accumulating literature implicating the immune system in NDDs, and transcriptomics of post-mortem brain tissue from individuals with NDDs has revealed an inammatory signal. We interrogated two RNA-sequencing datasets of ASD and TS (compared to age-matched controls) and identied the top 1000 differentially expressed genes, to explore commonly enriched pathways using an over-representation analysis through GO, KEGG and Reactome.


Background
Neurodevelopmental disorders (NDDs) such as autism-spectrum disorders (ASD) and tic disorders including Tourette syndrome (TS), are neurological conditions which commonly co-exist and have shared genetic contributions 1 . ASD is characterised by social communication and language de cits, and repetitive stereotypical behaviour. Tics are repetitive stereotyped movements (motor tics) or vocalisations (vocal tics), and when present for more than 12 months, ful ll a diagnosis of TS. Tics are present in 11-22% of children with ASD, while ASD is present in 12% of children diagnosed with TS [2][3][4] . Limited disease speci c treatments are currently available for NDDs, and management focuses on symptom mitigation and developmental support 5,6 .
The genetic aetiology of neurodevelopmental disorders is thought to be due to variants in multiple genes that converge on common pathways 7,8 . However genetic aetiologies in these disorders are unable to explain the wide phenotypic heterogeneity, instead the interaction between environmental and genetic factors are proposed to play an important role in pathogenesis of NDDs. In addition, immune dysregulation and in ammation have long been suggested to contribute to the pathophysiology, where early insults during gestation, such as maternal immune activation (MIA), can impact the development of the foetal brain [9][10][11][12][13][14][15][16] . MIA, encompassing maternal conditions such as infection, asthma, obesity, autoimmune disease, and psychosocial stress, are associated with increased incidence of NDDs in offspring, such as ASD and TS [17][18][19][20] . MIA is thought to act as a disease primer, which in addition to genetic predisposition, results in increased expression of neurodevelopmental disorders 21 . Studies have also shown dysregulation in proin ammatory cytokines such as IL-12, TNF, monocyte chemoattractant protein 2 (MCP-2), and IL-2 in the brains and peripheral blood of individuals with ASD and TS 22-25 .
Transcriptomic analyses (RNA sequencing) of post-mortem brains from individuals with ASD have shown upregulated genes involved in in ammation and microglial dysregulation 26,27 . Similarly, analysis of postmortem brain striatum from individuals with TS identi ed up-regulated genes in immune and in ammatory pathways, and implicated microglial activation as a primary source of in ammation 28 . In both the ASD and TS brain transcriptome studies, the downregulated genes were enriched in pathways involved in synaptic function and GABA neurotransmission, aligning with the genetic variation found in these disorders 26-28 . By contrast, the upregulated in ammatory ndings were considered more likely to be due to environmental factors or secondary 26-28 .
Given the shared genetic heterogeneity and comorbidity of NDDs, there is an increasing need to examine common disease pathways. As in ammation has been reported in brain transcriptomics in both ASD and TS, we examined for shared gene expression in order to improve our understanding of the pathophysiology of NDDs and provide future potential therapeutic targets 26-28 .

Methods
Data availability, and open-source bioinformatic analysis: Human brain transcriptome data (RNA-seq) from two independent published studies were obtained with authors permission from synapse.org and analysed for differential gene expression and pathway enrichment analysis 26, 28 . Unlike TS, where only one study interrogating the brain transcriptome exists, there are a number of studies investigating ASD brain transcriptome 29-31 . The current ASD dataset was chosen as it presented the largest cohort of samples 26,27 . The ASD data were downloaded from synapse.org (ID: syn8234507) as count les, and RNA-seq metadata of 42 ASD cases were matched with large sample size with matched controls. The TS data was downloaded as BAM les from synapse.org (ID: syn3158906), which included putamen and the caudate nucleus regions from 9 TS cases and 9 normal controls 28 . The bioinformatic work ow, including all utilised code and quality control gures can be found at https://github.com/sarahalshammery/ASDTS.

Data quality control
The ASD dataset were prepared and sequenced as described (www.doi.rog/10.7303/syn4587615), reads were mapped against the Genome Reference Consortium Human Build 37 (GRCh37, otherwise known as hg19). The TS dataset were mapped against GRCh37 (hg19), and gene level counts for REFSEQ genes were assessed using HTSeq-count 28 . The raw counts for each dataset were converted to the counts per million (cpm) scale and ltered by requiring each gene to have > 0.01 cpm of 50% within the samples.
The data was normalised as per the EdgeR guide using Trimmed Mean of M-values (TMM) normalisation 28 .

Differential gene expression analysis
The top 1000 genes following differential gene expression analysis of each dataset were considered differentially expressed genes (DEGs) in this investigation. The DEGs were identi ed by a quasi-likelihood (QL) negative binomial (NB) generalised log-linear model (glmQLF). Genes with a log fold change > 0 were considered to be up-regulated, and those below 0 were down-regulated. DEGs were visualised through a volcano plot using the ggplot 2 package 32 .

Pathway and Network enrichment analysis
Enrichments of the top 1000 DEGs were identi ed through an over-representation analysis using GO Biological Process, Reactome and the Kyoto Encyclopedia of Genes Genomes (KEGG), through the compareCluster function from the ClusterPro ler package (FDR < 0.05) [33][34][35][36][37][38][39] . Given the perceived more signi cant mechanistic insights of the Reactome results, they are presented in the main text, whereas GO and KEGG are presented in the supplementary material.
The protein-coding DEGs which were common to both the ASD and the TS top 1000 DGE analyses, were visualised using a protein-protein interaction (PPI) network through the Search Tool for the Retrieval of Interacting Genes/Proteins (STRING; https://string-db.org/), with an interaction score > 0.4 39 . The active interaction sources included experiments, co-expression, and gene fusion. The PPI network and functional enrichments and of GO, KEGG and Reactome from the DEGs common to both ASD and TS datasets were further imported into Cytoscape 40 . CytoHubba, an app for Cytoscape was used to identify hub genes by ranking nodes by network features through the MCC method 41 . The expression of the hub genes in the disease cohorts compared to controls were visualised using the ggplot 2 package 32 .

Transcriptional signatures
To identify relationships within the cases and their respective controls, we set out to explore differences based on transcriptome signatures. The ASD and TS cases were not observed to be transcriptionally distinct from their respective normal controls using hierarchal clustering analyses (Supplementary Figure  1 and Figure 2).

Differential gene expression analysis
Autism spectrum disorder: The top 1000 genes following DGE analysis within the PFC of ASD cases and normal controls were considered differentially expressed genes and used for further analysis. The differentially expressed genes consisted of 912 up-regulated genes and 88 down-regulated genes represented through a volcano plot ( Figure 1A). Results of the DGE analysis can be accessed in supplemental material (Supplementary Table 2).
Tourette syndrome: The top 1000 differentially expressed genes within the striatum of individuals with TS and normal controls consisted of 711 up-regulated genes and 289 down-regulated genes, as shown in the volcano plot ( Figure 1B). Results of the DGE analysis can be accessed in supplemental material (Supplementary Table 3).
Immune pathways are enriched in ASD and TS brain transcriptome Autism spectrum disorder: To explore enriched terms and pathways in the top 1000 ASD DEGs, overrepresentation analyses were conducted through three databases (FDR <0.05). The GO analysis revealed 754 terms, all enriched by up-regulated DEGs, and involved many immune response and in ammatory signalling terms (Supplementary Table 2, Supplementary Figure 3). The top 3 GO terms were "response to lipopolysaccharide", "response to molecule of bacterial origin" and "response to interferon-gamma". Overrepresentation analysis using KEGG revealed 55 pathways, 54 of which were enriched by up-regulated genes (Supplementary Figure 4). The top 3 KEGG pathways (based on FDR) were "Systemic lupus erythematosus", "Malaria" and "Neutrophil extracellular trap formation" (Supplementary Table 2 terms were "in ammatory response", "cytokine-mediated signalling pathway" and "myeloid leukocyte activation". Over-representation analysis using KEGG revealed 56 pathways, all enriched by up-regulated genes involved in the immune response and in ammatory signalling (Supplementary Table 3, Supplementary Figure 6). Of the 56 pathways, the top 3 KEGG pathways were "Staphylococcus aureus infection", "Cytokine-cytokine receptor interaction" and "Viral protein interaction with cytokine and cytokine receptor". Enrichment of the DEGs using Reactome revealed 28 pathways, all enriched by upregulated DEGs (Figure 2 B, D). Of the 28 pathways, the top 3 Reactome pathways (sorted by FDR) were "Neutrophil degranulation", "Interferon gamma signalling" and "Signalling by interleukins". Overall, 18/28 Reactome pathways were involved in the immune response consisting of cytokine signalling, innate and adaptive immune response pathways, 5/28 pathways were involved in signal transduction, 3/28 pathways were involved in the homeostasis pathway, and 1/28 pathway was involved in response to stimuli. The full list of pathways from the three databases can be found in supplemental material (Supplementary Table 3).
Differentially expressed genes common to ASD and TS Of the top 1000 DEGs from the ASD analysis, and the top 1000 DEGs from the TS analysis, 133 DEGs were found to be shared. The common protein-coding DEGs were mapped into a PPI network, and their expression in the ASD and TS cohorts was visualised ( Figure 3A). From this network, we identi ed the top ve hub genes using Cytoscape and CytoHubba, which consisted of Colony Stimulating Factor 2 Receptor Subunit Beta (CSF2RB), Lymphocyte Cytosolic Protein 2 (LCP2), Hematopoietic cell kinase (HCK), Hematopoietic Cell-Speci c Lyn Substrate 1 (HCLS1), and Pleckstrin (PLEK) 40,41 . The raw data are presented in log scale for the ve hub genes in cases compared to controls, shown for ASD ( Figure  3B) and TS ( Figure 3C). Inspection of the hub genes identi ed the rst four as tyrosine kinases or associated with tyrosine kinase activity and cell signalling (Table 1). A full list of the common DEGs can be found in supplemental material (Supplementary Table 4).
Common differentially expressed genes in ASD and TS enrich immune pathways As many of the enriched dysregulated pathways in ASD and TS overlapped, we set out to explore enriched pathways from the 133 DEGs common to both disorders, using overrepresentation analyses through KEGG and Reactome. The KEGG analysis revealed up-regulated genes enriched in 26 pathways in ASD and 25 pathways in TS, with the top three common pathways involving "Complement and coagulation cascades", "TNF signaling pathway" and "Malaria" ( Figure 4A). Using Reactome, the 133 DEGs common to ASD and TS involved up-regulated genes enriched in 14 pathways for ASD and 12 pathways for TS, with the top 3 common pathways implicating "Interleukin-4 and Interleukin-13 signaling", "Interleukin-10 signaling" and "Signaling by Interleukins" (Figure 4B). The full list of pathways from the two databases can be found in the supplementary material (Supplementary Table 4).

Discussion
In this study we investigated enriched immune and in ammatory pathways in post-mortem brain tissue of individuals with ASD and TS, as well as pathways common to both disorders. Differential gene expression of the PFC region in ASD revealed that the majority (912 genes) of the top selected 1000 DEGs were up-regulated compared to normal controls. Analogous to this, in the striatum of TS, the majority (711 genes) of the identi ed top 1000 DEGs were also up-regulated compared to normal controls. This analysis validates the previous studies of up-regulated genes in post-mortem brains of individuals with ASD and TS 28, 42 .
The identi ed dominant signal of immune response and in ammation from the ASD GO enrichment analysis, aligns with studies investigating brain transcriptome and pathology of individuals with ASD, and supports the involvement of astrocytes and activated microglia 26, 42,43 . Of interest, the top GO term (by FDR) was response to lipopolysaccharide (LPS), a TLR4 agonist which stimulates an aberrant innate immune response in preclinical and clinical studies of NDDs 44,45 . For example, studies have highlighted a differential innate immune response when monocytic cells from children with ASD are treated with LPS, characterised by dysregulated levels of cytokines including IL1B, vital in the neuro-immune crosstalk 46-48 .
The enriched pathways established by the KEGG and Reactome analyses in the ASD cases identi ed major cellular pathways with therapeutic potential. The differential expression of central immune genes comprising cytokines, and CD cell markers (such as IL1B, IL6, CD80, CD40), support the reports of dysregulated cytokine levels in brains of individuals with ASD 25, 49 . Next, involvement of complement genes vital in phagocytosis (C1QB, C1QC, C1R, C1S), which play a central role in immunity, response to infection as well as synaptic pruning, further implicate the involvement of the immune system in ASD [50][51][52] . In addition, the enrichment of histone subunits fundamental to gene expression and epigenetic regulation (H3C13, H3C7, H2BC11, H2BC3), supports the concept of potential association between epigenetic regulation and in ammation 53 .
Analysis of the TS differentially expressed genes using GO identi ed numerous enriched immune response and in ammatory signalling terms. The enriched pathways highlighted by the KEGG and Reactome analyses in TS identi ed up-regulated DEGs involved in the immune response such as cytokine signalling (IL1B, CXCL10, TNF, CCL2) 24 . In addition, pathways involving genes within major histocompatibility complexes II (i.e., ICAM1, HLA-DRB1, HLA-DOA) and the complement system (i.e., complement components C3, C1, and complement factor B) were enriched. These ndings were similarly observed in the original analysis of these TS cases 28 .
Given the substantial comorbidity and overlap between NDDs, we identi ed genes and pathways common to both ASD and TS. We identi ed 133 common DEGs, ve of which were determined hub genes: CSF2RB, HCK, HCLS1, LCP2, and PLEK, which were all up-regulated in both disorders. Interestingly, the rst four genes are either tyrosine kinases or associated with tyrosine kinase activity, and the fth is a protein kinase substrate, which are key regulatory proteins involved in cell signalling, and are therapeutically targetable 54,55 . The use of tyrosine kinase modulators in many oncologic diseases is well established, however these agents have also shown promise in preclinical models of non-oncologic neurological disorders such as Alzheimer's disease and Parkinson's disease, where in ammation and microglia are central to the pathogenesis 55-57 .
Our investigation has con rmed immune and in ammatory pathways are commonly enriched by upregulated genes in ASD and TS. To further explore these intersecting ndings, the 133 genes common to ASD and TS were analysed separately, which repeatedly identi ed enriched in ammatory pathways involving cytokine signalling and the complement system. These pathways involved immune genes (IL1B, ICAM1, and JAK3) and genes of the complement system (C1QB, C1QC, C4B), the latter of which is particularly relevant to NDDs, due to the importance of complement in neurodevelopmental processes such as neurogenesis and synaptic pruning 58 . We utilised this approach as it allowed for comparison of the same genes within both disorders, while employing the distinct p values from each analysis, offering insight into the strength of each disorder's signal.
Our current study identi ed commonly enriched in ammatory pathways, however, several questions regarding the involvement of the immune response in ASD and TS remain unanswered. The cause of the identi ed in ammatory signals is still ambiguous, in addition to its nature. Research investigating the source of in ammation in NDDs has suggested it is an environmental or secondary component, rather than genetic 28, 42 . In particular, the in uence of MIA, which could create a neuroin ammatory environment in offspring, may alter immune signalling pathways and epigenetic control of cell function during the critical periods of development 16 . In addition, the identi ed in ammatory signal might be casual and pathogenic, or alternatively reactive or protective in origin, which cannot be deduced from the current investigation. Further functional and mechanistic explorations of tissue from individuals with NDDs might elucidate the nature of this in ammation.
Despite our ndings, this study has a number of caveats. Firstly, our analysis involved different brain regions from the two disorders, prefrontal cortex for ASD, and caudate and putamen for TS, as corresponding brain region data was not available for the two disorders at the time of analysis.
Secondly, the majority of the samples within the two datasets were not children, as cohorts of paediatric post-mortem brain samples are scarce. Therefore, our analysis represents late-stage disease, and it is unclear if the ndings will be re ected in younger cohorts. It is not known whether the in ammatory signal seen in ASD and TS accumulates over the course of life, or is present in childhood.
Finally, the approach taken to identify differentially expressed genes differs from other statistical analysis often used in DGE analysis. Most cases of ASD and TS, along with other NDDs, are understood to involve the accumulation of many common risk variants in converging biological pathways, therefore we analysed the DGE of the top 1000 genes rather than employing a stringent statistical cut-off in order to unravel a wide range of genes. Although this approach may identify false positives where individual genes are not statistically dysregulated, it improves the cumulative power of pathway analysis, enabling many genes with small changes to be included.

Conclusions
The results from this study con rm the presence of in ammation and involvement of a dysregulated immune response within post-mortem brain of individuals with ASD and TS. Our ndings bring new evidence of shared dysregulation of immune response and in ammatory pathways in NDDs and indicates a role for in ammation as an important environmental factor in the expression of these disorders. Findings from this work, when considered within the broad literature, provide a rationale to treat in ammation. The implicated pathways involved in in ammation, cell signalling including tyrosine kinases, and epigenetic machinery, may open doors for treatment through drug repurposing of monoclonal antibodies, kinase inhibitors and epigenetic modulators such as HDAC inhibitors, respectively. Approval to obtain the RNA-sequencing datasets of autism spectrum disorders and Tourette syndrome were obtained with author's permission, data was accessed from synapse.org.

Consent for publication
Not applicable.

Availability of data and materials
The autism spectrum disorders dataset was downloaded from synapse.org (ID: syn8234507). The Tourette syndrome dataset was downloaded from synapse.org (ID: syn3158906). The code used to analyse these datasets can be found at https://github.com/sarahalshammery/ASDTS.

Competing interests
The authors declare that they have no competing interest.

Funding
Financial support for the study was granted by the Petre Foundation, Brain Foundation, NHMRC Investigator Grant (APP1193648).
Authors' contributions SA analysed, interpreted, and wrote the results of this investigation. SP, HFJ, VXH, WAG and RCD assisted in the interpretation and writing of the results. BSG assisted in the analysis and interpretation of the results. All authors read and approved the nal manuscript. Largely unclear -reported to mediate protein kinase C family, plays a role in phagocytosis and cytoskeletal organisation.   Protein network of common differentially expressed genes and expression of hub genes in ASD and Tourette syndrome. A) Protein-protein-interaction (PPI) network of genes found to be commonly differentially expressed in autism spectrum disorders (ASD) and Tourette syndrome (TS). The network consists of nodes (circles) and edges (lines) representative of predicted functional associations of the common protein-coding genes. Edge thickness is indicative of the strength of predicted evidence. Panels